- OpenAI data centers have lost former leader Chris Malone after a short tenure during an expensive and closely watched expansion.
- A reorganization of OpenAI data centers moved infrastructure oversight under vice president Sachin Katti before Malone’s departure.
- The exit joins a growing list of senior OpenAI leaders leaving while the company faces scrutiny over spending and an eventual IPO.
- Stargate’s huge US buildout makes steady infrastructure leadership central to OpenAI’s ability to deliver future models at scale.
Table of Contents
OpenAI data centers lose a pivotal operator
OpenAI data centers are dealing with a leadership exit at the worst possible moment for ambiguity. Chris Malone, who led the company’s data center work, left last week, according to reporting by The Wall Street Journal. One executive departure is usually unremarkable. Malone’s job was different: he was responsible for turning enormous compute promises into physical buildings, power contracts, cooling systems, and working GPU clusters. In AI, that makes for one of the most consequential jobs in the company.
Malone arrived at OpenAI in March 2025 after spending nearly five years at Meta and more than a decade at Google. That background matters. Google and Meta have spent years learning the unglamorous lessons of hyperscale computing: substations take time, water use sparks local fights, hardware delivery slips, and a data center plan is only as good as its access to electricity. AI labs are now discovering those same realities at warp speed.
OpenAI said it had recently reorganized its infrastructure group to match the scale and pace of its work. The company added that it has a deeply experienced data center team with clear leadership and the technical ability to carry out its plans. That may well be true. Still, corporate statements about organizational clarity tend to arrive when clarity has become a question.
The Journal reported that Malone had stopped reporting directly to OpenAI president Greg Brockman and instead reported to Sachin Katti, an OpenAI vice president who took over leadership of the infrastructure organization. Uday Ruddarraju leads the data center team, Brent Mayo runs data center build and delivery, and Spas Lazarov oversees data center engineering. That is a capable-looking roster, but it also signals that responsibility for OpenAI data centers has been distributed across several leaders rather than concentrated in one highly visible owner.
The Stargate buildout is where strategy meets concrete
OpenAI data centers are not an internal IT project. They are central to Stargate, the reported US infrastructure initiative involving OpenAI, Oracle, Nvidia, SoftBank, and Microsoft. The project has been described as a $500 million data center initiative, a figure that captures how the AI business has changed. Training a frontier model is no longer simply a matter of hiring elite researchers and ordering chips. It is increasingly an industrial project, closer to building a utility network than launching a consumer app.
That shift puts companies such as OpenAI in a strange position. Its public identity is still tied to ChatGPT and model releases, but its competitive future may hinge on whether it can secure enough megawatts and GPUs before rivals do. Microsoft has its Azure fleet. Google owns the full stack from custom TPUs to global cloud regions. Amazon has AWS, Anthropic as a major customer and partner, and its own chips. OpenAI has major partners, yet partnerships can be both a strength and a dependency.
My read is that Malone’s departure deserves attention because data center execution has become a bottleneck, not a back-office function. A delayed model launch is embarrassing. A delayed power connection can strand billions of dollars’ worth of Nvidia hardware. And if OpenAI data centers cannot secure sufficient compute, its researchers may have ambitious plans but no kitchen in which to cook.
The politics are hard to ignore. Large US data center developments now invite questions about power prices, water consumption, grid reliability, and tax incentives. The Trump administration’s support for Stargate gives the effort political heft, but it will not make transmission lines appear overnight. Local permitting boards, utilities, construction crews, and turbine manufacturers all operate on timelines that do not care about Silicon Valley’s quarterly expectations.
Executive churn is becoming part of the OpenAI story
The OpenAI data centers change lands amid a broader run of senior departures. Business Insider recently counted 13 executive exits during 2026, and several came in a short span. Denise Dresser was replaced as chief revenue officer after roughly eight months. Brad Lightcap, OpenAI’s longtime chief operating officer, said he was starting something new. Fidji Simo, who served as product and business chief and reported to Sam Altman, stepped away to address chronic illness while retaining an advisory role.
Elsewhere, Kate Rouch departed as chief marketing officer, also reportedly for health reasons. Chloé Bakalar, OpenAI’s head of ethics, left in July. The company reportedly disbanded its preparedness team, which had focused on assessing catastrophic risks from advanced models. Bill Peebles, who led the now-defunct Sora image-generation effort, also exited after that project was shut down.
People leave fast-growing companies. OpenAI is expanding through a brutal market for AI talent, paying for infrastructure on a historic scale, reshaping its organization, and operating under relentless public attention. Brockman has argued that OpenAI’s spotlight means every departure receives scrutiny that similar departures elsewhere would not. He has a point. Nobody writes a headline when a mid-level cloud executive leaves a quieter company.
But the counterargument is stronger when the exits touch revenue, operations, marketing, ethics, product leadership, safety, and now OpenAI data centers. Those are not random corners of the org chart. They are the functions that determine whether a company can build, sell, govern, and finance its technology. A recurring pattern does not prove a crisis, but it does make the company’s internal stability a legitimate business question.
An IPO will make the questions harder to avoid
OpenAI data centers will probably be judged less by executive titles than by results: how much usable capacity comes online, what it costs, how reliably it supports new models, and whether the company can convert that spending into durable revenue. That arithmetic is especially uncomfortable as OpenAI reportedly prepares for a public offering that may now arrive in 2027 rather than this year.
Public markets are far less patient with grand narratives once they can inspect the numbers every quarter. Investors will want to know how much OpenAI pays for compute, how concentrated its infrastructure relationships are, what margins it can sustain, and whether customers keep paying enough to justify the capex arms race. Nvidia’s earnings have made AI infrastructure look like a gold rush. Gold rushes, as history keeps reminding us, also produce plenty of expensive empty towns.
OpenAI can weather a leadership reshuffle if its new structure makes decisions faster and construction more predictable. But the company has reached the point where talent churn cannot be waved away as a side effect of fame. The real test is whether OpenAI can turn a sprawling cast of infrastructure partners and internal leaders into actual capacity before its rivals pull further ahead.
