What this is. Every Sunday we sort the 60 stocks and ETFs we cover into trend regimes and publish the three numbers that decide each trade: entry, invalidation, target. You have them before Monday opens, so the decision is made in advance and in writing rather than in the moment. This is a risk discipline first and a stock-picking service second.

What to expect. When a beaten-down market snaps back hard we will lag the names that gapped off the bottom, because the rules wait for a reclaim. In a bad tape that wait is the whole point. You do not get to skip the drawdown and catch the first bounce with one rule, and anyone selling you a framework that does both is selling hindsight.

How to check us. Every level we publish is marked afterwards, win or lose, in the scorecard at the end of each issue, along with the running record against the S&P. Nothing gets quietly dropped.

If you already know the framework, skip to This Week’s Playbook. If a level does not sit right with you, leave a comment: we read every one, and they shape what gets covered.

Market Conditions

The broad tape softened this week. The S&P 500 (SPY) closed at 765.72, Trend-Hold and down 1.4% on the week, while the Nasdaq-100 (QQQ) at 713.44 is Transitional and down 2.4% on the week. Beneath the surface, regime breadth deteriorated: Breakdowns rose from 7 to 8, Repairs rose from 5 to 9, and Trend-Holds fell from 30 to 20. Regime downgrades outnumbered upgrades 21 to 9, a clear risk-off tilt.

The selling concentrated in NBIS -21%, CRWV -17%, INTC -12%, RDW -11% on the downside, while XRPI +37%, ETHA +29%, COPX +10%, SLV +7% held up against the tape.

Fresh damage showed up in BABA, COST, CRWV and NOC, among others, dropping to Breakdown. The cleanest signal was a full reclaim of Trend-Hold in ADBE, ETHA, LLY and SLV. This reads as a defensive rotation rather than a systemic stress event, but it changes what a defined entry looks like.

With trends rolling over, most pullback prices now sit too close to the level that says the idea is wrong, so the triggers that survive are mostly on strength rather than on weakness.

What Changed From Last Week

• Reclaimed Trend-Hold: ADBE, ETHA, LLY, SLV. The exception in a week of downgrades, and worth more for it. The weekly chart still has to hold what the daily just reclaimed.

• Lifted out of Breakdown into Transitional: PALL, XRPI. Early bounces that still need to prove themselves.

• Up a step from Repair into Transitional: XRH0.L. Basing turned into an actual recovery attempt.

• First repair off the lows: LULU, NVO. Damaged names showing the first sign of basing.

• Cooled out of Trend-Hold into Transitional: AMD, AMZN, ASML, ITA, LMT, LRCX, NBIS, NVDA, PANW, PINS, QQQ, RTX. Leaders that lost their trend and now need to re-prove it.

• Lost the trend, into Repair: AIXA.DE, AVGO, INTC, ORCL. Downgraded a full step, treat these as damaged until reclaimed.

• Broke down: BABA, COST, CRWV, NOC, WMT. Fresh damage, the framework moves these to the avoid list. CRWV and NOC went from Trend-Hold to Breakdown in a single week, which is a three-regime fall and the sharpest kind of break the framework registers.

How Our Trade Plans Work

Each name is classified into a regime (Trend-Hold, Transitional, Repair or Breakdown) from the 1D EMA stack, cross-checked against the 1W chart, and every level traces back to a specific indicator. T1 is the first profit-take, T2 the structural objective, T3 the stretch target. Regime sets how much risk a name is allowed and when: Trend-Hold earns full risk but only on a pullback to support or a confirmed breakout; Transitional gets half or nothing; Repair gets a starter only after a reclaim; Breakdown gets none until a reclaim confirms.

Two stops, two jobs. The trade stop is the tactical level where a single trade is wrong and you cut it. The regime invalidation is the deeper level where the classification flips and the reason to own the name is gone. One manages risk, the other manages conviction.

Size by risk, not by dollars. Stop distances in this issue run from 2% to 35%, so the same dollar amount in two names can mean risking 19 times as much on one as the other. Decide what a single losing trade may cost you, say half a percent of the account, and divide it by the distance from entry to stop. A 2% stop carries 19 times the position of a 35% stop for identical risk. The wide-stop names are not off limits; they are simply small.

This update ranks 60 names. You are not meant to trade all of them, or even most of them. Most weeks a handful sit at a level worth acting on and the rest are there so you can check our work.

The Week Ahead

These names in the universe report during the plan week:

• Wednesday: NVDA

Earnings are binary and the framework does not trade into them. If one of these sits at a defined entry, either take it at reduced size or wait for the print and re-read the structure afterwards. A level that was valid on Friday is not wrong after a gap, it is simply no longer the level.

That one matters more than the rest of the list, because you have a live instruction on it this week: NVDA is tagged “Wait for breakout”. Take the print as the risk event it is, size for it, or stand aside until it is out.

This Week’s Playbook

Start here. Detailed setups for the top names follow; the watchlist covers the rest.

• Closest to a defined entry: TSM, SPY, DG, RDW, XOVR, MSFT. SPY can be taken on a dip. The rest have no pullback that clears their stop, so their only trigger is on strength above Friday’s close.

• Best leaders to buy only on weakness: PLTR, RBRK, XOM, OSCR, LLY. Clean trends with room to run, but wait for the pullback.

• Most dangerous chase setups: ETHA, TGT. Overbought leaders; the trend is real but the entry is not, wait for a reset.

• Avoid until a reclaim confirms: BABA, BIDU, COST, CRWV, META, NKE, NOC, WMT. No long exposure; not every dip is a setup, and cash is the position.

• Where the payoff justifies the stop. A wide stop is not a bad setup, it is a small one, but some names are not paying enough to be worth any size at all.

• Not paying for the risk, quarter size or skip: AMAT (16% stop for 1% to T1), XRH0.L (12% stop for 1% to T1), NFLX (14% stop for 1% to T1), AMD (13% stop for 1% to T1)

- No usable room between entry and stop, skip entirely: NBIS (0.4 ATR), ADBE (0.5 ATR), ASML (0.8 ATR), LRCX (0.9 ATR), GLD (1.0 ATR). A fill at the bottom of these bands is already at the exit.

Top 5 Actionable Setups

The highest-conviction names sitting closest to a defined entry, with full detail, levels, and both stops for each. The next tier follows in the Also Actionable table, and the rest of the universe in the full watchlist below it.

Taiwan Semiconductor Manufacturing Co. (TSM), Last close: 418.95

Setup: Taiwan Semiconductor Manufacturing Co. held its Trend-Hold, little changed on the week. Full Trend-Hold on both the daily and weekly, the cleanest structure in the framework, with first support at the 1D EMA20 (418.06). Daily RSI 50.4, MACD bullish fading; the trend is intact but uncommitted, favor a defined pullback.

• Market view: the foundry anchoring the AI supply chain, the most resilient chip structure; tracking the broad tape this week.

• Entries:

Read more

Key Takeaways

• Inside its own 52-week range, copper sits at 88 out of 100, gold at 53 and silver at 36. 3 metals, 3 completely different starting points.

• Published 2027 forecasts put gold 11% to 24% above the Thursday settle, silver 27% below to 40% above, and copper 20% below to 10% above.

• The Warsh Fed has removed forward guidance and 9 of 18 FOMC members now pencil in a rate rise this year. Jackson Hole, which opens Thursday, brings his first speech as chair.

• GLD and SLV closed 0.91 and 0.94 of the way up their volatility bands on Thursday and have run further since. COPX is the least extended of the 3.

• COPX holds copper miners, not copper. Its ratio to the metal ranged 10.3 to 15.9 over the past year.

A note on timing. Every level below is the Thursday August 20 settle, the last completed session, since Friday is still trading. Where Friday matters, it’s flagged as intraday.

Start with what’s actually driving them.

The Variable, and the Exception

Every argument about gold and silver runs through the real yield, and the real yield runs through Kevin Warsh. He took office as Fed chair on May 22 and used his first meeting on June 17 to hold at 3.50% to 3.75% while making the direction of travel clear. 9 of 18 policymakers now project at least 1 rate rise in 2026, against none in March. He has put noticeably more rhetorical weight on inflation than his predecessor, though the Fed’s July language still describes its commitment to both price stability and maximum employment as unwavering. Core PCE, the measure the Fed actually targets, ran 3.3% in June against 3.7% headline.

The removal of forward guidance matters more than it sounds. A Fed that telegraphs its intentions makes speeches boring. A Fed that doesn’t makes them tradeable. Warsh speaks at Jackson Hole, which opens Thursday, his first as chair, with no guidance to anchor expectations against.

The move in the precious pair came from the other side of that variable. On August 19 the Treasury announced it would at least double its long-dated buybacks, from $2 billion to $4 billion an operation between September 9 and November 4, aimed at the 10 to 30 year part of the curve, which has seen a buyers’ strike since June. The 10-year fell 5 basis points to 4.647% and the 30-year dropped 9 to 5.196%. Gold settled up 4.1% at $4,512 that day, silver up 6.4% at $66.97.

Then it unwound. Yields were higher the next session and the 10-year now sits near 4.72%, above where it closed on the announcement. Gold and silver kept their gain anyway and extended it. Read that as conviction or as a crowd that hasn’t marked itself back. Either way the trigger lasted about 24 hours and the price move didn’t.

Copper is a separate case moving at the same time, pulled physically rather than monetarily: LME spot traded as much as $545 a tonne above the 3-month contract in mid-August, the widest since the 2021 squeeze, on falling exchange stocks. That’s a signal about warehouses today, not a view on rates, and it explains why copper sits near the top of its range while silver sits near the bottom.

Where the forecasts point.

What Published Forecasts Imply

One caveat. These are published 2027 forecasts but not all the same animal: some are annual averages, some dated targets, some scenarios. The diamond marks the single most-cited central figure, not a computed mean. Treat the ranges as a map of disagreement rather than a distribution.

Gold is the only one of the 3 whose 2027 range sits entirely above the current price. UBS maps a path of $4,600 by December, $5,000 by March 2027 and $5,200 by June. The 2027 cluster runs $5,000 to $5,600, with JPMorgan and UBS at $5,400, Goldman $5,400 to $5,600 and Westpac at $5,000. Bank of America carries an $8,000 scenario explicitly not its base case. UBS still expects central banks to buy 750 to 1,000 tonnes a year. UBS’s own September figure, $4,400, is below spot, so even the constructive house sees the next leg lower before higher.

Silver has no dated waypoints, only annual averages, and they disagree violently. JPMorgan models a 2027 average of $63.90, below today. Commerzbank targets $80 by end-2027. Citi carries $95 on the upside and $50 on the downside. The physical story is better than the price and moving the right way: the World Silver Survey published in April puts the 2026 deficit at 46.3 million ounces, 15% wider than 2025’s 40.3 million, a sixth straight shortfall that has pulled 762 million ounces from above-ground stocks since 2021. Total supply is forecast down about 2%, with mine production essentially flat at minus 0.3% and recycling up 7%, so the tightening comes from the balance, not from mines closing.

What makes silver interesting isn’t the forecast though, it’s the ownership. Net speculative length was 23,646 contracts against 115,127 of open interest on August 11, about 20% of the total, with only 12 of the past 60 weeks lower. The physical squeeze that spiked the 1-month lease rate toward 39% was back in October 2025; silver made its $121 record in January, then gave up more than 40%. Both the squeeze and the crowd have gone. The deficit hasn’t.

Copper is where consensus conflicts most sharply with current price strength. The 2027 analyst poll averages $13,059 a tonne and Bank of America $13,501, both below the $14,111.50 LME 3-month traded at this week. Cochilco sits at $5.10 a pound, about $11,244, and Macquarie models a floor near $11,000 by the third quarter of 2027. But UBS runs the other way at $15,500 by June 2027, and BMI has the balance flipping to a 66,000-tonne deficit from 2027, widening every year after.

One wrinkle before comparing anything. COMEX copper trades at a premium to LME, roughly $365 a tonne, because US buyers are pricing a Commerce recommendation to phase in duties of 15% in 2027 and 30% in 2028. That tariff is recommended, not enacted. COMEX at $6.46 a pound is $14,242 a tonne against LME’s $14,111.50, so quoting the wrong one misjudges the gap by a couple of points.

How to trade it.

The Setups

GLD holds physical gold bullion, SLV physical silver, and COPX copper mining equities rather than the metal. Gold is about GLD multiplied by 10.89 and silver about SLV multiplied by 1.107, using a 20-day median because the futures-to-fund relationship wobbles more than 1% day to day, so metal equivalents are approximate. The daily plots show the 20-day and 50-day averages; the 100-day and 200-day are quoted below. The risk-reward panel measures the breakout trigger to the final objective against the stop.

GLD, $415.26 on Thursday (gold $4,516). Intraday Friday near $422.

Read more

The market has priced it as marketing software for years. Management now says it’s AI infrastructure, and it put that case in writing this month, closing on the line that the gap between those 2 views is “the entire repositioning opportunity.”

There’s real support for the claim. OpenAI, Palantir, Snowflake and Gap all signed on inside 6 months, and retention is the best the company has ever posted.

There’s also a test it hasn’t passed. Infrastructure businesses earn infrastructure margins, and Zeta’s gross margin is 59.5%, and it’s falling, not rising.

The harder problem is that the market has already granted most of the rerating anyway. At $27.85 the stock trades 37% above what Zeta’s own valuation slide said it was worth 3 weeks ago.

You’re not early to this one. You’re paying for the part that hasn’t happened yet.

Key Takeaways

• The quarter was strong on every headline. Revenue of $442.8M grew 43.5%. Adjusted EBITDA of $91.7M grew 56%. Free cash flow of $58.0M grew 73%. It was the 20th consecutive beat-and-raise.

• The underlying growth rate is slowing. Strip out the Marigold acquisition and political money and growth runs 29%, 28%, 23%, then 20% across 2026, on management’s own guide. The 39% headline is acquisition math.

• The gap between the 2 earnings figures is stock. GAAP earnings were $0.03 a share. Adjusted were $0.21. The difference is $52.1M of stock compensation, which is 11.8% of revenue.

• Gross margin is the weak point in the infrastructure claim. At 59.5% it sits below every company in the cohort Zeta names, and it’s fallen from roughly 62% a year ago.

• The company’s own valuation argument is largely spent. Its slide showing a 5% free-cash-flow yield against a 2% cohort implied about $20.34 a share. The stock is 37% above that.

• A securities case just cleared its first hurdle. On July 8 a federal court denied the motion to dismiss, so the matter proceeds to discovery.

Subscribe now

About the company

sells enterprises a single system for finding customers, keeping them, and measuring both. The asset underneath is an identity graph covering 535M individuals, built over 20 years from its own email infrastructure, its ad exchanges, and the Disqus and LiveIntent publisher networks. Athena is the layer on top, generally available since Q1, and it lets customers query that graph by asking rather than by building reports.

Trailing revenue is $1.57B and the market values the business at $6.99B. It recorded its first GAAP profit in June 2026 and still carries an accumulated deficit of $1.06B.

Q2 2026 overview

The AI adoption data is the real news

• Most companies claim AI traction. Zeta put numbers on it.

• More than 40% of super-scaled customers were monthly active on Athena within roughly 130 days of its enterprise launch.

• Customers who’ve adopted broadly across the product are about 20% of the base and roughly 70% of revenue.

• Those same customers grow about 4x faster than customers early in adoption, and their retention is 4% above the company average.

• Average super-scaled tenure has stretched to 56 months from 48.

Cross-selling backs it up. Customers using more than 1 use case rose 90%, those using 5 or more channels rose 50%, and cross-sell deals won rose 43%.

Here’s the operating leverage: quota-carrying salespeople increased by around 0.5% (1 person), while pipeline rose more than 60%. Management also said 89.6% of new code was machine-generated in the quarter, up from 75% in Q1, with internal AI costs under 1% of revenue.

Partnerships

Four new partnerships happened within 6 months:

1. OpenAI models power Athena’s voice and Zeta now works with OpenAI’s advertising operations, which management called the most instrumental partnership in its history.

2. Snowflake put the Zeta platform inside more than 100 customer environments.

3. Palantir rebuilt Zeta’s Data Cloud on Foundry under a 7-year agreement.

4. named Zeta its system of record.

That’s real validation, but very much forward-looking. Very little of it sits in the $1.57B of trailing revenue, which is where the narrative is still unsubstantiated.

Adjustments to consider

Reported growth of 43.5% is flattered by Marigold’s Enterprise Business, which added $48.1M in the quarter and roughly $199M across the year. Take it out, along with political money, and the picture inverts: 29% in Q1, 28% in Q2, a guided 23% for Q3 and 20% for Q4.

That’s a business decelerating to 20% while the headline says 39%. Although both numbers are true, only one describes the operational engine.

Subscribe now

Fundamental Overview

Revenue Segmentation

Reportedly, revenue is broken down as follows:

• Channel:

  • Direct platform revenue, meaning work that runs entirely on channels Zeta owns, was 72% of the mix, down from 75% a year ago. Integrated revenue, which leans on third-party channels and carries thinner economics, took the other 28%.

• Geography:

  • The US brought in $402.6M, up 34.9%. International did $40.2M versus previous $10.0M, and that near-quadrupling is almost entirely Marigold, which brought offices across the UK, Denmark, the Netherlands, France and Germany. International is now 9.1% of revenue (previously 3.2%). Zeta became a multinational business in 2 quarters, by purchase rather than by build.

• Source of increase:

  • Of the $134.3M of added revenue, $75.9M came from new customers and $58.4M from existing ones.

• Vertical:

  • Clients spread across 15 industries, with Consumer and Retail largest at 24% of 2025 revenue, then Travel and Hospitality and Insurance at 11% each and Telecom at 10%.

  • 8 of the top 10 grew more than 20% this quarter.

  • Concentration risk sits at the customer level rather than the sector level: 1 customer is more than 10% of revenue and of receivables.

Margin Breakdown

Cost of revenues rose 54.7%. Gross profit came in at $261.8M for a 59.1% margin, down from 62.1% a year ago, a fall of almost 3%.

The driver? $54.2M of incremental media costs, plus $5.9M more in people and $3.9M in technology. Media cost is what you pay to reach an audience through somebody else’s inventory, and it scales with the integrated business.

Below the gross, the things look better:

• Selling and marketing grew 20.4% to $104.0M, less than half the rate of revenue.

• General and administrative grew 22.1% to $75.9M, driven mostly by $14.5M more in technology and infrastructure.

• Research and development grew 38.1% to $42.2M, which is the right place to be spending.

• Depreciation and amortisation grew 30.2% to $22.7M on the acquired intangibles.

Total OPEX grew 35.8% against revenue growth of 43.5%, and that spread is the operating leverage. It carried operating income to a positive $16.9M against a $5.1M loss a year ago.

So these two forces run against each other: gross margin erodes on mix, operating margin expands because the cost base below it grows slower than sales.

Cash Utilization

Adjusted EBITDA reached $91.7M at a 20.7% margin, up 170 basis points. Free cash flow hit $58.0M at a 13.1% margin, up 220 basis points, converting 63% of EBITDA into cash.

Across the first half, operating cash flow was $118.9M against $76.8M, and note how it was produced: a net loss of $5.1M more than offset by $152.6M of non-cash addbacks, of which $105.1M is stock compensation. That’s real cash in the door. It’s also cash that exists because a large slice of pay is settled in shares rather than dollars.

Investing used $70.9M, mostly $50.8M of acquisition payments plus $12.3M of software development and $7.8M of capital spending. Financing used $58.6M, almost all of it $55.6M of share repurchases.

Net revenue retention reached 120% in 2025, its best on record, meaning existing customers spent 20% more than the year before. Management raised the full year across the board, to $1.818B of revenue, $405.2M of EBITDA and $255.3M of free cash flow. The cash flow raise outran the revenue raise, $20.3M against $33M.

Stock Compensation

Stock compensation ran $52.1M in the quarter, 11.8% of revenue, and $105.1M across 6 months. That single line is why GAAP earnings were $0.03 and adjusted were $0.21.

It isn’t spread evenly. Stock is 22.5% of the selling and marketing line and 29.4% of research and development. Those are the 2 functions the growth story depends on, and roughly a quarter of what they cost is paid in equity rather than cash.

The forward load is heavier: $381.4M still unrecognised over about 1.8 years, front-weighted at $95.0M for the rest of 2026 and $116.9M in 2027. And in March the board granted a further nearly 11 million performance units, payable up to 185% on full achievement, against revenue targets running from $2.0B to $3.0B through 2030.

Add it up. Unvested restricted stock, options and performance units total roughly 42.7M shares against 251.0M outstanding, about 17%, before that 185% multiplier. There’s a further $50M of Class A stock committed under a long-term partnership agreement sitting in the purchase obligations note.

The buyback runs in parallel. Zeta repurchased 3.2M shares for $55.6M in the first half at an average of about $17.40, including some at $14.59 in April.

Balance Sheet

Cash is $310.0M against $197.5M of borrowings, a net cash position of about $112.5M. The current ratio is 2.38 and working capital is $393.2M. In July it refinanced into a $1.0B facility, a $250M term loan and a $750M revolver that sits entirely undrawn.

Two things to note on the assets:

1. Goodwill and intangibles total $710.1M, or 48% of total assets. That’s the accumulated price of buying growth. Intangibles amortise on a 4.55-year life, so $29.5M more hits GAAP earnings this year and $50.3M in 2027 whether the acquired business performs or not.

2. Remaining performance obligations are $218.7M for the next 12 months. Set that against $1.818B of guided 2026 revenue and only about 12% of next year’s revenue is under contract today. Deferred revenue is $33.9M, under 2% of the annual run rate.

That’s the number that argues hardest against the infrastructure framing.

Infrastructure companies book long contracts and carry large backlogs.

Zeta’s revenue is consumption-based and re-earned every quarter. It has been re-earned 20 quarters running, which is a real achievement, but earned is not owed.

Relative valuation

At $6.99B you’re paying 3.85x guided 2026 sales, 17.3x guided EBITDA, and a 3.65% forward free-cash-flow yield. Against Palantir at 153x EBITDA or Datadog at 78x forward earnings, that’s cheap. Against The Trade Desk, which now trades at 7.5x EBITDA after its growth stalled, it isn’t.

