Google TPU Could Sell as Well as Nvidia Rubin

FT and others reported yesterday that Anthropic is committing to spend hundreds of billions of dollars on Google’s chips and cloud services to secure critical computing resources amid soaring demand for its AI tools. The company plans to use multiple gigawatts of Google’s TPU capacity, rivaling Nvidia’s GPUs, with about 3.5GW coming through a partnership with Broadcom starting next year. This deal will provide Anthropic nearly 5GW of new computing power, a capacity costing tens of billions of dollars. Broadcom will also supply custom TPUs for Google through 2031. For each TPU generation (since v4?) Google has had big TPUs for training and smaller configs for inference. The company previously disclosed that for the forthcoming v8 it would work with Broadcom for the training chip and MediaTek for the inference one. MediaTek has impressed me with its ability to cram more functionality in less die area than rivals. FT link: https://www.ft.com/content/28757ce7-0d9f-4ffb-bb91-16dc83f2cf6a?syn-25a6b1a6=1

Other contents

Phoronix Shows Nvidia Vera to be Among the Fastest Server Processors

Phoronix Shows Nvidia Vera to be Among the Fastest Server Processors

Bolt Confounds Skeptics, Tapes Out Zeus GPU

Bolt Confounds Skeptics, Tapes Out Zeus GPU

Matrix Multiplication Comes to x86

Matrix Multiplication Comes to x86

Blowout AMD & Intel Earnings, Google TPUv8, Astera Scorpio-X, RISC-V New, and More: It’s BWR Ep 13.

Blowout AMD & Intel Earnings, Google TPUv8, Astera Scorpio-X, RISC-V New, and More: It’s BWR Ep 13.

AMD floats Venice

AMD floats Venice

Alibaba steps up data-center AI accelerator cadence

Vera! What has become of you!?

Vera! What has become of you!?

Will Qualcom or Intel help world's largest chip startup take flight?

Will Qualcom or Intel help world's largest chip startup take flight?

Physical Design Trends for AI Accelerators in 2026

Physical Design Trends for AI Accelerators in 2026

Google TPUv8: Early Specs and Performance Gains

Google TPUv8: Early Specs and Performance Gains