Amazon Web Services CEO Matt Garman said customers are shifting from using AWS to train large AI models toward embedding those models in their own workflows, driving increased demand for inference compute even as some companies continue to run large training clusters. He called the AI business potential "very large" and said Amazon will keep raising capital spending to meet demand. Amazon last week said it now expects 2026 capital expenditure of $220bn, up from a prior $200bn forecast; the highe

2026-08-04

Amazon Web Services CEO Matt Garman said customers are shifting from using AWS to train large AI models toward embedding those models in their own workflows, driving increased demand for inference compute even as some companies continue to run large training clusters. He called the AI business potential "very large" and said Amazon will keep raising capital spending to meet demand. Amazon last week said it now expects 2026 capital expenditure of $220bn, up from a prior $200bn forecast; the higher plan reflects rising prices for storage chips and other data‑center components.