AWS Growth Ends Wall Street's AI Panic
For three weeks, AWS growth had been the number Wall Street was waiting on, the swing vote in a quiet revolt against the artificial intelligence trade. Every hyperscaler that reported earnings this summer was met with the same interrogation: where are the returns on all that spending? Meta's free cash flow had cratered. Microsoft was rationing its own cloud. The mood heading into Thursday evening was skeptical bordering on hostile, and Amazon walked into it carrying the heaviest capital budget in corporate history.
Then the AWS growth figure landed, and the argument shifted overnight. Amazon Web Services expanded 37 percent year over year in the second quarter, its fastest pace since 2021 and a clear beat over the 31 percent analysts had penciled in. The unit that funds Amazon's entire empire was not merely holding the line against Google and Microsoft. It was pulling away. Shares jumped more than 9 percent in after-hours trading, adding well over a hundred billion dollars of market value in a matter of minutes.
This was the AWS growth number the bulls had been waiting all year to see. It reframed a spending story that had looked reckless as an investment story that suddenly looked disciplined. For a market that had spent weeks pricing in disappointment, the reacceleration was less a data point than a verdict.
Anatomy of a Blowout Quarter
The headline results were emphatic. Amazon posted revenue of $200.6 billion, up 20 percent from $167.7 billion a year earlier, comfortably ahead of the roughly $196.5 billion consensus. Earnings came in at $5.75 per share against expectations closer to $1.82, the kind of gap that usually signals either a one-off gain or a business inflecting hard. In Amazon's case it was the latter. You can read the company's own breakdown through its investor relations portal, which lays out how the cloud unit did the heavy lifting.
AWS alone generated $42.2 billion in the quarter, ahead of the $40.5 billion analysts modeled. That AWS growth translated into the profit engine that makes the rest of Amazon's ambitions affordable. Advertising posted another record, and the North American retail operation kept margins expanding even as the company poured cash into data centers. The result was a rare quarter where nearly every segment moved in the same direction at once.
What made the report resonate beyond the numbers was context. Amazon had entered earnings season as the swing vote on the AI trade. If AWS growth had stalled, the bears would have had their proof that hyperscaler capex was outrunning demand. Instead the cloud division delivered its cleanest beat in years, and it did so while its two homegrown units, artificial intelligence services and custom silicon, each cleared a $25 billion annual revenue run rate. Chief executive Andy Jassy summarized the quarter in one word during the call. AWS, he said, is "booming."
The Backlog Tells the Real Story
The figure that should worry Amazon's competitors is not the current AWS growth rate. It is the backlog. Contracted future revenue, the money customers have already committed to spend, swelled from $244 billion at the end of 2025 to $364 billion by the close of the first quarter. A jump of that magnitude in a single quarter does not come from thousands of small enterprise deals. It comes from a handful of enormous ones.
The largest is OpenAI, whose expanded compute commitment accounts for much of the increase. That relationship turns Amazon into critical infrastructure for the most closely watched company in the AI race, and it locks in AWS growth years before the revenue actually books. Backlog of this size functions like a forward order book for a manufacturer. It converts speculative demand into contractual certainty, which is precisely what investors had accused the hyperscalers of lacking.
It also reframes the capital spending debate. A backlog approaching $364 billion is the demand signal that justifies the supply build. Amazon is not constructing data centers on faith and hoping tenants appear. It is racing to pour concrete fast enough to serve commitments already signed. For a company that spent the spring being lectured about speculative overbuilding, the backlog is the most persuasive rebuttal available, and it is measured in hundreds of billions rather than adjectives.
A $220 Billion Bet on the Buildout
None of this comes cheap, and Amazon made no attempt to pretend otherwise. Jassy told analysts the company now expects capital expenditure to reach roughly $220 billion this year, up from earlier guidance near $200 billion. That is one of the largest annual capital programs any company has ever attempted, and it is aimed almost entirely at the servers, chips, and power capacity that AWS growth depends on. Details on the company's infrastructure priorities sit on the AWS platform itself, which continues to expand its regional footprint at a punishing pace.
For most of the year, a number that large would have been read as a threat to margins. On Thursday it read as confidence. Operating income guidance of $20 billion to $24 billion for the coming quarter, up from $19.2 billion a year earlier, told investors that Amazon can carry a record capital load and still expand profitability at the same time. That combination is the entire bull case for the AI trade compressed into a single line of a press release, and it is the one thing skeptics had insisted was impossible.
The strategic logic is straightforward. Custom silicon is the lever. Amazon's Trainium and Graviton chips now sit among the top three data center silicon businesses on the planet, and every workload that runs on Amazon's own processors rather than merchant hardware improves the economics of AWS growth. That is how the company intends to spend $220 billion and still widen margins. It owns more of the stack than any rival, and it collects the savings.
What the Reacceleration Means for the AI Trade
Amazon's report will be studied less for what it says about Amazon than for what it says about the AI trade as a whole. The market had convinced itself that hyperscaler spending had decoupled from hyperscaler demand, that the industry was building capacity faster than customers could absorb it. The AWS growth reacceleration is the strongest counterargument yet. When the largest cloud provider posts its fastest expansion in four years while raising capex, the overbuilding thesis gets much harder to defend. Readers tracking the sector can follow the fuller picture at Blumefield.
There are caveats worth holding onto. A backlog is a promise, not a payment, and it assumes customers like OpenAI keep scaling on schedule. The $220 billion capital figure leaves little margin for a demand air pocket. And a single quarter of reacceleration does not erase the structural question of whether AI revenue will ever match AI infrastructure spending across the industry. Amazon answered the near-term version of that question convincingly. The long-term version remains open.
Still, the tone has shifted, and tone matters in markets driven by narrative. For weeks the burden of proof sat squarely on the hyperscalers to justify their spending. After Thursday, that burden has moved back to the skeptics. Amazon did not just report a good quarter. It reset the terms of the debate, and it did so with the one metric nobody could argue with. The AWS growth number spoke for itself.