Google TPU Veteran Amir Salek Joins Anthropic’s Compute Team

Anthropic has hired former Google TPU chief Amir Salek to lead its in‑house chip effort, signaling a shift from relying on external accelerators to building custom hardware as AI companies intensify competition over compute efficiency and cost.

Machine Heart
Machine Heart
Machine Heart
Google TPU Veteran Amir Salek Joins Anthropic’s Compute Team

From Reliance on External Chips to Betting on Compute Stack

Anthropic has traditionally depended on external suppliers for AI compute, purchasing Nvidia GPUs, Google TPUs, and Amazon AWS chips to train and run its Claude models.

With Claude’s rapid iteration, the company forecasts a $19 billion compute spend by 2026, and continued reliance on third‑party chips creates cost and supply‑chain pressure.

Hiring Amir Salek, who led Google’s TPU program from 2013 to 2022 and previously helped build Nvidia’s on‑chip systems, underscores Anthropic’s intent to develop its own chips to optimize hardware for its models, improve inference efficiency, and lower long‑term costs.

AI Companies Join Chip Race

Anthropic is not alone; OpenAI recently announced its own chip project, Jalapeño, co‑developed with Broadcom and Celestica, aimed at large‑language‑model inference.

Both firms see inference cost as a new competitive frontier, where dedicated chips can reduce operating expenses and boost service efficiency.

Salek’s addition signals that AI competition will extend from model capabilities to the underlying compute infrastructure.

Compute Supply Chain Becomes Next Battlefield

The AI chip market is reshaping the industry landscape. Nvidia still dominates, but leading AI labs are pursuing custom silicon and multi‑vendor strategies to lessen dependence on a single hardware stack.

Anthropic has already expanded its compute resources through partnerships with chip makers and data‑center operators, preparing for future training and deployment needs.

In the coming years, a company’s core strength will be judged not only by model size and algorithmic innovation but also by its ability to build stable, efficient, and controllable compute infrastructure.

From Claude models to custom chips, Anthropic is attempting to transform into an AI infrastructure provider, reflecting a broader industry shift where the next AI battleground is the hardware that powers models.

Reference link: https://www.bloomberg.com/news/articles/2026-08-21/anthropic-taps-google-chip-veteran-as-part-of-push-into-hardware

Original Source

Signed-in readers can open the original source through BestHub's protected redirect.

Sign in to view source
Republication Notice

This article has been distilled and summarized from source material, then republished for learning and reference. If you believe it infringes your rights, please contactadmin@besthub.devand we will review it promptly.

chip designAI hardwareAnthropiccompute infrastructureTPUAmir Salek
Machine Heart
Written by

Machine Heart

Professional AI media and industry service platform

0 followers
Reader feedback

How this landed with the community

Sign in to like

Rate this article

Was this worth your time?

Sign in to rate
Discussion

0 Comments

Thoughtful readers leave field notes, pushback, and hard-won operational detail here.