Fundamental conclusion: the operating business is inflecting and the proof is in the adoption data. First GAAP profit, 120% retention, 63% cash conversion, operating costs growing 8% slower than revenue, and a sales force that isn’t growing are all real. On the flipside, underlying growth falling to 20%, gross margin below every named peer and still sliding on media mix, 11.8% of revenue paid in stock, only 12% of next year’s revenue under contract, and a valuation that has already absorbed most of the rerating. At 3.85x sales you aren’t overpaying for a 25% grower. You’re paying in full for the infrastructure story to work.

Subscribe now

Technical Overview

Weekly chart

The long-term picture is a repair that has completed.

Price sits above the 20-week EMA at $21.73, the 50-week at $19.78 and the 200-week at $16.08. Those are averages of 20, 50 and 200 weekly closes, so roughly 5 months, 1 year and 4 years of price. Having all 3 beneath you in that order is what a restored uptrend looks like.

Weekly RSI is 65.1, firm and not yet stretched. Weekly ADX is 24.0, which measures how strong a trend is rather than its direction, and a reading in the mid-20s says the weekly trend is real but still building rather than exhausted.

The context matters. This stock hit an all-time high of $38.20 in November 2024, spent 16 months breaking down, bottomed at $14.36 in March and has doubled since. The weekly Kijun, a midpoint of the last 26 weeks that often acts as the trend’s floor, sits at $22.13.

So the weekly says the 2024 damage is undone and the primary trend is up. It also says the old high is still 37% above here, which is a lot of trapped supply to clear.

Daily chart

The daily is key for timing, and it’s cleaner than the weekly. Price at $27.85 sits above every daily moving average: the 20-day at $26.23, the 50-day at $23.51, the 100-day at $21.58 and the 200-day at $20.16. Bullishly stacked.

• ADX at 38.0 confirms a strong trend, meaning buyers hold clear control and sellers are barely pushing back.

• RSI is 61.4, which is firm, but still has room before it’s overbought.

• MACD is flattening, at 1.99 against a 1.88 signal line. Momentum is still positive but no longer accelerating, which usually means consolidation rather than reversal.

• ATR is $1.47, about 5.3% of price. moves more in a day than most move in a week.

Since mid-April, 39% of sessions have traded down to the 20-day at some point. This is a trending stock that still offers entries.

Technical conclusion: a clean Trend-Hold on both timeframes, and nothing in the price action argues against it. The weekly gives you permission to own it, the daily gives you a defined place to buy. The risk isn’t the chart, it’s the entry. Price sits 6.8% under a high set 9 days ago after a 94% run, so paying up here means buying the top of a 7-session range with no catalyst to break it until November.

Share

The Trade Plan

Zeta is consolidating and currently sits above every moving average that matters.

That makes this a trend-continuation setup rather than a catalyst trade. It also means the honest advice is that you’re being paid to be patient. The 20-day is rising toward price at the same time price is resting, and those 2 lines converge somewhere around $26.50 within a couple of weeks.

• Entry zone: $26.20 to $27.

  • This is the rising 20-day at $26.23 and the lower half of the post-earnings range.

• Stop: a daily close below $24.20.

  • That sits just under the pre-gap close of $24.26 and the daily Kijun at $24.48. Losing it means the entire earnings move has been given back, which is a clean thesis break rather than noise.

• Targets:

  • T1 is $29.90, the August 12 high and the 52-week high.

  • T2 is $32.30, where the September 2024 shelf at $32.11 meets the upper Bollinger band at $32.41.

    T3 is $34.10, the October 2024 high. 27.7% from entry, or 3.0 to 1.

What the valuation changes about this trade.

This stock may be fully priced objectively, but that isn’t necessarily a reason to avoid it. It is, however, a reason to know which trade you’re in.

• Size smaller than usual. There’s no valuation cushion underneath. A stock at 4x sales that disappoints re-rates to 3x, and that’s 22% lower before any support level gets a say.

• The stop carries more weight here. $24.20 isn’t a suggestion, it’s the entire risk framework. In a cheap stock you can average down. In this one you can’t, and 3x sales sits at $21.73 which is also the 20-week EMA.

• Take the targets as they come. T2 at $32.30 is 4.46x sales and 20x EBITDA for a business guiding 25% underlying growth. If you get there, take something off rather than holding out for $38.20 on the story.

If you’re buying to hold for years rather than to trade the trend, the honest answer is different. You’d be paying today for a repositioning the margins haven’t earned yet, and your entry is nearer $24 than $27.

If it breaks out instead of pulling back, a daily close above $29.90 clears the range and puts T2 in play directly. Don’t buy the first day through. Let it retest $29.90 and hold, because failed breakouts from tight ranges are common and the retest costs you almost nothing.

Subscribe now

Bottom line

Zeta is doing the hard part. The adoption data is specific and checkable, retention is the best it has posted, and it’s generating 60% more pipeline without adding salespeople.

What it hasn’t done is earn infrastructure economics. Gross margin is the lowest in the cohort it chose for itself and it’s falling on media mix, underlying growth goes to 20% by Q4, only 12% of next year’s revenue is under contract, and 11.8% of every revenue dollar goes out as stock.

The market has already granted most of the rerating the company asked for. Its own fair-value math pointed at $20 and the stock is at $27.85. That doesn’t make it expensive. It makes the easy money gone.

One number decides it: $24.20. Above it this is a strong uptrend with a real business behind it and a defined place to buy. Below it the earnings move is fully retraced, and you’re holding a 25% grower at 4x sales.

All content provided is for informational and educational purposes only and does not constitute investment advice, a recommendation, or an offer to buy or sell any security. The trade plans, levels, and scenarios discussed are illustrative frameworks based on market structure and are not guarantees of performance. Markets involve risk, and losses are possible. Past performance is not indicative of future results. Each reader is responsible for their own investment decisions, position sizing, and risk management.

At $543.67 Meta trades at roughly 16 times forward earnings, 31% below its high, with 62 analysts covering it and not one carrying a sell rating. The lowest target on the Street, $580, sits above the current price. On the multiple everyone uses, this is the cheapest the stock has looked in years.

The problem is the denominator: reported earnings rest on one unusually judgment-sensitive input, and that input is about to absorb the largest capital program in the company’s history. Advertising is strong. Free cash flow is not. And the distance between those 2 facts is an assumption about how long a graphics chip stays useful.

Key Takeaways

• Revenue grew 28% to $60.80 billion and advertising 27%, so the demand side of this business isn’t the issue. Operating margin fell from 43% to 31%.

• Free cash flow fell 91% to $784 million as capital spending of $31.08 billion consumed almost all operating cash. Meta guided 2026 capex to $130 to $145 billion against $72.2 billion in 2025.

• Meta extended the useful life of servers from 4 years to 5.5 in 2025, cutting depreciation by $2.9 billion. Amazon moved the opposite way, shortening its schedule.

• Meta has now gone 3 consecutive quarters without repurchasing a share, leaving stock compensation unhedged and the share count drifting up.

• A bellwether trial opened in Oakland on August 18, the first phase of litigation brought by 29 states. Damages are put near $200 billion; the $1.4 trillion figure is Meta’s own. The product remedies sought may matter more.

Start with what’s working.

The Advertising Business Isn’t the Problem

This isn’t a broken company and nothing below should be read that way. June-quarter revenue was $60.80 billion, up 28%, a figure the filings and aggregator feeds confirm independently. Advertising was $59.4 billion and grew 27%, with impressions up 14% and average price per ad up 12%, so volume and pricing are contributing together. Management guided September to $61 to $64 billion.

What changed is everything below the revenue line. Total costs rose 55%, research and development up 67% to $21.66 billion. Operating income fell 8% to $18.77 billion, the margin went from 43% a year ago to 31%, and net income fell 14% to $15.85 billion.

Be fair about the composition, because 2 items were quarter-specific: $2.4 billion of legal charges and $1.18 billion of severance tied to the May layoff of roughly 8,000 people. Strip both out and operating income would have risen 9%, putting the underlying margin near 37%. That’s still 6 points of compression, and those 6 points are the part that doesn’t reverse. Legal costs aren’t obviously a one-off here either. Reality Labs generated $431 million of revenue against a $4.62 billion operating loss.

Now for the denominator.

The Assumption Underneath the Multiple

Effective January 2025, Meta extended the estimated useful life of most servers and network assets from 4 years to 5.5. That revision reduced 2025 depreciation by $2.9 billion, flowing directly into reported profit. Management cited better hardware reliability and operational experience, a legitimate argument. Short sellers argued it overstated earnings by understating how fast a graphics chip becomes obsolete, also legitimate.

The reason to take the sceptical side seriously is that Amazon went the other direction. In February 2025 it shortened the useful life of a subset of servers and networking equipment from 6 years back to 5, explicitly citing the increased pace of technology development in AI. 2 hyperscalers facing the same acceleration in hardware economics reached opposite conclusions about useful lives. Both can be defensible, since fleets, workloads and upgrade cycles differ, but the divergence tells you how much judgment sits inside the number.

The argument isn’t that analysts missed the disclosure; they build forecasts off Meta’s stated policy. It’s narrower: if AI hardware loses economic usefulness faster than the accounting life assumes, reported and consensus earnings overstate sustainable economic earnings, and a multiple built on them is less cheap than it looks.

That would be a modest concern if the asset base were stable. It isn’t. Meta guided 2026 capital spending to $130 to $145 billion against $72.2 billion in 2025, close to a doubling. The outlook was raised from an original $115 to $135 billion to $125 to $145 billion in April, before the lower end was lifted again to $130 billion in July. Most of it will reach the income statement as depreciation once assets are placed in service, on varying schedules, landing on the very line the multiple divides by. The effect has started: depreciation and amortisation ran $6.36 billion in the June quarter against $4.34 billion a year earlier, up 46%, before most of the 2026 spending has arrived. Bank of America estimates capital spending will be a 1.6 point drag on operating margins this year against late 2024.

So the multiple looks cheap partly because the market hasn’t settled which company it’s buying: an advertising business with a temporary spending problem, or a capital-intensive infrastructure business never worth 25 times earnings.

Cash settles arguments that accounting can’t.

Cash Doesn’t Care What You Assume

Depreciation is an estimate. Cash isn’t, and the cash statement is where it gets uncomfortable.

Operating cash flow was $31.86 billion, up 25% from $25.56 billion a year ago, so the business is generating more cash, not less. Capital expenditure is what changed: $31.08 billion including finance-lease principal against roughly $17 billion a year earlier. That left free cash flow of $784 million on Meta’s own definition, against $8.55 billion a year ago, a 91% decline, with capex absorbing about 97% of the cash produced. You’ll see a second pair of figures quoted, since the statement line excludes finance-lease principal and gives $1.75 billion; the direction is identical either way.

Funding the gap has changed the balance sheet. Long-term debt rose to $83.66 billion from $58.74 billion at year-end, roughly $25 billion of it issued in the June quarter, against $90.26 billion of cash and securities.

That comparison understates what Meta has committed to, because the balance sheet doesn’t carry it. The June filing discloses about $279 billion of lease obligations for leases not yet commenced, covering data centres, colocations and network infrastructure beginning through 2036 on terms of up to 30 years. Separately it discloses $349 billion of non-cancelable contractual commitments, mostly third-party cloud capacity, servers, data centres and Reality Labs hardware, of which roughly $54 billion falls due this year and $82 billion next.

Those are 2 different categories and shouldn’t be added together or treated as debt. But they are contracted, enormous next to $83.66 billion of borrowings, and absent when you set that figure against $90.26 billion of cash. Hyperion in Louisiana shows the mechanism: Blue Owl funds own 80% of the venture and Meta 20%, so Meta isn’t the primary beneficiary under the accounting test and doesn’t consolidate it, having agreed an initial 4-year lease from 2029 with options extending to 20 years.

And the buyback is off. Meta has gone 3 consecutive quarters without repurchasing a share, having returned well over $10 billion a quarter that way as recently as early 2025. The $0.525 dividend continues, worth roughly $1.3 billion a quarter, but with repurchases at zero, share-based compensation is no longer being offset through buybacks, putting upward pressure on the share count. For a stock whose bull case is per-share earnings power, that direction matters.

The Trial Nobody Can Price

Opening arguments began in federal court in Oakland on August 18 before Judge Yvonne Gonzalez Rogers. Meta fell 4.4% that session on 1.6 times average volume while Pinterest rose 3% and Snap 2%, which suggests the selling was about Meta rather than advertising.

Be careful with the damages figure, because the widely quoted one isn’t the plaintiffs’. This is a bellwether phase with 4 states as plaintiffs, from litigation brought by 29 state attorneys general who argue Facebook and Instagram were designed to be addictive to minors and that children’s data was improperly collected. The states say damages could reach $200 billion, roughly Meta’s 2025 revenue. The $1.4 trillion number in circulation is Meta’s own characterization of what it could theoretically face. Both are ceilings rather than forecasts: in March a Los Angeles jury found Meta and Google liable in a separate case and awarded $6 million.

The money may not be the real exposure. The states are also seeking product remedies: bans on infinite scrolling, autoplay, like buttons and beauty filters for users under 18, and on engagement-optimised recommendation for minors. A financial penalty is a charge against a company generating roughly $60 billion of quarterly revenue and nearly $16 billion of quarterly net income. A court-ordered change to how ranking works for teenagers would reach the revenue line instead. The trial runs about 6 weeks; the 8-member jury is advisory and the judge decides.

What the price is doing.

The Price

Levels are based on the August 18 close of $543.67. Meta trades below every daily moving average, the 20-day at $587.82, the 50-day at $598.12 and the 200-day at $624.77, and closed below its lower volatility band. Sellers hold the directional readings decisively, though ADX at 13 says no trend has established itself, so this reads as sharp repricing rather than a mature downtrend. RSI at 36 is weak without being oversold, and the faster stochastic-RSI gauge near 55 is nowhere near a reset.

The level that matters is below: the 52-week low at $520.26, the 60-day low at $524.49 and the weekly 200-period average at $524.04 sit within $5 of each other.

A Trade Plan

Read more

Key Takeaways

• All 3 trade near 22x trailing and in single digits forward. The market is treating them as one AI-memory trade rather than as businesses with different exposures.

• Micron disclosed 16 take-or-pay agreements carrying about $100 billion of minimum revenue, backed by $22 billion of customer commitments including roughly $18 billion of cash already received.

• Sandisk disclosed $93.9 billion of minimum contracted revenue in floor-and-ceiling bands, backed by $16.5 billion, covering more than half of fiscal 2027 bits. Deeper coverage of its own book than Micron has.

• SK Hynix says it has closed long-term agreements with around 10 key customers, but hasn’t published the economics.

• On latest reported operating margin the order is Micron at 81.2%, Sandisk near 79% and SK Hynix at 76%, not the ranking most write-ups assume.

Start with what the market is saying.

The Market Says They’re One Trade

The convergence isn’t a coincidence. All 3 report the same phenomenon: DRAM contract prices rose about 90% in the first quarter and another 58% to 63% in the second, while enterprise NAND rose in the mid 50s in the first and 70% to 75% in the second. That’s the largest move in a decade, and it lands almost entirely in gross margin because the cost base barely moves.

Look at the shape rather than the levels. Micron’s GAAP gross margin went from 37.7% to 84.6% in 5 quarters, Sandisk’s from 22.5% to 78.3%. Neither invented a product in that window; Sandisk said 2 thirds of its sequential growth was pricing. This is a price event wearing an earnings costume, which is why the market refuses to capitalize it.

One caveat on those forward figures. They’re consensus as of the August 17 close and are being revised weekly against fast-moving prices; other providers currently put Sandisk nearer 8 times and SK Hynix in the high 6s. The precise number matters less than the pattern, which is single digits for all 3.

Now what they did about it.

All 3 Tried to Contract Away the Cycle

Between late June and early August, each told investors it had locked in future revenue. It’s the most important development in memory this year, and it’s been reported piecemeal.

Micron disclosed 16 strategic customer agreements on June 24, 14 carrying cumulative minimum revenue commitments of roughly $100 billion. Most run 5 years, they’re take-or-pay and can’t be cancelled, and customers put up over $22 billion of financial commitments including almost $18 billion of cash. They cover about 20% of DRAM volume and a third of NAND volume through calendar 2030.

Sandisk disclosed its own version on August 5: $93.9 billion of minimum contracted revenue across 8 customers, priced in bands with floors and ceilings, backed by $16.5 billion of cash deposits and financial instruments, with a weighted-average duration above 4 years. Supply and demand commitments are set annually and quarterly, and a customer that misses its quarterly obligation sees those guarantees convert into compensation. The agreements cover more than half its fiscal 2027 bits and roughly 2 thirds of fiscal 2028, deeper coverage of its own book than Micron has arranged.

SK Hynix said on July 29 it had finalised long-term agreements with around 10 key customers, most spanning 5 years, its president describing them as a way to reduce exposure to price swings. It also has a separately confirmed multi-year technology and supply partnership with Nvidia. It hasn’t published minimum revenue, coverage or guarantees.

So the popular framing, that one has visibility while the others are naked cycle bets, is wrong. What differs is the shape of the protection, not its existence.

What the contracts don’t cover.

What the Contracts Don’t Cover

None of this covers the whole book. Micron’s agreements leave roughly 80% of DRAM volume exposed. Sandisk’s pricing bands trade some upside for downside protection and revenue visibility, which is what a floor paired with a ceiling does. SK Hynix hasn’t given enough detail to judge.

That leaves the cycle argument, and it’s a real argument rather than a settled fact. TrendForce projected in late July that DRAM stays in structural shortage through 2028 with the gap widening in 2027, while NAND supply turns sufficient in the second half of 2027. Against that, Morgan Stanley expects the NAND crunch to persist through 2027, on the view that solid-state drives absorb work when HBM and DRAM run short. Kioxia says its own NAND is sold out probably through 2027. The bear case rests on consumer devices, where smartphone and notebook production are both forecast to fall this year.

What isn’t disputed is that pricing is decelerating. DRAM contract prices are projected to rise 13% to 18% in the third quarter against 58% to 63% in the second, NAND 10% to 15% against 70% to 75%. Prices are still climbing. The rate of change has collapsed. That’s why SK Hynix fell 11% on a record quarter and why Sandisk beat on both lines and still sold off.

There’s also a supply answer nobody controls. Micron alone is spending about $27 billion of capital this year, and China’s CXMT grew DRAM revenue 716% to 7% share, freshly public and expanding. All 3 are being paid today to build the capacity that ends the shortage.

Why the pick is Micron.

Why Micron

Micron’s May quarter did $41.46 billion of revenue, up 346% from $9.30 billion, at a record 84.9% non-GAAP gross margin and $25.11 per share. It guided the August quarter to $50 billion give or take $1 billion, roughly 86% gross margin and a record $31.00 per share. At $1,011.75 it carries a $1.14 trillion market value.

3 things separate it, and none is the contract book alone.

Margins first, because this gets stated backwards constantly. On reported operating margin, Micron’s fiscal third quarter was 80.4%, Sandisk’s most recent quarter about 78.5%, and SK Hynix 76%. Those aren’t perfectly comparable, since the first 2 are US GAAP and the third is a Korean filing, but the ranking survives either convention. Micron is the most profitable of the 3 on the latest numbers, not the second or third.

Second, breadth. DRAM was $31.3 billion of the quarter and NAND $9.9 billion, so flash is 24% of revenue rather than all of it, and Micron’s entire 2026 HBM output is committed. Whichever way the NAND argument resolves, Micron owns a diversified version of the answer. Counterpoint put its DRAM share at 25% in the second quarter against SK Hynix at 26% and Samsung at 39%, a single point where there used to be several.

Third, the form of the protection rather than its size. Both books are enforceable, but Micron’s disclosed backing is almost $18 billion of customer cash actually received, against Sandisk’s $16.5 billion of deposits and instruments. Cash in hand is the cleanest version of a commitment. Debt to equity is 0.06 and short interest 2.65% of float, the lowest of the 3. Consensus runs 41 buys, 5 holds and no sells across 46 analysts, though targets from $361 to $2,200 say nobody agrees what normalized earnings are.

On price, patience is warranted. Micron trades above every daily moving average, the 20-day at $914.61 and the 200-day at $626.18, but the faster stochastic-RSI gauge is pinned at its maximum, as overbought as that reading gets, and average true range of $72.16 is 7.1% of the price. The 60-day range runs from $732.20 in May to $1,255.00 on June 25. The next report lands September 23.

Why the other 2 rank behind.

Why the Other 2 Rank Behind

SK Hynix has the strongest competitive position in the business. It reportedly won about 70% of Nvidia’s HBM4 orders for the Vera Rubin generation, and Goldman expects it to hold more than half the HBM market. The share-loss story gets quoted too confidently: estimates vary by house and by whether share is counted in revenue or bits. Second-quarter revenue of 79.3 trillion won rose 257% and operating profit of 60.5 trillion won rose 557%.

What holds it back is disclosure and instrument. Its 76% operating margin is the lowest of the 3, it hasn’t quantified its long-term agreements, and its US listing is 27 sessions old.

SK Hynix listed on Nasdaq Global Select on July 10, raising $26.5 billion at $149 a share in the largest initial US share sale ever by a foreign company, each share representing a tenth of a Seoul-listed common share. It’s sponsored, SEC-registered and trades 43.5 million shares a day. But it went from $194.80 on July 14 to $124.80 on July 29, down 36% in 11 sessions, and back to $171.38. There’s no moving average worth computing, and the price is largely set in Seoul overnight.

Sandisk is a better business than its critics allow. Its fiscal fourth quarter, ended July 3, did $8.97 billion of revenue, up 372% on the year, with $43.97 per share and another $14 billion of buyback authorized, and its contract book genuinely de-risks fiscal 2027. What keeps it third is concentration and price. It’s single-segment, so the uncontracted half of its book sits on the one product where forecasters disagree about 2027, and its ceilings cap what it earns if the bulls are right. Short interest is 7.36% of float, nearly 3 times Micron’s, the next report isn’t until November 6, and it’s up 653% this year, closing above its upper volatility band with average true range at 9.5% of the price.

