Amazon CEO Andy Jassy says AI inference remains roughly 10 times too costly for widespread adoption in business software. Jassy gave the assessment in an a16z interview published Friday. Inference is the computation needed whenever an AI model responds to a question or executes a task. Jassy believes Amazon's own chips could help reduce inference costs.
Amazon plans $220 billion in capital expenditure this year. Amazon still cannot build enough infrastructure to meet AI demand. Jassy said major capacity commitments are effectively sold out for 2026. Jassy said capacity is largely booked for 2027, with a substantial portion of 2028 also sold out. Jassy believes demand would justify Amazon spending considerably more capital.
Jassy described AI demand as concentrated at two ends of the market. Jassy said AI laboratories consume computing resources on a scale unlike anything Jassy has previously seen. Jassy said businesses at the other end mainly apply AI to particular tasks, including customer service and fraud detection. However, everyday business software remains a largely untapped market for AI. Jassy believes everyday business software could eventually generate the most demand. Companies have little reason to redesign existing software around AI unless the benefits warrant the additional operating expense.
Jassy said Amazon encountered a similar challenge years earlier when Intel dominated processors for cloud servers. Amazon sought better computing performance per dollar but struggled to make that objective a priority for Intel. Amazon consequently developed the Graviton processor. Jassy said Amazon encountered a similar situation with AI.
Nvidia is the dominant supplier in the AI processor market. Amazon developed Trainium AI chips to give customers a less expensive option. Jassy did not identify Nvidia by name but emphasized that the dominant supplier remains a significant partner. Jassy said Trainium provides approximately 30% to 40% better price-performance than comparable processors, allowing customers to obtain more computing power for the same expenditure. Jassy described Amazon's chips business as an important part of reducing inference costs. Amazon has disclosed future Trainium revenue commitments exceeding $225 billion, including agreements with Anthropic and OpenAI.
Amazon will report third-quarter earnings on October 29. Traders expect Amazon's AI chip strategy to be a central topic during the earnings call. Kalshi traders put the probability that executives will mention Trainium at 96%. Kalshi traders assign a 95% probability to a mention of Amazon's Bedrock AI platform. Jassy predicts that AI will enable companies to operate with fewer employees. However, Kalshi traders assign only a 15% probability to executives mentioning "layoffs" or "job cuts" during the call. The market has drawn trading volume exceeding $257,000, indicating substantial interest in the discussion as Amazon increases AI spending. Kalshi and Benzinga have a data collaboration agreement.
Amazon CEO Andy Jassy on AI Inference Costs and Chip Strategy
BenzingaShare
Summary
Amazon CEO Andy Jassy said AI inference costs are still about 10 times too high for mass business use. He pointed to Amazon’s custom chips like Trainium as a way to cut these costs. On-chain analysis shows AI demand is split between two market ends, with most business software untouched. Amazon plans $220 billion in capex this year, with AI infrastructure already sold out through 2028. Jassy compared the AI chip market to Intel’s past cloud dominance, pushing Amazon to build its own solutions. Trainium delivers 30% to 40% better price-performance. Amazon has over $225 billion in future Trainium revenue. Traders watch for AI strategy updates during Q3 earnings on October 29.
Source:Show original
Disclaimer: The information on this page may have been obtained from third parties and does not necessarily reflect the views or opinions of KuCoin. This content is provided for general informational purposes only, without any representation or warranty of any kind, nor shall it be construed as financial or investment advice. KuCoin shall not be liable for any errors or omissions, or for any outcomes resulting from the use of this information.
Investments in digital assets can be risky. Please carefully evaluate the risks of a product and your risk tolerance based on your own financial circumstances. For more information, please refer to our Terms of Use and Risk Disclosure.