AWS Records Fastest Growth in Nearly Five Years Driven by AI Demand
Amazon Web Services (AWS) disclosed that artificial‑intelligence (AI) workloads are propelling its fastest growth rate in almost half a decade. CEO Matt Garman told Bloomberg Tech that the AI segment now operates at a $25 billion revenue run rate, a figure that includes model‑training revenue from large tech firms but is increasingly dominated by inference services such as Amazon Bedrock that power real‑world applications.
Capital Expenditure and Capacity Commitments
Garman highlighted that Amazon’s overall capital expenditure for the current year stands at $220 billion, which is $20 billion higher than the previous year. A substantial portion of this spending is earmarked for expanding AWS’s data‑center infrastructure to meet AI‑driven demand. He confirmed that most of AWS’s compute capacity is already committed through the end of 2027 and into 2028, with customers signing five‑year contracts to lock in the required compute resources.
Custom Silicon Strategy – Trainium and Graviton
AWS’s in‑house chip business, built around the Trainium and Graviton processors, underpins the $25 billion AI run rate. The company rents out capacity powered by these chips rather than selling the silicon outright. Trainium capacity is sold out through the end of next year, and the integrated hardware‑software stack delivers 20 % to 30 % lower inference costs compared with standard market options. While AWS continues to be one of Nvidia’s largest customers, offering Nvidia GPUs alongside its own silicon, Garman indicated that the firm may eventually explore selling chips directly to third parties.
Open‑Weight Models and Regulatory Stance
Addressing the emerging ecosystem of open‑weight AI models, Garman said AWS signed a recent open‑weights letter advocating for a balanced regulatory framework that treats closed‑source and open‑weight models equally, avoiding over‑legislation that could stifle innovation. He projected that creators of open‑weight models will increasingly adopt licensing structures when deploying their models on commercial cloud platforms, turning intellectual‑property rights into a monetizable asset.
Outlook
The interview captures AWS at a pivotal moment: record AI‑driven demand, capacity sold out years in advance, and a commitment to sustain elevated capex into the next fiscal year to build the necessary data‑center and chip infrastructure for global enterprise customers.