Anthropic is committing to its largest infrastructure expansion yet to secure the capacity behind its fast-growing AI business.
The company has expanded its partnership with Google and Broadcom to access multiple gigawatts of next-generation compute, expected to come online starting in 2027. Anthropic said its annualized revenue has reached $30 billion, up from about $9 billion at the end of 2025, driven by rising enterprise demand.
Inside the 3.5 gigawatt commitment
Anthropic said it signed a new agreement with Google and Broadcom for multiple gigawatts of next-generation compute capacity. The expansion builds on earlier partnerships and centers on Google Cloud’s tensor processing units, or TPUs, which Broadcom helped design and manufacture.
A Broadcom disclosure, reported by CNBC, put the scale of the agreement at about 3.5 gigawatts of compute capacity. The infrastructure is expected to come online starting in 2027.
“This groundbreaking partnership with Google and Broadcom is a continuation of our disciplined approach to scaling infrastructure,” said Krishna Rao, CFO of Anthropic.
“We are building the capacity necessary to serve the exponential growth we have seen in our customer base,” Rao added.
Most of the new infrastructure will be located in the United States, extending Anthropic’s earlier $50 billion commitment to domestic compute investment.
Demand is rising faster than capacity
Anthropic’s expansion comes as demand for its services continues to grow. TechCrunch stated that the company’s annualized revenue has surpassed $30 billion, up from about $9 billion at the end of 2025. More than 1,000 enterprises now spend over $1 million annually, a figure that has doubled in less than two months.
The growth is pushing the company to secure compute well ahead of need. Broadcom CEO Hock Tan said Anthropic’s demand could exceed 3 gigawatts in 2027, up from about 1 gigawatt in 2026, according to CNBC.
Anthropic is also spreading workloads across multiple chip platforms. The company noted that it runs its systems on Google TPUs, AWS Trainium, and Nvidia GPUs, allowing it to match AI workloads to the chips best suited for them.
Why custom chips are becoming the default
The expanded partnership highlighted the changes in how AI infrastructure is built. Instead of relying only on standard hardware, companies are working more closely with chipmakers and cloud providers to shape how chips are designed and deployed.
CNBC emphasized that Broadcom’s role in building Google’s TPU places it at the center of that shift, alongside cloud providers and model developers in a more tightly integrated ecosystem.
Other companies are taking similar steps. For instance, OpenAI is working with partners on custom silicon and large-scale GPU deployments, while cloud providers continue investing in their own chips to support growing workloads.
As access to compute becomes harder to secure, deals like Anthropic’s suggest that long-term capacity and hardware partnerships will continue to shape the industry.
Take a closer look at how Microsoft’s $10B investment is expanding AI, cloud, and cybersecurity infrastructure in Japan and what it means for the tech industry.