Amazon (AMZN) AWS re:invent 2025 Keynote summary

Key business highlights

• AWS achieved $132 billion in revenue, growing 20% year-over-year, with $22 billion in annual growth, surpassing the annual revenue of over half the Fortune 500.

• S3 stores over 500 trillion objects, and Graviton CPUs now power more than half of new AWS cloud capacity for the third consecutive year.

• Amazon Bedrock powers AI inference for over 100,000 companies, and the new Bedrock Agent Core SDK has surpassed 2 million downloads.

• AWS expanded its global infrastructure to 38 regions and 120 availability zones, adding 3.8 GW of data center capacity in the past year.

• AWS's private network grew 50% to over nine million kilometers of cable, supporting millions of customers across all industries.

AI and infrastructure innovation

• AWS introduced Ocelot, its first quantum computing chip prototype, reducing quantum error correction costs by over 90%.

• AWS AI Factories now allow customers to deploy dedicated AI infrastructure in their own data centers, supporting compliance and sovereignty.

• Trainium 3 Ultra servers, featuring the first 3nm AI chip in AWS, deliver 4.4x more compute and 5x more AI tokens per megawatt than previous versions.

• Trainium 4 is in development, promising 6x FP4 compute and 4x memory bandwidth over Trainium 3.

• Bedrock's model selection doubled, now including new open weights models from Google, NVIDIA, and Mistral, plus Amazon's Nova 2 family for multimodal and speech-to-speech AI.

Customer and partner stories

• Startups and enterprises, including 85% of Forbes AI 50 and CNBC Disruptor 50, rely on AWS; more than half of all VC-funded startups run on AWS.

• Sony leverages AWS for its Sony Engagement Platform and Data Ocean, processing 760 TB of data and deploying enterprise LLMs and agentic AI for operational efficiency.

• Adobe uses AWS to power Firefly, Acrobat Studio, and Experience Platform, integrating AI and agentic capabilities for creative and business workflows.

• Writer, a startup, uses AWS for large-scale model training and integrates Bedrock Guardrails for enterprise-grade AI compliance and flexibility.