Cloud’s Second Wind: What Amazon, Google, and Microsoft Just Told Us About AI Demand

Amazon, Microsoft, and Google just reignited the cloud race with massive AI-driven growth and record capex. Here’s what their latest earnings tell us about the future of cloud computing.

Every few quarters we get an earnings season that isn’t really about earnings. This one was about infrastructure. Amazon AWS AMZN 0.00%↑, Google Cloud GOOG 0.00%↑, and Microsoft Azure MSFT 0.00%↑ didn’t just print good numbers, they reset expectations for how big, how fast, and how expensive the AI-powered cloud is about to be.

Let’s walk through what they actually reported, what’s underneath it, and what it means for the next 6–24 months.

The numbers that matter

Amazon / AWS
Amazon came in with revenue up 13% to ~$180B, but the real line item was AWS: $33B, +20% YoY. That’s important for two reasons:

1. we’re back at a 2-handle after the post-2022 digestion phase, and

2. AWS did it while still throwing off double-digit billions of operating income. Even after legal and severance noise, AWS is the profit engine. That tells you AI workloads aren’t diluting the margin story yet. Financial Times

Alphabet / Google Cloud
Google passed $100B in quarterly revenue for the first time, but the real flex was Cloud: $15.2B, +34% YoY, >$3.5B in operating income, and this is the killer stat; a $155B backlog, up 46% QoQ. That backlog is a demand receipt for AI infra and large contracts. It means customers are locking in capacity ahead of time because they’re worried about supply. Constellation Research Inc.

Microsoft / Azure
Microsoft’s FY26 Q1 was the cleanest AI tell: Azure +40%, Microsoft Cloud +26%, Intelligent Cloud +28%. Satya basically said: we can sell AI across all workloads, but it’s heavy on GPU-rich infrastructure, so gross margin percent steps down while absolute gross profit steps up. That’s exactly what we saw: revenue soaring, cost of revenue up even faster because they’re scaling AI datacenters. MLQ

So: AWS 20%, Google Cloud 34%, Azure 40%. That is simultaneous acceleration across the Big 3. That doesn’t happen in a weak IT spending environment.

What’s actually driving it (it’s not just “AI”)

Saying “AI drove cloud” is like saying “rain made the river.” Technically true, not helpful. Let’s break it into 4 demand streams we can actually model:

1. AI infrastructure buys – customers paying for GPU/TPU-backed compute, vector databases, model hosting, fine-tuning. This is the most capex-hungry part, and it’s where Google and Microsoft are leaning hardest. Alphabet’s capex guide jumped to $91–93B for the year. Amazon is on a ~$100B capex year because it has to build both retail/logistics and AI infra. Financial Times

2. AI-adjacent storage & networking – if you train and serve models, you’re moving and storing a lot more data. That’s steady, high-margin, and sticky.

3. Application modernization – companies that paused cloud migrations in 2023–24 are resuming, but now they’re doing it because they want to plug into AI services. AI became the forcing function.

4. SaaS/ISV pull-through – as more software vendors ship AI features, their infrastructure bills go up and… right back to the hyperscalers.

AI is no longer a separate budget. It’s the reason cloud budgets are growing again.

The “arms-race capex” isn’t a bug

A year ago, when Google said it would spend $75B+ to support AI, the market flinched. Now Alphabet is talking $91–93B and nobody’s laughing because they just showed 34% cloud growth and a 46% backlog pop. That’s what real demand does: it makes high capex look rational. Bloomberg

Microsoft said basically the same thing in softer words: we’re seeing “growth across all workloads,” but cost of revenue was up 43% because AI infra is expensive. They also told us capex will be “higher for the year.” That’s code for: we’re not done building. Microsoft

Amazon? They’re cutting 14,000 corporate roles and spending ~$100B this year. That’s the tell. They’re slimming opex so they can pour more into infra. You don’t do that unless you’re seeing a multi-year runway. Financial Times

So the pattern is clear:

• Demand is front-loaded.

• Revenue recognizes over time.

• Capex is now.

• Margins compress a bit in the middle.

• Cash flow catches up later.

That’s the cloud-AI cashflow curve we’re in.

