OpenAI data center pivot highlights Wall Street’s IPO concerns

Speaking at BlackRock’s U.S. Infrastructure Summit earlier this month, OpenAI CEO Sam Altman acknowledged that his company faces a harsh reality in data centers.

“At this scale, it’s like so many things can go wrong,” Altman said during a fireside chat at a conference in Washington, D.C.

Altman cited the example of severe weather that temporarily “disrupted things” at the data center campus in Abilene, Texas. This facility will serve as OpenAI’s flagship site; oracle and SoftBank’s $500 billion Stargate project. Altman said the company also faces supply chain challenges and pressure to meet tight deadlines.

The stakes are rising as Altman aims to transform OpenAI, valued at $730 billion in a record funding round last month, from a darling of private markets to an investable asset for a more discerning class of public market fund managers. That means backing away from big spending plans, shelving certain ambitious projects, and embracing OpenAI’s role as a buyer of large amounts of cloud capacity rather than a builder of giant data centers.

“OpenAI realized that the market didn’t necessarily appreciate a reckless approach to growth and spending,” Futurum Group CEO Daniel Newman said in an interview with CNBC. “The market wants OpenAI’s revenue to grow at a pace that justifies the spending. In my opinion, the point is to try to show a little more fiscal responsibility.”

The strategic shift means OpenAI may have to settle for doing less while trying to compete with Anthropic. google Many other companies are developing AI models, apps, and features. OpenAI trains and runs AI models that require vast amounts of computational resources such as chips, processing power, memory, and energy. Altman and other OpenAI executives have long emphasized that computing is a major bottleneck for the company, which has raised an astronomical amount of funding, including $110 billion earlier this year, including $50 billion. Amazon.

In a November post on X, Altman wrote that OpenAI and other companies “face such severe compute constraints that we have to rate limit our products and not offer new features or models.”

Up until that point, the big story about OpenAI last year was the extraordinary lengths Altman went to to secure capacity. The company has completed a series of multi-billion dollar infrastructure deals with the following companies: Nvidia, advanced micro device and broadcom. Altman said in a November post that OpenAI is considering a commitment of about $1.4 trillion over the next eight years.

The deal roiled public markets, raised concerns about a potential AI bubble, and left many investors wondering how OpenAI could make such an eye-popping commitment with annual revenues of $13.1 billion.

OpenAI’s most notable announcements are: Nvidia. The chipmaker, which is also the world’s most valuable company, agreed in September to invest up to $100 billion over many years in the startup, with capital allocation tied to building OpenAI and using Nvidia’s technology. OpenAI said it plans to deploy at least 10 gigawatts of Nvidia systems, and that the initial $10 billion investment will coincide with the completion of the first gigawatt, a unit of power roughly equivalent to the electricity consumption of a mid-sized city.

In a press release, the partnership states that “OpenAI will be able to build and deploy at least 10 gigawatts of AI data centers.”

Analysts told CNBC at the time that the deal was reminiscent of the vendor financing that fueled the dot-com bubble in the late 1990s. Altman repeatedly dismissed concerns about OpenAI’s ambitious infrastructure plans, suggesting that revenue could grow to hundreds of billions of dollars by 2030.

But in recent months, as the company has geared up for a potential IPO later this year, OpenAI has tempered expectations and outlined a more cautious strategy. OpenAI told investors in February that it aims to bring total computing spending to about $600 billion by 2030. This number is more directly tied to expected revenue growth.

The company emphasizes discipline in other areas of its business as well. In December, OpenAI declared a “code red” to focus on improving its ChatGPT chatbot in the face of increasing competition from Google and Anthropic.

Fiji Simo, OpenAI’s CEO of applications, held an all-staff meeting about the enterprise business earlier this month and said the company is “aggressively working” on high productivity use cases.

“What’s really important for us right now is to stay focused and execute very well,” Simo said, according to a partial transcript of the meeting reviewed by CNBC.