The Answer

The evaluation favors Micron. Not because it’s the only one with contracts, because it isn’t, and not because it’s the cheapest, because it isn’t that either. Micron ranks first in our assessment because it earns the highest margin of the 3 on the latest reported figures, spreads that across DRAM, HBM and NAND instead of one product, backs its contracts with $18 billion of customer cash already banked, and trades through an instrument with years of history rather than weeks.

The gap is narrower than the market’s similar multiples suggest. SK Hynix has the stronger franchise and the cheaper valuation, and greater disclosure around its contract economics could materially strengthen its case. Sandisk, meanwhile, has protected a larger share of its own book than either.

For us, that leaves Micron with the best overall balance of profitability, diversification, contractual visibility and investability among the 3 today.

One caution applies to all 3. Each is short-term overbought, and this cycle has already produced drawdowns of 36% and 44% in these names within a single month this summer. Identifying the strongest setup does not remove entry-price risk.

This is research and commentary, not personal investment advice. Levels discussed are illustrative; size positions to your own risk tolerance and time horizon. The author may hold positions in names discussed.

That news, landing alongside a $600 million settlement of a long-running US investigation and reported progress on access to Nvidia silicon, helped produce an 11% single-session gain on nearly 39 million shares and a 35% recovery from the June low. So Thursday isn’t really about whether the losses are shrinking. It’s about where that money goes, and management has been direct about the answer. Alibaba still pays a dividend, but the marginal dollar isn’t heading back to shareholders. It’s going into data centers.

Key Takeaways

• Operating income collapsed from RMB 35.0 billion in the June 2025 quarter to negative RMB 0.8 billion by March 2026 while revenue barely moved. Thursday laps that RMB 35.0 billion comparison, the highest of the last 5 quarters.

• Cloud revenue grew 38% last quarter, external growth 40%, AI products 30% of the mix and an 11th straight quarter of triple-digit AI growth. Consensus looks for roughly 45% Thursday.

• Regulatory pressure, not a winner, is de-escalating the delivery war. Draft rules published June 17 curb sustained subsidies, and all 3 major operators publicly backed them within an hour.

• The freed cash has a destination. Capex hit RMB 126.1 billion in fiscal 2026, roughly a third of the entire 3-year AI commitment in 1 year, and management says it will overshoot the plan anyway.

• Washington cleared Alibaba to buy Nvidia’s H200. Beijing has restricted imports, leaving actual shipments minimal. Cloud growth may increasingly be limited by silicon rather than demand.

Start with where the money comes from.

The Boom Sector and the Slump Sector

Alibaba isn’t one business facing one market. It’s 2 businesses facing economies pointed in opposite directions, and that gap explains almost everything about the stock.

The consumer side is weakening on volume even as prices recover, and that split matters. Retail sales fell 0.6% in May, the first decline since December 2022, then managed 1.0% in June and 0.6% in July, reported this morning against 1.5% expected. Cars fell 17.0%. HSBC had already cut its full-year forecast from 5.2% to 2.8%, and the property slump is in its fourth year, which matters because housing is where Chinese household wealth lives. The one genuine improvement runs the other way: the GDP deflator turned positive at 1.6% in the second quarter, ending 12 straight negative readings, and producer prices rose for the first time since late 2022. Prices have stopped falling. People just aren’t buying as much. That’s the ground Taobao and Tmall are standing on, and it’s why China e-commerce revenue grew 6% last quarter.

The cloud side operates in something closer to a mania. Alibaba’s cloud unit grew 38% last quarter, with external revenue up 40%, AI products at 30% of the mix and an 11th consecutive quarter of triple-digit AI growth against an annualized AI run rate near $5.2 billion. According to Omdia, Alibaba took 38.1% of China’s AI cloud market across full-year 2025, up from 35.8% in the first half, in a market Omdia sizes at RMB 56.7 billion. Volcano Engine, Baidu, Tencent and Huawei trail well behind. Research firms define that market differently, so treat it as one house’s view.

Then there’s the third market, the one that did the damage. Instant retail turned into a subsidy war with Meituan and JD.com that cost the industry something extraordinary. Morgan Stanley puts Alibaba’s fiscal 2026 losses there near RMB 86 billion. Meituan guided to a full-year loss as large as RMB 24.3 billion against a RMB 35.8 billion profit the year before, and JD’s new business unit lost RMB 46.6 billion. Alibaba wasn’t new to delivery, having owned Ele.me since 2018, but it went from also-ran to roughly 36% of the market in about a year. The price was most of its operating profit.

Now what the fight cost.

What Actually Broke

The damage doesn’t show up in revenue, which is why headline numbers have misled all year. Quarterly revenue has hovered between RMB 236 billion and RMB 285 billion for 5 straight quarters. Income from operations, over the same stretch, went RMB 28.5 billion, RMB 35.0 billion, RMB 5.4 billion, RMB 10.6 billion, then negative RMB 0.8 billion. Adjusted EBITA, the profit measure Alibaba emphasizes, fell 84% in the March quarter alone, to RMB 5.1 billion.

Some of that is presentation rather than deterioration. Merchant subsidies that used to sit in sales and marketing expense are now booked against customer management revenue, which cuts reported revenue and gross margin without changing the underlying economics. Adjusting for it, China e-commerce customer management revenue grew 8% on a like-for-like basis, and group revenue grew 11% once the divested Sun Art and Intime businesses are stripped out of the comparison. The business isn’t shrinking. It’s being spent.

This is also why the valuation deserves a second look. At $123.81 the stock trades near 19x trailing earnings and roughly 2.0x sales, on fiscal 2026 revenue of RMB 1.02 trillion, about $148 billion. Forward multiples run near 18x depending on the provider, not the low teens some screens show. The trailing figure also flatters the operating business: fiscal 2026 net income attributable to shareholders was RMB 105.9 billion against income from operations of just RMB 50.2 billion, because Alibaba holds large investment stakes whose gains land below the operating line. Headline P/E is a poor proxy for what the core businesses are worth when investment gains inflate net income by that much.

Where the recovered profit goes.

The Savings Already Have a Destination

Here’s what makes Thursday interesting rather than merely positive. The delivery war is winding down, and not because anyone won it. China’s market regulator flagged it as a priority enforcement case in January, summoned 7 operators in February, and published draft rules on June 17 restricting sustained large-scale subsidies and forced cost-sharing with merchants and riders. Meituan, JD and Taobao’s instant unit all publicly backed the rules within an hour. Management’s own target is to halve those losses in fiscal 2027, with unit economics guided positive by the end of it.

That’s a genuine profit recovery. But investors shouldn’t assume it drops cleanly to the bottom line or comes back to shareholders. Alibaba committed RMB 380 billion to cloud and AI infrastructure over 3 years. It spent RMB 126.1 billion in fiscal 2026 alone, a pace that over 3 years lands almost exactly on the full commitment, and RMB 26.9 billion in the March quarter. Chief executive Eddie Wu has said the company will overshoot the plan anyway, with reports of a revision toward RMB 480 billion. The company’s own explanation for falling free cash flow names the 3 culprits in order: quick commerce, user acquisition for the Qwen app, and cloud infrastructure. Management has said plainly that margin is secondary right now.

And it’s funding this partly by selling what no longer fits. Sun Art and Intime are already gone. On Monday, 3 days before the print, Alibaba agreed to sell its gaming arm Lingxi Games to Trustar Capital in a deal valuing the studio above $1.5 billion, with Reuters reporting a figure above $2 billion. That’s a coherent strategy. It’s also a different company than the one many investors think they own: less a cash-generative marketplace returning capital, more a conglomerate recycling mature assets into an infrastructure bet.

The part no boardroom decides.

The Constraint Alibaba Doesn’t Control

The bull case assumes cloud growth is demand-limited. It may be supply-limited instead, and the direction of the constraint matters, because the popular version has it backwards. Washington approved the export licenses, clearing about 10 Chinese companies including Alibaba to buy Nvidia’s H200. It’s Beijing that has since restricted the imports. Customs officials held up shipments and component suppliers paused H200 production, leaving actual deliveries minimal. By July a US trade official was describing the volume shipped as very few.

The restriction reads as deliberate industrial policy. Beijing wants domestic silicon adopted, with Huawei among the clearest beneficiaries as Nvidia’s China position has weakened. So the number to watch Thursday isn’t just cloud growth. It’s whether management sounds capacity-constrained, and what it says about the domestic chips it may have little choice but to use. No quarterly result resolves that.

What the price says.

The Price Going In

Levels are based on the August 14 close of $123.81. The stock bottomed at $91.99 on June 26 and ran to $132.32 by August 10, a 44% advance, before giving back 6%. It sits above the 20-day and 50-day EMAs ($122.05 and $119.61), just under the 100-day at $123.94, and well below the 200-day at $129.76. Daily ADX near 30 marks an established trend with buyers in control of the directional readings, while the weekly reading near 13 says no trend exists on the longer frame. The faster daily stochastic-RSI gauge has reset to 16, unwinding the overbought condition, though the weekly version sits near 95. Average true range is $3.84, about 3.1%.

I’m not publishing a trade plan here, and the reason is the setup rather than the company. A binary event lands in 3 days, the operationally significant news was already telegraphed at that briefing, and the stock has traveled 35% off its low absorbing it. Consensus makes the point better than I can: 38 of 40 analysts rate it a buy at a mean target near $190, roughly 53% above the price. That gap isn’t a level structure you can trade against. It’s an unresolved argument. Short interest under 2% of float removes the squeeze mechanic too. The levels worth marking are $132.32 as the August high and the $115 to $120 shelf where the 50-day EMA meets the rally’s base. Everything between is noise until Thursday.

Bottom Line

The thesis has quietly changed, and I don’t think the market has fully repriced what it changed into. Alibaba is no longer a cheap Chinese e-commerce business waiting for the consumer to recover. It’s an AI infrastructure buildout funded by an e-commerce business whose customers are buying less, and increasingly funded by selling off the pieces of itself that don’t serve that goal. Both halves of that sentence matter. The AI half is genuinely strong, with a leading share, accelerating growth and real revenue behind it. The funding half is running against retail sales that grew 0.6% last month. Thursday’s June quarter laps RMB 35.0 billion of year-ago operating income, the highest of the last 5 reported quarters, so a weak-looking comparison isn’t the same as a deteriorating business. What I’d watch isn’t the headline. It’s capex guidance, what management says about chip supply, and whether the money saved on delivery shows up anywhere other than the data center budget. My expectation is that it won’t, and that’s the thesis, not the complaint.

This is research and commentary, not personal investment advice. Levels discussed are illustrative; size positions to your own risk tolerance and time horizon. The author may hold positions in names discussed.

No shared multiple survives that gap, so I won’t force one. What they share is a moment. Each just proved its slice of the AI buildout is bigger than the market had modeled, and each now faces the same question in a different form.

Both have shown they can raise money. What neither has shown is what the next leg of growth costs the people who already own the shares.

Key Takeaways

• SMCI closed fiscal 2026 at $39.1 billion of revenue against $22.0 billion the prior year, and guided fiscal 2027 to $65 to $72 billion, well above a consensus running between $52.5 and $54.4 billion by provider.

• June-quarter gross margin hit 17.5% against 9.5% a year earlier, on enterprise mix roughly doubling to half of revenue. But the full fiscal year came in at 10.8%, and management guided September back to 10.4% to 10.8%.

• Fiscal 2026 consumed $6.81 billion of operating cash as receivables and inventory swelled, though the June quarter itself generated $747 million of positive operating cash flow.

• WULF signed a 20-year lease with Anthropic for roughly 401 MW and now draws 71% of revenue from HPC leases, though total revenue still fell 6% YoY and the operating loss was about $140 million.

• Neither lacks access to capital. SMCI raised billions through equity and convertibles in June; WULF holds about $3.0 billion of cash. The open question is what the next round costs.

SMCI: The Margin Changed. The Working-Capital Burden Didn’t Disappear.

Fiscal 2026 closed at $39.1 billion of revenue against $22.0 billion in fiscal 2025, with net income of $2.2 billion and adjusted diluted EPS of $3.63 versus $2.06. The June quarter landed at $11.1 billion, near the low end of its $11.0 to $12.5 billion guide but 91% above the $5.8 billion a year earlier, with non-GAAP EPS of $1.70 (GAAP $1.62) against a consensus of $0.88 to $0.92 depending on the provider. One disclosure to carry through: the company labels these figures preliminary and unaudited, with the fiscal-year close incomplete and a board review underway related to certain export-control transactions.

Revenue was never the doubt here. Margin was. The bear case has always been that Supermicro assembles other people’s silicon into racks at whatever price the market allows, and the filed quarterly gross margins supported it, bottoming at 6.3% in December and 9.9% in March. June reported 17.5% GAAP and 17.6% adjusted, against 9.5% a year earlier and guidance of 8.2% to 8.4%, which management attributed to mix, with enterprise roughly doubling to half of revenue.

Then the qualifier that decides what it’s worth. Full-year fiscal 2026 gross margin was 10.8%, slightly below fiscal 2025’s 11.1%, and management guided September to 10.4% to 10.8%. Supermicro isn’t claiming 17.5% is the new normal. So the question isn’t whether the margin was real. It’s whether June began a mix shift or was an unusually favorable quarter inside a business that still runs near 11%.

The top-line guide extended rather than walked back: September at $14.5 to $15.5 billion, full fiscal 2027 at $65 to $72 billion, with record backlog and more than $60 billion of new orders booked in the fourth quarter alone.

Then the balance sheet. Fiscal 2026 consumed $6.81 billion of operating cash, overwhelmingly through receivables and inventory, against $7.5 billion of cash and roughly $8.7 billion of bank debt and convertible notes at June 30. That isn’t distress; it’s what happens when a company nearly doubles revenue in a business that buys silicon and carries inventory before it can bill anyone. The June quarter itself turned, generating $747 million of positive operating cash, so the drain isn’t constant. It is structural, which is why the company raised billions through equity and mandatory convertible preferred in June to fund components against those orders.

That dashboard is computed through March 31, so it predates the June quarter and shows trailing revenue of $33.70 billion rather than the $39.1 billion fiscal-year figure. Several tiles read n/a because the 2 sources disagreed, itself a signal on a company with SMCI’s filing history.

Technical levels are based on the August 14 close. SMCI gapped from $31.60 on August 11 to $37.61 the next session and has reached $39.84, roughly 26% higher in 3 sessions and about 70% above its July 17 low of $23.38. Price sits above every daily EMA (20-day $31.99, 50-day $31.08, 100-day $31.36, 200-day $33.19), with buyers well ahead on the directional readings. It’s also stretched: RSI at 70.7 is overbought, the faster stochastic-RSI reading is pinned at its maximum, and Bollinger %B above 1.0 means price closed outside the upper volatility band. Consensus hasn’t followed, at a hold across 16 analysts and a $42.38 mean target 6% above the price.

June is the most interesting number Supermicro has printed in years, and its own September guide says don’t extrapolate it. Price has already traveled 26% while consensus sits 6% away, so the market has paid for the quarter without yet paying for the thesis.

Now the other end of the chain.

WULF: The Contract Book Changed. The Income Statement Didn’t.

TeraWulf’s June quarter produced $44.8 million of revenue, of which $31.9 million came from high-performance computing leases, up 52% sequentially and now 71% of the mix against roughly zero a year earlier. That transformation is the entire story. The number underneath is less flattering: total revenue fell 6% YoY, because the bitcoin mining business it’s exiting shrank faster than leasing grew. The reported net loss of $939.9 million reads as catastrophic until you look at composition, where a $755.7 million non-cash fair value charge tied to Google warrants accounts for most of it. Strip that out and the business still ran an operating loss near $140 million, so the accounting explains the headline, not the economics.

The contracts are the asset. A 20-year lease with Anthropic covering roughly 401 MW of critical IT capacity turns a volatile mining operation into something closer to a landlord with a 2-decade tenant, and management reaffirmed a target of contracting an incremental 250 to 500 MW each year. TeraWulf entered the third quarter with 102 critical MW operating at Lake Mariner, which frames the task: the contracted book is several times the operating base, and the gap has to be built first.

Building it is a financing question, though not an access question. The capital structure carries roughly $2.5 billion of convertible notes at the parent and $3.2 billion of senior secured notes at the operating subsidiary, against about $3.0 billion of cash and restricted cash earmarked for contracted development. That is what funding a buildout ahead of its revenue looks like. TeraWulf also agreed to sell its 50.1% Abernathy joint-venture interest for roughly $530 million, recycling capital toward projects it controls outright.

Technical levels are again based on the August 14 close. WULF trades at $17.38, about 42% below its June 22 high of $29.84, and below every daily EMA (20-day $18.08, 50-day $19.96, 100-day $19.98, 200-day $17.74).

Sellers lead the directional readings, with ADX near 28 marking an established downtrend, though momentum has stopped deteriorating. The weekly picture is washed out, its stochastic-RSI reading near zero. Against that, consensus is unusually bullish: a strong buy across 18 analysts and a $36.64 mean target 111% above the price, with 26.2% of the float short as of July 31.

The contracts changed what this company is, and the price hasn’t agreed for 2 months. That gap is either the opportunity or the warning, and it resolves on execution, on whether contracted megawatts become operating megawatts on schedule and on budget.

What both reports left open.

The Question They Both Left Open

Strip out the difference in scale and the 2 reports rhyme. Each proved demand: Supermicro with more than $60 billion of new orders in a quarter, TeraWulf with a 20-year lease from one of the better-funded buyers of AI capacity in existence. Each has also proved it can raise capital. What neither has shown is that the next leg can be funded without a real cost to the people who already own the shares, through working capital, leverage, financing expense or dilution. For Supermicro that cost arrives before the server ships, as cash disappears into inventory and receivables. For TeraWulf it arrives before the megawatt starts paying rent. Same constraint, opposite mechanics.

How to hold either one.

Setups

These are 2 unrelated positions with unrelated risks, and nothing below is a comparison between them. They’re illustrative levels, not instructions.

SMCI

1. Pullback zone: $35.50 to $37.50, covering the 50% retracement at $35.44. Invalidation: a close below $33, under the 200-day EMA, which puts price back into the pre-earnings range.

2. Continuation: a close above $41, with the 38.2% retracement at $39.21 beneath it as first support. Objectives: $45, then $51.40 at the June high, then $58.78 at the 52-week high. Invalidation tightens to $37.50.

3. Price outside its upper band with the faster oscillator maxed makes chasing here poor risk and reward. The next report is currently estimated for early November.

WULF

1. Support zone: $15.00 to $16.00, at the lower volatility band above the July 29 low of $14.81. Invalidation: a close below $14.50, which breaks that low.

2. Reclaim: a close above $20, clearing the 50-day and 100-day EMAs that sit within 2 cents of each other. Objectives: $21.43 is less than 1 average day above the trigger, so treat it as a waypoint; $23.41 and $29.84 are the levels worth sizing against. Invalidation tightens to $18.

3. At $17.38 the stock sits between both setups, which is itself the answer for today. A 9.8% average daily range with 26.2% of the float short means size matters more than entry.

Bottom Line

Treat these 2 as separate answers to separate questions. Supermicro printed its strongest margin quarter in years and then guided the next one back to 10.4% to 10.8%, which makes the durability of that mix the whole investment question, at 7.5x forward earnings with a hold consensus and a target 6% away. TeraWulf is increasingly valued like a digital-infrastructure landlord, on megawatts it hasn’t built, with analysts 111% above a price that has fallen for 2 months. What links them isn’t a multiple, it’s a constraint. Both told you the buildout is larger than the models carried. Both still have to fund the next leg, and in each case the bill arrives before the revenue does.

This is research and commentary, not personal investment advice. Levels and trade plans are illustrative; size positions to your own risk tolerance and time horizon. The author may hold positions in names discussed.

Welcome to IWP’s weekly trade plan update.

What this is. Every Sunday we sort the 60 stocks and ETFs we cover into trend regimes and publish the three numbers that decide each trade: entry, invalidation, target. You have them before Monday opens, so the decision is made in advance and in writing rather than in the moment. This is a risk discipline first and a stock-picking service second.

What to expect. When a beaten-down market snaps back hard we will lag the names that gapped off the bottom, because the rules wait for a reclaim. In a bad tape that wait is the whole point. You do not get to skip the drawdown and catch the first bounce with one rule, and anyone selling you a framework that does both is selling hindsight.

How to check us. Every level is scored afterwards from the entry as published, win or lose, in the scorecard at the end. Last week lost money and the full record is down there with the workings.

If you already know the framework, skip to This Week’s Playbook. If a level does not sit right with you, leave a comment: we read every one, and they shape what gets covered.

Market Conditions

The broad tape steadied this week. The S&P 500 (SPY) closed at 776.34, Trend-Hold and roughly flat on the week, while the Nasdaq-100 (QQQ) at 731.07 is Trend-Hold and up 1.1% on the week.

Beneath the surface, regime breadth improved on balance: Breakdowns rose from 6 to 7, Repairs fell from 7 to 5, and Trend-Holds rose from 27 to 30. Regime upgrades outnumbered downgrades 13 to 9, though the movement was underneath the surface rather than in the indexes.

The biggest moves went to NBIS +48%, OSCR +17%, CRWV +16%, SMSN.L +16% on the upside, against AVGO -8%, LULU -7%, AMAT -6%, BIDU -6% on the downside. Fresh damage showed up in LULU, NVO and PALL, dropping to Breakdown.

The cleanest signal was a full reclaim of Trend-Hold in AMD, CRWV, FTI and LRCX, among others. The index trend is intact but it did not extend this week, and the churn underneath cut both ways. That is a rotation inside an uptrend rather than a new leg of it.

What Changed From Last Week

• Reclaimed Trend-Hold: AMD, CRWV, FTI, LRCX, NBIS, NOC, OSCR, SMSN.L. The strongest signal of the week. Several got there in a single week, so the weekly chart still has to hold it.

• Lifted out of Breakdown into Transitional: WMT. An early bounce that still has to prove itself.

• Up a step from Repair into Transitional: COST, NFLX, ORCL. Basing turned into actual recovery attempts, one notch short of a trend.

• First repair off the lows: XRH0.L. A damaged name showing the first sign of basing.

• Cooled out of Trend-Hold into Transitional: AMAT, AVGO, BRK.B, LLY. Leaders that lost their trend and now need to re-prove it.

