Nvidia and MHI Take Aim at AI's Heat Problem

Nvidia and Mitsubishi Heavy Industries explore cooling and power cooperation for gigawatt-scale AI data centers.

28 Jul 2026

Nvidia and Mitsubishi Heavy Industries are circling one of AI infrastructure's fastest-growing headaches: heat. The two companies are exploring cooperation on cooling and power systems built for next-generation GPU platforms. It marks a shift in how hyperscale data centers get engineered across North America.

MHI is rolling out 10MW-class centrifugal chillers designed for gigawatt-scale AI deployments, a scale that would have sounded absurd a few years back. At Data Center World 2026 in April, MHI executive Shinichiro Gomi argued in a keynote that AI data centers must move to industrial-scale infrastructure to stay viable. As GPU density keeps climbing, that argument lands harder every quarter.

Nvidia's Rubin platform makes the case concrete. Its architecture assumes industrial cooling will already be in place, so operators cannot retrofit older facilities and expect the numbers to hold up. That dependency turns partnerships like this one into strategic necessities rather than side deals, and it pulls equipment suppliers into the AI roadmap earlier than they've ever been.

Both companies showed up prominently at Nvidia's GTC 2026, a signal that the collaboration carries real weight beyond a handshake agreement. Operators planning large AI builds should start treating cooling capacity as a budget line item alongside chip generation, not an afterthought bolted on later. Mechanical engineering and compute architecture are converging into a single design conversation.

The financial stakes are real. Gigawatt-scale facilities represent a new tier of capital investment, and suppliers who can deliver at that threshold gain outsized leverage in future deals. Will every data center operator need an industrial partner like MHI eventually? Given where GPU density is headed, that looks less like a question and more like a certainty.

With Nvidia setting the platform standard and MHI supplying the thermal backbone, the pairing positions both firms well as AI compute demand keeps climbing.