Amazon’s custom silicon operation just crossed a $20 billion annual revenue run rate. Growth exceeds 100% year over year. CEO Andy Jassy calls it one of the top three datacenter chip businesses worldwide.
That’s no small feat. The lineup—Graviton CPUs, Trainium AI accelerators, Nitro security chips—powers AWS’s edge in the AI race. Jassy laid it out during the Q1 2026 earnings call: “If our chips business was a standalone business and sold chips produced this year to AWS and other third parties as other leading chip companies do, our annual revenue run rate would be $50 billion.” The Register broke the story first.
AWS hit $37.6 billion in Q1 revenue, up 28%—its fastest growth in 15 quarters. Overall company sales reached $181.5 billion, up 17%. Net income jumped to $30.3 billion, boosted by $16.8 billion in pre-tax gains from Anthropic investments. But the chip unit steals the show.
Trainium2 offers 30% better price performance than comparable GPUs. It’s sold out. Trainium3, shipping since early 2026, boosts that by 30-40% and is nearly fully subscribed. Trainium4? Reservations are piling up, with broad availability 18 months away. Commitments total over $225 billion. OpenAI pledged roughly two gigawatts of Trainium capacity through AWS, ramping in 2027. Anthropic locked in up to five gigawatts across current and future generations.
Meta signed on for tens of millions of Graviton cores to handle agentic AI workloads. Why? “As AI systems shift from answering questions to taking actions, and as post training and inference scale up, the compute required pulls heavily on CPUs,” Jassy explained. Graviton delivers up to 40% better price performance than x86 rivals. Ninety-eight percent of the top 1,000 EC2 customers use it. Uber’s deploying Graviton4 and Trainium3 for ride and delivery AI.
AWS AI Momentum Accelerates
AWS AI revenue run rate tops $15 billion—nearly 260 times the $58 million from its first three years. Bedrock processed more tokens in Q1 than in all prior years combined. Customer spending there rose 170% quarter over quarter. It now previews OpenAI’s GPT-5.4, with GPT-5.5 incoming, plus Anthropic’s Claude Opus 4.7. A Cerebras tie-up delivers the fastest AI inference yet. Bedrock AgentCore deploys agents every 10 seconds.
But here’s the kicker. Amazon’s deploying over 2.1 million AI chips in the past year, more than half Trainium, paired with over a million Nvidia GPUs. This mix gives customers options. Jassy: “Nobody has a better set of chips across AI and CPU workloads than AWS with Trainium and Graviton.” Power remains the main bottleneck, not chips.
Reuters confirmed the doubling from $10 billion earlier this year, tying it to hyperscalers ditching Nvidia dependence. Reuters. Business Insider noted Jassy floating third-party Trainium sales. Business Insider.
Capex? Amazon plans around $200 billion for 2026, much locked in AI infrastructure. Free cash flow tanked 95% to $1.2 billion trailing twelve months—AI’s price tag. Yet margins should expand. Trainium at scale saves tens of billions yearly in capex, adds hundreds of basis points to operating margins versus third-party chips.
Chips Reshape Cloud Wars
Amazon’s not alone in custom silicon. Google TPUs, Meta MTIA—all real, all scaled internally. But none match Amazon’s external traction. Nvidia’s data center dominance holds via CUDA lock-in, yet Trainium chips frontier labs like Anthropic and OpenAI. That’s a crack.
Recent X chatter amplifies it. Analyst Brian Sozzi highlighted the $225 billion Trainium commits and Meta’s Graviton scale. X post by Brian Sozzi. Others note AWS backlog at $364 billion, perishable sales up 40x year over year.
Jassy’s shareholder letter crystallized the shift. AWS at $142 billion run rate. Chips could go external—racks to third parties. Demand overflowed; two partners sought all 2026 Graviton capacity. Amazon couldn’t fill it.
Risks linger. Execution on Trainium4. Power constraints. Nvidia partnerships continue; AWS buys plenty. But vertical integration widens the moat. Customers get choice—Trainium for training, Graviton for inference, Nvidia where needed.
The stock dipped post-earnings. Why? Guidance baked in heavy spend. Long-term? This positions Amazon as AI infrastructure’s quiet powerhouse. $50 billion standalone potential. Triple-digit growth. Locked-in labs. Chips aren’t a hedge anymore.
They’re the engine.
WebProNews is an iEntry Publication