Chris Malone, the executive who oversaw construction of OpenAI’s data centers, has left the company, according to people familiar with the matter. An OpenAI spokesperson confirmed that Malone is no longer with the company and said its infrastructure division was reorganized earlier this year. The departure is the latest change in a unit that has become central to OpenAI’s strategy and to its balance sheet.
Malone joined OpenAI at a moment when the company was shifting from renting server capacity to building its own. Training frontier models requires clusters of graphics processors that are scarce, expensive, and booked years in advance, and the company concluded it could not rely on third parties to supply the computing power its road map demanded. Its infrastructure organization grew from a small procurement team into a large operation managing construction contracts, power agreements, and equipment purchases worth tens of billions of dollars.
People familiar with the matter said OpenAI is restarting some data-center plans, particularly projects involving leasing entire facilities. The shift, they said, reflects a reassessment of how fast the company can build versus how fast it needs capacity. Building from the ground up takes years and ties up capital; leasing an existing facility can put chips to work in months. The two approaches have different economics, and OpenAI’s leadership has been weighing both as it races to keep up with demand for its models.
The restructuring of the infrastructure division comes as OpenAI’s footprint expands on several fronts. The company has committed to large campuses of its own, has signed long-term power contracts that lock in electricity for years, and is a partner in the Stargate venture with Oracle and other investors. At the same time it continues to buy cloud capacity from Microsoft, its largest backer, and from others. Managing that mix, and the trade-offs between speed and ownership, has become one of the most difficult jobs in the company.
Executive turnover at OpenAI has been a recurring theme of its rapid growth. The company has seen departures across research, policy, and product teams as it has scaled from a research lab into a global corporation with tens of thousands of employees. The infrastructure group, in particular, has churned, as veterans of construction and energy industries adapt to the unusual rhythm of an AI company, where plans can change with each new model release.
Analysts said the departure is unlikely to slow OpenAI’s build-out on its own, but it adds uncertainty to a part of the business where delays are expensive. Data-center projects are measured in billions of dollars and years; a gap in leadership at the top of the program can ripple through procurement, construction, and power contracts. OpenAI’s ability to keep its commitments, from new model launches to enterprise service levels, depends on the infrastructure group performing smoothly.
The company has moved quickly to fill the gap. According to people familiar with the matter, responsibilities for data-center projects have been divided among existing executives, with construction overseen by one team and power and equipment procurement by another. OpenAI has also stepped up hiring of executives with backgrounds in large-scale industrial construction, a pool that has grown as traditional data-center operators compete for the same talent.
The leasing question is the most immediate test. Renting entire facilities, rather than buying chips in bulk from cloud providers, is a newer approach for OpenAI, and it requires different contracts and different relationships with landlords and power suppliers. People familiar with the company’s plans said several lease negotiations are underway, focused on facilities that can be retrofitted quickly and that sit near reliable power.
The stakes extend beyond OpenAI. The company’s spending has become a major force in the data-center market, and its decisions influence prices for power, land, and chips across the industry. A slowdown in its construction program would ease pressure on constrained markets; an acceleration would tighten them further. Malone’s exit, and the reorganization that preceded it, will be read by suppliers as a signal of which way the company is leaning.
The episode also illustrates how the company’s structure has changed. OpenAI began as a nonprofit research lab with a handful of employees; it now operates a commercial arm, a sprawling product portfolio, and an industrial-scale supply chain. The infrastructure group, once a cost center, is now where the company’s biggest bets are made and where its largest risks are managed.
For now, OpenAI says its infrastructure plans stand. The company continues to describe its computing expansion as among its top priorities, and it has given no indication that the leadership change will alter the pace of construction. Whether the reorganization that cost Malone his role delivers the discipline the company is looking for will be measured in megawatts, and in quarters, over the next two years.
