Amazon Web Services (AWS) is unveiling a new AI supercomputer, Project Rainier, built using its own Trainium machine learning chips.
Set to launch in 2025, it aims to rival Nvidia’s GPUs, which currently dominate the AI chip market. The supercomputer will be one of the largest created for training AI models.
In collaboration with AI startup Anthropic, which is valued at $18 billion, AWS will use Project Rainier to advance AI capabilities. Amazon has invested $8 billion to improve its Trainium chip technology.
AWS’s push to challenge Nvidia is part of a broader strategy to provide more chip options for AI workloads. The company has committed to investing $100 billion in AI infrastructure over the next decade to stay competitive in the fast-growing AI market.
In addition to Project Rainier, AWS is partnering with Nvidia on Project Ceiba, a supercomputer using over 20,000 Nvidia Blackwell GPUs. AWS is also launching an Ultraserver, a cloud server designed for AI programs using advanced Trainium chips, with Apple adopting the chips for its own AI initiatives.
