Broadcom discloses USD 21 B Anthropic order for Google’s 7th‑gen TPUs
Broadcom’s latest filing reveals a $21 billion purchase from Anthropic, the AI startup that has been building its own large‑language models on Google’s cloud platform. The size of the order signals more than just a one‑off transaction; it underscores how tightly intertwined hardware and software providers have become in the race to scale generative AI. While Anthropic will run its workloads on Google’s seventh‑generation tensor processing units, the chips themselves never leave Broadcom’s fabs.
The partnership stitches together Google’s design expertise with Broadcom’s manufacturing capacity, turning a theoretical architecture into physical silicon that can be deployed at scale. For investors and industry watchers, the deal offers a concrete data point on how much capital is flowing into the infrastructure that underpins today’s AI services. It also raises questions about supply‑chain dynamics and pricing models as cloud providers lean on external foundries to meet soaring demand.
Now in their seventh generation, TPUs are available to customers through Google Cloud and power many of Google's internal systems, including training and deployment of the Gemini family of models. Google designs the TPU architecture, while Broadcom converts those designs into manufacturable silicon.
Now in their seventh generation, TPUs are available to customers through Google Cloud and power many of Google's internal systems, including training and deployment of the Gemini family of models. Google designs the TPU architecture, while Broadcom converts those designs into manufacturable silicon and handles volume production. The relationship mirrors Google's long-standing strategy of controlling key AI hardware design while relying on semiconductor partners for fabrication expertise.
Anthropic, a long-term user of TPUs, recently announced plans to significantly scale its infrastructure. The company intends to deploy one million TPUs, backed by more than one gigawatt of new compute capacity coming online in 2026. This represents one of the largest dedicated AI compute buildouts in the industry.
Several other companies have also confirmed their use of TPUs, including Meta, Cohere, Apple and Ilya Sutskever's new startup, Super Safe Intelligence (SSI). A report from The Information indicates that Meta is evaluating the deployment of TPUs in its data centres starting in 2027. The growing adoption of TPUs stems from their power efficiency and tight optimisation for AI training and inference, creating increasing competitive pressure on NVIDIA's GPU dominance.
Broadcom said it now has five TPU/XPU (custom AI accelerator) customers--with Google and Anthropic named on the call.
Is this the biggest single AI silicon contract yet? Broadcom’s announcement puts Anthropic’s TPU spend at $21 billion, a figure that dwarfs most chip deals announced this year. The $10 billion order arrived in the quarter before the earnings call.
A massive order. Followed by an $11 billion commitment for delivery in late 2026, according to CEO Hock Tan. Meanwhile, a $73 billion backlog of AI product orders sits on Broadcom’s books, slated for shipment over the next several years.
Seventh‑generation TPUs are already accessible through Google Cloud and underpin many of Google’s internal workloads, including training the Gemini model family. Google supplies the architecture; Broadcom turns those designs into silicon. Yet the timeline for fulfilling the new Anthropic orders remains vague, and it is unclear whether the backlog reflects sustained demand or a temporary surge.
The scale of the contracts suggests confidence in Broadcom’s manufacturing capacity, but whether the market can absorb such volume without price pressure is still uncertain. Only the forthcoming shipments will reveal how these numbers translate into real‑world impact.
Further Reading
- Broadcom Surges After Disclosing $21 Billion Order for Google AI Chips Used by Anthropic - Nasdaq
- Anthropic and Google’s Massive TPU Deal Highlights New AI Infrastructure Arms Race - Silicon Valley Business Journal
- Google Bets Big on Anthropic With Up to One Million TPUs in Expanded Cloud Partnership - The Register
- Anthropic’s Google Cloud Expansion: What the One‑Million‑TPU Deal Means for AI Compute Economics - SemiAnalysis
- Chipmakers, Cloud Giants, and AI Labs: How Google and Anthropic’s TPU Deal Reshapes the Supply Chain - Financial Times
More in Business & Startups
Harness secures USD 240 M E round, led by Goldman Sachs, for AI software delivery
Harness just closed a $240 million Series E, with Goldman Sachs heading the round, to fund its...
Time names AI power brokers 2025 Person of Year as firms spend billions
Time’s latest Person of the Year list puts the people who build and fund artificial‑intelligence...
Tomb Raider unveils Legacy of Atlantis remake on Unreal Engine 5
Square Enix’s Lara Croft franchise is expanding again, this time with a pair of titles that will...
Cohere's Rerank 4 quadruples context window cuts errors, improves search accuracy
Why does this matter? Cohere’s latest Rerank 4 model expands the context window to four times the...
Artists reflect on a decade of AI collaborations in Gradient Canvas
A decade has passed since Google’s Gradient Canvas first paired artists with machine‑learning...
Common Questions Answered
What is the total value of Anthropic's order for Broadcom's seventh‑generation TPUs?
Anthropic placed a $21 billion order with Broadcom for seventh‑generation Tensor Processing Units. The deal includes an initial $10 billion purchase and a subsequent $11 billion commitment slated for delivery in late 2026.
How does Google’s role differ from Broadcom’s in the production of the seventh‑gen TPUs used by Anthropic?
Google designs the architecture of the seventh‑generation TPUs and makes them available through Google Cloud, while Broadcom converts those designs into manufacturable silicon and handles the volume production. This division of labor lets Google control hardware design while relying on Broadcom’s fab capabilities.
What AI models are powered by the seventh‑generation TPUs mentioned in the article?
The seventh‑generation TPUs power many of Google’s internal systems, including the training and deployment of the Gemini family of large‑language models. Anthropic will also run its own generative‑AI workloads on these TPUs in Google’s cloud.
How does the $21 billion Anthropic order compare to Broadcom’s overall AI product backlog?
The $21 billion Anthropic order dwarfs most chip deals announced this year and represents a significant portion of Broadcom’s $73 billion AI product backlog. The backlog covers orders slated for shipment over the next several years, indicating Broadcom’s extensive involvement in AI silicon.