OpenAI has informed its investors this week that its early and aggressive investment in computing resources has provided a crucial advantage over its competitor, Anthropic PBC. This communication comes as Anthropic considers an initial public offering (IPO). According to a memo sent to certain investors, the developer of ChatGPT stated that by "rapidly and consistently" increasing its computational capacity to support broader adoption of its software, OpenAI has pulled ahead of Anthropic. The memo was issued following Anthropic's launch of a more powerful AI model named Mythos.
OpenAI indicated in the memo that its ambitious infrastructure build-out, despite past criticism over high costs, has positioned the company to better keep pace with growing demand for AI products. The company wrote, "This gap is critical because compute has now become a product constraint." OpenAI declined to comment further.
Anthropic, which was founded by former OpenAI employees, has at times struggled to keep its services online in recent months, impacted by demand surges following new product releases and public support related to its stance with the Pentagon. In its investor memo, OpenAI referenced a report by Stratechery blogger Ben Thompson, which suggested that compute limitations may have influenced Anthropic's decision to release Mythos only to select partners.
OpenAI stated it has 1.9 gigawatts of available computing capacity for 2025, triple the amount from the previous year. The company projects this figure will grow into the "low double-digit" gigawatt range next year and reach approximately 30 gigawatts by 2030. In contrast, OpenAI estimates that Anthropic will possess 1.4 gigawatts of capacity by the end of 2025 and between 7 to 8 gigawatts next year. OpenAI claimed in the memo, "Even at the high end of Anthropic's projected range, our growth rate remains significantly faster and the gap is widening."
In recent months, Anthropic has increased its investment in the physical infrastructure for its AI services, including a commitment to spend $50 billion on data center construction in the United States. Furthermore, Anthropic has expanded strategic collaborations with Broadcom (AVGO) and Alphabet's (GOOGL) Google, securing access to approximately 3.5 gigawatts of computing capacity starting in 2027. Anthropic also maintains a diversified supplier portfolio, working with all three major cloud providers: Google, Microsoft (MSFT), and Amazon.com (AMZN).
In response to a request for comment, Anthropic referenced a previous statement from its Chief Financial Officer, Krishna Rao, regarding the deals with Broadcom and Google. Rao said, "This groundbreaking collaboration with Google and Broadcom continues our prudent approach to scaling infrastructure. We are making our most significant compute commitment to date to keep pace 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. Nonetheless, the speed of OpenAI's expenditures faces scrutiny, given the company's expectation that it will not be profitable for several years. On Thursday, OpenAI announced it would pause an infrastructure project in the UK, citing energy costs.
In an interview late last year discussing his company's spending plans, Anthropic CEO Dario Amodei stated, "As a company, we manage things as responsibly as we can. I think some players are in a 'YOLO' mode." In its investor memo, OpenAI characterized Amodei's view as a misjudgment of the market's demand for more AI products. The company stated, "In hindsight, this caution looks less like discipline and more like underestimating the speed at which demand would arrive."
Comments