A senior executive responsible for overseeing the construction of OpenAI's data centers has departed the company. Chris Malone, who held the position, left last week, according to sources familiar with the matter. A company spokesperson confirmed his departure but did not specify a reason or name a single successor.

The spokesperson stated that the infrastructure organization was restructured earlier this year to support the scale and pace of its work. They noted that a deep and experienced data center team is in place with clear leadership and the technical capability to execute the plan. Malone has not responded to requests for comment.

From Stargate to a Narrower Role

Malone assumed the role of data center lead in March 2025, shortly after OpenAI announced the Stargate project with Oracle and SoftBank, which aimed to meet computing expansion needs through self-built facilities. After a rocky start for that plan, OpenAI shifted more toward signing agreements with cloud providers and leasing chips rather than developing its own campuses. The company has since revived some internal data center arrangements, but now does so by leasing entire facilities instead of just renting computing power by the chip from cloud vendors.

According to insiders, this work is now being led by other executives and is no longer under Malone's purview.

Restructured Reporting Lines and Divided Responsibilities

Before joining OpenAI, Malone spent about five years at Meta as a distinguished engineer in charge of data center strategy. Earlier, he spent roughly twelve years at Google in senior technical roles focused on data center technology. Upon joining, he reported to President Greg Brockman until this year's infrastructure reorganization.

Following the restructuring, Vice President Sachin Katti took over a broader organization handling compute and Stargate-related efforts, still reporting to Brockman. Malone and Adrian Caulfield became co-leads of the data center technical engineering and design team. Katti previously worked at Intel.

In March, OpenAI split its compute and infrastructure operations into three segments: designing facilities, negotiating commercial deals with cloud providers and chip companies, and managing existing facilities. In January, Spas Lazarov, a former Apple data center engineering director, joined as head of data center engineering. In July, Uday Ruddarraju was promoted to chief technology officer for compute capacity, reporting to Brockman. Brent Mayo, who joined from xAI earlier this year, reports to Ruddarraju and is responsible for keeping compute projects on schedule. Both previously worked on Musk's Colossus supercomputer project in Memphis.

Other personnel changes in the infrastructure unit include Keith Heyde and Peter Hoeschele, who had managed the Stargate joint venture. The company has not linked any project cancellations or partner changes to Malone's departure.

Executive Changes Amid IPO Prep and Rising Compute Spending

OpenAI is preparing for a potential IPO expected in 2027 and is competing with Anthropic for enterprise clients. Several senior executives have left in recent weeks, including Chief Revenue Officer Denise Dresser, Chief Operating Officer Brad Lightcap, and Fidji Simo, who previously served as a close deputy to CEO Sam Altman. Chief Financial Officer Sarah Friar told employees at an internal meeting that the company plans to go public by 2027, potentially earlier if growth accelerates.

Compute spending projections have also been revised upward. In July, the company raised its expected compute expenditure through 2030 from approximately $600 billion to $750 billion. Earlier this month, OpenAI signed a lease with SoftBank's SB Energy for a 10-gigawatt data center in Ohio, with Nvidia providing partial financial guarantees. Currently, the company is pursuing three parallel paths: leasing entire facilities, cloud agreements, and the self-built Stargate project. Which layer will ultimately be responsible for delivery timelines remains dependent on whether the organizational structure is adjusted again.