In an increasingly competitive environment for artificial intelligence, compute power has emerged as the primary battleground. OpenAI's and Anthropic's CFOs, Sarah Friar and Krishna Rao respectively, are under pressure to secure the necessary infrastructure to meet soaring demand from users and businesses alike.
Both companies are racing to enhance their compute capabilities, which Friar identifies as a key competitive advantage. In an interview with Bloomberg TV on May 15, she pointed out that the tech industry is facing an impending shortage of compute resources, particularly by 2026. OpenAI, which has over 900 million weekly active users of ChatGPT, is struggling to maintain sufficient computing capacity as its enterprise sales team is inundated with requests from companies eager to implement its models.
Rao shares this view, stressing that the compute resources Anthropic acquires are essential to its operations. On a recent episode of the Invest Like The Best podcast, he stated, "The compute that we procure, it’s the lifeblood of our business—it is the most important thing in the company." He spends a significant portion of his time—between 30% to 40%—making compute-related decisions, highlighting the importance of these choices.
Both executives face the challenging task of determining how much compute power to secure as their firms grow. Rao referred to these decisions as “some of the most consequential and hardest decisions” a company can make. Finding the right balance is crucial; securing too much capacity risks unnecessary costs, while too little could lead to unmet customer demand. "You have to really think ahead to plan this," he added.
To meet their compute needs, OpenAI and Anthropic utilize a mix of suppliers. Anthropic primarily sources its GPUs from Nvidia, along with Google TPUs and Amazon’s Tranium chips. OpenAI also partners with Nvidia and Amazon but diversifies its supply chain by incorporating AMD chips and developing in-house hardware with Broadcom. Recently, Cerebras Systems, a compute provider, went public on May 14, increasing competition for AI infrastructure.
The high costs associated with AI infrastructure have prompted significant fundraising efforts within the sector. Rao noted that establishing a sustainable AI business is capital-intensive; when he joined Anthropic, the startup was completing a $750 million funding round at a $15 billion valuation. The company is now reportedly seeking to raise $30 billion, which could value it at an astonishing $900 billion, potentially exceeding OpenAI's own valuation of $852 billion achieved in March following a $122 billion cash injection.
https://www.youtube.com/watch?v=wEEZPpx8qow
As competition intensifies, OpenAI is also exploring additional fundraising strategies. While Friar acknowledges that their recent capital raise provides ample flexibility, she remains open to securing further funds, stating, "We want to be able to, over the long run, raise money across the whole spectrum—and the public markets are significantly bigger."
Despite its established position, OpenAI faces a rapidly advancing competitor in Anthropic. Founded in 2021 by former OpenAI employees, Anthropic’s revenue run rate surpassed $30 billion in April, a significant jump from $9 billion at the end of 2025. In comparison, OpenAI recorded an annualized revenue of $24 billion as of March, although these figures are not entirely comparable since OpenAI does not include revenues from its cloud partners in its total.
As both companies pursue compute power and funding aggressively, the implications for the AI sector are substantial. The decisions made by Friar and Rao will not only shape their respective firms but also influence the broader trajectory of AI development and deployment in the coming years.
