Amazon Web Services (AWS) posted a 20% year-over-year revenue increase to $33 billion in Q3 2025—its fastest growth since 2022—driven by surging demand for AI compute and cloud infrastructure. CEO Andy Jassy said AWS added more than 3.8 gigawatts of capacity over the past year, outpacing all other cloud providers. AWS segment operating income reached $11.4 billion, up from $10.4 billion a year earlier.
Amazon highlighted a string of AI infrastructure developments, including Project Rainier—a 500,000-chip AI cluster built on its custom Trainium2 processors, which are now fully subscribed. AWS also launched EC2 P6e-GB200 UltraServers powered by NVIDIA Grace Blackwell Superchips and new Graviton4 and custom Intel Xeon 6 instances designed for high-performance workloads. The company expanded Amazon Bedrock’s model library with new foundation models from OpenAI, Anthropic, DeepSeek, and others.
AWS continues to build its regional footprint, adding a new Region in New Zealand and planning 10 additional Availability Zones and three new Regions. Its growing AI ecosystem includes Transform, an agentic migration tool saving hundreds of thousands of developer hours; Connect, an AI-enabled contact center service now generating $1 billion in annualized revenue; and AgentCore, infrastructure primitives for secure, scalable AI agents. Amazon also expanded Project Kuiper’s satellite broadband network to 150+ satellites, achieving downlink speeds above 1 Gbps.
• AWS revenue rose 20% YoY to $33.0 billion; operating income reached $11.4 billion
• Added 3.8 GW of new power capacity in past 12 months, more than any other cloud provider
• Launched Project Rainier, a 500,000-Trainium2 AI cluster for Anthropic’s Claude models
• Introduced EC2 P6e-GB200 UltraServers and new Graviton4, Intel Xeon 6, and AMD EPYC instances
• Expanded Bedrock with new foundation models and released AgentCore for scalable AI agents
• Grew Connect to $1B annual run rate and Transform to 700,000 developer hours saved
• Added AWS Region in New Zealand with 10 more Availability Zones in the pipeline
“We continue to see strong demand in AI and core infrastructure, and we’ve been focused on accelerating capacity – adding more than 3.8 gigawatts in the past 12 months,” said Andy Jassy, President and CEO, Amazon.
🌐 Analysis: AWS’s re-acceleration underscores the scale of AI infrastructure spending reshaping the cloud landscape. The 3.8-GW capacity build signals Amazon’s return to aggressive capital deployment, rivaling Google’s and Microsoft’s recent expansions in AI-dedicated data centers. With Trainium2 adoption accelerating and Bedrock’s open-model strategy broadening, AWS is positioning itself as both a platform and silicon leader in the race to power foundation model development.
