How Companies Are Using NVIDIA DSX to Break the Power Bottleneck

NVIDIA says power is now the single biggest bottleneck in AI infrastructure, pointing to more than $300 billion in backlogged equipment and over 200 gigawatts of projects waiting in U.S. interconnection queues. Its answer is DSX, a platform meant to plan, simulate and run the grid connection, the building and the racks as one system.

GE Vernova, Hitachi Energy, Schneider Electric with AVEVA and ETAP, and Siemens Energy all announced work aligned with it. When the companies that build transformers and switchgear start designing to a chip company's reference architecture, the center of gravity in energy planning is shifting.

We like the core idea. Treating an AI facility as a load that can flex with grid conditions, rather than a fixed block of demand that must be served at all times, is the kind of engineering that could get more capacity online without waiting years for new transmission. Simulating generators, cooling and switchgear together before breaking ground is just good practice, and it is overdue.

The numbers deserve a careful eye, though. Phaidra reports about 10% more compute inside the same power envelope, which is meaningful and specific. Emerald AI's figure of up to 100 gigawatts unlocked on the existing U.S. grid is described as modeled, and we're treating it as a ceiling from a model, not capacity anyone has delivered.

What I'm watching is who this flexibility really serves. "Tokens per watt" is the metric in this announcement, and that's the data center's metric. Utilities and ratepayers care about reliability and who pays for upgrades. If flexible AI factories genuinely shed load during peaks, everyone benefits. If flexibility becomes a talking point that helps projects jump interconnection queues while households absorb grid costs, the politics will turn fast.

There's also a concentration point. A "DSX Ready" label on energy hardware, like Hitachi Energy's battery storage qualifying for it, puts NVIDIA in the role of setting design requirements for power gear. That can speed standardization. It also means one chip vendor's roadmap, here the Vera Rubin generation, shapes how substations and storage get specified. Useful, probably inevitable, and worth keeping honest.

Interconnection queues are the waiting lists for new power projects and large loads to connect to the grid. AI data center demand has collided with those queues, making power access, not chips, the gating factor for many builds. DSX components named in the announcement include DSX Sim for simulation, DSX Flex for syncing demand with grid conditions, and DSX MaxLPS and DSX Exchange for efficiency and data sharing.