Amazon CEO Andy Jassy said high AI inference costs limit wider enterprise adoption. He cited inference computing as the main cost driver. Jassy also named a shortage of skilled staff as another barrier. He spoke during Amazon’s third-quarter earnings call. Inference already makes up over half of AWS AI revenue. AWS revenue grew 39.3% to $46B in the quarter. Jassy said custom silicon like Trainium chips can cut inference costs. He called cost-effective inference critical for broader use.
- 55% of AI-optimized IaaS spend on inference in 2026
- $23.3B projected inference spend in 2026
- 17-fold increase in Bedrock token volume
- 7x higher Bedrock spend by Fortune 500
Sources