OpenAI calls out to investors: computing power has become a product "hard constraint", we run faster and harder than Anthropic

This week, OpenAI revealed to its investors that, thanks to its early aggressive investment in computational resources, the company has gained a critical advantage in its competition with rival Anthropic PBC. Currently, Anthropic is considering an initial public offering (IPO). According to a memo sent to some investors, the developers of ChatGPT stated that OpenAI has surpassed Anthropic by rapidly and continuously increasing its computational capacity to support wider adoption of its software. This memo was released after Anthropic launched a more powerful AI model called Mythos. OpenAI pointed out in the memo that despite being criticized for high costs, its ambitious infrastructure development allows it to better keep up with the growing demand for AI products. OpenAI wrote in the memo, "This gap is crucial as computational power has now become a product constraint." OpenAI declined to comment. Founded by former OpenAI employees, Anthropic has struggled in recent months to maintain online services due to increased demand sparked by new product launches and public support in its standoff with the Pentagon. In a memo to investors, OpenAI referenced a report by Ben Thompson, author of the Stratechery blog, which suggested that computational limitations may have influenced Anthropic's decision to release Mythos only to select partners. OpenAI stated that it currently has 1.9 gigawatts of available computational capacity in 2025, triple the previous year's capacity (1 gigawatt is roughly enough to power 750,000 U.S. households). The company expects this number to increase to "low double-digit" gigawatts range next year, reaching around 30 gigawatts by 2030. In comparison, OpenAI estimates that Anthropic will have 1.4 gigawatts of capacity by the end of 2025, with 7 to 8 gigawatts next year. OpenAI wrote in the memo, "Even at the upper end of Anthropic's predicted range, our growth rate remains significantly ahead and the gap is widening." In recent months, Anthropic has increased its investment in physical infrastructure for AI services, including a commitment to invest $50 billion in building data centers in the United States. Additionally, Anthropic has expanded its strategic partnerships with Broadcom Inc. and Alphabet Inc.'s Alphabet Inc. Class C, enabling access to approximately 3.5 gigawatts of computational capacity starting in 2027. Anthropic also has a diversified supplier mix, collaborating with the three major cloud service providers: Alphabet Inc. Class C, Microsoft Corporation, and Amazon.com, Inc. In response to requests for comment, Anthropic quoted its Chief Financial Officer Krishna Rao's previous statement regarding transactions with Broadcom Inc. and Alphabet Inc. Class C. Rao said, "This groundbreaking collaboration with Alphabet Inc. Class C and Broadcom Inc. continues our prudent approach to expanding infrastructure. We are making our most significant computational commitment to date in order to keep up with this unprecedented growth." Historically, Anthropic's spending strategy has been more conservative than OpenAI's. OpenAI plans to invest approximately $600 billion in data centers and chips by 2030, and recently raised $122 billion to support these commitments. However, considering that OpenAI is not expected to be profitable in the coming years, its spending speed is still being questioned. On Thursday, OpenAI announced it would suspend an infrastructure project in the UK, citing energy cost issues. CEO of Anthropic, Dario Amodei, mentioned the company's spending plan at the end of last year, stating, "As a company, we will manage responsibly. I think some players are 'YOLO-ing.' In the investor memo, OpenAI described Amodei's viewpoint as a misjudgment of the market's demand for more AI products. The company stated, "In hindsight, this caution, rather than self-discipline, seemed to have underestimated the speed of demand arrival."