How To Power a Data Center (Text Version)

This is the text version of the video How To Power a Data Center.

Large, unpredictable, and rapidly changing loads—such as data centers, artificial intelligence, and crypto centers—present new challenges for the electric grid. The National Laboratory of the Rockies (NLR) used a combination of energy resources to smoothly power such difficult demand. This video describes the demonstration, which was made possible by the platform for Advanced Research on Integrated Energy Systems (ARIES).

[Music starts, narrator speaks]

In less than a second, a data center's power demand can drop or grow fivefold, sometimes unpredictably.

It's like flicking a giant light switch up and down on 10,000 homes.

How can the electric grid or a company that owns a data center power such a large, rapidly changing load?

Answering that question requires a real power system that can simulate these exact scenarios at high fidelity.

And to do it, the U.S. Department of Energy has the ARIES platform.

[Text on screen: ARIES by National Laboratory of the Rockies, Advanced Research on Integrated Energy Systems]

In a project sponsored by the Office of Electricity, power system experts showed that using a mix of energy resources can tackle the challenge of powering data centers.

[Text on Screen: U.S. Department of Energy Office of Electricity]

See how they did it.

First, they modeled a power grid.

They built a scaled-down data center and its power sources—identical to what exists in cities across the country.

Next, they used a generator that can replicate gas, nuclear, or geothermal power along with real batteries and fuel cells for fast responses to large load swings—and, if needed, backup power.

Then, to simulate dynamic power demand, they leveraged energy data from a leading AI data services company.

Finally, they initialized the grid scenarios and pressed "go."

The first thing they noticed was wear on the generator.

The data center's cyclic demand put torsional strain on the shaft, which is not built for rapid start-and-stop power.

If this were to continue unassisted, the generator would soon wear down.

But when it is combined with other technologies, like batteries and fuel cells, these more responsive resources supplied the pulsating loads so the generator could spin smoothly, providing steady baseload power.

The researchers then went a step further and simulated a generator failure.

When this happened, the battery and fuel cell temporarily provided reserve power to keep the system running.

The combination of baseload generation integrated with other technologies could be the key to low-cost, high-reliability power for data centers, AI, and more.

And that's just one of the many experiments made possible with ARIES.

This is your platform: Find out how your company can advance power solutions with ARIES.

[Text on screen: ARIES by National Laboratory of the Rockies. National Laboratory of the Rockies. Visit our website to learn more.]

[Narration ends, music stops]