OpenAI's Data-Center Chief Was Just the First to Leave

The Departure Nobody Expected to Notice

Chris Malone joined OpenAI in March 2025 as head of data centers — a role that sits at the very center of the company’s strategy. He left last week. No explanation was offered beyond a corporate line about reorganization.

That silence speaks volumes. In a company that has lost more than a dozen executives since the start of the year, Malone’s exit is the most consequential because his domain is the one thing that separates OpenAI from every rival.

Anthropic is profitable. Google has deep pockets and a trillion-dollar cloud business already optimized for AI. Mistral is lean. OpenAI’s advantage, if you can call it that, is its willingness to spend recklessly on compute infrastructure. That advantage is now being quietly dismantled from the inside.

Who Malone Was Replacing

Malone reported to Greg Brockman, OpenAI’s co-founder and president. Brockman was relieved of most management duties in 2019 when Sam Altman became CEO. He returned after the Blip crisis of 2023, sat out 2024 on sabbatical, and now controls the infrastructure and product teams that make OpenAI’s models actually work at scale.

According to Thibault Sottiaux, who leads the API and app offerings, “everyone reports to Greg at the end of the day.” That phrasing is telling. It means Brockman is no longer coordinating between teams — he is the team.

The reorganization that followed moved Malone’s data-center function under Sachin Katti, vice president of infrastructure. Malone, once a direct report to the president, now answers to a vice president. That is not a promotion ladder. That is a demotion disguised as restructuring.

OpenAI’s official explanation was rehearsed and thin: a realignment of the infrastructure team. The implication is plain — whoever stays after the shuffle is the person Brockman trusts to build the compute stack the company claims to be prioritizing.

The Hardware-Software Schism

OpenAI’s internal tension is not new. It is the structural contradiction at the heart of the company. On one side: the research team building ever-larger language models, pushing GPT-5.6 to market, growing its desktop app subscriber base by 15 million in two months. On the other: the infrastructure team burning billions to train those models, building the data centers and chip supply chains that make the models possible.

Malone was the hardware side made flesh. His departure — coming just nine months after he joined — suggests the company is deprioritizing the very capability that differentiates it. This is not a minor personnel change. It is a strategic signal.

Brockman, before he returned to OpenAI, built Stripe’s business. His instinct is commercial scaling, not frontier compute investment. The infrastructure teams now reporting to him may reflect a company that is pivoting from building the future to selling the present.

The IPO Shadow

OpenAI filed confidentially for an IPO in June. The filing does not need to go public until 2027. Most companies that file confidentially hit the trading floor within five months. SpaceX took less than two. By that standard, OpenAI’s timeline is unusually stretched — and the stretch itself is revealing.

The company is reportedly seeing losses grow alongside revenue. Anthropic, its closest rival, is already profitable. The pressure to reach the public markets without terrifying investors with a bleeding balance sheet is enormous. Every executive departure that trims headcount is a move toward that goal.

But there is a second, subtler pressure. An IPO requires disclosure. OpenAI’s financial statements will be read side by side with Anthropic’s — and the comparison will not be flattering. The company’s leadership knows this. The departures are, in part, damage control for an audience that has not yet been invited into the room.

Altman’s Bad Year

Sam Altman has publicly acknowledged a “bad past 12 months” at OpenAI. The reorganizations, the departures, the quiet removal of expensive side projects — all of it fits a familiar pattern. Brilliant founders create the product. Experienced operators scale it and prep it for public markets. Fidji Simo and Kevin Weil were brought in as veterans of that second act. Now they are gone too.

The cycle repeats. The question is who writes the next chapter.

Brockman’s return to full operational control changes the answer. He was present at OpenAI’s founding. He understands the technology in a way Altman, the product visionary, does not. And he has proven willing to disrupt — seeding internal rivalries during the Blip crisis, taking a sabbatical, and returning to consolidate power.

What Comes Next

OpenAI’s user base is growing. GPT-5.6 is competitive. The desktop app has added millions of subscribers. The product is working. But product excellence alone does not prepare a company for public markets. Revenue discipline does. Cost control does. And the executive departures suggest OpenAI is choosing the latter over the former.

The company that emerges from this reshuffling will be leaner, more commercial, and less willing to bet billions on compute infrastructure that may not pay off for years. That is not necessarily a bad thing for an IPO roadshow. It is a risky thing for a company that claimed its advantage was the size of its ambition.

Malone’s departure is a symptom, not the disease. The disease is a company that cannot reconcile its identity as a research lab with its obligation to become a public business. Until that tension is resolved — and it will be resolved by whoever controls the organization chart — the departures will continue. The question is whether OpenAI will still have a compute advantage left by the time it files its prospectus.