Amazon CEO Andy Jassy disclosed that AWS cloud computing capacity is sold out through 2027 with substantial 2028 commitments already booked, driven by AI infrastructure demand. Amazon plans $220 billion in capital expenditure for 2026 and has secured over $225 billion in future Trainium chip revenue commitments. Amazon's Trainium chips deliver 30% to 40% better price-performance than comparable processors. Amazon reports third-quarter results on October 29, 2026.
AWS capacity constraints
- ▪Amazon is scheduled to report its third-quarter results on October 29, 2026
- ▪Amazon Web Services offers cloud infrastructure through services including Elastic Compute Cloud (EC2), which provides virtual servers
- ▪Amazon plans to spend about $220 billion on capital expenditure in 2026
- ▪Amazon Web Services has sold out most of its cloud computing capacity for 2027
- ▪Much of Amazon Web Services' 2026 cloud capacity is already booked
- ▪Andy Jassy spoke on a video podcast with Andreessen Horowitz co-founder Ben Horowitz and general partner Erik Torenberg
- ▪A substantial portion of Amazon Web Services' 2028 cloud computing capacity is already committed
AI demand dynamics
- ▪The middle of the market, where everyday business software has yet to adopt AI widely, remains largely untapped
- ▪Businesses are using AI for specific tasks such as customer service and fraud detection
- ▪Andy Jassy expects the middle market segment could eventually become the largest source of AI demand as businesses expand their use of AI
- ▪Andy Jassy described AI demand as a barbell, with AI laboratories consuming computing resources on an unprecedented scale at one end
- ▪Amazon Web Services offers Amazon Bedrock, which gives customers access to AI models
Inference cost barriers
- ▪Amazon is developing its own chips to improve performance and reduce computing costs
- ▪Andy Jassy said AI inference costs are still an order of magnitude too expensive for AI to be used as pervasively as possible
- ▪Andy Jassy suggested AI inference costs need to fall roughly tenfold for AI to become more widely used
Trainium chip development
- ▪Amazon's Trainium revenue commitments include agreements involving Anthropic and OpenAI
- ▪Amazon Trainium chips offer 30% to 40% better price-performance than comparable processors, according to Andy Jassy
- ▪Amazon has disclosed more than $225 billion in future Trainium revenue commitments
- ▪Andy Jassy credited Amazon's investment in Anthropic with supporting Amazon's chip efforts
- ▪Andy Jassy compared Amazon's Trainium chip development effort with the company's earlier decision to develop Graviton processors when it sought better performance per dollar from server chips
Annapurna Labs acquisition
- ▪Andy Jassy described the acquisition of Annapurna Labs as potentially the best deal in Amazon's history
- ▪Annapurna Labs helped develop both Amazon Trainium and Amazon Graviton chips
- ▪Amazon acquired chip design firm Annapurna Labs more than a decade ago
Story comments
Loading comments…