“This is a race.”

OpenAI currently does not own any data centers and may not in the near future, said people familiar with the matter who asked not to be identified because they were not authorized to speak publicly.

Instead, we have chosen to rely heavily on partners like Oracle. microsoft And Amazon is trying to piece together as much capacity as possible.

A year ago, the landscape for OpenAI was very different. In January 2025, President Donald Trump announced the Stargate project at a White House event with Altman, SoftBank CEO Masayoshi Son, and Oracle Chairman Larry Ellison. The companies have committed $500 billion over four years to building new AI infrastructure in the United States.

According to a blog post at the time, OpenAI will be in charge of project management, while SoftBank will be in charge of finance. Oracle and Nvidia have been named as key initial technology partners.

“Oracle, Nvidia, and OpenAI will work closely together to build and operate this computing system,” the release states.

Upon breaking ground on Stargate, OpenAI was poised to develop much of the project itself and was looking to directly lease or own some data center campuses, according to a report in The Information. But after the company faced real construction problems and struggled to get support from lenders, it changed direction.

Oracle is leasing Stargate’s data center campus in Abilene and has assumed tens of billions of dollars in debt to fund construction.

OpenAI and Nvidia said in a September release that the first gigawatt-scale Nvidia systems will be deployed in the second half of 2026. Experts said the timeline would be tight even in the best of circumstances.

Walid Saad, an engineering professor at Virginia Tech, said building a 1-gigawatt data center could take three to 10 years from start to finish. Challenges can arise at every step, from locating a site, securing the proper permits and permissions, accessing power, building the physical structure, delivering the hardware, to finally getting it online.

“There are regulations, there are permits, and different locations have different processes,” Saad said. “There are processes that they can’t control. We don’t know what’s going to happen.”

Gartner AI analyst Arun Chandrasekaran told CNBC in an interview that these obstacles are becoming very real for OpenAI.

“They’re starting to say, ‘You know, let’s try to get as much capacity as possible from providers that are willing to give us that capacity,'” Chandrasekaran said.

OpenAI has not commented on this matter.

As part of OpenAI’s $110 billion funding announcement last month, the company agreed to consume approximately 2 gigawatts of Trainium capacity through Amazon Web Services infrastructure. Trainium is a custom AI chip from AWS. Amazon announced the latest version, Trainium 3, in December.

Nvidia also contributed to OpenAI’s funding round, investing $30 billion. As part of the deal, OpenAI said it has agreed to expand its collaboration with Nvidia and use 3 gigawatts of dedicated inference power and 2 gigawatts of training capacity on Nvidia’s upcoming Vera Rubin system.

“OpenAI is doing what it needs to do: gain access to large-scale computing,” said Futurum Group’s Newman, adding: MetaAnthropic and Google are doing similar things. “This is racing.”

Nvidia’s investment comes after months of speculation about the status of a major infrastructure deal that the two companies announced in September. The chipmaker disclosed in its November quarterly report that the $100 billion deal may not materialize, and the Wall Street Journal reported in January that the deal was “on ice.”

Nvidia said in a February filing that there is “no guarantee” that the company will “enter into an investment and partnership agreement with OpenAI or that the transaction will be completed.”

At a conference earlier this month, NVIDIA CEO Jensen Huang further tempered expectations, saying the opportunity to invest $100 billion in OpenAI is probably “not in the cards.”

The investment has nothing to do with any deployment milestone and differs from the deal structure the two companies touted six months ago. Huang said “this may be the last time” Nvidia invests in OpenAI ahead of its IPO.

“To their credit, they’ve built an incredible growth story, but the rest of the journey isn’t free,” Newman said of OpenAI. “And with such a high cost structure, the path to profitability will be scrutinized at every step.”

–CNBC’s Kate Rooney contributed to this report

Spotlight: OpenAI refocuses enterprise with all-hands meeting amid IPO drive