OpenAI mulls new revenue from AI infrastructure

OpenAI is exploring a potential new revenue stream: renting out its AI-ready data centers and infrastructure to companies needing massive computing power.

The idea mirrors Amazon’s early move into cloud computing nearly two decades ago, when it began offering excess capacity to outside businesses. That experiment evolved into Amazon Web Services (AWS), now a multibillion-dollar powerhouse and a cornerstone of the modern internet.

For OpenAI, the logic is similar. The company has poured resources into cutting-edge chips, servers, and cooling systems to power its large-scale AI workloads. Allowing others to rent that infrastructure could open the door for startups and smaller firms to access high-performance computing without building it themselves—while creating a lucrative business line for OpenAI.

Still, Chief Financial Officer Sarah Friar emphasized the idea remains speculative. With demand for ChatGPT and other AI products soaring, OpenAI’s immediate priority is securing enough capacity for its needs. In a recent interview, Friar confirmed the company isn’t actively pursuing the plan yet but sees it as a possible opportunity in the future.

CFO eyes infrastructure leasing down the road

OpenAI has become good at building data centers tailored for artificial intelligence. This know-how could be productized. Friar said that OpenAI is trying to take more control of its infrastructure design instead of relying on external vendors, cautioning that if the company only purchased equipment from others, it would risk giving away its intellectual property.

The company has raised tens of billions for ultrapowerful AI chips and facilities. Its Stargate project with SoftBank and Oracle promises to construct some of the largest data centers in the world, in the United States and beyond.

Excitement is running high after CEO Sam Altman promised ambitious plans, saying that people should expect OpenAI to spend trillions of dollars on infrastructure shortly. He added that the company is working on an interesting new financial instrument to fund these mega-projects, though he did not provide further details.

Until now, OpenAI has depended on Microsoft and the technology company Oracle to cover most of its infrastructure costs. But Friar said banks and private equity firms are now piling in, providing debt financing. In addition to debt, the company is considering innovative financing mechanisms.

But even with such aspirations, OpenAI still loses money. Data centers and GPUs are very expensive. But growth is rapid. In July, the firm surpassed a $1 billion monthly revenue for the first time — buoyed by surging global demand for ChatGPT and its enterprise tools.

Altman frames bold plans, warns of AI hype

Despite his push for massive spending, Altman has also acknowledged that AI may be in a bubble. In an interview last week, he compared the current enthusiasm to the dot-com era, noting that smart people often go overboard on a kernel of truth.

Still, expansion is moving ahead. The project, Stargate, is a $500 billion joint venture and is already starting with projects in the US, Norway, and the UAE. The aim is to secure AI capacity to a level no company has ever achieved.

Altman has suggested we need 100 million AI GPUs to power future models. It could cost $3 trillion. For reference, that’s more than the GDPs of many countries.

According to recent accounts, OpenAI is reportedly gearing up for a second stock sale. It would enable current and former employees to cash out their shares at an estimated valuation of around $500 billion. That would be almost double its last disclosed valuation of $300 billion, in a financing round by SoftBank.

Investor appetite remains strong. OpenAI secured $41 billion in its latest round, exceeding its initial goal of $40 billion because of strong demand.

Want your project in front of crypto’s top minds? Feature it in our next industry report, where data meets impact.