Google unveils new TPUs positioned for the ‘agentic era’
The company says the new TPU designs target faster frontier-model training and improved efficiency across the AI lifecycle.
Google introduced two new TPU designs intended to support increasingly agentic AI workloads. The company says the chips and accompanying system designs are meant to reduce training time for frontier models and improve efficiency across training and inference.
The announcement underscores Google’s push to pair in-house silicon with its AI platform strategy.
Read the original reporting at Ars Technica.
