AWS CEO: Inference Now Dominates AI Demand

AWS CEO Matt Garman highlighted a significant shift in AI demand at the AWS re:Invent 2025 conference. He stated that AI inference now accounts for roughly two-thirds of all AI compute demand, up from about one-third in 2023.

Inference is the process of making AI models think, as opposed to training them to learn. Garman described inference as 'a new Lego' in computing, suggesting it's becoming a basic building block that enterprises will use to automate tasks and run AI agents.

AWS is putting its hardware behind this bet by showcasing its Trainium3 chips, which are purpose-built for inference workloads rather than training-optimized GPUs.

This shift has significant implications for the crypto mining sector. Several major miners have pivoted toward offering AI and high-performance computing services after tightening margins post-April 2024 halving.

However, if Garman's prediction is correct that 80-90% of enterprise AI value will come from inference-powered agents, the hardware requirements change. Training demands massive parallel processing across clusters of top-tier GPUs running for weeks or months, whereas inference needs lower latency and higher throughput per query.