Google Debuts AI Chips with 4X Performance Boost, Secures Anthropic Megadeal Worth Billions

Why It Matters

The Anthropic megadeal secures a massive revenue stream for Google and validates its custom‑silicon strategy, positioning TPUs as a credible, cost‑effective alternative to Nvidia GPUs. This could reshape AI infrastructure economics and accelerate the industry-wide transition to inference‑scale deployments.

Summary

Google Cloud unveiled its seventh‑generation Tensor Processing Unit, Ironwood, which offers more than four‑times the training and inference performance of the previous generation and scales to pods of 9,216 chips with 1.77 petabytes of high‑bandwidth memory and optical circuit‑switching. Anthropic announced a commitment to access up to one million Ironwood TPUs, a multi‑year contract valued in the tens of billions of dollars, marking one of the largest AI infrastructure deals ever. Alongside the new TPU, Google expanded its Arm‑based Axion processor lineup with N4A and C4A instances, promising up to 2× price‑performance for general‑purpose AI workloads. The rollout is framed as an “age of inference,” shifting cloud focus from model training to serving billions of real‑time requests and intensifying competition with Nvidia’s GPU dominance.