Today's Market Update For You

AWS Grew 37% to $42.2 Billion in Q2 — Its Fastest Rate in 18 Quarters — While Amazon Raised Full-Year Capex to $220 Billion Because Memory Costs Rose and Andy Jassy Said Capacity Will Still Fall Short of Demand Through 2027

Amazon Web Services reported second-quarter revenue of $42.2 billion, up 36.7% year over year — its fifth consecutive quarter of acceleration and the fastest growth rate in 18 quarters — generating $16.6 billion in operating income at a 39.4% margin, a 650 basis point expansion from the year-ago period. Amazon's total consolidated revenue reached $200.6 billion for the quarter, up 20%, with operating income rising 43% to $27.5 billion. On the same earnings call, CEO Andy Jassy raised the full-year capital expenditure forecast from approximately $200 billion to $220 billion, citing elevated high-bandwidth memory costs as the primary driver of the additional $20 billion — and told analysts that even at that spending level, Amazon will not have enough infrastructure to meet customer demand in 2026 or 2027. The AWS contracted backlog reached $496 billion, up $132 billion in a single quarter, with demand already committed through 2028.

The two most important numbers in the release sit in tension with each other in a way that captures the operational paradox of the AI infrastructure cycle. The 650 basis point operating margin expansion — occurring simultaneously with a 76% year-over-year increase in capital expenditure from the prior year's $125 billion — provides the clearest available evidence that Amazon's Trainium custom chip strategy is generating real financial results. Rather than purchasing Nvidia GPU capacity at market prices that reflect the same memory shortage adding $20 billion to Amazon's capex bill, Trainium allows Amazon to capture more of the infrastructure economics internally. Both Amazon's AI revenue and its custom chip business independently crossed a $25 billion annualized run rate in Q2, each growing at triple-digit percentages year over year. Jassy stated that Anthropic and OpenAI — which he described as the two leading AI labs globally — have each made multi-year, multi-gigawatt capacity commitments on Trainium, providing the blue-chip validation that establishes the platform rather than the revenue run rate alone.

AWS Q2 2026 — The Key Numbers

AWS Q2 Revenue

$42.2B (+37%)

Fastest growth in 18 quarters; 5th consecutive quarter of acceleration

AWS Operating Margin

39.4% (+650bps)

650bp YoY expansion — occurring simultaneously with a major capex increase

AWS Contracted Backlog

$496B

+$132B in a single quarter; demand already committed through 2028

2026 Capex (Raised)

$220B

Up from $200B; $20B increase driven by elevated HBM memory costs

The Trainium Thesis — Why Custom Silicon Drives Margin While Capex Rises

Buying Nvidia GPUs at Market	Building Trainium In-House

Market pricing reflects HBM shortage — inflated input costs flow directly to AWS	Trainium3 delivers ~30% better cost-performance than comparable Nvidia GPU instances
Nvidia captures the margin on every accelerator unit sold — not AWS	At full scale, Jassy projects "tens of billions" in annual capex savings and hundreds of bps of margin improvement
GPU supply subject to Nvidia's production timeline and export controls	Trainium annualized run rate $25B+; Anthropic and OpenAI both committed multi-year, multi-GW capacity
No third-party revenue from the chip itself	External Trainium sales "highly likely" — could add a third revenue stream Nvidia does not share
The 650bp margin expansion while capex rose 76% is the clearest available evidence that the in-house chip strategy is already producing financial results — not a future promise.

The structural insight embedded in Jassy's demand framing — a "barbell" of AI labs and startups on one end, and enterprises still in the early stages of production inference on the other — is that the AWS growth story has not yet reached the enterprise demand cycle that historically drives the cloud industry's largest revenue phase. The first wave of hyperscaler AI spending was driven by model developers building and training at scale; the second wave, which Jassy described as enterprises moving workloads to the cloud and integrating AI into production applications, has only begun. AWS's contracted backlog growing by $132 billion in a single quarter and demand visibility extending to 2028 provides a revenue floor that the current growth rate, impressive as it is, does not fully capture.

Sources: CNBC · AWS About Amazon · MLQ.ai · Fierce Network · TechTimes