Amazon is making one of the largest AI infrastructure bets in corporate history, tying roughly $200 billion in 2026 capital spending to AWS expansion, custom silicon, and surging AI demand.
In a shareholder letter dated this week, Amazon President and CEO Andy Jassy outlined the company’s strategy to invest aggressively through the current AI cycle, alongside a reprinted 1997 shareholder letter from founder Jeff Bezos that reinforces Amazon’s long-standing “Day 1” philosophy. Together, the documents frame today’s AI buildout as a continuation of Amazon’s historical approach—prioritizing scale, experimentation, and long-term market leadership over near-term financial optics.
Jassy said AWS AI revenue reached a run rate of more than $15 billion in Q1 2026 and noted that demand continues to outstrip available capacity. He positioned Amazon’s custom silicon portfolio—Graviton CPUs, Trainium AI accelerators, and Nitro infrastructure—as central to improving price-performance and controlling AI economics at scale. Beyond AWS, the letter highlights robotics, satellite broadband, same-day fulfillment, grocery, and AI-enhanced services such as Alexa+ as parallel growth vectors, while acknowledging that elevated capital expenditures are currently weighing on free cash flow.
- Amazon plans approximately $200 billion in capex in 2026.
- The majority of spending targets AWS infrastructure, including data centers, power, chips, servers, and networking.
- Jassy said these investments are backed by committed customer demand.
- AWS AI revenue run rate exceeded $15 billion in Q1 2026.
- AWS added 3.9 GW of new power capacity in 2025.
- Amazon expects AWS to double total power capacity by the end of 2027.
- AWS continues to face capacity constraints and unserved demand.
- Amazon’s chips business (Graviton, Trainium, Nitro) exceeds a $20 billion annual revenue run rate.
- Trainium2 has largely sold out.
- Trainium3 is nearly fully subscribed after initial 2026 shipments.
- A significant portion of Trainium4 capacity is already reserved ahead of broad availability.
- Amazon Bedrock runs most inference workloads on Trainium.
- Custom silicon could save tens of billions annually in capex over time, according to Jassy.
- Amazon reported 2025 revenue of $717 billion, up 12% year over year.
- AWS revenue grew 20% to $129 billion in 2025.
- Operating income rose 17% to $80 billion.
- Free cash flow declined from $38 billion to $11 billion due to increased infrastructure investment.
- Amazon has deployed more than one million robots in fulfillment centers.
- The company has committed over $4 billion to expand its rural delivery network.
- Amazon’s low Earth orbit satellite initiative, Amazon Leo, has more than 200 satellites in orbit.
- Leo is scheduled for commercial launch in mid-2026.
- Delta Air Lines plans to deploy Leo connectivity across 500 aircraft starting in 2028.
- Amazon’s grocery business exceeded $150 billion in gross sales in 2025.
- Jassy highlighted AI-driven re-architecture efforts including Bedrock’s “Mantle” inference engine and the rebuilt Alexa+.
“We’re not going to be conservative in how we play this—we’re investing to be the meaningful leader, and our future business, operating income, and FCF will be much larger because of it.”
🌐 Analysis: Jassy’s letter positions Amazon’s AI strategy as a full-stack infrastructure play spanning silicon, power, networking, and cloud services. The emphasis on Trainium, Graviton, and Bedrock indicates a deliberate effort to control both performance and cost across the AI lifecycle, rather than relying solely on third-party accelerators.
The inclusion of Bezos’s 1997 letter reinforces Amazon’s long-term investment doctrine, framing current AI spending as consistent with its original approach to scaling infrastructure ahead of demand. As competitors such as Microsoft, Google, and Oracle expand their own AI infrastructure footprints, Amazon is signaling that its combination of hyperscale cloud, custom silicon, logistics, and emerging platforms like Leo could provide a differentiated systems-level advantage.