Amazon Is Making Returns from AI – A Whopping US$15bn In Just One Quarter!

It is that time of year again when Amazon releases its annual shareholder letter (along with the 1997 letter all over again) and the bottom line was that Amazon is making returns from AI – real returns, $15bn in a quarter to be exact! In our view, that number reframes both Amazon’s trajectory and the competitive dynamics of the global tech sector.

Yes, Amazon has entered the monetisation phase of the AI cycle earlier, and at greater scale, than most observers expected. The company is not simply participating in the AI boom; it is shaping the infrastructure, economics, and pace of the entire sector.

What are the Best ASX Stocks to invest in?

Check our buy/sell tips

Amazon Is Making Returns From AI: Take Its Word For It!

The most commercially significant disclosure is that AWS’s AI revenue run rate surpassed US$15bn in Q1 2026, roughly 260 times the size of AWS’s entire business at the same age. This is not a forward projection or a directional claim. It is realised revenue, booked inside a single quarter, and it signals that AWS has become the default environment for enterprise‑grade AI workloads.

Jassy attributed this to four structural advantages: breadth of model‑building tools, lower‑cost inference via custom silicon, proximity to enterprise data, and the operational reliability that AWS has spent two decades institutionalising. These are not new talking points, but the difference now is that they are converting into cash flow at scale.

The letter also makes clear that AWS’s growth is constrained not by demand but by power and datacentre capacity. The business added 3.9GW of new power in 2025 and expects to double total capacity by 2027, yet still faces unserved demand. Two large customers reportedly attempted to buy all available Graviton capacity for 2026. That detail alone illustrates the imbalance between AI demand and global compute supply.

The chips business becomes a pillar and capex becomes a competitive moat

Amazon’s custom silicon strategy was once viewed as a margin‑protection exercise but has become a standalone economic engine. Trainium2 sold out, Trainium3 is nearly fully subscribed and Trainium4 is already partially reserved 18 months ahead of launch. Jassy estimates the chips business now carries a US$20bn revenue run rate, rising to ~US$50bn if monetised like a traditional semiconductor vendor.

The strategic implication is straightforward. If Amazon controls the economics of inference at scale, it controls the most expensive layer of the AI stack. That advantage compounds: lower inference costs attract more workloads, which drive more data gravity, which increases switching costs, which strengthens AWS’s position as the default AI platform.

The letter is unusually explicit about the capital cycle. Amazon expects to deploy roughly US$200bn in capex in 2026, much of it tied to AI infrastructure. Importantly, Jassy notes that a substantial portion of this spend is already backed by customer commitments, including the widely reported US$100bn OpenAI agreement.

This is the part of the letter that matters most for investors. Amazon is signalling that the capex curve is steep but predictable, and that the free cash flow trough is temporary. The company has been through this cycle before with the first cloud wave. The difference now is that the revenue opportunity is larger, the customer commitments are firmer, and the competitive barriers are higher.

Reinvention as operating doctrine

Jassy uses the phrase “straight line was a lie” to describe Amazon’s history of non‑linear progress. It is a useful framing for the company’s willingness to reset even successful products. The Bedrock team rebuilt its inference engine from scratch in 76 days. Alexa underwent a full cognitive rewiring to become a generative‑AI‑native assistant. Retail interfaces are being redesigned from first principles.

These examples are not anecdotes; they are signals. Amazon is preparing for a world in which AI reshapes every customer interaction, and it is willing to disrupt its own products before competitors do.

What it means for Amazon

The letter positions Amazon at the centre of the AI economy. Retail remains a US$600bn business with structural room to grow. AWS is a US$142bn run‑rate platform with accelerating AI demand. The chips business is scaling into a new profit pool. And the company’s willingness to invest through the cycle, supported by customer‑backed capex, gives it a multi‑year visibility that few peers can match.

The most important takeaway is that Amazon is no longer investing in AI for strategic optionality. It is investing because the returns are already material. US$15bn in a quarter is not a promise; it is a baseline.

What it means for the broader tech and AI sector

Jassy’s letter also functions as a macro signal. The AI cycle is entering its industrial phase. The bottleneck is no longer model quality but compute, power, and datacentre throughput. Companies with the balance sheet, operational discipline, and silicon strategy to scale infrastructure will define the next decade of technology.

For hyperscalers, the message is clear: the AI market will not be won by model labs alone. It will be won by those who can deliver inference at scale, at predictable cost, with global reliability. For enterprises, the implication is that AI adoption is shifting from experimentation to production. The economics are becoming clearer, the use cases more repeatable, and the infrastructure more standardised.
And for investors, the letter marks a pivot point. The AI narrative is moving from promise to profit. Amazon is the first major proof point.