---
title: "Anthropic’s Move into In-House Chips: Poaching Google TPU Founder for a ‘De-NVIDIA’ Gamble on the Eve of a $2 Trillion IPO"
type: "News"
locale: "en"
url: "https://longbridge.com/en/news/296665741.md"
description: "Anthropic is shifting its strategic focus from \"scrambling for compute\" to \"building compute.\" The company has hired Amir Salek, founder of Google’s custom chip business, to establish an in-house chip team, while collaborating with Fractile, Google, Broadcom, and Samsung to secure approximately 3.5GW of TPU capacity by 2027. This dual-track strategy of \"in-house development + diversified procurement\" aims to reduce reliance on NVIDIA and compress inference costs"
datetime: "2026-08-22T04:09:38.000Z"
locales:
  - [zh-CN](https://longbridge.com/zh-CN/news/296665741.md)
  - [en](https://longbridge.com/en/news/296665741.md)
  - [zh-HK](https://longbridge.com/zh-HK/news/296665741.md)
generator: "portal-rs"
---

# Anthropic’s Move into In-House Chips: Poaching Google TPU Founder for a ‘De-NVIDIA’ Gamble on the Eve of a $2 Trillion IPO

As it approaches the critical window for an IPO valued at $2 trillion, Anthropic is shifting its strategic focus from "scrambling for compute" to "building compute," attempting to take direct control of the underlying infrastructure supporting its explosive growth.

**On Friday, the developer of Claude announced that it has officially hired Amir Salek, founder of Google’s custom chip business, to build an autonomous AI custom chip design team.**

**At the same time, the company is aggressively advancing external compute procurement:** It has reached an initial purchase agreement worth approximately $250 million with UK chip startup Fractile, expanded collaborations with Google and Broadcom to lock in about 3.5GW of next-generation TPU capacity by 2027, and plans to jointly design custom hardware with Samsung.

This series of moves sends a clear signal: Leading large AI model companies are accelerating their efforts to break free from absolute dependence on single chip suppliers like NVIDIA.

Through a dual-track strategy of "in-house development + diversified external procurement," Anthropic not only aims to alleviate compute supply bottlenecks but also hopes to achieve structural optimization in inference costs. Autonomous control over compute is becoming a core variable determining the valuation ceiling for large model companies.

## Poaching Talent from Google to Complete the Final Piece of the In-House Chip Puzzle

Amir Salek’s arrival marks a key step in Anthropic’s entry into autonomous semiconductor manufacturing.

Salek previously spent eight years at NVIDIA, where he founded and led the system-on-chip design team, participating in the development of GPU and Tegra chips. In 2013, Google recruited him to build its custom chip business from scratch, where he spearheaded the research and development of the first seven generations of TPU chips—a product line that remains the backbone of Google’s AI infrastructure to this day.

After leaving Google in 2022, Salek served as a Senior Managing Director at Cerberus Capital Management, focusing on investments in the semiconductor and AI sectors.

**In his new role at Anthropic, Salek will report to James Bradbury and be responsible for building and leading the AI custom chip design team. In addition to in-house development, Anthropic plans to collaborate with Samsung on hardware and model design to further enhance speed and efficiency.**

This layout aligns closely with industry trends. Competitor OpenAI has jointly developed the Jalapeno chip, designed specifically for inference workloads, with Broadcom; Google DeepMind has long relied on its own TPUs; and Meta is advancing its MTIA accelerators.

By developing its own chips, Anthropic can precisely tailor hardware design to its model architecture, establishing barriers in underlying compute capabilities and reducing reliance on external general-purpose chips.

## Billions Spent on "Hoarding Chips" to Build a Compute Network Independent of Single Suppliers

Before its in-house chips come to fruition, Anthropic has shown strong aggression in expanding external compute capacity.

**On the chip procurement front,** in addition to the initial $250 million order from Fractile, Anthropic has significantly deepened collaborations with established giants—expanding the scale of cooperation with Google and Broadcom to lock in approximately 3.5GW of next-generation TPU capacity by 2027.

This figure is roughly equivalent to the electricity consumption of a medium-sized city and does not include the 1GW of capacity scheduled for delivery in 2026 under the Google Cloud agreement signed last October. **Broadcom plays a key role here: It is responsible for developing and supplying custom TPUs and has committed to supplying network components for Google’s next-generation AI racks through 2031.**

On the data center front, Anthropic’s contracted scale has reached hundreds of billions of dollars, with partners covering traditional cloud providers, chip giants, and emerging compute suppliers, including capacity agreements signed with Riot Platforms and Volta Infra Holdings.

**Notably, market estimates suggest that more than half of Anthropic’s long-term compute is hosted on TPUs via Google Cloud. Against the backdrop of slowing growth in Google’s total compute capacity, Anthropic has had to accelerate its external search for chips.**

## The Ledger Behind Compute Autonomy: Solidifying the Physical Foundation for a $2 Trillion IPO

Anthropic’s aggressive investment in chips and compute directly corresponds to its rapidly improving financial model.

As of the end of July, the company’s annualized revenue exceeded $65 billion, with preliminary Q2 revenue surpassing $11.5 billion, a year-over-year increase of more than 14 times. The gross margin for inference infrastructure jumped from 38% to over 70%, and adjusted operating profit turned positive for the first time.

