Google Cloud launches eighth‑gen TPU 8t and 8i chips
Get the Tech newsletter
Daily tech — startups, AI labs, chips, the launches that shape the next decade. Free.
- Google Cloud unveiled its eighth‑generation TPU lineup, featuring the TPU 8t for training and the TPU 8i for inference, slated for general availability later in 2026.
- Google positioned the new chips for the “agentic era,” aiming to accelerate both large‑scale model training and real‑time inference workloads.
- Google introduced the Virgo Network, a megascale data‑center fabric designed to support the higher bandwidth demands of the new TPUs.
- Google framed the rollout as a direct challenge to Nvidia, highlighting cost and performance advantages for enterprise AI workloads.
- Bloomberg and other outlets (e.g., CNBC, WSJ) emphasize the chips’ potential to “speed up AI” and “take on Nvidia,” indicating a consensus on their competitive significance.
Why it matters: Enterprises that run AI models on Google Cloud will gain access to faster, more cost‑effective training and inference hardware, while Nvidia’s market share in AI accelerators could be eroded as Google pushes a home‑grown alternative. The rollout also signals a shift toward specialized, agentic‑era chips across cloud providers.
