OpenAI、投資家に自社インフラがAnthropicに対する優位性をもたらすと説明

OpenAI is pitching investors on the idea that its early infrastructure buildout gives it a decisive advantage over Anthropic. Meanwhile, the company is pausing its UK data center project, and Anthropic is exploring custom AI chips.

The competition between OpenAI and Anthropic for AI dominance is increasingly playing out at the infrastructure level. In an investor memo sent out this week, according to Bloomberg, OpenAI argues that its early and aggressive buildout of compute capacity gives it a critical edge over rival Anthropic.

OpenAI says it has outpaced Anthropic by adding compute quickly and consistently, according to the memo. The ambitious infrastructure push - which critics had called too expensive - has allowed OpenAI to keep up with surging demand for AI products more effectively.

The memo likely came in response to Anthropic's announcement of a more powerful AI model called Mythos. The model will initially only be available to select partners through Project Glasswing for safety reasons. Some observers, however, question whether Anthropic could even roll out such a large model effectively with its current compute capacity.

There's speculation that Mythos is the first 10-trillion-parameter model. That speculation is based in part on public statements from Elon Musk that xAI is currently training a model of that size. It also draws on benchmarks shared by Nvidia CEO Jensen Huang showing how quickly a new Vera Rubin system could train upcoming 10-trillion-parameter models. Following this logic, OpenAI would have an advantage because it could more easily handle demand for powerful models like the upcoming "Spud." In practice, though, most customers would likely get access to distilled, smaller versions of these models.

Anthropic is exploring custom AI chips to reduce its dependency

Anthropic isn't sitting still. According to Reuters, the AI lab is considering designing its own chips. The plans are still in an early stage - there's no dedicated team and no concrete design yet, according to one of three sources cited. The company could ultimately decide against the effort and continue buying chips from other suppliers.

The move is driven by the ongoing shortage of AI chips. Anthropic currently uses a mix of Google's Tensor Processing Units (TPUs) and Amazon chips to develop and run its chatbot Claude. Just this week, Anthropic signed a long-term deal with Google and Broadcom that builds on a commitment to invest $50 billion in US compute infrastructure.

Designing an advanced AI chip costs around half a billion dollars, according to industry sources. Meta and OpenAI are reportedly pursuing similar efforts, according to Reuters. Anthropic declined to comment.

OpenAI pauses UK Stargate project over regulation and energy costs

While OpenAI is touting its infrastructure lead to investors, the company is putting a key data center project in the UK on hold. According to Reuters, an unfavorable regulatory environment and high energy costs are to blame.

The so-called Stargate UK project launched in September 2025 in partnership with Nvidia and Nscale, coinciding with a visit by US President Donald Trump to the UK that generated a total of 150 billion pounds in investment commitments. The project was meant to strengthen the UK's sovereign compute capacity and accelerate AI adoption across the country.

OpenAI said it sees enormous potential in the UK's AI future and plans to resume the project once regulatory and energy cost conditions support long-term infrastructure investment. London remains home to the company's largest international research hub.

AI News Without the Hype – Curated by Humans

Subscribe to THE DECODER for ad-free reading, a weekly AI newsletter, our exclusive "AI Radar" frontier report six times a year, full archive access, and access to our comment section.

Subscribe now