• Lost the trend, into Repair: ETHA, GOOG. Treat these as damaged until reclaimed, and note GOOG fell two regimes in one week rather than one.

• Broke down: LULU, NVO, PALL. Fresh damage, the framework moves these to the avoid list.

How Our Trade Plans Work

Each name is classified into a regime (Trend-Hold, Transitional, Repair or Breakdown) from the 1D EMA stack, cross-checked against the 1W chart, and every level traces back to a specific indicator. T1 is the first profit-take, T2 the structural objective, T3 the stretch target. Regime sets how much risk a name is allowed and when: Trend-Hold earns full risk but only on a pullback to support or a confirmed breakout; Transitional gets half or nothing; Repair gets a starter only after a reclaim; Breakdown gets none until a reclaim confirms.

Two stops, two jobs. The trade stop is the tactical level where a single trade is wrong and you cut it. The regime invalidation is the deeper level where the classification flips and the reason to own the name is gone. One manages risk, the other manages conviction.

Size by risk, not by dollars. Stop distances in this issue run from under 1% to over 40%, so the same dollar amount in two names can mean risking sixty times as much on one as the other. Decide what a single losing trade may cost you, say half a percent of the account, and divide it by the distance from entry to stop. A 4% stop carries eight times the position of a 32% stop for identical risk. The wide-stop names are not off limits; they are simply small.

This update ranks 60 names. You are not meant to trade all of them, or even most of them. Most weeks a handful sit at a level worth acting on and the rest are there so you can check our work.

The Week Ahead

These names in the universe report during the plan week:

Tuesday: BIDU | Wednesday: TGT | Thursday: BABA, WMT

Earnings are binary and the framework does not trade into them. If one of these sits at a defined entry, either take it at reduced size or wait for the print and re-read the structure afterwards. A level that was valid on Friday is not wrong after a gap, it is simply no longer the level.

Those matter more than the rest of the list, because you have a live instruction on them this week: BABA is tagged “Wait / half size”, TGT is tagged “Buy dips” and WMT is tagged “Wait / half size”. Take the print as the risk event it is, size for it, or stand aside until it is out.

Just beyond the plan week: NVDA on Wednesday August 26. Worth knowing now if you are sizing a position you intend to hold through it.

This Week’s Playbook

Start here. Detailed setups for the top names follow; the watchlist covers the rest.

• Closest to a defined entry: DG, AMZN, SPY, TSM, PINS, QQQ, plus 2 more tagged Near entry below. Price is sitting on the 1D EMA20, so the trigger is near and you are not chasing.

• Best leaders to buy only on weakness: MSFT, PANW, LMT, TGT, SHOP, RBRK, plus 6 more tagged Buy dips below. Clean trends with room to run, but wait for the pullback.

• Most dangerous chase setups: none this week. Nothing in the universe is overbought enough to warn you off, which tells you the leaders have been resting rather than running.

• Avoid until a reclaim confirms: BIDU, LULU, META, NKE, NVO, PALL, XRPI. No long exposure; not every dip is a setup, and cash is the position.

• Where the payoff justifies the stop. A wide stop is not a bad setup, it is a small one, but some names are not paying enough to be worth any size at all.

• Tight and paying: RDW (5% stop for 19% to T1), PINS (2% stop for 7% to T1), XOVR (4% stop for 9% to T1), NOC (2% stop for 4% to T1)

- Wide but paying, so size small: PLTR (9% stop for 15% to T1)

- Not paying for the risk, quarter size or skip: ORCL (23% stop for 2% to T1), INTC (19% stop for 3% to T1), AIXA.DE (23% stop for 3% to T1), WMT (7% stop for 1% to T1)

Top 5 Actionable Setups

The highest-conviction names sitting closest to a defined entry, with full detail, levels, and both stops for each. The next tier follows in the Also Actionable table, and the rest of the universe in the full watchlist below it.

Dollar General Corp. (DG), Last close: 123.28

Setup: Dollar General Corp. held its Trend-Hold, down 3% on the week. 1D Trend-Hold inside a Transitional 1W; 1D EMA20 at 123.05 and EMA50 at 120.53 are the structural supports. Daily RSI 51.9, MACD bearish stabilizing; constructive but unhurried, let price come to the levels.

• Market view: deep-discount retail, a read on low-end consumer health; trailing the broad tape this week.

Read more

Lam Research sits 23.1% below its own high, set the very same day. So the frame for everything below: the market isn’t arguing with these earnings. It’s arguing with how long they last.

Key Takeaways

• AMAT posted record Q3 revenue of $9.12 billion, up 25% YoY, with record non-GAAP EPS of $3.50, and guided Q4 to $10.25 billion, which against last year’s $6.80 billion works out to roughly 51% growth.

• LRCX posted record June-quarter revenue of $6.72 billion, up 30% YoY, at a record 38.4% non-GAAP operating margin, and guided September to $8.10 billion, roughly 52% above the year-ago quarter.

• Both stocks peaked on June 30 and have been marked down since, so this is multiple compression against rising fundamentals, not deteriorating businesses.

• Nearly identical market caps ($424 billion versus $422 billion) buy you very different things: AMAT carries about 33% more revenue, LRCX converts it far better, and they meet at roughly 29x forward earnings.

• Buying both provides less diversification than it looks like: same cycle, same customers, same peak date.

The one that just got marked down first.

AMAT: A Record Quarter, Sold Anyway

The quarter was a genuine beat on the line that matters. Revenue of $9.12 billion grew 25% YoY and 15% sequentially, non-GAAP EPS of $3.50 rose 41% and cleared the $3.45 consensus, non-GAAP gross margin was 50.4%, and non-GAAP operating income of $3.10 billion (34.0% of revenue) was a record. Semiconductor Systems did $7.04 billion at a 37.7% operating margin, services added $1.78 billion at 30.1%, and operating cash flow of $3.04 billion was also a record.

The guide is the part worth sitting with. Management pointed Q4 at $10.25 billion give or take $500 million, with non-GAAP EPS of $4.02. Against the year-ago October quarter’s $6.80 billion that implies roughly 51% growth, and it’s worth remembering AMAT was shrinking a year ago, at -3.5% and -2.1% in the October and January quarters.

Applied again raised its Semiconductor Systems outlook for calendar 2026, now to growth exceeding 30% against a prior expectation above 20%, expects to outgrow the market this year, and sees another strong growth year in 2027. Revenue came slightly below the $9.18 billion consensus in my data feed, though estimates varied by provider, and no gross margin guidance accompanied the outlook.

That’s the thread sellers pulled: the stock fell about 5% after hours to roughly $508 despite the beat and the raise.

That dashboard is one quarter stale, since the feeds haven’t ingested Thursday’s filing; trailing-12-month revenue of $29.02 billion becomes about $30.84 billion once the new quarter rolls in, pulling price-to-sales from 14.6x to roughly 13.8x at Thursday’s close, or near 13.1x after hours.

The forward multiple of 29.9 is the number that matters; trailing P/E of 51.5 sits on a different earnings basis, so the 2 aren’t a clean growth bridge. Consensus is willing to underwrite the forward figure, at a $633 mean target across 35 analysts, though those estimates predate Thursday’s guide.

Technical levels below are based on the August 13 close, before the after-hours reaction. AMAT went into the print pinned to its 20-day EMA at $534.63, above the 50-day at $529.59, the 100-day at $486.77 and the 200-day at $412.42, with the daily ADX at 10.6 showing no established trend and RSI at a neutral 49.5.

The weekly timeframe still reads as a strong uptrend with buyers leading sellers, but weekly momentum is fading, and at an after-hours price near $508 it sits below the 50% retracement at $528.16, with the next shelf at the 100-day EMA and the 61.8% retracement at $478.25. With a daily average true range of $35.59, about 6.7% of the price, a single session can travel most of that span.

The business accelerated and the stock got cheaper, the setup value investors say they want and rarely buy. This isn’t a falling knife either, since the guide went up rather than down and the drop lands on a defined shelf instead of a fresh low. The complaint is margin visibility, not demand.

Now the better business.

LRCX: The Higher-Quality Compounder, Already Re-Rated

Lam’s June quarter was the stronger print of the 2, on the metrics that separate good semicap from great. Revenue of $6.72 billion grew 30% YoY and 15% sequentially, non-GAAP gross margin of 52.0% came in above the high end of its own guidance on pricing and mix, and a 38.4% non-GAAP operating margin was a record. Non-GAAP EPS of $1.82 grew 41% against fiscal 2025. Growth has accelerated in each of the last 2 quarters, from 22.1% to 23.8% to 30.0%, with gross margin climbing 49.6%, 49.8%, 51.7% alongside it on the filed figures. Rising volume alongside expanding margins points to unusually strong operating leverage, helped by pricing, efficiency and mix.

The September guide is where it separated: $8.10 billion give or take $400 million, roughly $1 billion above consensus, with gross margin guided to 52% and operating margin to 39.5%.

Against the year-ago September quarter’s $5.32 billion that’s about 52% growth. Lam also raised its calendar 2026 wafer fab equipment forecast to the low $150 billion range from $140 billion, and framed 2026 as its third consecutive year of outgrowing that market. A supplier raising the size of its own end market mid-cycle is a different signal than one merely beating a quarter.

Here the 2 companies invert. LRCX carries the richer valuation, 18.2x trailing sales against AMAT’s 13.8x, and the higher trailing P/E at 56.5, but it earns it: a 50.5% trailing gross margin against 49.0%, a record 38.4% operating margin last quarter against AMAT’s 34.0%, and a 65.1% return on equity against 39.8%. Same market value, about a quarter less revenue, converted far more efficiently.

On forward earnings the 2 land within a rounding error of each other, 29.2x versus 29.9x, which is the market saying these are the same asset priced through different mixes. The catch is upside: consensus sits at $369, only about 9% above Thursday’s close, versus 18% for AMAT.

LRCX is the exact inverse of AMAT’s setup. It bottomed at $250.50 on July 29, its own report date, and has since run 34.5% to $337.01, closing above every daily EMA (20-day $314.67, 50-day $318.48, 100-day $300.29, 200-day $257.43) with buyers leading sellers and momentum still building.

It’s also stretched: the faster stochastic-RSI reading near 98 is about as overbought as that gauge registers, Bollinger %B at 0.91 has price pressed against the upper volatility band, and Thursday’s $345.12 high stopped at a triple confluence with the upper band at $343.84 and the 38.2% retracement at $345.13.

The better business, already paid for the quarter, sitting overbought at a defined ceiling. Quality doesn’t exempt it from needing an entry.

The risk neither number shows.

The Cyclical Trap, Named

In semicap, the P/E lies at both ends. A low multiple on peak earnings is the classic value trap; a high multiple on trough earnings is usually the buy. Roughly 29x forward on earnings visibly inflecting upward is neither. It’s a bet this cycle extends beyond 2027 rather than topping in it. The evidence for extension is recent: both raised outlooks rather than merely beating, and Lam raised its estimate of the market’s size. The evidence for caution is the 6-week derating itself, which suggests investors are discounting duration, valuation and AI-capex cycle risk despite improving near-term fundamentals.

Which brings up concentration. The mix isn’t identical, Lam skews toward memory while Applied is broader with a large services base, but both sell wafer fab equipment to the same handful of customers and both peaked June 30. Owning both gives you less diversification than 2 tickers suggest, so size it closer to 1 position split across 2 names than 2 independent ideas.

How to act on it.

Post-Earnings Setup

The 2 names sit at opposite ends of the same reaction cycle: AMAT is being marked down as you read this, LRCX has already run 34% off its low into resistance. That difference, not company quality, drives the entries. They’re illustrative levels, not instructions.

AMAT

1. Support test: $486 to $510, where the 100-day EMA sits just above the 61.8% retracement. The after-hours print lands at the top of it. Invalidation: a close below $470, under that retracement, which reopens the May range.

2. Reclaim confirmation: a close back above $535, over the 20-day and 50-day EMAs it just lost. Objectives: $564 at Thursday’s high, $584 at the upper band, then $633 at the analyst mean. On a reclaim entry, invalidation tightens to a close below $510.

3. Estimates predate the guide and will be revised; let the shelf prove itself first.

LRCX

1. Support test: $315 to $325, the 20-day and 50-day EMA cluster with the 50% retracement at $316.29 underneath. This is the entry the current price doesn’t offer. Invalidation: a close below $300, at the 100-day EMA, which would undo the post-earnings advance.

2. Breakout confirmation: a close above $346, clearing the upper band and the 38.2% retracement together. Objectives: $369 at the analyst mean, $396 at the weekly upper band, then $438 at the June high. On a breakout entry, invalidation tightens to a close below $324, Thursday’s low, which would mark it a failed break.

3. With the faster oscillator near overbought and price at a triple-level ceiling, chasing here pays up for a move already made. Patience is a position.

Bottom Line

Your read is right on the businesses and needs 1 adjustment on the trade. Both are executing at records, both raised, and both are cheaper than 6 weeks ago while earning more, which is multiple compression against rising fundamentals rather than a broken story. The adjustment: these aren’t 2 opportunities.

They’re the same AI-capex cycle wearing 2 tickers, peaking the same day and priced within a rounding error of each other on forward earnings. If forced to choose today, AMAT offers the better entry mechanics, a fresh markdown into a defined shelf with 18% to consensus, while LRCX offers the better business, higher margins and returns and an end market it just told you is bigger than everyone thought, at a price that already reflects it.

The markdown currently offers the cleaner setup: AMAT is moving toward an entry zone, while LRCX is asking investors to pay for a move that has already happened.

This is research and commentary, not personal investment advice. Levels and trade plans are illustrative; size positions to your own risk tolerance and time horizon. The author may hold positions in names discussed.

Key Takeaways

• CRWV revenue hit $2.575 billion, up 112.5% YoY, and management raised full-year guidance to $12.4 to $13.2 billion in revenue and $35 to $39 billion in capex, evidence the demand story isn’t a one-quarter beat.

• NBIS revenue was $582.3 million, up 454% YoY, decelerating from Q1’s 684% almost exactly as predicted, but adjusted EBITDA swung to a $236.2 million profit from a $21 million loss, margin jumping from 32% to 41%.

• On trailing sales, CRWV is roughly flat at about 7.8x versus 7.9x a week ago; NBIS is cheaper at an implied 49x versus 54.4x, genuine compression despite the bigger stock move.

• CRWV’s $104 billion revenue backlog and its total liabilities both grew this quarter; the balance sheet expanded alongside the growth, it didn’t get safer.

• Michael Burry added to his NBIS short at $247 the same day it closed at $259.20, flagging that Nebius stretched its server depreciation schedule from 4 to 5 years.

CRWV: The Multiple Held, and the Guidance Raise Is the Real News

Last week’s question was simple: hold triple-digit growth and show contracted capacity keeps arriving ahead of financing. Revenue answered at $2.575 billion, up 112.5% YoY and 23.9% sequentially. More consequential: management raised full-year revenue guidance to $12.4 to $13.2 billion and capex guidance to $35 to $39 billion.

The quarter was adjusted-operating-income positive at $128 million, even though GAAP operating income was still a $49 million loss, an improvement from Q1’s $144 million loss; net loss narrowed 15.4% sequentially to $626 million, even as it more than doubled YoY against an easier $290 million comparison. CoreWeave’s own revenue backlog, its defined metric for enforceable customer commitments, sits at roughly $104 billion against a market value near $59 billion.

The balance sheet didn’t get safer, whatever the stock price says. Total liabilities jumped 41.8% in a quarter, from $50.8 billion to $72.0 billion, while cash grew even faster, from $2.2 billion to $5.5 billion.

A P/S multiple alone can hide that, so pair it with an enterprise value: using the most recent isolated debt figure available, $35.1 billion from Q1’s breakout, against the fresh cash and market cap gives an EV near $89 billion, about 11.7x revenue, cheaper than the roughly 13.2x the same math implied a week ago, since the growing cash pile more than offset the equity re-rating. Last week’s line still applies: this is a bet that contracted AI demand outlasts a balance sheet built like a leveraged utility.

Tuesday’s print funded another round of that bet, on better terms than the stock price alone suggests.

On trailing-12-month revenue of $7.59 billion (up from $6.23 billion pre-print) against the $59.4 billion market cap, the multiple works out to about 7.8x sales, a hair below last week’s 7.9x.

That’s a narrower, more defensible claim than saying the market refused to pay more: trailing multiples reset mechanically as a new quarter rolls in, and this one simply held. The mean analyst target, $138.51 across 38 analysts, is essentially unchanged from last week, still implying nearly 29% upside, but built on estimates the print hasn’t reached yet.

Technical levels below are based on the August 12 close. The stock gapped from Tuesday’s $90.32 close to a $108.70 open and traded as high as $111.82, on volume nearly 3 times average, blowing straight through last week’s $95 breakout trigger and $109 first objective in one session.

Price now sits 14% to 25% above every daily EMA (20-day $85.90, 50-day $89.40, 100-day $93.18, 200-day $94.18), extended and should be read that way; RSI at 65 has room before overbought, but the faster stochastic-RSI reading above 93 says the bounce is stretched short-term. The old $93 to $95 ceiling is now the support to watch, and price spent the entire week before earnings chopping just underneath it, so this isn’t unfamiliar territory to hold.

The business grew into last week’s skepticism on both counts, the good part (guidance, margins, backlog) and the uncomfortable part (a $21 billion larger balance sheet in one quarter). The multiple says it’s roughly a wash; the enterprise-value math says maybe a touch cheaper. The structure says don’t chase the gap, let it come back to test what used to be resistance.

Now the one that broke its own pattern.

NBIS: Growth Decelerated on Schedule, but Profitability Inflected Hard Enough to Matter More

Nebius carried an uncomfortable precedent into this print. In May, Q1 revenue grew 684% and adjusted EPS beat sharply, but the stock held those gains only until DA Davidson downgraded it to Neutral purely on valuation, even while raising its target from $200 to $250, and NBIS fell almost 9% the following Monday on no change to the business. That’s the real risk here: a downgrade needs nothing but a valuation opinion to erase a week of gains. Wednesday didn’t repeat it.

Revenue came in at $582.3 million, up 454% YoY, decelerating from Q1’s 684% on schedule now that the year-ago base has stopped being tiny, and about 46% sequential, itself a step down from Q1’s roughly 75% pace. The headline growth rate slowed on both counts, exactly as the arithmetic demanded. The market rewarded it anyway, because the number that matters more than the growth rate is profitability: adjusted EBITDA swung to a $236.2 million profit from a $21 million loss a year ago, margin jumping from 32% to 41%. The AI cloud segment alone did $575 million of revenue, up 514% YoY, at a $286 million adjusted EBITDA and a 50% margin. Growth decelerated on schedule; profitability inflected hard enough to overwhelm that story.

Management reaffirmed 2026 guidance of $3.0 to $3.4 billion in revenue at roughly a 40% adjusted EBITDA margin. Michael Burry’s short, disclosed 6 days before the print, didn’t step aside; he added to it at $247 the same day the stock closed at $259.20, still underwater on both entries.

His argument is sharper than a generic valuation call: Nebius extended its server depreciation schedule from 4 to 5 years, which he frames as a bet the GPUs decay slower than the roughly 50% annual rate its own deal economics seem to assume. That’s the valuation risk in one sentence, unresolved by this print.

The valuation question is where this quarter’s real news sits. Trailing 12-month revenue is now $1.355 billion against a market cap of $65.8 to $70 billion, depending on share-count convention. Either way the multiple compressed, from 54.4x a week ago to roughly 49x to 52x, because the denominator grew faster than the price. Analyst targets are scattered post-print, from the low $240s to $286 by provider, itself a signal the revision cycle hasn’t settled. Short interest sits near 27% to 28% of float, essentially unchanged, since FINRA’s reporting lags too far behind Wednesday to show any post-earnings covering.

Technical levels below are based on the August 12 close. NBIS gapped from Tuesday’s $193.23 close to a $226 open and traded as high as $259.44 intraday, clearing last week’s entire level ladder, the $207 trigger, the $236 first objective, and nearly the $255 second, in one session. Price is now about 27% above its 20-day EMA ($204.53) and roughly 15% from its $299.86 52-week high.

ADX at 13.6 remains weak, a reminder that one powerful session doesn’t build trend strength, but the buyer-versus-seller readings flipped decisively toward buyers, reversing last week’s near-even split.

Growth decelerated exactly as required, and the market looked past it because profitability inflected hard and the multiple got cheaper doing it. Healthier than the raw 34% headline suggests, but a stock 15% from an all-time high, on an unconfirmed trend, with a credible short seller freshly adding, isn’t one to add to blind. Let it breathe.

What to do with both now.

Post-Earnings Setup

Both level ladders from last week triggered and were exceeded in a single session, so the question now is whether old resistance holds as new support on the first pullback. They’re illustrative levels, not instructions.

CRWV

1. Support test: $95 to $100, the former breakout trigger. Holding it on the first pullback confirms the move; failing it wouldn’t erase the quarter, but would say the gap got ahead of itself. Invalidation: a close below $90, which gives back the entire earnings gap.

2. Continuation confirmation: a close above $112, over Wednesday’s high. Objectives: $118 as the first round-number pause, $138 near the spring swing high (almost exactly the $138.51 analyst mean), then $153 at the 52-week high. Invalidation tightens to a close below $105, Wednesday’s low.

3. Estimates aren’t revised yet and the balance sheet is materially larger than a week ago; both the bull and bear case got new evidence, size accordingly.

NBIS

1. Support test: $216 to $226, Wednesday’s low-to-open range and the prior resistance shelf. Invalidation: a close below $205, under the 50-day EMA, erasing most of the re-rating.

2. Continuation confirmation: a close above $260, over Wednesday’s high. Objectives: $275 as a round-number waypoint, then the $299.86 52-week high, with anything beyond untested territory this stock has never traded. Invalidation tightens to a close below $238, roughly Wednesday’s own midpoint.

3. Scattered, fast-moving analyst targets, a still-weak ADX reading, and a credible short seller who just added: that combination argues for patience over chasing the next leg.

Bottom Line

Last week’s frame holds up, but its 2 halves resolved differently.

