AWS +37%, a $1 Trillion Vision, and the CapEx the Market Keeps Getting Wrong

Three months ago, Amazon raised its capital expenditure guidance and the market punished the stock. The logic was familiar: more spending today means less profit today, and analysts who model spreadsheets quarter by quarter responded accordingly.

Q2 2026 answered that logic with data.

The Numbers That Matter

AWS grew revenues by 36.7% year over year in Q2 2026 — its fastest growth rate in eighteen quarters, and the fifth consecutive quarter of acceleration. In a single quarter, AWS added more than $4.6 billion in incremental revenue, which Andy Jassy described as approximately 80% more than Amazon's previous best quarterly revenue addition in its history.

The operating margin for AWS reached 39.4%. That is not a software margin accidentally achieved by a hardware business. It is the margin profile of a business with genuine pricing power, high switching costs, and a cost structure that scales more efficiently as it grows.

The remaining performance obligation — the contractual backlog of future revenue already committed by customers — stands at $496 billion, growing at triple digits. That number represents years of visibility into future cash flows that no quarterly earnings model can adequately capture.

The $1 Trillion Statement

Andy Jassy did not hedge his language. He stated that Amazon has always believed AWS could become a several-hundred-billion-dollar business and that he now believes it will be at minimum double that — and very possibly a trillion-dollar annual revenue business with attractive free cash flow and return on invested capital.

To contextualise that ambition: AWS currently generates revenues at roughly a $120 billion annual run rate. Reaching $1 trillion would require roughly an eightfold increase from current levels. Given that the cloud market itself is in its early stages of penetration — with the majority of enterprise workloads still running on-premises — and that AI inference is beginning to create an entirely new category of cloud demand that did not exist three years ago, the trajectory is credible in a way it would not have been a decade ago.

Jassy added that demand already committed for 2028 is, in his word, surprising. Companies are not just signing contracts for next quarter. They are locking in multi-year cloud commitments because they understand that the infrastructure they need will be constrained and that securing capacity now is a competitive advantage.

Why the CapEx Is Not the Problem

Amazon raised its 2026 capital expenditure guidance to $220 billion, up from an earlier estimate of approximately $200 billion. The increase was driven in part by higher memory costs — which, incidentally, validates the thesis behind owning the leading memory supplier in the AI stack.

The critical statement from Jassy: even with $220 billion in CapEx, Amazon will not have enough capacity to satisfy all the demand it has for 2026. And he expects that dynamic to persist into 2027.

This is not a story about reckless spending. It is a story about structural undersupply in an industry where demand is growing faster than the fastest-spending company in the world can build. When you cannot build fast enough to meet the demand you already have in hand, capital expenditure is not a risk factor. It is a bottleneck on growth.

Fernando Sánchez's framework is useful here: these hyperscalers are transitioning from asset-light businesses — valued for their capital efficiency — to asset-heavy infrastructure businesses valued for the same reason Buffett values railroads and pipelines. The moat is no longer the algorithm. It is the physical infrastructure that took years and hundreds of billions of dollars to build and that cannot be replicated in any competitive timeframe.

The total estimated investment across the major hyperscalers in AI infrastructure reaches approximately $743 billion by end of 2026, with projections toward $1.1 trillion by 2027. Against that investment, the compute capacity deficit is expected to persist for four to five years — meaning pricing power is structurally embedded for the foreseeable future.

Two Business Lines Worth Watching

Beyond the cloud narrative, two elements from Jassy's Q2 commentary deserve attention.

Trainium chips. Amazon's proprietary AI chip business has already reached a $25 billion annualised revenue run rate, growing at triple digits year over year. Anthropic and OpenAI have both signed multi-year commitments to use AWS with Trainium. And Jassy explicitly confirmed that Amazon is actively exploring the possibility of selling Trainium chips independently — outside of the AWS cloud context — to third-party customers who want the hardware without the platform. If that product line launches, it transforms Amazon from a cloud infrastructure provider into a chip supplier competing in the same addressable market as Nvidia.

Grocery. Amazon is now the second-largest food retailer in the United States. Monthly active customers for perishable products have grown more than 50% since the beginning of 2026. Six of the twenty top-selling items on Amazon.com are now fresh food products. Same-day delivery orders are averaging more than three times as many units per order compared to the prior period. This is not a peripheral business. It is a frequency driver that brings customers back to Amazon's commerce ecosystem multiple times per week — and it converts them to higher-spending Prime members in the process.

Advertising. The advertising business grew 26% in Q2. The strategic advantage of Amazon's ad inventory is conversion: an ad on Amazon is not an impression that might eventually lead to a purchase. It is a purchase opportunity that the customer is already in the process of considering. Prime Video sports viewers spend 12% more and place 17% more orders. Alexa's conversational shopping is becoming a meaningful channel. The advertising business is approaching the scale where it belongs in the same sentence as Google and Meta.

The Recurring Pattern

Q2 2026 is the third consecutive quarter in which a hyperscaler delivered exceptional cloud growth numbers and watched its stock price initially punished by a market focused on capital expenditure. In each case, the punishment was followed by a reassessment once the numbers were properly understood.

The CapEx is not the problem. The CapEx is how you build a business that generates 39% operating margins on $120 billion in annual revenues while growing at 37% per year. The market occasionally forgets this. The underlying economics of the business do not.

---

This analysis draws on Amazon Q2 2026 earnings commentary from Andy Jassy and Fernando Sánchez's analysis on Invertir desde Cero. For informational purposes only — not financial advice.