Annual letter: Andy Jassy says AWS' AI revenue has hit a $15B annual run rate as of Q1 and that Amazon's internal chips business is generating $20B+ per year

Andy Jassy's new letter to Amazon shareholders is a data-heavy defense of the tech giant's biggest bets — from AI and custom chips …

GeekWire 2026-04-09 Todd Bishop

Context & Ripple Effects

Amazon has progressively reframed generative AI from a strategic priority into AWS’s primary growth case: its 2024 shareholder message characterized the technology as a major platform transition, while Jassy later raised AWS’s long-range sales ambition on AI demand. The new disclosed run rates put current revenue behind that narrative rather than leaving it solely as an investment thesis.

The letter also pairs AI revenue with Amazon’s internal-chip business and a roughly $200B 2026 capital-expenditure plan. That makes the case for vertically linking cloud demand, compute capacity and custom silicon, after Amazon had already signaled a sharp AI-led increase in 2025 capital spending.

First-order effects

AWS gains a clearer commercial proof point for AI services, while Amazon can present custom chips as a meaningful business alongside the cloud infrastructure they support.
Amazon’s stated 2026 capex plan commits it to continued buildout of AI capacity; its cloud customers gain access to a larger supply of AWS-hosted AI infrastructure and services.

Second-order effects

Microsoft and Google face added pressure to demonstrate that their own AI infrastructure spending is translating into durable cloud revenue, not only model development and capacity commitments.
A larger internal-chip contribution gives Amazon more incentive to steer AI workloads toward its own silicon, potentially sharpening price and performance competition for cloud compute.

Third-order effects

If AI-service revenue continues to scale alongside bespoke silicon, hyperscalers may increasingly compete as vertically integrated AI infrastructure operators rather than primarily as sellers of general-purpose cloud capacity.
The durability of the capex cycle will increasingly depend on whether enterprise AI workloads sustain utilization and revenue growth fast enough to support ever-larger infrastructure commitments.

The trend: This is a data point in the shift from AI experimentation to a capital-intensive, vertically integrated cloud race in which revenue, utilization and proprietary compute must reinforce one another.

Discussion

@edzitron.com Ed Zitron on bluesky
Two years in and $200bn+ capex and you're at $3.75bn a quarter? Fire Andy Jassy! Fire everybody involved in this pointless escapade! This is a terrible business! [embedded post]
@bwnash Brian on bluesky
Oh, look! Only a $200b annual expenditure to generate $15b in revenue. — How fortunate we are to share an era with a true business genius like Jassey! [embedded post]
@wado1 Luther Yungblood on bluesky
So AI revenue is 15b a year — But AI expenses are at least 200b a year — Am I reading that right? Dmh [embedded post]
@downingark Frank Downing on x
Jassy's shareholder letter is making Amazon's return on AI CAPEX even more clear: - Most of the AWS CAPEX for 2026 is already backed by customer commitments - AWS could be growing even faster than 24% with more capacity. Multiple customers have asked to buy all Graviton CPU
@0xmentalillness @0xmentalillness on x
Some interesting bits from Jassy [image]
@ljkawa Luke Kawa on x
Amazon's most consistent, fundamental corporate strategy, the one that's made the e-commerce/cloud giant into the powerhouse it is today, is also its AI strategy: $AMZN https://www.aboutamazon.com/ ... [image]
@benbajarin Ben Bajarin on x
I liked the $AMZN shareholder letter. I take point 4 with a grain of salt (trn is largely AWS as a customer and Anthropic). Points 5 and 6 are most important for the “durability” of the cycle debate. https://www.aboutamazon.com/ ...

Chronicles