Anthropic expands compute contracts with Google and Broadcom amid revenue growth to $30 billion

Anthropic is expanding its computational capacity agreements with Google and Broadcom amid sharp business growth: the company's annualized run-rate revenue has reached $30 billion. This is one of the most impressive growth indicators among generative AI companies in recent years. Anthropic's partnership with Google began in 2023, when the tech giant initially invested several hundred million dollars, then increased total commitments to several billion dollars.

Along with capital, Google provided access to its cloud infrastructure, particularly tensor processing units (TPU) — which today form the foundation for training and inference of Claude family models. Expanding the current agreement means Anthropic will be guaranteed additional computing capacity as demand grows. Broadcom acts as a specialized chip supplier in this configuration.

The company develops ASIC solutions for major tech customers and designs custom accelerators for specific AI workloads. Deepening collaboration with Broadcom signals that Anthropic is seriously diversifying its hardware base — reducing dependence on standard NVIDIA GPUs amid persistent supply shortages and pricing pressure in the market. The $30 billion run-rate metric indicates that Anthropic's revenue is growing faster than even optimistic analysts expected.

In mid-2024, figures around $1 billion ARR were cited, while by year-end estimates had grown to $3-4 billion. A 7-8x increase over several quarters is the result of mass enterprise adoption of Claude API, active use of Claude in Amazon products (another major investor), and growing Claude.ai audience among individual users.

Compute is a central expense line for any AI company. Training large language models requires thousands of accelerators running continuously for months; inference at a growing user base costs no less. Expanding compute agreements is the standard response to rising demand.

What matters more is this: Anthropic is building a diversified strategy, relying simultaneously on Google's TPUs and Broadcom's specialized chips rather than being tied to a single supplier. This approach reflects a trend visible across all major AI labs. Microsoft is building its own Maia chips for Azure, Meta is developing MTIA, Google is improving TPU now in its sixth generation.

Dependence on NVIDIA is high for everyone, but leading players are investing in alternative hardware. Anthropic is following the same path — and at current growth rates, it has both the financial resources and strategic motivation to do so proactively. Expanding compute agreements amid $30 billion run-rate revenue is a signal: Anthropic is not just meeting current demand, but reserving capacity for the next wave of growth.

In an industry where computational scarcity directly limits product development speed, such a move makes strategic sense.