Industry implications

Let’s talk beyond the Big 3.

a) It just got harder to compete.
When three companies are each dropping $70–100B a year on infrastructure, you can’t “out-efficient” your way to parity as an independent cloud. The moat is becoming sustained capex + AI ecosystem, not just price. This hurts late entrants and sovereign-wannabes that don’t have scale. (They’ll survive, but mostly in regulated or geo-bound niches.)

b) Chips and power become the true choke points.
All three talked about AI infra demand; behind that is a power, land, networking, and supply-chain story. This is great for the ecosystem around them, GPU and custom accelerators, grid-friendly data centers, optical networking, even backup power.

c) Vertical and sovereign clouds get pulled into hyperscalers, not away from them.
Because the hyperscalers are the only ones that can reliably source enough AI compute, even governments and highly regulated industries will have to do “sovereign on hyperscaler hardware.” That’s favorable to MSFT and GOOGL, slightly less unique for AWS because they were already strong in gov. Google’s huge backlog suggests they’re winning large, complex, often AI-first deals; exactly this category. Constellation Research Inc.

d) Software pricing will shift to “AI + infra awareness.”
If cloud providers’ costs are spiking due to AI, you’ll see that show up in how they price high-intensity services. Expect more tiered/committed/prioritized capacity models. Google’s backlog jump is already a signal: customers are committing to get capacity, not shopping spot.

Near, medium, long term

Near term (next 2–3 quarters)

• Growth stays hot: Azure at 40% won’t stay 40%, but it doesn’t have to, anything north of low-30s is a monster at their base. Google Cloud in the low/mid-30s is now credible given the backlog. AWS at 20% is healthy and still the biggest absolute dollar grower. Financial Times

• Capex stays elevated: all 3 have telegraphed that 2025 isn’t the peak. That will keep investor questions on “AI margins” alive.

• More AI platform announcements: the demand is there; now they’ll race to make it easier to use it (agents, managed RAG, turnkey fine-tuning, vertical AI starters). That improves attach and stickiness.

Medium term (12–24 months)

• Usage monetization catches up. Right now, a lot of AI interest is infra-first. Medium term, the money shifts to applications; Copilot-style, Gemini/Workspace-style, Bedrock-style wrappers on top of models. That’s higher-margin revenue on top of the same infra.

• Customers push on TCO. Once AI pilots become production, CIOs will ask “why is my cloud bill 40% AI?” This is where AWS’s operational discipline and Microsoft’s E5-style bundling are advantages. Google’s answer is “we’re fastest and we can prove it”, good for AI-native workloads.

• Partners matter more. SI/consulting channels will be key because enterprise AI is still services-heavy. Hyperscalers will share economics to accelerate adoption.

Long term (3–5 years)

• Cloud = AI runtime. We stop talking about “cloud vs AI.” Cloud becomes the place you operate AI systems, multi-agent apps, inference farms, data governance, safety layers. Whoever owns that runtime owns the margin.

• Hardware abstraction returns. Once capacity stops being physically scarce, cloud providers will try to re-abstract away the GPU brand and sell you “performance classes.” That’s how they win back gross margin.

• Consolidation at the edge. To reduce latency and egress costs for AI apps, more work happens closer to users; CDN, operator edge, private 5G. Hyperscalers will push managed edge to keep workloads in their orbit.

What each player just signaled

AWS (steady, profitable, building):
Message was: “We’re growing again, we’re still the cash machine, and yes, AI is big enough that we’ll keep spending.” They don’t have to be the fastest grower if they remain the most used and most profitable. They’re playing the long game of AI services on top of Bedrock + custom silicon. Financial Times

Microsoft (distribution monster):
They’re the best at turning AI infra into AI revenue because they can sell it through Azure, through M365, through GitHub, through Dynamics. That’s why the market forgives the margin pressure: it believes in the attachment engine. Azure +40% proves they’re not just riding the same clients; they’re expanding workloads. Microsoft

Google (the surprise right now):
This was the quarter that said: “We belonged in the big cloud conversation all along.” +34% revenue, $3.6B operating income, $155B backlog, that’s not a sideshow, that’s a scaled cloud with visibility. The high capex looks aggressive, but now it’s backed by demand. Constellation Research Inc.

Bottom Line

Cloud isn’t slowing. It’s changing shape. The first decade was about moving servers; the next decade is about feeding models. That’s why the three biggest clouds are spending like it’s 2011 again. The good news: demand is real and visible. The bad news: it’s capital-intensive and favors giants.

This analysis is for informational and educational purposes only. It is not investment advice.