OpenAI Says Microsoft Can’t Keep Up With Its Data Center Plans

OpenAI is planning to grow its data center footprint beyond Microsoft’s infrastructure, with the artificial intelligence startup’s leadership indicating its hyperscale backer is moving too slowly to meet its skyrocketing computing needs.

OpenAI CEO Sam Altman and Chief Financial Officer Sarah Friar told employees and shareholders this month that the firm plans to take a greater role in securing data centers and the chips housed inside of them, reducing the company’s reliance on Microsoft’s infrastructure, The Information reported. 

OpenAI’s leadership expressed impatience with Microsoft’s ability to provide the digital infrastructure needed to stay ahead of a growing pack of rival artificial intelligence firms that includes Anthropic and Elon Musk’s xAI, according to the report. 

The comments come less than a week after OpenAI closed a $6.6B funding round that included significant capital investment from Microsoft and Nvidia as well as major funds like Softbank and Abu Dhabi's state-backed investment vehicle MGX. 

Microsoft and OpenAI’s partnership, codified through Microsoft’s $13B investment in the ChatGPT creator in early 2023, has elevated both firms to the head of the pack in Big Tech’s escalating AI arms race. 

The terms of the deal give Microsoft a significant cut of OpenAI’s profits, the right to use OpenAI’s technology in perpetuity and mandate that Microsoft remain the AI firm’s exclusive cloud infrastructure provider — a position Microsoft has leveraged into growing market share for its Azure cloud products. 

For OpenAI, the partnership firmly established the company as the benchmark for a successful AI startup and allowed it to rapidly scale up its computing and data center infrastructure almost entirely on Microsoft’s dime. The bulk of the company’s computing power is housed in Microsoft facilities near Phoenix. 

But in recent months, OpenAI has begun chafing at Microsoft’s pace of infrastructure investment, which the firm’s top executives feel is too slow to meet the company’s long-term growth and product development targets, according to The Information. While Microsoft is developing new data center campuses at a clip that would have been unimaginable just two years ago, OpenAI wants to get bigger faster. 

Microsoft’s relative sluggishness when it comes to capital investments in data centers and other digital infrastructure to support AI is due, at least in part, to fears that Wall Street will punish the tech giant’s share price for AI capital expenditure that isn’t tied to near-term revenue. As Bisnow has reported previously, the world’s largest cloud and social media firms have gone to great lengths to show investors that their record Capex isn’t a long-term moonshot but will have immediate bottom-line impact. 

OpenAI’s capital expenditure strategy, by contrast, is focused beyond the immediate horizon. The company has adopted a maximalist view of its future AI infrastructure needs, pitching White House officials last month on the need for multiple five-gigawatt AI data centers throughout the U.S. 

According to The Information, OpenAI’s urgency to expand its computing infrastructure is also being driven by the firm’s rivalry with xAI, whose founder, Elon Musk, also co-founded OpenAI. Last month, xAI announced it had launched one of the largest clusters of AI processors in the world, the firm’s so-called Colossus data center cluster near Memphis, Tennessee.

OpenAI has already started pursuing data center deals with infrastructure providers besides Microsoft. In June, OpenAI signed an agreement to rent hundreds of megawatts of server capacity at a data center being built in Abilene, Texas, by Oracle, a Microsoft competitor. While Microsoft was technically a party to the deal, it was negotiated directly between OpenAI and Oracle. 

According to The Information, OpenAI is now exploring the possibility of leasing the entire complex from Oracle, eventually giving it access to as much as 2 gigawatts of capacity at what will be one of the world’s largest AI clusters. The site is expected to have around a gigawatt of power by 2026.

A month after announcing the Oracle deal, OpenAI’s Altman floated the idea of the company self-developing its own data centers. As Bisnow reported in July, Altman told tech industry executives he intends to establish one or more companies that will use outside capital to acquire land and power and fund the construction of new data centers for OpenAI, part of a broader effort to develop and design proprietary artificial intelligence chips. 

Data centers aren't the only asset class in which OpenAI is rapidly expanding its global real estate portfolio. The company announced Wednesday it is opening new offices in multiple cities around the world, adding New York City, Seattle, Paris, Brussels and Singapore to its global footprint. OpenAI already has offices in San Francisco, London, Dublin and Tokyo. 

As Bisnow reported Tuesday, OpenAI’s New York offices will occupy 90K SF at the Puck Building in the SoHo neighborhood of Manhattan. The building is owned by Kushner Cos. and houses an office of Joshua Kushner’s Thrive Capital, a venture capital firm that has invested more than a billion dollars in OpenAI.