Amazon CEO Andy Jassy used his annual shareholder letter to outline the company’s aggressive push into AI infrastructure, emphasizing growing demand for its in-house chips and defending plans to spend roughly $200 billion in capital expenditures by 2026. The strategy centers on expanding Amazon Web Services (AWS) capacity as competition intensifies across the AI sector.
Jassy highlighted Amazon’s Trainium AI chips as a key differentiator, positioning them as a lower-cost alternative to hardware from Nvidia. He said demand is so strong that capacity for upcoming Trainium3 and even future Trainium4 chips is nearly sold out. The Trainium business has already reached a $20 billion annual revenue run rate, according to Jassy, underscoring AWS’s growing role in AI model training and deployment.
The CEO also pointed to Amazon’s Graviton CPUs as a challenge to Intel, noting that 98% of the top 1,000 EC2 customers now use the chips. At the same time, Amazon continues to invest in broader infrastructure, including satellite internet through Project Kuiper and expanding data center capacity tied to long-term agreements with partners such as OpenAI.
Jassy framed the spending surge as a calculated response to sustained demand rather than speculative investment, pushing back against concerns of an AI bubble. He cited multiple large, and in some cases undisclosed, customer commitments for AWS capacity as evidence that enterprise appetite for AI infrastructure remains strong despite market skepticism.
