Microsoft, as it turns out, does not have enough data center capacity to support all of OpenAI’s needs. According to The Information, AI companies intend to aggressively scale their computing capabilities.
After OpenAI raised $6.6 billion in funding in October, CFO Sarah Friar spoke to employees saying the company would be more aggressive in its search for data centers and AI chips, according to reporters. She previously told shareholders that Microsoft was too slow to work with an AI startup. For example, Elon Musk’s xAI quickly launched a large AI data center in Memphis when the need arose.
OpenAI is gradually moving away from Microsoft, which remains a key stakeholder in the company. In June, a deal was announced with Oracle to lease servers in a data center currently under construction in Abilene (Texas). Although Microisoft is mentioned in the press release, the IT giant is not too involved in the project. Instead, it will actually rent servers and provide them to OpenAI. This allows us to say that the company is still the exclusive cloud provider of the latter.
Image source: Mariia Shalabaieva/unsplash.com
Now OpenAI is negotiating the lease of this entire data center, by mid-2026 its capacity will be almost 1 GW. If Oracle manages to obtain more energy, the site will grow to 2 GW. Additionally, OpenAI could pursue its long-promised 5 GW Stargate data center, but without Microsoft – if the AI company can find funding for a data center worth more than $100 billion and supply it with energy.
According to OpenAI CEO Sam Altman, who spoke to employees this week, costs could theoretically be reduced by using chips developed by OpenAI itself, but it is not yet clear at what stage of implementation this project is. Earlier it was reported that Altman is trying to convince the US authorities of the need to build a 5-GW data center network in the country.