AWS re:Invent 2025 Recap for Automotive and Manufacturing

This article recaps AWS re:Invent 2025 announcements relevant to automotive and manufacturing industries, organized by strategic workload areas.

• Trainium3 UltraServers deliver 4.4x compute performance for autonomous vehicle AI training and inference
• Amazon S3 object size increased to 50 TB; S3 Vectors scales to 2 billion vectors with 90% cost reduction
• Amazon OpenSearch Service adds GPU acceleration for vector databases, 10x faster at lower cost
• Zoox trains multimodal autonomous models achieving 95% GPU utilization across 64+ GPUs
• Graviton5 CPU offers 25% higher performance and 5x larger cache than previous generation
• Nissan accelerates SDV development with 75% faster testing on unified cloud platform
• Lambda Managed Instances and Durable Functions enable serverless compute flexibility
• BMW modernized Connected Drive backend, achieving 60% faster feature time-to-market
• Amazon Nova 2 models support speech-to-speech, reasoning, and multimodal AI interactions
• Amazon Bedrock AgentCore enables secure enterprise agent deployment with quality controls
• Toyota built RAG-driven dealer assistant handling 7,000+ interactions monthly
• Audi uses AI quality inspection for near-100% weld coverage with reduced manual effort
• AWS Transform modernizes legacy applications; Mercedes-Benz converted 1.3M COBOL lines to Java
• AWS Parallel Computing Service reduced HPC setup from six weeks to 30 minutes

AWS highlights AI agents, compute optimization, and modernization tools enabling automotive and manufacturing digital transformation at scale.