Anthropic is moving to take greater control of the computing infrastructure that powers its AI models, pursuing data center leases that would place the company in direct control of operating its own facilities. The effort could secure vast amounts of power-intensive capacity as the company positions itself for its initial public offering.

Infrastructure Expansion

Anthropic has assembled more than a dozen preliminary agreements tied to U.S. data centers representing more than 1 gigawatt of capacity, according to The Information and Reuters. The company is exploring ways for Alphabet Inc.'s (NASDAQ:GOOG)(NASDAQ:GOOGL) Google to help support the financial commitments associated with those leases. Under the arrangements under discussion, Google could provide guarantees tied to Anthropic's lease obligations, helping the startup shoulder the enormous costs of securing computing infrastructure. The Information also reported that Google co-designs some of the server chips Anthropic may ultimately deploy at those facilities.

Strategic Positioning

The expansion underscores the escalating race among leading AI developers to lock down the computing power needed to train and run increasingly sophisticated models. Surging demand for Anthropic's Claude models has already pushed the company to pursue additional capacity through large-scale infrastructure agreements. Earlier this month, Anthropic disclosed that it had confidentially filed paperwork for a U.S. initial public offering. The company has not yet determined how many shares it would sell or what price it would seek in an offering.

The Billion-Dollar Race For AI Computing Power

The infrastructure push comes as Anthropic's valuation continues to climb. Anthropic's latest funding round valued the startup at $965 billion, surpassing OpenAI as the most valuable startup, after raising $65 billion in Series H. Altimeter Capital, Dragoneer, Greenoaks and Sequoia Capital led the funding round. In April, Anthropic also said Alphabet was committing up to $40 billion in the company, Bloomberg reported.

"This funding will help us serve the historic demand we are experiencing, stay at the research frontier, and bring Claude to more of the places where work happens," said Krishna Rao, CFO of Anthropic.

Earlier this week, Anthropic launched Claude Fable 5, its most capable consumer-facing AI model to date. The company said Fable 5 excels at coding, research and other knowledge-intensive tasks, particularly over long and complex workloads. If Anthropic goes public at a $1 trillion valuation, it would instantly rank among the most valuable companies globally and could become the second- or third-largest IPO in history, trailing SpaceX and Saudi Aramco.

An OpenAI employee has withdrawn his signature from an amicus brief supporting Anthropic in its legal battle against the U.S. Department of Defense, citing a recent policy reversal by the AI company. Brian Fioca, a member of OpenAI's Go to Market Staff, announced his decision on X, stating he could no longer truthfully argue that Anthropic is not a supply chain risk. The withdrawal follows Anthropic's decision to abandon a controversial safeguard that would have secretly hindered competitors from using its newest AI model, Claude Fable 5.

In March, more than 30 employees from Google, OpenAI, and Google DeepMind filed the amicus brief to support Anthropic's lawsuit against the Department of Defense. The brief urged the court to recognize the ethical commitments of AI developers as contributions to good governance rather than obstacles. However, Anthropic's recent policy shift has prompted Fioca to reconsider his stance, underscoring the tensions between AI safety measures and competitive transparency.

Anthropic initially implemented safeguards in Claude Fable 5 to curb high-risk uses of advanced AI, such as cybersecurity or biological weapons research. While some measures aligned with its safety strategy, a hidden provision aimed at reducing the model's effectiveness for frontier AI research drew criticism. Researchers argued the undisclosed restrictions amounted to a covert handicap on competitors using Claude to develop rival models.

Following backlash from researchers and industry observers, Anthropic reversed the policy, acknowledging it had "made the wrong tradeoff." The company apologized for the lack of transparency and pledged to make the safeguards visible. Critics, however, noted that the incident raised concerns about Anthropic's willingness to embed covert performance degradation into its models, despite terms of service already prohibiting customers from using its systems to train competing AI products.

The controversy has implications for Anthropic's legal standing and its coalition of supporters. Dean Ball, a senior fellow at the Foundation for American Innovation, remarked that defending Anthropic against government overreach would have been more difficult had the policy reversal occurred earlier. The incident highlights the challenges AI companies face in balancing safety, transparency, and competitive dynamics in a rapidly evolving industry.