Chief Financial Officer Krishna Rao described this as the company’s "most important investment in computing to date," aimed at meeting the exponential growth of its customer base.

For Anthropic, which is poised to go public at a valuation of $2 trillion or higher, controlling pricing power and supply chain initiative in compute is not just about costs; it is crucial for delivering on that steep growth curve to investors in the next phase of the AI race.

### Related Stocks

- [NVDS.US](https://longbridge.com/en/quote/NVDS.US.md)
- [ANTH.NA](https://longbridge.com/en/quote/ANTH.NA.md)
- [NVDA.US](https://longbridge.com/en/quote/NVDA.US.md)
- [GOOG.US](https://longbridge.com/en/quote/GOOG.US.md)
- [AVGO.US](https://longbridge.com/en/quote/AVGO.US.md)
- [GOOGL.US](https://longbridge.com/en/quote/GOOGL.US.md)
- [GGLL.US](https://longbridge.com/en/quote/GGLL.US.md)
- [NVDU.US](https://longbridge.com/en/quote/NVDU.US.md)
- [NVDX.US](https://longbridge.com/en/quote/NVDX.US.md)
- [GGLS.US](https://longbridge.com/en/quote/GGLS.US.md)
- [AVGX.US](https://longbridge.com/en/quote/AVGX.US.md)
- [NVDQ.US](https://longbridge.com/en/quote/NVDQ.US.md)
- [AVL.US](https://longbridge.com/en/quote/AVL.US.md)
- [GOOW.US](https://longbridge.com/en/quote/GOOW.US.md)
- [NVDL.US](https://longbridge.com/en/quote/NVDL.US.md)
- [AVGU.US](https://longbridge.com/en/quote/AVGU.US.md)
- [NVDY.US](https://longbridge.com/en/quote/NVDY.US.md)
- [AVGG.US](https://longbridge.com/en/quote/AVGG.US.md)
- [NVD.US](https://longbridge.com/en/quote/NVD.US.md)
- [NVDB.US](https://longbridge.com/en/quote/NVDB.US.md)
- [NVDD.US](https://longbridge.com/en/quote/NVDD.US.md)
- [NVDG.US](https://longbridge.com/en/quote/NVDG.US.md)
- [NVDO.US](https://longbridge.com/en/quote/NVDO.US.md)
- [NVDW.US](https://longbridge.com/en/quote/NVDW.US.md)
- [NVYY.US](https://longbridge.com/en/quote/NVYY.US.md)
- [NYYY.US](https://longbridge.com/en/quote/NYYY.US.md)
- [DIPS.US](https://longbridge.com/en/quote/DIPS.US.md)
- [MAGX.US](https://longbridge.com/en/quote/MAGX.US.md)
- [GOOY.US](https://longbridge.com/en/quote/GOOY.US.md)
- [AVGW.US](https://longbridge.com/en/quote/AVGW.US.md)
- [AVS.US](https://longbridge.com/en/quote/AVS.US.md)
- [GOU.US](https://longbridge.com/en/quote/GOU.US.md)
- [GOOX.US](https://longbridge.com/en/quote/GOOX.US.md)
- [GOOP.US](https://longbridge.com/en/quote/GOOP.US.md)
- [SMH.US](https://longbridge.com/en/quote/SMH.US.md)
- [SOXX.US](https://longbridge.com/en/quote/SOXX.US.md)
- [SOXL.US](https://longbridge.com/en/quote/SOXL.US.md)
- [SOXQ.US](https://longbridge.com/en/quote/SOXQ.US.md)
- [XSD.US](https://longbridge.com/en/quote/XSD.US.md)
- [PSI.US](https://longbridge.com/en/quote/PSI.US.md)
- [FTXL.US](https://longbridge.com/en/quote/FTXL.US.md)
- [07388.HK](https://longbridge.com/en/quote/07388.HK.md)
- [07788.HK](https://longbridge.com/en/quote/07788.HK.md)
- [09388.HK](https://longbridge.com/en/quote/09388.HK.md)
- [SOXS.US](https://longbridge.com/en/quote/SOXS.US.md)
- [SSG.US](https://longbridge.com/en/quote/SSG.US.md)
- [USD.US](https://longbridge.com/en/quote/USD.US.md)
- [SOXW.US](https://longbridge.com/en/quote/SOXW.US.md)
- [SOXM.US](https://longbridge.com/en/quote/SOXM.US.md)

## Related News & Research

- [Google Is Reportedly Working with AMD for Its New TPU. What This Means for AMD Stock.](https://longbridge.com/en/news/296528134.md)
- [Nvidia Q2 Preview: Strong Earnings and Low Valuation Set the Stage for Upside](https://longbridge.com/en/news/296688005.md)
- [Google Lost About $186 Billion After 4 of Its Top AI Researchers Left, but Sundar Pichai Says Google Will Be ‘A Founding Investor’ Anyway](https://longbridge.com/en/news/296636401.md)
- [How to Play MRVL Stock Now as Marvell Technology Expands Its Relationship With Google](https://longbridge.com/en/news/296708030.md)
- [Google Gemini is getting a dedicated student hub](https://longbridge.com/en/news/296390917.md)

---
> **Disclaimer: This article is for reference only and does not constitute any investment advice.**