For a decade, the knock on Amazon was that it was a low-margin retailer cosplaying as a tech company. Q2 2026 retired that argument for good. AWS grew 36.7% year over year to $42.2 billion — the cloud's fastest quarter in 18 — and Amazon responded by telling Wall Street it plans to spend an almost incomprehensible $220 billion on capital expenditures this year. The spending is the story. But so is this: the bet is already paying off.
| AWS Growth | 36.7% YoY to $42.2B (fastest in 18 quarters) |
| Total Revenue | $200.6B (+20% YoY) |
| AWS Operating Margin | 36.8% (record, vs Google Cloud 35.6%) |
| 2026 CapEx | ~$220B (raised from $200B) |
| AWS Backlog | $496B, growing triple digits |
The headline numbers land like a freight train. Total revenue hit $200.6 billion, up 20% year over year. Operating income surged 43% to $27.5 billion. And the crown jewel delivered: AWS operating income of $16.6 billion on a 36.8% operating margin — a record, and richer than Google Cloud's 35.6% in the same stretch. Nearly 61% of Amazon's entire operating profit now comes from AWS. The e-commerce giant is, at this point, an AI infrastructure company with a very successful retail side hustle.
The $220 Billion Question
Let's not sugarcoat the other side of the ledger. Amazon raised 2026 capex guidance to roughly $220 billion — up from the $200 billion it guided just a quarter ago, partly because memory prices are inflating every data center bill on earth — and it burned $7.6 billion of free cash flow over the trailing twelve months. That's the bear case in two numbers: a company spending like there's no tomorrow.
Here's the counterweight. AWS finished Q2 with a $496 billion backlog, growing triple digits, with $57 billion of new commitments added in a single quarter. That's not hope. That's booked, contracted demand for compute that hasn't been built yet. When customers sign ten-year agreements for capacity that doesn't exist, the capex isn't a gamble — it's fulfillment of orders already on the books.
Chips, Claude, and the Flywheel
The most underrated part of this quarter: Amazon's custom silicon is no longer a science project. Trainium and Graviton each serve more than 100,000 customers, and the AI and chips businesses each now run at an annualized pace above $25 billion. Anthropic — the company training the models everyone else is chasing — committed to spend more than $100 billion on AWS over the next ten years, locking in up to five gigawatts of Trainium-powered capacity. Over a million Trainium2 chips are already training Claude, and Trainium3, coming online this year, pairs with NVIDIA's NVLink Fusion in a unified Graviton-Trainium instance — Amazon building its own AI stack with NVIDIA as a supplier, not a gatekeeper. On the software side, Bedrock now runs Claude for more than 100,000 customers, and tools like Claude Cowork and Bedrock AgentCore are quietly turning AWS into the default operating system for enterprise AI agents.
Scoreboard: AWS vs. Azure vs. Google Cloud
Yes, Azure grew 43% and Google Cloud is sprinting off a much smaller base. But here's the math that matters: AWS just grew 36.7% on a $148 billion trailing-twelve-month base. Nobody in the history of enterprise software has grown that fast at that scale. Synergy Research pegs AWS at roughly 31% of the cloud market, against Azure's 21% and Google Cloud's 14%. And in Q2, Amazon locked up arrangements with both OpenAI and Meta — two of the biggest AI consumers on the planet. The private-lab deals are stacking up like dominoes.
The Transformation Is Real
Strip away the noise and here's the actual story: Amazon has completed the most profitable pivot in corporate history — from retail giant to landlord of the AI economy. Advertising chipped in $19.8 billion, up 26%, and retail keeps printing cash to feed the machine. The soft Q3 guide ($197–$202 billion against a $204 billion consensus) is a calendar artifact — Prime Day moved from July to June — and Amazon says growth would be roughly 400 basis points higher without that shift.
Wall Street got the memo. At least 15 brokerages raised their targets after the print, with JPMorgan and Barclays both moving to $365, and the stock remains a consensus strong buy with a mean target around $313. The market finally believes the returns are real.
The risk is real too, and it's simple: Amazon is spending more than any company in history, and if AI demand hiccups, the biggest spender feels it first. But the backlog, the ten-year contracts, and the fastest AWS growth in 18 quarters all point the other way — the demand isn't slowing, it's accelerating. Amazon isn't just winning the AI cloud war. It's building the infrastructure the winner will need for the next decade.
Disclosure: The Signal holds no position in AMZN. Positions may change. This is not financial advice.




