Amazon's Trainium Chip Powers 1.4M Deployments, Targets Nvidia
Get the Tech newsletter
Daily tech — startups, AI labs, chips, the launches that shape the next decade. Free.
- Amazon's Trainium has 1.4 million chips deployed across three generations, with Anthropic's Claude running on over 1 million Trainium2 chips, per the AWS chip team.
- AWS agreed to supply OpenAI with 2 gigawatts of Trainium computing capacity as part of a $50 billion investment deal, and became the exclusive cloud provider of OpenAI's new AI agent builder, Frontier.
- Trainium3, released in December 2025 and built on TSMC's 3-nanometer process, works with new Neuron switches in a mesh configuration that Amazon says cuts costs up to 50% versus classic cloud servers for comparable performance.
- Project Rainier — one of the world's largest AI compute clusters — went live in late 2025 with 500,000 Trainium2 chips dedicated to Anthropic workloads.
- Amazon's chip team claims Trainium now supports PyTorch with 'basically a one-line change' to port applications, directly targeting the switching costs that have locked developers into Nvidia's GPU ecosystem.
- Apple publicly praised the AWS chip team in 2024 for its Graviton ARM-based CPU, Inferentia inference chip, and Trainium — a rare public endorsement from Apple.
- AWS announced a partnership with Cerebras Systems this month, integrating Cerebras inference chips on Trainium-equipped servers for low-latency AI performance.
Why it matters: With 1.4 million Trainium chips already deployed and 2 gigawatts of new capacity pledged to OpenAI, Amazon is no longer a secondary player in AI infrastructure — it's offering enterprises a credible Nvidia alternative at up to 50% lower cost, and the one-line PyTorch port could accelerate developer migration away from CUDA-locked codebases.
Ask SkimNews
