Amazon lifted its full-year capital expenditure forecast to $220 billion on July 30, driven by higher memory costs and surging demand for AI compute, as AWS clocked its fastest revenue growth in 18 quarters.
Wall Street has spent months nervously watching hyperscaler capex balloon, waiting for some executive to blink. Andy Jassy didn't blink. Amazon's Q2 2026 earnings report, released Thursday, pushed the company's full-year spending forecast up by $20 billion, from $200 billion to $220 billion, making it the single largest capital expenditure commitment disclosed by any hyperscaler this earnings season. The stock jumped more than 8% in after-hours trading. That is not the market flinching at a bubble. That is the market concluding the spending is working.
AWS revenue hit $42.2 billion in the quarter, up 36.7% year over year and well ahead of analyst expectations for 31% growth. The operating margin came in at 39.4%, producing $16.6 billion in operating income, up from $10.2 billion a year earlier. Amazon's AI and chips businesses each crossed annualized revenue run rates of more than $25 billion, more than doubling from a year ago, according to Jassy. Total company net sales reached $200.6 billion, up 20%, Amazon's first $200 billion revenue quarter.
Here's the thing about the revised number: Jassy said explicitly that even $220 billion will leave Amazon short of what its customers want to buy. He expects the capacity constraint to persist into 2027. That is a significant admission, and it says more about the state of AI infrastructure demand than any analyst projection. You don't keep raising your capex forecast by tens of billions if the economics are going soft. You do it when your sales team keeps turning away business.
The $20 billion increase from Amazon's previous $200 billion guidance was attributed in part to higher memory costs, which reflects a broader reality in the GPU supply chain: the components feeding AI data centers remain scarce and expensive, and the hyperscalers are absorbing the cost rather than slowing down. Nvidia's H100 and successor chips, plus high-bandwidth memory from SK Hynix and Samsung, have seen sustained pricing pressure throughout 2026, and Amazon's revised forecast is a direct consequence of that dynamic playing out at scale.
For context, Microsoft is running at a roughly $190 billion annual capex pace in fiscal 2026, including a $37.5 billion quarter in Q2 FY2026, up 66% year over year. Google's parent Alphabet has guided toward $175 to $185 billion for the year. Meta, which caused its own brief investor panic earlier this year when it raised AI spending guidance sharply, is projected in the $115 to $135 billion range. Amazon's $220 billion sits at the top of that stack, and the combined annual AI capex across the four largest hyperscalers is now tracking close to $725 billion for 2026, according to analysis from Value Add VC. A year ago the combined figure was roughly $410 billion. That is not gradual scaling. That is a sprint.
AWS is closing the perception gap with Azure #
AWS has long been the cloud revenue leader by absolute dollars, but Azure has spent the past two years claiming the narrative around AI growth rate. Microsoft's Azure posted 39% growth in its most recent quarter, which edges out AWS's 36.7%. The gap, though, is narrowing, and AWS's current trajectory is its strongest since 2021. Jassy has been direct about the goal: the $220 billion commitment is framed explicitly around closing the competitive distance on AI workloads, where Azure's deep integration with OpenAI's models gave Microsoft an early positioning advantage.
The Anthropic factor complicates simple comparisons. Amazon's Q2 net income reached $62.6 billion, or $5.75 per diluted share, against $1.68 a year earlier, but $53.4 billion of that came from non-operating pre-tax income largely attributed to its Anthropic investment. Strip that out and the underlying business still beat expectations handily, but the headline profit figure overstates the operational picture. Anthropic, in which Amazon has invested billions, is also a supplier of frontier AI models running on AWS, so the relationship is strategic as much as financial.
What makes the $220 billion figure remarkable is not just the size but the investor reaction to it. Earlier this year, when Meta raised its own AI capex guidance, the stock dropped as markets worried about returns on the spending. Amazon got the opposite response. That divergence reflects something real: AWS is producing visible, accelerating revenue from AI workloads right now, which makes the spending look like fuel rather than a bet. When a cloud division grows 37% year over year at a $169 billion annualized run rate, the capex required to sustain it reads differently than an aspirational number attached to a product still finding its market.
Frankly, the capacity constraint Jassy described is the most telling detail in the entire report. A company spending $220 billion and still turning away demand is not describing a bubble. It is describing a backlog. Whether that backlog holds through 2027 is the real question investors should be sitting with now.
Also read: Travis Kalanick raises $1.7 billion for Atoms as a16z bets on specialized industrial robots over humanoids • Together AI raises $800 million at an $8.3 billion valuation as enterprises abandon closed AI models • A federal judge says the government's case for banning Anthropic has gotten worse