Google unveils TPU 8t and 8i AI chips
Get the Tech newsletter
Daily tech — startups, AI labs, chips, the launches that shape the next decade. Free.
- Google unveiled the latest generation of its Tensor Processing Unit (TPU) chips, named TPU 8t for AI training and TPU 8i for inference.
- Google scheduled general availability of the TPU 8t and TPU 8i chips for later in 2026.
- Google positioned the new TPU chips for the "agentic AI" era, targeting AI agents.
- Google introduced a major networking upgrade called Virgo Network with the TPU 8t/8i launch, the biggest upgrade since TPU v4.
- Google marketed the chips as cost‑cutting, aiming to slash AI compute expenses and directly challenge Nvidia's hardware.
- Google committed $750 million to accelerate partner development of agentic AI on its cloud platform.
Why it matters: Google Cloud customers and AI developers gain faster, cheaper compute for training and inference, while Nvidia faces heightened competition in the AI‑chip market; the Virgo Network upgrade could also boost data‑center bandwidth, reshaping performance benchmarks for next‑gen AI workloads and could accelerate deployment of agentic AI services.