CRWV proved demand and execution while making its capital structure bigger, not smaller, and its trailing multiple sits almost exactly where it did a week ago, roughly 7.8x sales, with the enterprise-value math suggesting it may even be a shade cheaper once the fresh cash is counted.

NBIS proved growth decelerates exactly as scale arithmetic demands, but profitability inflected hard enough that the multiple compressed, from 54.4x to roughly 49x, despite the bigger headline move. CRWV’s multiple held; NBIS’s got cheaper; both grew into the rally by different mechanisms rather than the market simply deciding to pay more.

CRWV still has to prove its capital structure can outrun its own growth; NBIS still has to prove Burry is wrong about how fast the hardware under 41% margins actually ages. Neither question got answered this week. Both got new evidence, and the price moved to reflect it inside a single session.

This is research and commentary, not personal investment advice. Levels and trade plans are illustrative; size positions to your own risk tolerance and time horizon. The author may hold positions in names discussed.

The company raised its full-year revenue guide. What got sold was the other side of the P&L: a GAAP loss, a 13-point drop in gross margin, and a trimmed ceiling on the adjusted EBITDA guide. The market isn’t questioning whether HIMS can generate revenue. It’s questioning the incremental economics of generating it.

Key Takeaways

• Q2 revenue of $753.2 million grew 38.2% YoY, roughly 31% excluding the $40 million Eucalyptus contribution, and 23.9% sequentially; full-year revenue guidance rose to $3.1 to $3.3 billion, now including the acquisition.

• Profitability drove the sell-off: a GAAP loss of $0.37 per share against a $0.17 profit a year ago, and a $25 million trim to the top of the full-year adjusted EBITDA guide.

• Gross margin printed 63.8% versus 76.4% a year ago, but gross profit dollars grew 15.5% and gross profit per subscriber held roughly flat, which reframes the margin scare as a mix question.

• Positioning is loaded: about 30% of the float short at 4.5 days to cover, a hold consensus with a $29.38 mean target below the pre-print price, and $225 million remaining on the buyback.

• A 6% to 7% gap is roughly one ATR for this stock, so the pre-report level map survives the reaction. The setup section maps it.

The Print: Growth Came Back

The scare in this story was never Q2. It was Q1, when YoY growth collapsed to 3.8% against a brutal comparison, and the market had to ask whether the growth era was over. The quarterly path tells it: +110.7%, +72.6%, +49.2%, +28.4%, then the +3.8% stumble, and now +38.2% reported.

Composition matters here: Eucalyptus contributed roughly $40 million, so organic growth was about 31%, with U.S. revenue up 16% and Rest of World revenue jumping from $7.5 million to $131.4 million. The sequential move, +23.9%, is what no base effect explains away.

Guidance went up too, from $2.8 to $3.0 billion to $3.1 to $3.3 billion, though the old range excluded Eucalyptus and the new one includes it; management told Reuters the existing businesses were running ahead of the prior outlook even before the acquisition.

Zoom out and the machine is intact: fiscal 2025 revenue grew 59.0% to $2.35 billion, the 3-year CAGR runs at 64.5%, and the trailing 12 months including the new quarter sum to roughly $2.58 billion, about 2.9x sales at Monday’s close.

One housekeeping note on the dashboard below: its window is stamped pre-Q2 (through March 31), because the aggregator hasn’t ingested last night’s filing yet. That’s why its growth tile prints +3.8% and its TTM revenue $2.37 billion; the fresh filed numbers are the ones above.

Now the bill.

The Bill: What the Growth Costs

The GAAP print was ugly: a net loss of $86.3 million, $0.37 per diluted share, against a $42.5 million profit a year ago, with operating income swinging from +$26.7 million to -$97.2 million. But the composition softens it.

The quarter absorbed $47.5 million of legal contingencies, $28.8 million of acquisition and transaction costs, and $4.6 million of restructuring; the adjusted net loss was $20.8 million, and Q2 adjusted EBITDA of $60.3 million actually beat management’s own $35 to $55 million guide.

The guidance change was a trim, not a slash: the full-year adjusted EBITDA floor stayed at $275 million while the ceiling came down from $350 million to $325 million. The cleaner bearish datapoint is cash generation: the quarter used $35.9 million of operating cash and $68.2 million of free cash flow, and trailing operating cash flow of roughly $263 million is now trending the wrong way.

The balance sheet is adequate rather than alarming, and worth stating precisely. Quarter-end cash of $609.8 million plus $231.2 million of short-term investments gives about $841 million of liquidity against $1.365 billion of convertible notes.

Reported working capital is negative by about $99 million, but that figure carries a $537.4 million deferred acquisition payable, and the year-ago cash balance was swollen by convertible proceeds, so neither the “cash halved” arithmetic nor the working-capital line means what it appears to at first glance.

Liquidity buys time; the burn rate decides how much. The Street was braced before the print: a hold consensus across 15 analysts with a $29.38 mean target below the pre-report price and a $21 to $40 range. Positioning is the loaded part: about 30% of the float short per the mid-July data, 4.5 days to cover, and $225 million remaining under the buyback authorization, which management frames as capacity for exactly this kind of disconnect.

So the fork, and it’s wider than a simple mix-versus-price-war coin flip. Gross margin fell 13 points, yet gross profit dollars grew 15.5%, from $416.2 million to $480.8 million, and the disclosed per-subscriber figures make it concrete: monthly revenue per average subscriber rose from $76 to $92, which at the respective margins works out to roughly $58 of monthly gross profit per subscriber last year and roughly $59 now. Flat, not collapsing. That leaves 3 readings:

The bad one: competitive pricing is eroding what each customer is worth. The neutral-to-good one: branded weight-loss drugs and international revenue structurally carry lower margins but more revenue per customer and similar gross-profit dollars.

The genuinely good one: management is choosing margin dilution to buy retention, cross-selling and lifetime value, and it says openly that gross margins will stay below historical levels as those businesses scale. So the test isn’t whether the old margin comes back; it’s the pairing.

Watch gross margin alongside gross-profit dollars per subscriber, acquisition spend, adjusted EBITDA conversion and free cash flow. If margin falls while per-customer gross profit and cohort economics hold, that’s mix.

If both deteriorate, that’s erosion.

The Structure Before the Open

Technical levels below are based on the August 10 close, computed with standard settings and verified against TradingView defaults; the premarket print comes from this morning’s trade, not the regular session.

Before the report, HIMS was a textbook coil: $31.77 sitting just above an unusually compressed moving-average stack, with the 20-day EMA at $31.08, the 50-day at $30.71, the 100-day at $29.81 and the 200-day at $31.14, the whole stack packed between $29.81 and $31.14.

ADX at 13.7 says there’s no trend, the directional readings are dead even at 21.2 versus 21.5 (neither buyers nor sellers in control), RSI sits at 51.4, and the MACD line at -0.43 holds just above its signal at -0.56, momentum flat but curling up. The 60-day range frames it: $21.53 in mid-May to $39.05 in early July.

That’s what makes the gap interesting rather than alarming. The premarket price around $29.80 lands directly onto the support shelf, the 100-day EMA at $29.81 and the August 6 low at $29.53, rather than through it.

With average true range at $2.55, about 8% of the price, the move is roughly one ATR arriving all at once: it changes the price without, so far, changing the structure. The weekly timeframe leans mildly constructive underneath: price above the weekly 20-period EMA at $29.77, weekly MACD positive, and the weekly 200-period EMA far below at $25.83.

A trendless, coiled stock is getting knocked into its own support cluster by a print the market graded C-minus on profitability, while about 30% of the float is short. Gaps like this resolve on the follow-through sessions, not at the open. The levels below do the sorting.

How to trade the sorting.

Post-Earnings Setup

Day-1 notes: because the gap is roughly one ATR, the pre-report levels still govern, and the crowded short side means moves in either direction can overshoot. They’re illustrative levels, not instructions.

1. Support test: $29.00 to $29.80, where the 100-day EMA, the August 6 low and the weekly pivot shelf converge, and where the premarket happens to land. Holding this zone through the first 2 sessions is the constructive outcome. Invalidation: a daily close below $28.00, under the weekly support at $28.63, which opens the halfway retracement of the year’s range at $26.39 and the lower band at $25.51.

2. Reclaim confirmation: a daily close back above $32.30, over last week’s highs and the pivot resistance, which would mean the market re-graded the print overnight. Objectives: $34.80 at the weekly retracement level, $36.90 at the upper band, then $39.05 at the July high. On a reclaim entry, invalidation tightens to a daily close back below $30.90, Monday’s low; risking $4 to the crash shelf for a $2.50 first objective isn’t a trade.

3. With about 30% of the float short at 4.5 days to cover, and buyback capacity aimed at weakness, downside overshoots can reverse violently. Let the first 2 sessions pick the side. Patience is a position here.

Bottom Line

The intellectual center of this print is the difference between chosen losses and suffered losses, and the quarter offered evidence for both: 13 points of gross-margin compression on one hand, flat gross-profit dollars per subscriber and an adjusted EBITDA beat on the other.

The real question isn’t whether HIMS can restore its old gross margin; management has already said it won’t in the near term. It’s whether the lower-margin growth generates enough gross-profit dollars, retention and operating leverage to justify the capital being deployed, and Q2’s negative free cash flow says that proof is still pending.

You don’t need to predict the open to trade it: either the $29 shelf holds through the week, or $32.30 gets reclaimed, or $28 goes and the market has voted erosion. The next report isn’t on the calendar yet, though the cadence points to early November. The market didn’t sell the growth. It sold the bill. The next 2 quarters decide what the bill was buying.

This is research and commentary, not personal investment advice. Levels and trade plans are illustrative; size positions to your own risk tolerance and time horizon. The author may hold positions in names discussed.

Bargain is a forecast about the earnings. TTD has earned the first label, and Friday’s report is precisely what puts the second in doubt: revenue growth of just 3.0% YoY, followed by Q3 guidance of at least $650 million, a floor that implies roughly a 12% YoY revenue decline. The screen is real. So is the reason it triggered.

Key Takeaways

• TTD dropped 21.9% on Friday to $13.80, a fresh 52-week low intraday, after Q2 revenue grew just 3.0% YoY and the company guided Q3 to at least $650 million, roughly 12% below last year.

• The value screen is real: 16.4x trailing earnings, about 8x EV/EBITDA, 2.17x sales, more than $1 billion of net liquidity, and the $434 million booked as debt is mostly operating leases.

• The slowdown is leaking into profitability: adjusted EBITDA fell 11% YoY with 5 points of margin compression, and Q2 free cash flow of roughly $136 million was down sharply from a year ago.

• The sell side isn’t calling it a bargain: a hold consensus, mean targets clustered in the mid-$14s near the price, and 4 same-week downgrades including Citi to sell at $11.

• With the next report about 87 days away and roughly 18% of the float short, nothing forces a decision on day 1. The levels below define when this becomes investable rather than merely cheap.

The Screen That Says Bargain

Give the bargain case its due, because parts of it are genuinely right. At $13.80, TTD trades at 16.4x trailing earnings, about 8x EV/EBITDA, and 2.17x trailing sales for a business running a 76.9% gross margin. The balance sheet is the strongest part of the story: $1.1 billion of cash plus roughly $360 million of short-term investments, while the $434 million classified as debt is mostly operating-lease liabilities rather than borrowed money.

Call it what it is, an effectively debt-free company with more than $1 billion of net liquidity and free cash flow north of $800 million over the trailing year. This isn’t a broken company; it’s a profitable, high-gross-margin business that can fund its own turnaround without asking anyone’s permission.

The screen also isn’t picking up some accounting mirage. Fiscal 2025 revenue grew 18.5% to $2.9 billion, the 3-year revenue CAGR runs at 22.4%, and trailing-12-month revenue stands at $2.99 billion with a 13.6% net margin. The stock, meanwhile, is down roughly 74% over 12 months against a 21.7% gain for the S&P 500, from a 52-week high of $56.77. Fiscal-2025 EPS grew 15.4% while the price collapsed, which is another way of saying the multiple did all the falling. That’s the setup value investors dream about.

Until you ask why the multiple fell.

What the Screen Misses

A single-quarter stumble doesn’t take a stock from $56.77 to $13.80. A trend does, and TTD’s growth trend is 2 years of deceleration: +27.3%, +22.3%, +25.4%, +18.7%, +17.7%, +14.3%, +11.8%, and now +3.0%, one brief uptick in early 2025 and nothing since. Q2 revenue of $715.1 million came with net income of $64.4 million, down 28.6% YoY, and GAAP diluted EPS of $0.14.

Then came the guide: Q3 revenue of at least $650 million against $739 million a year ago, a floor that concedes roughly a 12% decline. That’s company guidance, not analyst pessimism.

The slowdown is also already leaking into operating leverage. Adjusted EBITDA fell 11% YoY, from $271 million to $241 million, with the margin compressing 5 points from 39% to 34%, and non-GAAP net income dropped from $203 million to $158 million.

That’s cleaner evidence than the GAAP decline, because nobody can wave it away as tax or stock-comp noise. Free cash flow tells the same story in softer form: still strong over the trailing year, but the quarter’s own contribution, roughly $136 million, was down sharply from a year ago. Strong in absolute terms; no longer immune.

Which raises the question the whole bargain debate turns on: is the reset cyclical, execution, or structural? Management pointed to a tough environment for large advertisers, consumer packaged goods and autos in particular, and the reporting around the print adds pressure from closed ecosystems, pricing and execution questions, and the Publicis dispute.

Against that, CTV and audio kept growing at double-digit rates, with international CTV strong. If the $650 million floor is a kitchen-sink number, macro plus Publicis disruption plus internal cleanup, then sub-8x EV/EBITDA with a billion of net liquidity is genuinely interesting. If TTD is instead losing pricing power or share to Amazon and the closed ecosystems, then 16x trailing isn’t cheap at all, because normalized earnings power sits below trailing earnings. That fork, not the multiple, is the investment decision. The screen sees a 16x multiple on trailing earnings; the trajectory says those trailing earnings may be the high-water mark, not the base.

The noise around the print doesn’t settle the fork either. The leadership team was heavily reshaped over the summer, with a new CFO, CMO and chief commercial officer arriving since June, though Jeff Green remains CEO. The forward multiple is a moving part by construction: the dashboard prints 13.2x, but estimates are being revised as you read this, and quotes run into the mid-teens depending on the provider.

When the E is falling, a low P/E is a moving target. The sell side reflects it: a hold consensus, mean targets clustered in the mid-$14s, essentially at the price, and 4 same-week downgrades, Citi to sell with an $11 target plus Raymond James, Truist and Susquehanna. Latest reported short interest runs about 73.6 million shares, roughly 18% of the float, positioning color that can fuel bounces but isn’t a thesis.

Now the price structure.

The Structure After a 22% Day

Technical levels below are based on the August 7 close, computed with standard settings and verified against TradingView defaults; readings elsewhere may differ slightly. Friday’s anatomy: the stock closed Thursday at $17.67, opened at $12.87, printed a fresh 52-week low at $12.83, and clawed back to $13.80 on more than 5x average volume. That leaves price 24% below its 20-day EMA at $18.07 and beneath the entire volatility envelope, under the lower Bollinger Band at $15.64.

Daily RSI at 28.4 is oversold, but the MACD line at -0.56 dropped below its signal at -0.33 on Friday itself, momentum flipping negative in a single session, and the directional readings are one-sided: -DI at 45.1 versus +DI at 18.0, sellers in complete control. The weekly timeframe removes any ambiguity about context, with ADX at 34.9 marking an established, strong downtrend and every weekly EMA stacked overhead from $21.11 to $60.05. Average true range has expanded to $1.30, close to a 10% daily swing at this price.

One detail argues the bulls aren’t extinct: Friday closed 7.2% above its open, so buyers did show up at the low. They just haven’t proven they can hold the ground, and oversold is a description, not a signal, the day after a guidance reset.

The map overhead is simple because Friday left a vacuum: the first hurdle is Friday’s $14.57 bounce high, then nothing but gap until $17.39, then the 20-day zone around $18.07 to $18.23. The next earnings report is currently expected on November 5, about 87 days out, though the company hasn’t formally confirmed the date. There’s no catalyst between here and there, only estimate revisions and whether the low holds.

The screen says cheap and the screen isn’t wrong. But every trend measure on the page says the market hasn’t finished repricing the growth reset, the first reclaim test at $14.57 hasn’t been attempted, and there’s no catalyst for roughly 87 days. Cheap plus falling is a watchlist entry, not a portfolio entry.

What would change that.

Post-Earnings Setup

Day-1 rules after a guidance crash: fresh lows have no memory, gaps act as ceilings until reclaimed, and time is on the buyer’s side when the next catalyst is a quarter away. They’re illustrative levels, not instructions.

1. Base setup: $12.90 to $13.60, and only after the $12.83 low has held for at least a week or 2 of sideways trade. There’s no established support in that zone, only Friday’s low, so time is the confirmation. Invalidation: a daily close below $12.50, which puts price in territory with no visible support from the past year of trading.

2. Reclaim confirmation: a daily close above $14.60, over Friday’s bounce high. Objectives: $15.60 at the lower Bollinger Band, $17.40 at the top of the gap, then approximately $18.20 at the 20-day zone. On a reclaim entry, invalidation tightens to a daily close back below $13.60; risking $2 to the crash low for a $1 first objective isn’t a trade.

3. With roughly 18% of the float short and somewhere between 4 and 6 days of average volume needed to cover, any stabilization can travel fast, which is exactly why the reclaim trigger beats anticipation. Patience is a position here.

Bottom Line

The bargain case is half right, and the half it gets right matters: an effectively debt-free, cash-generating business at 16x trailing earnings and about 8x EV/EBITDA can fund its own turnaround. But a bargain requires the earnings to hold, and everything new from Friday, the 3.0% growth print, the $650 million floor, the margin compression, the downgrades, argues the estimates are still finding theirs.

The real decision is the fork: if this is macro plus Publicis plus cleanup, the price is a gift; if it’s share loss to closed ecosystems, trailing earnings overstate normal. You don’t have to answer that today. Watch for the reset to prove itself: estimates stabilizing, the $12.83 low holding for weeks rather than days, and a reclaim of $14.60 as the first mechanical confirmation.

If the guide turns out to be a kitchen-sink number, there will be time and room to act above $14.60 with the crowd still short. Cheap is a fact about the price. Bargain is a forecast about the earnings. Wait for the forecast to earn its evidence.

This is research and commentary, not personal investment advice. Levels and trade plans are illustrative; size positions to your own risk tolerance and time horizon. The author may hold positions in names discussed.

Market Conditions

The broad tape broke higher this week. The S&P 500 (SPY) closed at 773.26, Trend-Hold and up 3.5% on the week, while the Nasdaq-100 (QQQ) at 723.03 is Trend-Hold and up 5.1% on the week. Beneath the surface, regime breadth improved: Breakdowns fell from 16 to 6, Repairs fell from 12 to 7, and Trend-Holds rose from 15 to 27. Regime upgrades outnumbered downgrades 27 to 4, a broad advance that carried the indexes with it. The advance was led by the highest-beta growth and AI names: RDW +58%, PLTR +40%, SHOP +29%, CRWV +26% to the upside, against OSCR -11%, XRPI -4%, FTI -3%, XRH0.L -2% on the downside. Fresh damage showed up in BIDU and XRH0.L, dropping to Breakdown. The cleanest signal was a full reclaim of Trend-Hold in AMAT, ASML, COPX and GLD, among others. That is the confirmation the prior week was waiting on: the reclaim now runs from the individual names all the way up to the indexes themselves. Broad participation is the strongest version of this signal, but it also means the easy part of the move is behind us. The framework still engages at defined levels: chasing extended names here is how a good tape turns into a bad entry.

What Changed From Last Week

• Reclaimed Trend-Hold: AMAT, ASML, COPX, GLD, LLY, NVDA, PLTR, QQQ, RBRK, RDW, SHOP, TSM, XOVR. The strongest signal of the week. Several got there in a single week, so the weekly chart still has to hold it.

• Lifted out of Breakdown into Transitional: AIXA.DE, CRWV, ETHA, INTC, LRCX, LULU, NOC, NVO, PALL, SLV, SOFI. Early bounces that still need to prove themselves.

• First repair off the lows: ASTS, NFLX, ORCL. Damaged names showing the first sign of basing.

• Cooled out of Trend-Hold into Transitional: OSCR. It lost its trend and now has to re-prove it.

• Lost the trend, into Repair: NBIS. Downgraded a full step, treat it as damaged until reclaimed.

• Broke down: BIDU, XRH0.L. Fresh damage, the framework moves these to the avoid list.

How Last Week’s Setups Played Out

We publish the levels in advance so they can be judged afterwards. That cuts both ways, so here is the whole scorecard, not the flattering half of it.

The featured five did their job.

1. SPY. The plan wanted the breakout above 755.58. It triggered on Monday and tagged the 758.45 first target the same day, closing the week at 773.26.

2. AVGO. Monday’s low at 374.61 carried price back through the 381.00 to 387.70 pullback band, and from there it cleared the first target at 407.52 and the second at 416.02 by Tuesday, closing at 427.76.

3. PANW. The pullback band at 327.25 to 330.27 filled on Monday’s low of 330.00, the 368.80 first target was reached on Wednesday, and price closed at 363.86.

4. BRK.B. The breakout above 516.66 triggered and price closed at 521.80. The 526.44 first target is still live, with the week’s high at 525.44.

5. ITA. The honest one of the group. The 251.49 breakout doubled as the first target, so it triggered and immediately hit, ran to 256.60, then gave the move back to close at 250.75. When entry and target are the same level, you get a round trip rather than a trade, and that is a weakness in the setup rather than a feature of it.

Four of five reached a first target and none came near an invalidation, which averaged out to about 5.9% against the S&P’s 3.5%. Worth keeping in proportion: in a week this strong, a basket of long setups reaching T1 is a good deal closer to what the tape was handing out than to genius. The real test of this section is the first flat week, not this one.

Now the part that cost you. The framework’s avoid list was the best-performing group in the universe. Those sixteen names averaged roughly 15.6% on the week, and four of them, RDW at 58%, PLTR at 40%, SHOP at 29% and CRWV at 26%, were the biggest gainers we cover.

