Google
Four systems in one generation
On November 9, Google announced TPU 8tx and TPU 8i Max. They join the previously announced training-focused 8t and inference-focused 8i.
TPU 8tx focuses on secure, confidential inference with sensitive data. TPU 8i Max is designed for large context windows and greater compute density per rack.
Designed for different workloads
Google attributes this separation to the security needs of regulated organizations and agents' growing memory demands. The announcement does not mean every model or existing cloud account is automatically moving to the new hardware.
OpenAI