Google Unveils Eighth-Generation TPUs, Launches Gemini Enterprise Agent Platform

Google has long used its in-house TPUs to power AI models including Gemini, while competing with Nvidia GPUs in the cloud computing market. Its eighth-generation lineup further separates training and inference workloads. Meanwhile, enterprises face fragmented agent tools and complex governance and deployment, making an integrated platform critical to broader adoption.

At Cloud Next 2026 in April 2026, Google unveiled the TPU 8t and TPU 8i, designed respectively for large-scale AI training and high-efficiency inference. It also launched the Gemini Enterprise agent platform, consolidating Vertex AI-related services, and committed $750 million to help consulting and service partners drive enterprise adoption.