Nvidia hires former OpenAI data center head Chris Malone as VP, DSX Platform

Chris Malone, OpenAI's former head of data centers, has joined Nvidia as vice president, DSX Platform, he said in a LinkedIn post. Data Center Dynamics reported the move on October 22, two months after his departure from OpenAI was reported.

Malone spent 18 months at OpenAI, arriving from Meta, where he spent nearly five years as a distinguished engineer. Before that he spent nearly 12 years at Google focused on data center infrastructure and IT hardware R&D, according to the report.

Nvidia first showed the DSX AI Factory Platform at GTC San Jose in March 2026. It pitched the platform as a guide to building data centers around its CPU-GPU racks and Spectrum-X Ethernet networking and storage. This week the company claimed the platform will let customers build AI data centers that cut water consumed by cooling by up to 100 percent while using up to 30 percent less power for the same output.

Nvidia also launched DSX POD at GTC Berlin this week, a modular version of the reference design for Vera Rubin-powered AI data centers. The units are built in containers that are integrated and validated before delivery. Nvidia says they can be connected to a prepared site's power, cooling, networking and other supporting infrastructure to reduce deployment time from six months to 14 days.

Malone described the job in operational terms. “AI is fundamentally changing how data centers are designed, built, and operated,” he wrote, citing scale, power density, cooling requirements and speed of deployment. “By bringing computing, power, cooling, and operations together in a common platform, we can help partners build AI factories faster, use energy more efficiently, and get more useful computing from every megawatt.”

A reference design fixes the power and cooling assumptions a developer hands its engineer, and Nvidia's own racks are what the DSX guide is written around. Malone's pitch is megawatt efficiency, but the 14-day figure covers fitting out a site that is already prepared, not creating the connection that prepares it.

The container does not shorten the interconnection queue, a point this publication made when Schneider standardized its factory-built power modules in September.

The numbers Nvidia is quoting are ceilings — up to 100 percent less water, up to 30 percent less power — and how much of either a given project captures depends on its cooling design and its power contract, neither of which ships in the box. Malone's own framing, more useful computing from every megawatt, is a claim about the building rather than the grid that feeds it.

At OpenAI, Malone worked alongside head of infrastructure Keith Heyde, a fellow Meta alum, according to Data Center Dynamics. Peter Hoeschele, who ran the Stargate joint venture, left around the same time and brought several senior AI infrastructure leaders with him. The team was then consolidated under former Intel executive Sachin Katti, who left earlier this month.

Former xAI lead Uday Ruddarraju now runs OpenAI's infrastructure and compute business and has brought xAI staffers with him, including Brent Mayo, named head of data center build and delivery.