A rule that waits for a reclaim will never put you in the first leg off a bottom, and in a violent snapback week that is exactly what it costs. The other side of the ledger is why the rule exists at all. RDW and CRWV had sat on the avoid list for six straight weeks before this one, and over that stretch they fell 67% and 48% from their summer highs. The rule that kept you out of those six weeks is the same rule that kept you out of this one. You do not get to skip the drawdown and catch the first bounce with a single rule, and anyone selling you a framework that does both is selling hindsight.

One outright bad call. OSCR sat on last week’s Trend-Hold list tagged “buy dips” and fell 10.6% to 27.90, straight through the 27.81 to 28.08 stop band. If you took it, the stop did its job and you are out. It is now Transitional and has to re-prove itself before it earns a place on any list again.

How Our Trade Plans Work

Our framework is rules-based and level-driven. We classify each name into a regime (Trend-Hold, Transitional, Repair, or Breakdown) using the 1D EMA stack, cross-checked against the 1W chart. Every entry, target, and invalidation traces back to a specific indicator level. T1 is the first profit-take layer, T2 is the structural objective, T3 is the stretch target on full continuation. Sizing follows the regime: Trend-Hold names take normal size but only on a pullback to support or a confirmed breakout; Transitional names take half size or wait; Repair names take a starter only after a reclaim; Breakdown names are avoided until a reclaim confirms.

Two stops, two jobs. The trade stop protects the position; the regime invalidation changes the thesis. The trade stop is the tactical level where a single trade is wrong and you cut it. The regime invalidation is the deeper level where the entire classification flips to a lower regime and the reason to own the name is gone. They differ on purpose: one manages risk, the other manages conviction.

Cash is a position, and waiting is part of the plan. This update ranks 60 stocks and ETFs: the highest-conviction, near-entry names get a full setup, and the rest are tracked in the watchlist table so nothing in the universe goes uncovered.

This Week’s Playbook

Start here. The detailed setups follow for the top names, and the full watchlist table covers the rest.

• Closest to a defined entry: GOOG, AMAT, XOM, LLY, TSM, ASML, plus 4 more tagged Near entry below. Price is already sitting on the 1D EMA20, so the trigger is near and you are not chasing. Near is not the same as cheap, so read the stop on each one before sizing. One caveat on that list. AMAT, LLY, TSM and ASML reclaimed their trend from well below, so the EMA200 invalidation sits 13% to 24% away. The entry is close, the thesis stop is not. Size those to the tactical stop rather than the regime level, or take half.

• Best leaders to buy only on weakness: PANW, AMZN, LMT, TGT, SHOP, XOVR, plus 1 more tagged Near entry below. Clean trends with room to run, but wait for the pullback.

• Most dangerous chase setups: MSFT, RTX. Overbought leaders; the trend is real but the entry is not, wait for a reset.

• Avoid until a reclaim confirms: BIDU, META, NKE, WMT, XRH0.L, XRPI. No long exposure; not every dip is a setup, and cash is the position.

Top 5 Actionable Setups

The highest-conviction names sitting closest to a defined entry, with full detail, levels, and both stops for each. The next tier follows in the Also Actionable table, and the rest of the universe in the full watchlist below it.

Alphabet Inc. (GOOG), Last close: 353.47

Read more

Key Takeaways

• CRWV’s latest reported quarter (Q1, ended March 31) delivered $2.08 billion of revenue, up 111.6% YoY, on roughly $25 billion of funded debt and total obligations approaching $35 billion.

• CRWV generates roughly $6 billion of trailing operating cash flow, but capital spending has run above $16 billion, leaving the business dependent on external capital despite that cash engine.

• NBIS grew Q1 revenue 683.9% YoY to $399 million, holds $9.4 billion of cash against $9.6 billion of debt, and trades at roughly 7 times CRWV’s trailing-sales multiple.

• Latest reported short-interest data show roughly 19% of CRWV’s float and 30% of NBIS’s sold short, and options imply roughly a 13% CRWV earnings move against a daily range near 8%.

• Neither stock needs to be chased before the prints. Both sit at well-defined technical decision zones, and earnings will decide which risk the market is willing to tolerate.

The leveraged one first.

CRWV: Scale Now, Balance Sheet Later

CoreWeave is the scaled one. The revenue arc across the last 9 reported quarters runs from $189 million in early 2024 to $2.08 billion in Q1 2026, an 11x ramp in 2 years, with fiscal 2025 revenue up 167.9% and the latest quarter still growing 111.6% YoY (reported). Trailing 12-month revenue is $6.23 billion at a 69.4% gross margin, and the quarter’s GAAP results still land red: a $144 million operating loss and a $740 million net loss, or $1.40 per diluted share.

The financing behind that ramp is the story. CRWV carries roughly $25 billion of funded debt, with total debt and lease obligations approaching $35 billion, against $2.3 billion of cash and a current ratio of 0.32. The more telling number is the cash-flow gap: roughly $6 billion of trailing operating cash flow against capital spending above $16 billion, which leaves the business dependent on external capital despite enormous operating cash generation.

At 7.9x trailing sales, the multiple looks tame next to Nebius. Analyst targets remain bullish on average, but the enormous ranges on both names, $36 to $303 here, underline how unsettled the terminal economics remain, and it doesn’t help the mood that a CEO stock sale of $28.2 million rounds out the pre-earnings headlines. What Tuesday needs to show is simple: revenue holding triple-digit growth, and evidence that new capacity keeps getting contracted before it gets financed.

This is a bet that contracted AI demand outlasts a balance sheet built like a leveraged utility.

Technical levels here and below are based on the August 7 close. From a June high near $132, CRWV lost more than half its value to a $60.55 low, and the rebound to $90.67 has reclaimed the 20-day EMA at $82.33 and the 50-day at $88.60.

Momentum has improved, with the MACD line at -1.16 back above its signal at -4.13 and the downside momentum from June now fading, but price is pressing directly into the $93 to $95 ceiling formed by the 100-day EMA, the 200-day and the upper Bollinger Band, while ADX at 18.8 says no new trend has formed yet.

An average true range of $7.58 is about 8% of the price in an ordinary day; it’s the kind of number that says Tuesday’s report can overwhelm every level on the page overnight.

The business is executing. The equity remains a leveraged claim on that execution. And 3 days before earnings, price has rallied directly into the zone where that tension gets tested.

Now the expensive one.

NBIS: A Cash-Rich Balance Sheet at a Nosebleed Multiple

Nebius is the inverse, and one sentence carries the whole section: the balance sheet has bought Nebius time; the valuation has removed its margin for error. The growth is even steeper, revenue of $50.9 million, $100.7 million, $146.1 million, $227.7 million and $399 million across the last 5 quarters, a 683.9% YoY jump in Q1 and 479% for fiscal 2025.

The reported $621 million GAAP profit overstates the underlying business, though. It was driven primarily by a roughly $781 million revaluation gain tied to ClickHouse, while the operating line still lost $128 million. The dashboard’s 93.1% trailing net margin deserves the same asterisk.

The time comes from liquidity: $9.6 billion of total debt almost fully offset by $9.4 billion of cash, with a current ratio of 8.3. That liquidity is real, but it shouldn’t be confused with internally generated free cash flow; recent customer advances and external financing contributed materially to the cash build.

The missing margin for error is the multiple: 54.4x trailing sales on about $0.87 billion of trailing revenue. Side by side, the pair reduces to one line: similar market cap, one-seventh the trailing revenue, roughly 7 times the sales multiple. The bull case is that the sequential ramp (+75% quarter-over-quarter in Q1, at a 72.1% gross margin) compounds into that multiple within quarters, and what Wednesday needs to show is equally simple: another steep sequential step, because at this valuation deceleration is the event.

Positioning adds the accelerant. Roughly 30% of the float is sold short per the latest reported data, some of which may reflect convertible hedging rather than outright bearish positioning, and Michael Burry has disclosed a bearish NBIS position, adding a high-profile skeptic to an already crowded setup. The stock is still up 167.6% over 12 months against 21.7% for the S&P 500, so plenty of momentum capital sits on profits it may want to protect into Wednesday.

The price structure is a stock still digesting a blowoff. From a $299.86 June high, NBIS fell 51% to $145.80 in July and now sits at $187.97, below its 20-day EMA at $200.96 and 50-day at $205.06, resting right on the 100-day at $188.42, with the 200-day far below at $155.56.

The MACD line at -5.54 holds above its signal at -7.58, so the decline is decelerating rather than accelerating, but ADX at 12.8 shows there’s effectively no trend to lean on, and a daily average true range near $25, about 13% of the price, is the kind of volatility that turns ordinary position sizes into oversized ones. That’s the backdrop 4 days before the print.

The growth is real and the liquidity is substantial. But at more than 50x trailing sales, growth itself is the margin of safety. Wednesday decides how much of that growth the market is still willing to prepay.

How to act on it.

Post-Earnings Setup

The honest plan for both names is the same shape: let the reports pick the setup. Any pre-earnings exposure carries materially higher gap risk than a post-confirmation entry, and pre-positioning means accepting binary catalyst risk before price has resolved direction. They’re illustrative levels, not instructions.

CRWV

1. Pullback setup: $82 to $86, around the 20-day EMA and this week’s support. Invalidation: below $76, under the mid-Bollinger shelf at $78.

2. Breakout confirmation: a daily close above $95, clearing the 100-day, 200-day and upper-band cluster in one move. Objectives: $109 (a computed retracement shelf sits at $108.57), then the $132 June high, then approximately $142 at the consensus median.

3. Options imply roughly a 13% earnings move against an 8% ordinary day; assume Tuesday redraws the level structure.

NBIS

1. Pullback setup: $150 to $162, around the 200-day EMA and the July support zone above the $145.80 low. Invalidation: below $143, which reopens the air pocket beneath.

2. Reclaim confirmation: a daily close above $207, back through the 20-day and 50-day EMAs and the daily pivot resistance. Objectives: $236 at the upper band, $255 at the consensus median, then approximately $300 at the June high.

3. Into Wednesday, patience is a position here.

Bottom Line

If forced to own one through earnings, I still prefer CRWV. Not because the balance sheet is safer, it clearly isn’t, but because the valuation already embeds considerably more skepticism and the revenue base is proven at scale. That discount exists for a reason: CRWV’s capital intensity leaves far less room for financing stress than the sales multiple alone suggests. The cleaner setup is to let earnings determine which thesis earns confirmation rather than requiring either thesis to work beforehand. Same price tag, opposite risks. CRWV must prove that demand and financing can support its capital structure. NBIS must prove that growth can catch its valuation. One has to grow into its liabilities; the other has to grow into its price.

This is research and commentary, not personal investment advice. Levels and trade plans are illustrative; size positions to your own risk tolerance and time horizon. The author may hold positions in names discussed.

But the harder question is valuation. Even after the post-earnings pullback, SpaceX is still worth roughly $1.5 trillion while spending more than $18 billion a quarter on capex.

With less than 2 months of trading history, the technicals can wait. For now, the real story is whether the fundamentals can grow fast enough to justify the price.

Key Takeaways

• Q2 revenue of $7.8 billion grew 92% YoY with all 3 segments growing. Adjusted EBITDA nearly tripled to $3.5 billion and the quarterly net loss narrowed by almost half to $541 million.

• AI is the swing factor. Segment revenue rose 247% YoY to $2.6 billion on $14.1 billion of newly contracted cloud compute sales, and segment adjusted EBITDA turned positive at $1.1 billion.

• Backlog sits at $47.5 billion, with roughly 56% expected to convert to revenue within 12 months. Management is guiding to a $100 billion annualized revenue run rate by December.

• The balance sheet was rebuilt in a single quarter: $85.7 billion of net IPO proceeds, $25 billion of investment-grade notes, $100 billion in cash and securities, no material debt maturities before 2031.

• The risks are concentration and capex. 2 customers drove nearly 38% of Q2 revenue, major disclosed cloud agreements include 90-day termination rights, and capital spending is running at $18 billion a quarter.

The Quarter: 3 Businesses, 1 Direction

Consolidated revenue for the quarter ended June 30 came in at $7.81 billion, up 91.9% from $4.07 billion a year earlier (reported). The net loss was $541 million against $1.01 billion in Q2 2025, and adjusted EBITDA of $3.54 billion was up 191%. For the first half, revenue grew 54% to $12.5 billion while the net loss widened to $4.8 billion, but that half-year figure carries a $1.5 billion loss on debt extinguishment from retiring the expensive merger-era debt, a cost of cleaning house rather than of running the business.

Connectivity is the profitable core. Starlink revenue reached $4.3 billion, up 66% YoY and 32% sequentially. Subscribers doubled YoY to 12.0 million, with 1.7 million net additions in the quarter, the best in the company’s history.

ARPU (average revenue per user) held at $66 per month versus Q1, though it’s down from $85 a year ago as international expansion and cheaper plans dilute the blend. The interesting move is upstream: enterprise and government revenue grew 108% YoY to $1.8 billion, with deals signed or rolling out across American, Southwest, Virgin Atlantic, Iberia and Aer Lingus, and more than $6 billion in multi-year U.S. government Starshield awards, primarily from 2 major Space Force programs. Segment operating income rose 79% to $1.7 billion, expanding margins by nearly 3 points, and segment adjusted EBITDA hit $2.6 billion. This is the cash engine that pays for everything else.

AI is the growth engine. Revenue of $2.6 billion was up 247% YoY and 213% sequentially, driven by new Cloud Services Agreements at the Colossus data centers, contracts worth $14.1 billion that contributed $1.6 billion of revenue in their first partial quarter.

Grok and X subscriptions added $258 million of YoY growth; advertising slipped to $367 million from $426 million during an ad-system overhaul. The operating loss narrowed 49% sequentially to $1.3 billion, and segment adjusted EBITDA flipped positive at $1.1 billion, versus a $609 million loss just a quarter earlier. Hold that number loosely: with $15.8 billion a quarter going into GPUs, depreciation follows with a lag, so operating income is the more conservative scoreboard here than adjusted EBITDA.

Installed compute reached 1.4 gigawatts of nameplate draw, up from 0.4 a year ago, with management targeting more than 2 gigawatts by year-end and, per Musk on the call, closer to 10 gigawatts than 5 by the end of 2027, built exclusively on NVIDIA hardware. The $60 billion all-stock Cursor acquisition is expected to close this quarter.

Space is the smallest segment by revenue and the point of the whole enterprise. Revenue of $962 million grew 29% YoY on 10 customer launches, while the operating loss widened to $542 million as segment R&D, which is mostly Starship, climbed to $1.1 billion for the quarter.

The company flew 78 missions and delivered 1,041 metric tons to orbit in the first half, most of it internal Starlink deployment that produces no segment revenue at all. The 2 successful Starship V3 flights matter more than the P&L here: Musk now calls the heatshield problem solved (his assessment, not yet an independently established fact) and a catch attempt of the ship itself could come as early as this month. Space loses money on paper partly because it’s the delivery mechanism for the other 2 segments.

Now the order book.

Backlog, Contracts, and the Run Rate Math

Backlog stood at $47.5 billion at quarter-end, of which $14.3 billion already sits on the balance sheet as deferred revenue. The conversion schedule is front-loaded: roughly 56% is expected to become revenue within a year, 34% between years 1 and 3, and 10% after that. Note the definition: backlog represents amounts under enforceable customer agreements, subject to the company’s stated exclusions, including optional purchases and constrained variable consideration.

The contract flow behind it was dense even by this company’s standards. Beyond the $14.1 billion of cloud compute sales and the $6 billion of Starshield awards, the FCC approved the EchoStar spectrum transfer, 65 MHz that anchors the Starlink Mobile build-out, for total consideration of about $19.6 billion, mostly in stock. On the call, Shotwell framed the mobile ambition against a roughly $600 billion U.S. wireless market dominated by the Big 3 carriers, with satellite service targeted for the end of next year. And in the first weeks of Q3, SpaceX contracted an additional $6.7 billion of cloud services revenue over a 6-month term that begins ramping in October, deals management says pay back their compute capital in under a year.

That’s the bridge to the headline claim: a $100 billion annualized revenue run rate by December, measured as December’s revenue times 12. To be clear, that’s guidance, not a reported number, and it’s aggressive.

Q2 revenue implies a monthly pace around $2.6 billion; the $100 billion run rate needs roughly $8.3 billion a month, more than triple, inside 6 months. The components are at least identifiable: the Google and Anthropic compute ramps starting this quarter and next, the new $6.7 billion of agreements, Cursor consolidating into the numbers, and the Starlink subscriber machine compounding underneath.

Musk went further and said the internal projection for $1 trillion in annual revenue has moved up from 2031 to 2030. I’d file that second claim under ambition rather than modeling input, but the December checkpoint is near-term, specific, and falsifiable. That makes it the most useful thing management said all quarter.

The money to attempt all this is real.

Balance Sheet: From Strained to Fortress in 90 Days

It’s easy to forget how stretched this company was in March: a $20 billion bridge loan, legacy X term loans, xAI notes coupon-ed at 12.5%, and a technical covenant default (waived) triggered by the xAI merger debt. One quarter later the picture is unrecognizable. The IPO raised $85.7 billion net.

Read more

That was enough: the first sub-midpoint quarter in its public history cost the stock 20% in a day and a fresh 52-week low, with the call adding the phrase investors least wanted to hear, model improvements “lighter than normal,” and a Q3 guide without the usual cushion.

What’s left trades near 26x trailing GAAP earnings and about 16x 2027 consensus EPS, and that gap is the whole debate. Cheap is a condition, not a signal.

A 0.3% midpoint miss ended the perfection premium. What remains is a value multiple attached to a 53% grower, and a question.

AppLovin reported Wednesday, August 5 after the close, and Thursday’s session did the grading: down 19.7% to $335.67, with an intraday print at $332.19 that marked a new 52-week low, in a stock that traded at $745.61 in February. Our July 19 piece called this setup in advance, concluding the trend was “broken until 460 is reclaimed”; 460 was never reclaimed, no entry was ever signaled, and readers following that framework were flat through Thursday’s gap. Prices are based on the August 6, 2026 close, a complete reaction day; results are from the release, the call, and the filed quarter.

Key Takeaways

• The quarter grew 53%: revenue of $1.92B landed inside the company’s own guidance range but 0.3% below its midpoint, adjusted EBITDA of $1.61B missed its midpoint by about 1%, GAAP EPS was $3.76, and margins printed 77.7% operating and 66% net, filed numbers, not adjusted ones.

• The blemish was historic by this company’s standard: per Piper Sandler, 30 basis points below the revenue midpoint and 100 below the EBITDA midpoint made this AppLovin’s first sub-midpoint quarter since going public.

• The wound that mattered was qualitative: management described the quarter’s AXON model improvements as “lighter than normal,” which touches the exact durability question that has hung over the story all year.

• The result is a valuation anomaly: about 26x trailing GAAP earnings and roughly 16x 2027 consensus EPS for a 53% grower, with the caveat that the estimates themselves are now the question.

• The map: decision zone 320 to 336 (the stock closed inside it, on a fresh 52-week low), invalidation at a daily close below 298, repair at 392, references 420, 443, and 460.

The Quarter, on Its Own Terms

The filed numbers describe a machine still running hot. Revenue of $1.92B grew 53% from a year earlier. GAAP operating income was $1.49B, a 77.7% operating margin, and net income of $1.27B put the net margin near 66%, margins with almost no comparison in public software, let alone advertising technology.

The stumble that triggered everything was tiny: revenue landed inside the company’s own $1.915B to $1.945B guidance range but 0.3% below its midpoint, and adjusted EBITDA of $1.614B came in $1M under the low end of its range, about 1% below the midpoint. In most stocks, in most months, that’s rounding.

AppLovin hasn’t been most stocks: its premium was built across 2 years of beating without exception, and Piper Sandler flagged this as the first quarter since its IPO below its own guidance midpoints. A streak that long becomes part of the price. Wednesday it ended.

Why 0.3% Cost 20%

The proximate math is expectations; the deeper story is the moat. AppLovin’s advertising engine, AXON, compounds by shipping model improvements that lift advertiser returns, and the bull case has always rested on that cadence continuing. 3 things got repriced at once. The miss broke the streak.

The model comment, improvements “lighter than normal,” made investors question whether the streak can restart. And the Q3 guide offered no relief: $2.055B to $2.085B centers at $2.070B, slightly below the $2.074B to $2.083B the Street had penciled in, a guide without the cushion AppLovin usually builds. Beneath all 3, a quieter reset: BofA wrote that it no longer models AppLovin becoming the large-scale channel for web advertisers it once projected, trimming the e-commerce expansion that carried part of the premium.

Management’s rebuttal deserves equal print: the next model improvement landed just after quarter-end, and the company said it sees no weakening in advertiser demand and no change in the competitive environment. It’s worth being precise about what didn’t happen: growth didn’t stall and margins didn’t crack. The market repriced the second derivative, not the level; the worry isn’t that growth disappeared, it’s that the rate of improvement in the engine driving future growth may be slowing. Our July 19 comparison piece put that exact question at the center of the premium; Thursday was the market marking the probability down.

The Valuation Case, Stated Honestly

Here’s the formulation that tells the story. At $335.67, AppLovin trades near 26x trailing GAAP earnings, the last 4 quarters of diluted EPS sum to $13.01, but only about 16x 2027 consensus EPS, which sits near $21 on the feed our dashboards use, struck at the August 6 close.

That gap is the entire debate: the market is still pricing a roughly 65% earnings ramp from here to there; it has simply stopped paying a premium multiple on top of the growth assumption. Nothing else we’ve covered this season pairs numbers like these; the closest multiple, Qualcomm’s 15.5x, is attached to negative growth.

The catch is the denominator. That $21 estimate already embeds continued AXON gains, gaming strength, and the consumer-advertising expansion the Street is now re-marking. Move the estimates down 15% and 16x becomes about 19x with the price standing still.

A 20% single-day repricing on a 0.3% revenue-midpoint miss says the market isn’t arguing with the multiple; it’s arguing with the E underneath it. The re-marking has already begun: within a day, Piper Sandler cut its target to $385 from $665 with a downgrade to Neutral, BofA went to $430 and Needham to $500, and post-earnings coverage reported further cuts at Wells Fargo (to $357, with a downgrade) and Goldman (to $465), a first-day wave of reductions running roughly 20% to 40%.

