Google Cloud Focus: Launch TPU 8T + 8I, All In AI Agent?
Google Cloud just launched its 8th-gen TPU 8t for training and TPU 8i for inference, its first clear split between the two, and rolled more enterprise AI tools into the new “Gemini Enterprise” push at Cloud Next 2026.
That sharpens the debate: is Google finally attacking AI from both sides — cheaper inference and enterprise agents — or still playing catch-up to Nvidia in chips and OpenAI/Anthropic in the app layer? Which matters more here: TPU share gains, or whether Gemini Enterprise actually gets adopted?