Google unveils two new TPUs designed for the “agentic era”
Summary
Google introduces TPU 8t (training) and TPU 8i (inference) to power an 'agentic era' of AI, with large-scale pods, increased memory, and higher compute efficiency. The dual-chip design emphasizes end-to-end hardware-software optimization, ARM-based hosts, and broad framework support, signaling a shift in AI hardware competition and potential impacts on enterprise AI deployments.