Notice what even the freshly slashed targets have in common: all of them sit at or above the price. Our feed’s $656 consensus mean hasn’t absorbed the wave; treat it as an artifact. The next report, expected in November though not yet confirmed by the company, is the next hard data point; until then, the argument is between the filed margins and the estimate path.

The July Framework, Scored

Our July 19 map for AppLovin said the stock was broken until it closed back above 460, offered a stabilization zone at 400 to 414, and set invalidation at a daily close below 392. Score it: 460 was never reclaimed, so the framework never signaled an entry, and the invalidation brushed its line on July 24 with a $391.98 close, 2 cents through it, which the disciplined would treat as fired. Either reading produced the same position into Wednesday’s report: none. The stock then lost 20% in a session. This is what the “broken until repaired” construction is for; the repair never came, so neither did the exposure.

The Map From a 52-Week Low

One possible framework, not a promise, and the honest preamble first: the stock closed below its lower Bollinger band on a fresh 52-week low with a daily momentum reading of 29 (oversold), and its 14-day average true range is 8.8% of price. There is no price structure beneath a 52-week low, only round numbers and older history, which makes the downside rules blunter than usual. All triggers are daily closes.

• Decision zone: 320 to 336. The stock closed inside it, at the fresh low. A hold here on daily closes is the first evidence the repricing has found its level.

• Invalidation: a daily close below 298, under the round $300 the market is already watching. Below that, the market isn’t debating the estimate path anymore; it’s pricing the model itself as broken, and no multiple argument overrides that.

• Repair: a daily close above 392 reclaims the old invalidation line, now the overhead test, with references at 420 (the pre-report shelf), 443 (the 61.8% retracement of the recent leg), and 460, the level the July piece named as the trend repair, which still stands.

• For fresh capital (illustrative): this is the same shape as Meta last week, a falling knife with a nearby round-number floor, and the same discipline applies. The patient routes are a hold of 320 to 336 across several daily closes, or the 392 repair for those who’d rather pay up for confirmation. A fill near 328 against the 298 line risks about 9%, and in an 8.8%-range stock the gap risk runs well past plan; size so a 15% to 20% adverse move stays inside your portfolio risk budget.

• Between now and the next report, expected in November, the evidence is any advertiser-facing data on AXON performance, where the target cuts settle once the first-day wave finishes, and whether 320 to 336 holds on closes.

Bottom Line

The numbers say this is now the cheapest growth story we cover: 53% growth, 78% operating margins, 16x the 2027 consensus. The price action says the market has stopped trusting the denominator, and a sub-1% miss, softer model progress, and a guide without the usual cushion were enough to end a perfection premium that had survived 2 years. Both can be true, which is why this isn’t a bottom-call; it’s a framework. If 320 to 336 holds and the estimate revisions come through modest, the anomaly resolves upward and 392 is the confirmation. If 298 goes, the market is repricing the machine, not the quarter. We’d let the level and the revisions answer it, in that order, before the multiple tempts anyone into answering it early.

This is research and commentary, not personal investment advice. Levels and trade plans are illustrative; size positions to your own risk tolerance and time horizon. The author may hold positions in names discussed.

The challenger is winning share. The market is charging it for the cost of winning.

AMD reported Tuesday, August 4 after the close, into a rally: the stock rose 7% that session to $518.58, capping a 21% rebound over 4 sessions, then gave it back on Wednesday, closing at $482.05, down 7% on the reaction. That lands it exactly at the top of the thesis zone our July 18 trio piece drew, after a month in which the stock first broke out above that piece’s 535 trigger, then crashed through its 432 failure line, then V-recovered into the print, a round trip that deserves an honest accounting below. Prices are based on the August 5, 2026 close, a complete reaction day; results are from the release, the call, and the filed quarter.

Key Takeaways

• The quarter delivered the challenger thesis: revenue of $11.5B grew 50%, data center revenue of $6.7B more than doubled year over year and is now 58% of the company (42% a year ago), adjusted EPS of $1.66 beat the $1.62 consensus, and the Q3 guide of $13.0B at the midpoint, against a $12.5B consensus, implies roughly 41% growth.

• The disclosed commitments are specific: Anthropic up to 2 gigawatts of MI450 compute in Helios racks, Meta up to 6 gigawatts, OpenAI 6 gigawatts, and Microsoft deploying Helios and 6th-generation EPYC at scale across Azure, without a public gigawatt figure.

• The margin story, stated precisely: non-GAAP gross margin came in at 56% (54% on a GAAP basis) and the September quarter is guided to roughly 56% again. Nothing missed; nothing expanded. Against a rally priced for incremental margin leverage, flat was the disappointment.

• Scoring our July map: the 535 breakout fired and briefly paid, the sector crash then forced exits at the 432 failure line, and the stock V-recovered into the print. The stop did its job against an unknowable decline; the rebound made it expensive in hindsight. The process change we adopted: half size on breakout entries in names this volatile.

• The new map: pullback zone 455 to 470, invalidation at a daily close below 424, breakout 531, references 548, 585, and 650.

The Quarter: What Winning Share Looks Like

Strip the reaction away first. Revenue of $11.5B grew 50% on the EDGAR-filed numbers, and the mix shifted decisively: data center, at $6.7B, more than doubled year over year and is now 58% of AMD, versus 42% a year earlier and less than a third of the company 3 years ago. The product cycle behind it is the broadest AMD has shipped, Helios rack-scale systems, MI450-series accelerators, 6th-generation EPYC, and Pensando networking, and the disclosed commitments now read: up to 2 gigawatts from Anthropic, up to 6 gigawatts from Meta, 6 gigawatts from OpenAI, and Azure-scale deployment from Microsoft. The guide put the next marker down: $13.0B at the Q3 midpoint, half a billion above consensus, implying roughly 41% year-over-year growth. One bookkeeping note: GAAP EPS was $1.38 against the $1.66 adjusted figure, the gap being the usual non-GAAP exclusions.

The Margin Question

The reaction wasn’t about a miss, because there wasn’t one. AMD delivered a 56% non-GAAP gross margin and guided the September quarter to roughly 56% again. That’s the issue, stated precisely: flat. Into a stock that had rebounded 21% in 4 sessions, investors were positioned for evidence that the revenue ramp had started producing incremental margin expansion, and the guide said not yet.

Management acknowledged on the call that data-center AI margins remain slightly below the corporate average, with the path from here depending on product and customer mix, which is the careful version of: the early Helios and MI450 ramp dilutes until scale and mix mature. The long-term frame is unchanged and ambitious, more than 35% revenue growth compounded, non-GAAP operating margin above 35%, and EPS above $20, all within 3 to 5 years; at $482, that endpoint is 24x. The market’s question, as it has been all season, is the path.

This is the third variation of the pattern on our pages this summer: GE Vernova fell on margin conversion inside a record demand quarter, Meta fell on costs arriving before returns, and AMD now falls on margins arriving flat when the rally had priced them rising. Futurum’s Shay Boloor put the positioning plainly: “It was priced for something much closer to a blowout... this was not an exceptional result.”

Scoring the July Map

The July 18 trio piece mapped AMD at $495.76: thesis zone 465 to 482, breakout above 535, structural failure below 432. The map’s first call worked: the breakout fired on July 21 and paid for 3 sessions, with closes of $544.43, $552.33, and $539.69. Then the sector-wide selloff that took roughly $1.3 trillion out of semiconductors reversed everything in its path; AMD fell through the zone and, on July 29, closed at $429.56, below the failure line, where the rules forced exits. 2 sessions later the rebound began that reached $518.58 by the eve of earnings.

Here’s the honest scorecard, without flattering it or flogging it. The 432 line existed because a close below it meant the uptrend structure was gone, and on July 29, mid-crash with no floor in sight, it was; nobody rings a bell announcing which breakdowns become the 40% kind, and the same discipline kept the AST SpaceMobile and Meta breakdowns from becoming far larger losses this summer.

In July’s case the market turned 2 sessions after the exit, so the insurance was paid for and went unused. That’s the standing bargain with stops: most months they’re what keeps a drawdown survivable, and occasionally the premium comes due. The genuine lesson isn’t that the map was wrong; it’s a sizing rule we’ve now made formal: breakout entries in 8%-daily-range names get half size from the start.

What the Price Assumes Now

At $482.05, AMD trades at 34.6x next-year consensus earnings on the feed our dashboards use, struck at the August 5 close, down from 37x when the trio piece ran. Depending on the provider and EPS convention, the trailing GAAP multiple sits above 100x and isn’t especially useful for judging the forward ramp.

On the same feed, the Street’s mean target is $579, about 20% above the price, with the median at $600 and a range from $320 to $1,250, the widest top end we track, which tells you how unsettled the out-year story is; per CNBC’s August 5 analyst roundup, JPMorgan raised its target from $385 to $550 while staying at Neutral, which captures the ambivalence neatly.

The bull case is arithmetic: if the $20-plus EPS target lands on schedule, today’s buyer paid 24x it. The bear case is the same arithmetic with a delay: every quarter margins run flat while revenue compounds, the multiple has to carry more of the story. Wednesday’s 7% was the market repricing the carry, not the destination. Worth remembering the trio thesis too: AMD’s print was always going to move the group, and NVIDIA, scheduled to report August 26, now answers with AMD’s margin evidence on the table.

The Map After the Whipsaw

One possible framework, not a promise, and sized for the volatility it lives in: AMD’s 14-day average true range is 8.4% of price, and the last 6 sessions include a 21% rebound and a 7% reaction fade. All triggers are daily closes.

• Pullback zone: 455 to 470, the August 3 low ($455.30), the old zone floor, and the area the rebound built its base in.

• Invalidation: a daily close below 424, under the July 29 panic low ($424.03) and the 100-day average near 429. Below that, the whipsaw resolves downward and the challenger premium is repricing in earnest.

• Breakout: a daily close above 531 clears the entire earnings-week range (the August 4 high was $530.13), with references at 548 (the July breakout shelf), 585 (the June intraday record area; the record close was near $580.91), and 650, the projected range extension carried from our July map.

• For fresh capital (illustrative): a fill near 462 against the 424 line risks about 8.2%, and July just demonstrated a gap can double that. Size so a 15% to 20% adverse move stays inside your portfolio risk budget, and per July’s lesson, any breakout entry above 531 deserves half size with its own stop back under 505.

• Between now and the next report, currently listed for November 3, the evidence is MI450 margin commentary, any new gigawatt-scale commitments, and NVIDIA’s August 26 print, which now doubles as a referendum on whether AMD’s share gains are coming out of the incumbent’s growth or the market’s.

Bottom Line

The quarter proved the share story: 50% growth, a doubled data-center business, and disclosed commitments from Anthropic (up to 2 gigawatts), Meta (up to 6), and OpenAI (6). The reaction refined the season’s rule rather than repeating it: demand is assumed, and now even holding margins flat gets charged when the rally into the print had priced them rising.

July whipsawed our map in both directions, and the scorecard above owns it along with the sizing rule it produced. From here, 455 to 470 is the first zone where the thesis can be evaluated against defined structural risk, 424 marks formal invalidation, and 531 would signal confirmation that the margin question is answered. We’d let one of those 3 speak before adding to the story.

This is research and commentary, not personal investment advice. Levels and trade plans are illustrative; size positions to your own risk tolerance and time horizon. The author may hold positions in names discussed.

• Revenue nearly tripled.

• Backlog hit a record.

• Adjusted EBITDA turned positive for the first time.

It also ended up on the Space Force’s Andromeda programme and picked up 18 missile-tracking spacecraft through L3Harris.

The stock closed Tuesday at $13.96, down 22% this year and 70% below its high.

Subscribe now

Key Takeaways

• Q1 was genuinely good. Revenue of $186.7M was around 3x from one year ago ($62.5M). Gross profit also went from $6.7M to $30.1M, and adjusted EBITDA hit positive $2.7M, the first ever.

• Guidance is the problem. Management targets $900M to $1B for 2026. That requires Q2 through Q4 to average $238M to $271M, so the record quarter has to be beaten by another 27% to 45%, 3 times in a row this year - which isn’t an easy accomplishment.

• Cash fell from $582.6M to $231.6M last quarter, after the $800M Lanteris purchase. The company has since raised funds twice.

• SpaceX completely changed the game with its listing. It listed in June at $135, hit $225.64 in the same week, and now trades at $125.33, below its own IPO price. Every small-cap space name lost its scarcity premium when that happened.

• The framework says avoid until it reclaims. Price sits below every daily moving average with sellers in control. There’s one level worth watching, and it’s below here.

Subscribe now

About the Company

Intuitive Machines builds and flies lunar landers, satellites, and the ground network to communicate with them. It’s the company that put the first US commercial lander on the Moon.

Today, it’s far different from what it was 18 months ago. The acquisition of Lanteris Space Systems shifted it from a mission-services company to a vertically integrated manufacturer, and a pending deal for Goonhilly Earth Station and COMSAT adds global ground stations. Management calls the result a next-generation prime contractor. Trailing revenue is $328.2M with market cap around $2.24B.

Its customer base is almost evenly split: 38% civil, 35% commercial, and 27% national security.

Subscribe now

News and Pipeline

The contracts are real but mostly not revenue yet

The Q1 haul was substantial. $428.9M in new awards, driven by the Space Development Agency’s Tranche 3 tracking layer and a $180.4M CLPS award from NASA in March. Backlog jumped $842M in a single quarter to $1.1B.

Since then, 2 things landed:

• The Space Force selected Intuitive Machines for Andromeda, the programme replacing its GSSAP surveillance satellites. The company cites a $6.24B ceiling, but that amount is allocated across 14 companies.

• 18 spacecraft for missile tracking, ordered through L3Harris under the SDA’s accelerated Tranche 3 tied to the Golden Dome initiative, built on the company’s IM-300 platform. The stock rose 5% on the news.

• Still pending: awards for the lunar terrain vehicle, where the company has submitted both an uncrewed design (Moon RANGER) and a crewed one (Moon RIDER). Management expected decisions “in the coming weeks” back in May.

• NASA has also expanded CLPS 1.0 from $2.6B to $4.2B and added a $6B CLPS 2.0 for heavier cargo beyond 2028, against what the company frames as a $20B Moon Base opportunity split across 2 phases.

The pipeline is good, but the issue is converting it on schedule, with enough cash to get there.

The Funding Gap

Cash went from $582.6M at year-end to $231.6M by the end of Q1. Most of that is the Lanteris purchase, but the operating business also burned cash. LUNR burned around $54.8M in Q1, plus $9.9M of capex, for negative free cash flow of $64.6M.

The company has gone to the market twice. A $175M private placement in February to fund Lanteris, and a $500M at-the-market equity programme disclosed in June, which is roughly 9% dilution against about 160M shares outstanding.

That’s the trade-off in plain terms. You’re being asked to fund a land grab. Whether that’s a good deal depends entirely on whether the backlog converts.

Subscribe now

The SpaceX Effect

For years, owning or was one of the few ways to own the space economy. That ended when listed.

It peaked at $225.64 within 4 days. It now trades at $125.33 ($112 in pre-market today, down a further 10%), below the IPO price, with a market value near $1.65 trillion.

When the category leader is available directly, capital doesn’t need a proxy. And when that leader is itself down 22% on the year, the smaller players get hit. Rocket Lab is off 2% this year, down 16%, Intuitive Machines down 22%. is the only one up by about 18%.

Subscribe now

Fundamentals

Gross profit went from $6.7M to $30.1M year over year, and gross margin on the quarter reached roughly 16%, well above the 8% trailing figure. Adjusted EBITDA turned positive at $2.7M. For a company that lost $19.1M on that measure just one quarter earlier, that’s a genuine inflection.

• Operating loss widened to $(39.2M) from $(10.1M) a year ago. SG&A of $50.7M includes $20M of one-time acquisition costs and $6.3M of stock compensation tied to Lanteris, so the underlying number is better than the headline, but it’s still a loss.

• TTM figures are not great but they predate the acquisition. Revenue $328.2M, net loss $109.3M, operating cash flow negative $88.5M, free cash flow negative $134.0M.

• Stockholders’ equity is negative $334.3M, and has been negative in all 12 reported quarters. Some of that is the capital structure it inherited from going public through a SPAC. It still means there’s no book value cushion underneath you.

• The current ratio is 1.22, so short-term assets cover short-term bills, but not by much.

How does it compare to peers?

Well, Rocket Lab runs a 37% gross margin and grew 64%. The market pays $46.5B for it.

Redwire runs a 13% gross margin on 58% growth for $2.5B.

Intuitive Machines is closest to Redwire on quality and size, and it trades at $2.2B.

The stock is not expensive when compared to management guidance. At $2.24B against the $950M guidance midpoint, that’s roughly 2.4x 2026 sales.

Rocket Lab trades at many multiples of that. The thing is that Rocket Lab’s revenue exists and Intuitive Machines’ is a forecast requiring 3 consecutive record quarters.

Fundamental conclusion: the business inflected and the balance sheet did not. First positive EBITDA, record backlog and a credible move into national security all argue the story is working. Negative equity, a 60% cash drawdown, 2 capital raises in 6 months and guidance demanding 27% to 45% sequential growth from a record base all argue you’ll be asked to fund it again. At 2.4x guided sales you’re paid for that risk. At 6.8x trailing sales, you’re paying for the guide to be right.

Subscribe now

Technicals

There’s no way to make this chart look good.

Price at $13.96 sits below every daily moving average: the 20-day at $14.69, the 50-day at $19.04, the 200-day at $19.42 and the 100-day at $20.79. When the 50-day, 100-day and 200-day all sit 36% to 49% above spot, the damage is structural, not a dip.

• ADX at 36.5 means a strong trend is in force, and the direction is down, with sellers at -DI 28.9 against buyers at +DI 17.7.

• RSI is 40.7, weak but nowhere near washed out.

• MACD is improving, at -2.31 against a signal of -2.77, so the histogram has turned positive at +0.46. That’s deceleration, not a reversal.

• ATR is $1.46, which is 10.5% of the share price. This name moves more in a day than most move in a month, and any stop has to respect that.

The weekly chart holds the only level that matters. Price is below the weekly EMA20 at $20.24 and the weekly EMA50 at $18.64, but it remains above the 200 week at $12.85. That average has defined the floor of this entire structure, and it sits just under the swing low at $11.29.

Technical conclusion: a Breakdown, not a base. Our framework’s rule for names in this condition is to stay out until price reclaims something, and nothing has been reclaimed. The one constructive detail is that the weekly 200-day at $12.85 is still holding, roughly 8% below spot. That is where a speculative buyer has defined risk. Everywhere else is guessing.

The Trade Plan

Earnings land Thursday, August 13, so everything below carries the same warning we give every name into a print. This one is riskier than most, because a 10.5% daily range means a bad number can move this stock 20% before you’ve finished reading the release.

The framework’s default answer here is no position. Price is below every average with a strong downtrend confirmed. That said, here are the levels that would change the picture.

• Speculative starter: $12.85 to $13.40. This is the weekly 200-day, the last structural support in the name. Starter size only, and only if you accept that you’re buying a downtrend.

• Stop: below $11.20. That sits under the swing low at $11.29. Losing it opens the 52-week low at $7.78, and there is nothing in between.

• The reclaim that matters: a daily close above $15.00. Above there, the 20-day flips from resistance to support and the first real objective is the weekly 50-day at $18.64.

• Targets:

  • T1 at $14.69 at the 20-day

  • T2 at $18.64

  • T3 at $20.79 where the 100-day sits

• If it gaps down through $11.20 after earnings, there’s no trade. That’s a fresh leg lower and the next reference is $7.78.

The stop sits around 15% below the entry zone, because a name with a 10.5% ATR cannot carry a tight stop. Position size has to be small enough that a 15% loss is survivable. If it isn’t, the trade is too big.

Bottom Line

Intuitive Machines is a genuinely improving business inside a broken chart.

The bull case is straightforward. Backlog of $1.1B, first positive EBITDA, a place on a Space Force programme with a multi-billion ceiling, 18 spacecraft ordered through L3Harris, and a Moon economy NASA is funding at scale. At 2.4x guided 2026 sales, none of that is priced as a certainty.

The bear case is equally straightforward. Guidance requires beating a record quarter by up to 45%, 3 times consecutively. Equity is negative, cash fell 60% in a quarter, and the company has raised twice in 6 months with a $500M programme still open. Meanwhile SpaceX’s arrival removed the reason anyone had to own the small caps by default.

The analyst mean sits at $39.22 against a $13.96 price. When the gap between the analysts and the market is that wide, one of them is badly wrong.

One number decides it: $12.85. Above the weekly 200-day, this is a beaten-down growth story with defined risk. Below it, the last support in the structure is gone and $7.78 comes into view.

We don’t predict. We prepare.

All content provided is for informational and educational purposes only and does not constitute investment advice, a recommendation, or an offer to buy or sell any security. The trade plans, levels, and scenarios discussed are illustrative frameworks based on market structure and are not guarantees of performance. Markets involve risk, and losses are possible. Past performance is not indicative of future results. Each reader is responsible for their own investment decisions, position sizing, and risk management

Both report tomorrow, Wednesday August 5.

Both sell the same molecule class into the same demand. And for the first time since we started covering them together, they are not telling the same story.

Subscribe now

Key Takeaways

• Both companies raised guidance. Only one raised into growth. Lilly went to $82B - $85B of revenue, 28% growth, and lifted EPS guidance to $35.50 - $37.00. Novo’s raise took it from -5%/-13% to -4%/-12% adjusted sales growth.

• Novo’s headline quarter was not what it looked like. Reported sales of DKK 96.8B, up 32%, include a one-off non-cash reversal of a USD 4.2B rebate provision. Strip it out and sales were DKK 70.1B, down 4%.

• Lilly is winning the open market. Its data puts US incretin share at 60.1% against Novo’s 39.4%, in a market growing 30%.

• You pay 39.6x for Lilly and 11.1x for Novo. That gap isn’t a mispricing to arbitrage. It’s growth priced against decline, and both can be right.

• See the trade plan below for thoughts on how to play this before and after earnings.

