Amazon is one of the world's leading corporate purchasers of carbon-free energy, with over 700 projects globally totaling more than 40 gigawatts-enough to power 12.1 million U.S. homes. These investments add new carbon-free energy sources to the grid, modernize infrastructure, stabilize electricity costs, and create thousands of jobs. Amazon's leadership in this space, recognized by BloombergNEF, is helping drive corporate momentum toward carbon-free energy procurement, which benefits entire communities by adding low-cost clean energy to the grid while supporting America's technological innovation.
AWS powers AI inference-the split-second process where trained AI models take your question and generate an answer-through a four-stage pipeline (tokenization, prefill, decode, detokenization) that typically completes in under two seconds, with inference accounting for 80-90% of AI computing power. The rise of AI agents, which coordinate multiple models to accomplish complex multi-step tasks rather than simple single predictions, creates new challenges including massive compute demands, critical latency requirements, exponential costs, and reliability needs that traditional cloud infrastructure wasn't built to handle. AWS addresses these challenges through innovations like Project Mantle (a new inference engine powering Amazon Bedrock), custom Trainium chips providing up to 40% cost reduction, intelligent request prioritization across different urgency levels, continuous checkpointing through Journal that allows failed requests to resume without restarting, and unified network architecture that eliminates delays between storage and processing-making AI faster, more reliable, and more affordable at scale.
AWS has announced the availability of Claude Sonnet 4.6 and Opus 4.6, Anthropic's latest AI models, in Amazon Bedrock, with Sonnet 4.6 delivering near-Opus intelligence at a more accessible $3/$15 price point for coding, agents, and enterprise workflows. While Opus 4.6 is Anthropic's most intelligent model setting new standards for autonomous coding and multi-step enterprise tasks that compress multi-day projects into hours, both models can process extensive documents and codebases and represent a significant leap in agentic AI capabilities by functioning as expert virtual collaborators. These models enable enterprises to build stronger spreadsheets, financial models, and specialized workflows while freeing senior talent for strategic work, with Anthropic pointing toward a future where AI systems become increasingly capable partners handling routine analysis, coordinating across departments, and managing complete workflows with less oversight.