Amazon Web Services' broad AI portfolio, anchored by Amazon Bedrock for managed foundation models, SageMaker for custom ML, and Q for enterprise assistants. By
AWS is the enterprise AI default cloud: Bedrock for managed frontier/open models (including Claude), SageMaker for ML engineering, Trainium/Inferentia silicon, and the deepest infrastructure catalog, where most regulated AI workloads ultimately land.
Pricing: Pure usage-based across services; Bedrock bills per token by model; enterprise agreements at scale (published pricing, mid-2026).
Bedrock buys AWS-native governance (IAM, VPC, logging) and multi-model switching; direct APIs sometimes lead on newest features. Regulated enterprises lean Bedrock; startups often go direct.
Model routing by task, prompt caching, provisioned throughput where steady, and tagging/budget alerts from day one: token spend is the new S3 bill surprise.