When an AI company's appetite for computing power is measured in gigawatts, you know the artificial intelligence race has entered a completely different league. Anthropic, the maker of the Claude AI models, announced Monday that it has signed a major new agreement with Google and Broadcom to dramatically expand its processing and compute capacity — and the scale of this deal signals just how explosively the demand for AI infrastructure is growing.
The new agreement expands Anthropic's use of Google Cloud's tensor processing units, or TPUs, which are Google's advanced AI chips. This is not a fresh relationship. The deal builds on a partnership the companies struck in October 2025 for more than a gigawatt of compute capacity. But the new commitment dwarfs that earlier arrangement.
While Anthropic did not publicly specify the exact size of the expansion, a recent Broadcom SEC filing reveals that the deal includes 3.5 gigawatts of compute. To put that in perspective, 3.5 gigawatts is roughly the output of three large nuclear power plants — all dedicated to running AI models.
This new compute capacity is expected to come online in 2027, meaning Anthropic is planning well ahead for a future where its AI models will require even more infrastructure than they do today.
Why Does Anthropic Need So Much Power?
The short answer: its business is exploding. Anthropic's run rate revenue has reached $30 billion, a drastic jump from the $9 billion the company recorded at the end of 2025. That kind of revenue growth in just a few months is almost unheard of, even in the fast-moving tech industry.
The company now has more than 1,000 business customers spending more than $1 million on an annualized basis. Enterprise adoption of Claude models has clearly moved from experimentation to serious, large-scale deployment — and every enterprise customer running AI workloads means more demand for the computing infrastructure that makes it all work.
Anthropic's CFO Krishna Rao described the partnership as a continuation of the company's disciplined approach to scaling infrastructure, stating that they are building the capacity necessary to serve the exponential growth in their customer base while also enabling Claude to define the frontier of AI development.
Investing Billions in American Infrastructure
This deal is not just about raw computing power. It is also a statement about where that power will be located. The majority of the new compute capacity will be housed in the United States and represents an extension of Anthropic's $50 billion commitment to invest in U.S. compute infrastructure.
That commitment puts Anthropic alongside other major AI players who are pouring tens of billions of dollars into building data centers across America. The AI infrastructure buildout has become one of the largest capital investment trends in the technology sector, with companies competing not just on model quality but on who can secure enough chips, power, and physical space to train and run the next generation of models.
A Company on an Unstoppable Trajectory
Anthropic's compute expansion comes during a period of remarkable momentum for the company, though not without challenges. The company recently closed a $30 billion Series G funding round that valued it at $380 billion. That valuation places Anthropic among the most valuable private companies in the world and reflects investor confidence that the demand for its AI models will continue to accelerate.
This growth has continued despite the U.S. Defense Department's labeling of Anthropic as a supply chain risk, a controversy that the company has been actively fighting in court. Anthropic won an injunction against the Trump administration over the Defense Department saga, signaling that the legal challenges have not slowed its commercial trajectory.
The Bigger Picture
The sheer scale of this deal — 3.5 gigawatts of compute from Google and Broadcom — underscores a reality that is becoming impossible to ignore. Building frontier AI models is no longer just about clever algorithms and talented researchers. It is increasingly an infrastructure game, where success depends on securing enormous amounts of computing power, energy, and capital.
For Anthropic, Monday's announcement is a bet that demand for Claude will continue growing at its current breakneck pace. For the rest of the AI industry, it is a signal that the compute arms race is only accelerating — and the companies that cannot keep up with infrastructure may find themselves left behind, no matter how good their models are.
