Andy Jassy Reveals AWS AI at $15B Run Rate, Amazon Chips at $20B+ in Annual Letter

Amazon CEO Andy Jassy just dropped his annual shareholder letter, and the AI numbers are enormous. AWS AI revenue has hit a $15 billion annual run rate, Amazon's custom chips business is generating over $20 billion annually, and the company is spending approximately $200 billion on capex in 2026 — almost entirely on AI infrastructure.

AWS AI: $15 Billion and Growing

AI revenue within AWS has reached a $15 billion annual run rate as of Q1 2026. That represents roughly 10% of AWS's total $142 billion revenue run rate — and it's the fastest-growing segment within Amazon's cloud division.

This puts AWS's AI business alone at a scale that would make it a significant standalone company. For comparison, many enterprise AI companies would consider $15 billion in annual revenue a moonshot target.

Amazon's $20B+ Chips Empire

Perhaps the most surprising revelation is the scale of Amazon's custom silicon business. The company's three chip lines — Graviton (general compute), Trainium (AI training), and Nitro (security and networking) — together generate over $20 billion in annualized revenue, growing at triple-digit pace year over year.

Jassy went further, suggesting that if the chips business were standalone and sold externally like Nvidia, the annual run rate would be roughly $50 billion. This positions Amazon as a serious competitor to Nvidia in the AI chip space — not just a customer.

"Not on a Hunch"

The letter's most quotable line addresses the elephant in the room: Amazon's massive AI spending. "We're not investing approximately $200 billion in capex in 2026 on a hunch," Jassy wrote, defending the spending that has compressed Amazon's free cash flow from $38 billion to $11 billion.

The $200 billion capex figure puts Amazon in the same league as Meta's $115-135 billion and OpenAI's $500 billion Stargate project in the AI infrastructure arms race. The TSMC packaging constraints make this spending even more strategic — companies that lock up compute now will have advantages for years.

Taking Aim at Nvidia

Jassy's letter took direct aim at Nvidia's dominance by suggesting Amazon could eventually sell its AI chips to external customers. This would transform Amazon from Nvidia's biggest customer into its direct competitor — a move that could reshape the $50+ billion AI chip market.

Amazon's Trainium chips are already used internally for training large AI models, and the company has been steadily improving performance with each generation. If Trainium can close the performance gap with Nvidia's GPUs, external sales could become a major new revenue stream.

Project Leo Satellite Update

The letter also revealed that Amazon's Project Leo satellite constellation already has over 200 satellites in orbit ahead of its commercial launch. This Starlink competitor adds another dimension to Amazon's infrastructure ambitions, connecting AI cloud services with global satellite connectivity.

Frequently Asked Questions

What is AWS AI's current revenue?

AWS AI revenue has hit a $15 billion annual run rate as of Q1 2026, representing roughly 10% of AWS's total $142 billion revenue run rate and growing as the fastest segment.

How big is Amazon's chip business?

Amazon's three custom chip lines (Graviton, Trainium, and Nitro) together generate over $20 billion in annualized revenue with triple-digit year-over-year growth. Jassy says if sold externally like Nvidia, the run rate would be about $50 billion.

How much is Amazon spending on AI in 2026?

Amazon plans approximately $200 billion in capital expenditures in 2026, primarily for AI infrastructure. This compressed free cash flow from $38 billion to $11 billion last year.

Will Amazon sell AI chips externally?

Jassy's letter suggested Amazon could eventually sell its Trainium AI chips to external customers, which would put Amazon in direct competition with Nvidia in the AI chip market.