Google Unveils 8th-Gen TPUs to Rival Nvidia
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: the TPU 8t for training and TPU 8i for inference, with general availability slated for later this year.
- Google also introduced Virgo Network, a megascale data center fabric that one analyst called the biggest networking upgrade since TPU v4, signaling the launch is as much about infrastructure as silicon.
- Google Cloud committed $750 million to accelerate partners' agentic AI development, framing the broader platform push around what it calls "the agentic era."
- The new TPUs were developed in partnership with Broadcom, underscoring Google's continued use of external fabrication rather than going fully in-house.
- Sundar Pichai headlined Cloud Next '26 with a message of "momentum and innovation at Google scale," positioning the TPU launch as the centerpiece of a wider agentic-AI platform rollout.
- Coverage across outlets — from Bloomberg to CNBC to VentureBeat — frames the launch as a direct shot at Nvidia, with VentureBeat arguing Google's custom silicon lets it "avoid the Nvidia tax."
Why it matters: Google is offering purpose-built alternatives to Nvidia's general-purpose GPUs for both training and inference, backed by a $750 million partner fund and a new networking fabric — a full-stack infrastructure play aimed at pulling AI workloads off Nvidia-centric hardware.
