Google Cloud launches 8th‑gen TPUs for AI training
Get the Tech newsletter
Daily tech — startups, AI labs, chips, the launches that shape the next decade. Free.
- Google Cloud unveiled its eighth‑generation TPUs, adding the TPU 8t for training and the TPU 8i for inference.
- Google said the new TPUs will be generally available later in 2026.
- Google positioned the chips for the "agentic era," emphasizing AI workloads that need both training and inference.
- Google introduced the Virgo Network, a megascale data‑center fabric to support the new TPUs.
- CNBC highlighted the launch as a direct challenge to Nvidia’s AI‑chip dominance.
Why it matters: Enterprises adopting Google Cloud's new TPUs gain access to dedicated training and inference hardware optimized for the agentic era, potentially lowering compute costs and accelerating AI deployment, while Nvidia faces heightened competition in the AI accelerator market and may shift cloud AI pricing dynamics.