What happened since our last article

On May 18, we said rewards trend-following and rewards patience on entry with discipline on the stop.

Lilly: Our pullback zone was $945 to $975. The lowest print since was $978.87, just above the pullback zone.

Price then hit our $1,025 breakout zone trigger. And from there, it hit our T1 at $1,065 and the T2 at $1,210. If you bought above the pullback zone or bought the breakout, you did well (24% return from pullback to T2 / 18% return from breakout to T2).

Novo: Our buy zone was $42.50 to $43.50 which was hit. Sadly, the price then closed below the $42.50 stop and the trade was over. It bottomed at $41.00, then recovered and hit our T1 on June 25.

See our previous coverage of LLY and NVO here:

Subscribe now

Lilly: an oral launch and a raised bar

is a trillion dollar business that turned one molecule into a franchise across diabetes, obesity, sleep apnea and heart failure. Trailing revenue is $72.25B.

The quarter it reported April 30 reset expectations. Revenue of $19.80B grew 56%, adjusted EPS of $8.55 grew 156%, operating income grew 142%.

The breakdown matters as much as the headline numbers:

• Volume added 65%. Price subtracted 13%. Lilly is growing by selling many more units at lower prices.

• Zepbound did $4.2B, up 80%. Mounjaro holds 51% of US prescriptions in its category.

• Beyond the franchise, Ebglyss grew 141% and Jaypirca 79%.

• The US grew 43%. Rest of world grew 178%.

Then there’s Foundayo, the oral. It hit US retail April 9 2026, roughly 35% of launch volume is coming through telehealth, 2 of 3 major PBMs confirmed access by mid-May, Medicare opened July 1, and full consumer promotion only starts this quarter. Wednesday’s earnings will give the first clean look at that ramp.

Management raised revenue guidance to $82B - $85B and EPS guidance to $35.50 - $37.00.

Here’s the detail most previews miss. Consensus for full-year EPS sits at $34.53, below the floor of management’s own raised range. Lilly has beaten 4 quarters running, by 25.9%, 5.2%, 19.2% and 12.5%.

Novo: a price cut, a pill, and an accounting artifact

Novo Nordisk is a $208B Danish company that built this category with semaglutide and is defending a lead it no longer holds in the US. Trailing revenue is DKK 327.8B, ($50.5B).

Its Q1 looked like a blowout and wasn’t.

Reported sales of DKK 96.8B were up 32%. But DKK 26.8B of that, (~USD 4.2B), is a one-off non-cash reversal of a rebate provision. It boosts revenue, gross profit and operating profit, and it does not repeat.

On Novo’s own adjusted basis: sales fell 4%, operating profit fell 6%, and gross margin was 80.6% rather than the reported 85.9%.

The commercial picture underneath is mixed rather than broken. The Wegovy pill has reached more than 1 million people in the US and runs about 207,000 weekly prescriptions. Novo still holds roughly 55% of weekly injectable GLP-1 volume globally, and Wegovy is in more than 55 countries.

But pricing is what shapes the next 2 years. From January 1, 2027, Novo cuts US list prices to a flat $675 per month for Wegovy, Ozempic and Rybelsus, roughly 50% off Wegovy and 35% off Ozempic. That’s why a company with a growing pill franchise guides its sales line lower.

CagriSema is submitted in the US with a decision due in the fourth quarter. It’s the 2027 swing factor, and it already read as inferior to Zepbound in Novo’s own comparison.

Fundamentals

On the income statement they’re very similar. Lilly runs an 82.8% gross margin, 47.3% operating margin and 35.0% net margin. Novo runs 81.9%, 45.3% and 37.2%.

Everything else diverges.

Cash conversion is key.

Lilly turned $20.48B of operating cash flow into $10.37B of FCF, spending 49% on the buildout. Novo generated the equivalent of $18.2B and spent 74% of it, leaving $4.8B as free cash flow. Novo is spending harder, relative to what it earns, to defend a franchise that is shrinking on an adjusted basis.

Balance sheets differ too.

Lilly’s return on equity is 101.3% against Novo’s 66.4%, but Lilly carries debt-to-equity of 1.39 versus 0.72, and Novo’s current ratio of 0.79 means short-term obligations exceed short-term assets.

On valuation, Lilly trades at 39.6x trailing earnings with a 1.04% free-cash-flow yield. Novo trades at 11.1x with a 2.28% yield (Novo reports in kroner and trades in dollars. Converted properly at 6.49 kroner per USD, the real multiple is 11.1x).

The analysts have already picked a side. Lilly carries 22 buy or strong-buy ratings against 4 holds and 2 sells, mean target $1,277, about 14% above spot. Novo carries zero strong buys, 4 buys and 10 holds, mean target $47.76, which is 1.4% above spot.

Fundamental conclusion: nearly identical margins, completely different futures priced in. Lilly is a compounder at a growth multiple with a thin cash yield and real leverage. Novo is a cash machine at a value multiple whose cash is being eaten by capex while adjusted sales decline. The 39.6x against 11.1x gap is the market’s honest read, not an error.

Technicals

Lilly closed at $1,121.36, sitting exactly on its lower Bollinger band. On the daily chart, It’s below the 20-day at $1,171.09 and the 50-day at $1,137.27, but still above the 100-day at $1,085.53 and the 200-day at $1,019.86. RSI is 41.1. ADX at 18.2 means no strong trend in force, just drift. Williams %R at -90.1 says the last 14 sessions have been one-way traffic down.

The weekly says don’t panic. Price holds above the EMA20 at $1,094.05, the 50 at $1,003.34 and the 200 at $752.77, with weekly RSI at 59.0 and ADX at 31.0. That’s a strong long-term trend taking a sharp short-term correction.

Novo closed at $47.09, pinned between its own averages. Below the 20-day at $49.00 and the 50-day at $47.35, above the 100-day at $46.19, and below the 200-day at $48.59 which still caps the whole downtrend. RSI is 43.3 and ADX 22.4, a stronger trend reading than Lilly’s, with sellers more firmly in control. It’s also oversold, with Williams %R at -84.8.

Its weekly is the difference. Novo holds above the weekly 20 at $46.19 but sits far below the weekly 50 at $50.39 and the weekly 200 at $65.15. This is a stock trying to build a base, not one in an uptrend.

Same sector, same week, same oversold readings. One is correcting inside an uptrend, the other is trying to end a downtrend.

Our trade plan

Read more

closed Friday at $36.80.

That’s 52% below the $76.87 high from last November and 48% below the $70.71 it printed in early June. It’s also 27% above the $28.93 low it touched on July 29.

None of those moves were really about the business.

Today it’s above $39.00

Here’s why the name is worth your time:

owns 5 GW of secured grid power, it’s converting that power into AI compute under multiyear contracts with , , Perplexity, and Figure AI.

In June 2026, it showed it could finance the buildout at an A credit rating (instead of just selling more stock).

Key Takeaways

• July’s low was a liquidity event, not an earnings event. A leveraged fund holding 11.7 million shares got margin-called and sold its entire book to Citadel. The stock rallied 30% the day that seller vanished.

• The real de-rating came 4 weeks earlier, on July 1, when Meta said it would lease out spare compute. That turned a hyperscaler from customer into competitor, and it hasn’t been un-said.

• Bitcoin mining is still 76.8% of revenue and shrinking fast. It fell to $111.2 M last quarter from $167.4 M, while AI Cloud nearly doubled to $33.6 M. The handoff is real but early.

• The funding bear case got a serious answer in June. A $3.65 B facility rated A by Fitch covers 96% of the Microsoft GPU spend at a 6% blended cost, with customer prepayments worth roughly 45% of GPU capital spend on newer deals.

• Diluted shares are up 53% from a year earlier and capital spending ran at 3.7x revenue over 12 months. This isn’t a self-funding business yet, and you’re paid to wait only if delivery lands on schedule.

Before the numbers, the story of the last 5 weeks:

Subscribe now

What Actually Happened in July

The sequence matters here, because almost none of this movement came from IREN.

July 1

Meta confirmed it would sell access to spare data centre capacity. The reaction was immediate and brutal for anyone renting out GPUs: a company everyone modelled as a future customer had just become a possible competitor. CoreWeave fell about 14% and Nebius about 17% in the same session. IREN went from $43.32 to $38.82 the following day. The AI infrastructure complex has not recovered that ground.

Mid-July

The de-rating widened into the whole group. IREN slid from $59.96 on June 18 to $33.62 by July 17, a 44% drawdown in a month, with peers down similar amounts. Nothing company-specific happened. The market was simply repricing what a megawatt of contracted GPU capacity is worth when the buyer list might be shorter than assumed.

July 20

IREN announced $2.8 billion of new customer contracts and raised its year-end target. The stock jumped 19.6% on 93.3 million shares, its heaviest session of the move.

July 29 and 30

This is the part that matters. Leopold Aschenbrenner’s Situational Awareness fund, which had run leverage reported as high as 400%, took margin calls from Goldman Sachs, JPMorgan and Bank of America. The fund fell 67% in July and shrank from roughly $45 billion to roughly $10 billion. It sold its whole public equity book to Citadel. IREN was its 4th-largest position at 11.7 million shares.

On July 29 IREN fell 13.6% to $29.31, touching $28.93. On July 30, once the fund’s book had changed hands and the forced seller was gone, it closed at $38.26, up 30.5%.

Subscribe now

The Business and Last Earnings

IREN spent 7 years buying grid-connected land and power in renewable-rich regions, then filling it with computers. For most of that history those computers mined Bitcoin. Now they’re being swapped for GPUs, and the company sells compute by the hour instead of hashes.

Power is the point. Anyone can order GPUs. Almost nobody can energize 480MW this year and another 730 MW next year, because substations, interconnection queues and grid connections take years to permit and build. That head start is the asset, and it’s why NVIDIA signed a partnership covering up to 5GW of infrastructure rather than just selling IREN some silicon.

The quarter ended March 31 shows exactly where the transition sits. Revenue was $144.8 M, flat against the same period a year earlier, with a net loss of $247.8 M and reported diluted earnings per share of -$0.74. Adjusted earnings before interest, taxes, depreciation and amortization came in at $59.5 M, a 41% margin, down for a 3rd straight quarter from $121.9 M.

Underneath that flat revenue line, 2 businesses are moving in opposite directions:

• Bitcoin mining revenue of $111.2 M, down 33.6% from the prior quarter on a lower average Bitcoin price and hardware retired ahead of GPU installation

• AI Cloud revenue of $33.6 M, up 94.2% from the prior quarter

• Impairments of $140.4 M, mostly decommissioned mining hardware, with management saying plainly to expect more of them

• Depreciation of $121.2 M, up from $99.2 M, as the GPU fleet starts hitting the income statement

The top line looks flat because a shrinking business is masking a doubling one. That single fact explains why the reported numbers look so much worse than the contracted ones, and why the stock trades on announcements rather than results.

On contracted revenue, the company lifted its year-end 2026 annualized run-rate target from $3.7 B to more than $4 B on July 20, and said roughly 85% is now under contract. The customer list runs from Microsoft and NVIDIA through Perplexity, Figure AI, Together AI, Fluidstack, Fireworks AI, Fal AI and Hume AI. Agreements average about 4 years. Deals signed since June carry prepayments equal to roughly 45% of the associated GPU capital spend. Cash stood at approximately $7.6 B at June 30, including $1.7 B restricted against the Microsoft financing.

Concentration deserves a mention too. Of the $3.1 B that was under contract as of May, roughly $1.9 B was Microsoft and $0.7 B was NVIDIA. The July additions diversified that meaningfully, which is the reason the announcement mattered more than the dollar figure suggested.

But 1 customer still anchors the plan.

Subscribe now

Fundamental Analysis

Over the TTM, IREN generated $757.1 M of revenue and spent $2.83 B on capital projects. That’s capital spending at 3.7x revenue. FCF was negative $2.13 B, a free cash flow yield of -16.3%.

That ratio tells you what you’re actually buying. This is a construction project with a revenue stream attached, and it stays that way through 2027 at least.

The valuation multiples need care, because 1 of them is broken:

• Price to sales of 17.3x, price to book of 4.93x

• Enterprise value to EBITDA of 31.2x

• Gross margin 53.6%, operating margin -29.2%, return on equity 3.1%, return on assets 1.4%

• Debt to equity of 1.49, current ratio of 3.72

• A trailing price to earnings that screens anywhere from 48x to 171x depending on whose earnings you use, and which you should ignore either way, because trailing net income of $76.9 million comes almost entirely from a single quarter of non-cash fair-value gains rather than operations

IREN shows positive trailing net income and a net margin above 10%. The September 2025 quarter booked a $665 M non-cash gain on financial instruments, which flipped a losing year into a nominally profitable one. Operating margin of -29.2% is the key number.

Set the trailing multiples aside and the forward math gets more interesting. Enterprise value is roughly $14.9 B. Against the company’s own $4 B run-rate target that’s 3.7x, and against consensus revenue of $3 B for the fiscal year ending June 2027, about 5.0x. Neither is expensive for infrastructure growing this fast. Both assume delivery lands.

Against peers, the balance sheet is what stands out:

• : debt to equity 1.49, gross margin 53.6%, flat reported growth, up 123% over 52 weeks

• CoreWeave : debt to equity 7.39, gross margin 69.4%, revenue up 112%, down 32% over 52 weeks

• Nebius : debt to equity 1.32, gross margin 72.1%, revenue up 684%

• Cipher : debt to equity 6.42, gross margin 12.4%, revenue down 29%

• Applied Digital : debt to equity 1.36, gross margin 25.9%, revenue up 407%

Every name in that group burns cash, and heavily. IREN is the only one whose reported growth is flat, purely because mining is falling faster than AI Cloud fills in. But it carries the second-lowest leverage in the peer set, and that’s a direct result of how the buildout is being paid for.

IREN closed $3.65 B of investment-grade GPU financing covering 96% of $5.81 B of GPU capital spend for the $9.7 B Microsoft contract. Blended cost of debt was 6%. Fitch rated it A and DBRS rated it A(low), the highest public rating achieved on GPU collateral and the first such deal done in the US private placement market.

That means that debt markets will now underwrite GPUs as investment-grade collateral, at rates that look nothing like venture funding. A year ago that was an open question and much of the bear case rested on the answer. The chief financial officer put it directly on the last call: roughly 95% of Microsoft GPU capital spend is expected to be funded through prepayments and GPU financing at an average interest rate of about 3%.

Diluted share count went from 218.7 million to 333.7 million in 12 months, up 53%. Convertible notes stand at $3.69 billion. NVIDIA’s right to buy 30 million shares at $70.00 would add another 8.4% if exercised, though with the stock 47% below that strike it’s a problem for a much better day.

Consensus is unusually one-sided. Across 15 analysts the mean target is $81.73 implying 122% upside. Even the lowest published target, $41.00, sits above the current price, which is rare and slightly uncomfortable.

Fundamental conclusion: the assets are real, the contracts are real, and the financing is materially better than the market gave credit for. What you’re underwriting is a 9-month, 30-fold scale-up in the AI business while the legacy business shrinks underneath it. Everything in the bull case is a contract or a target. Everything in the bear case is a reported number. That asymmetry is the whole investment.

Subscribe now

Technical Analysis

Weekly Chart

The weekly trend has rolled over. Price sits below its 20-week and 50-week averages at $44.48 and $41.18, and below the weekly Ichimoku cloud, a band running from roughly $41.00 to $47.76 that marks where the market last agreed on value. Downside pressure remains in control.

The structure since November is a sequence of lower highs, $76.87 then $70.71, and July’s $28.93 undercut March’s $30.76. Lower highs plus a lower low is a downtrend.

The 61.8% retracement, the level that usually decides whether a correction is a pause or a reversal, sits at $31.93. Price broke it on July 29 and reclaimed it the very next session. A break that fails immediately, on the heaviest volume of the entire move, is a liquidation low rather than trend continuation. That distinction is the technical heart of this setup.

Daily Chart

On the daily timeframe the damage is visible and the repair is early. At $36.80, price was below all 4 daily moving averages, with the 1D EMA20 at $38.99 and the 1D EMA50 at $44.13.

With today’s session, price hovering around $39, IREN finally reclaimed the 20-day EMA.

The 200-day average has just flattened and started to roll over for the first time in this cycle, which is the clearest evidence that the multi-year uptrend has genuinely paused rather than merely dipped.

Momentum tells a better story than price does. RSI at 44.1 has climbed from 30.3 on July 29 without yet reaching the midpoint, so there’s room in either direction. ADX 28.0 says a real downtrend is in place, though it’s cooling after peaking at 33.8 on July 17. The MACD line at -3.29 sits below its signal at -3.68, so momentum is still negative, but the gap between them has narrowed steadily since July 21, which means selling pressure is fading even while price goes sideways.

There’s also a mild positive divergence. Price made a lower low into July 29, at $28.93 against $32.22 on July 17, while RSI held roughly flat at 30.3 versus 30.7. Momentum refused to confirm the new low.

The short-term picture is more stretched. StochRSI at 92.4 marks the bounce as overbought on a very short leash, and on the 4-hour timeframe price is pinned right at its 20-period average around $36.89 with a reading of 17.8 on the trend-strength gauge, which is the technical definition of no trend at all. That’s consolidation, and it usually resolves with a fast move.

Below, $34.88 to $35.46 is near-term support from the late-July lows. Then $31.93, the line that decides the bigger picture. Then $28.93, the low that cannot break.

One number frames the risk in all of it. The average daily range is $4.40, roughly 12% of the share price, against a beta above 4. Sizing here isn’t optional, it’s crucial.

Technical conclusion: IREN is basing. The weekly trend is still down and price sits under every daily average, so this isn’t a buy-and-forget setup and calling it a bottom would be wishful. What changed is the character of the selling. July’s flush was forced supply, that supply has been transferred to a buyer who wanted it, and the failed break of $31.93 hands you a defined line to trade against. Patience above $28.93, aggression only above $41.30.

Subscribe now

Trade Plan

• Pullback entry, $32.00 to $34.00. The retest zone. It sits above the July low and just over the $31.93 retracement that failed to break, so you’re buying where the last sellers gave up rather than chasing a bounce that’s already run 27%.

• Breakout entry, above $41.30 on a daily close. That reclaims the post-announcement shelf and the 50% retracement at $40.51 in one move. Below it, every rally is still selling into a downtrend, and the burden of proof sits with the bulls.

• Stop, $27.90. Not arbitrary padding. A close below $28.93 means the failed breakdown didn’t fail after all, the weekly retracement is genuinely lost, and the single best reason for owning this is gone.

• Target 1, $44.50. Where the 50-day average, the upper volatility band and the 61.8% retracement of the June decline sit on top of each other. Expect friction.

• Target 2, $54.50. The base of the daily cloud and the 38.2% retracement, and the first place the longer-term trend could actually flip rather than just bounce.

• Target 3, $66.50. The upper weekly band and June’s supply zone. This one needs the delivery story working, not just the price.

Rolling the stop. At Target 1, lift the stop to your entry, since holding above $44.50 establishes a higher low and confirms the base has held. At Target 2, stop trailing to fixed prices and trail beneath the 1D EMA20 instead, because by then the trend is doing the work and a rigid level will shake you out of the move you waited for.

Sizing. Your risk is the distance from entry to stop, not the dollar value of the position. From $33.00 down to $27.90 that’s about 15%, and it’s wider from a $41.30 breakout unless you raise the stop with the entry. Decide what you’re willing to lose on the idea as a whole, then divide by that distance to get your position size. A position that feels comfortable on a quiet day won’t feel that way when the daily range opens to 12%, and this stock has done that repeatedly.

What Would Change the Read

There are 3, in order of how much they’d matter.

AI Cloud revenue at the next print. Not the target, the actual line. It was $33.6 M. If it doesn’t roughly double again, the ramp is behind schedule and the $4 B target starts looking like a 2027 number.

Commissioning at Horizon. The Microsoft handoff was scheduled for this quarter. Slippage there pushes revenue right and pressures the financing economics, since the debt costs money from the day it’s drawn.

Any equity raise. The June financing bought credibility precisely because it wasn’t equity. A large share issue would tell you the funding architecture isn’t holding, and with the count already up 53% the market’s tolerance is thin.

Bottom Line

is actionable, but not at any price and not in size.

The bull case is stronger than the tape suggests. Power is the binding constraint in AI infrastructure, IREN has 5GW of it secured against a 480MW draw this year, the customer list now runs from Microsoft and NVIDIA down to frontier labs, and June’s financing showed the buildout can be funded with A-rated debt and customer prepayments rather than endless equity. At 3.7x the company’s own run-rate target, you aren’t paying a premium for any of that.

The bear case is simpler, and it’s made entirely of reported numbers. Roughly 77% of revenue still comes from Bitcoin mining and it’s falling. Capital spending is 3.7x revenue. Share count is up 53% in a year. More impairments are coming, because management said so. And measured from that last reported quarter, the AI business has to grow roughly 30-fold in 9 months for the target to be hit.

Both are true at once. That’s why a company with 5GW of secured power and a $9.7 billion contract from Microsoft trades at half its high while the lowest analyst target on the street sits above the share price.

The next real information arrives with fiscal 2026 results. Calendars currently point to Thursday, August 27, about 24 days out, though the company hasn’t formally confirmed the date and some services carry mid-September instead. Either way, what matters in that release is the AI Cloud revenue line and the commissioning schedule at Horizon, not the headline loss, which will be ugly whatever happens.

If you’re already holding, the plan stays live above $28.93 and you don’t need to do anything else. If you’re looking to start, wait for $32.00 to $34.00 or make the stock prove itself above $41.30. Chasing it here, mid-range with the weekly trend still pointing down and the bounce already 27% old, is the one thing that doesn’t pay.

The single level that decides it all is $28.93. Above it, this is a base being built by a company whose contracts are arriving faster than its revenue. Below it, it’s a downtrend that hasn’t finished, and the story can wait.

All content provided is for informational and educational purposes only and does not constitute investment advice, a recommendation, or an offer to buy or sell any security. The trade plans, levels, and scenarios discussed are illustrative frameworks based on market structure and are not guarantees of performance. Markets involve risk, and losses are possible. Past performance is not indicative of future results. Each reader is responsible for their own investment decisions, position sizing, and risk management.