Back
CEOs of artificial-intelligence companies want to spend hundreds of billions of dollars building their energy-gobbling data centers, but that can’t happen without the necessary electricity supply. And they want to move way faster than electric utilities are used to.
One idea gaining traction is to allow data centers to come online more quickly if they agree to occasionally pull less power from the grid when demand is high, a concept endorsed by none other than Energy Secretary Chris Wright in a rulemaking proposal filed Thursday. The massive computing facilities could accomplish such flexibility with the help of on-site renewables and batteries, but precious few projects using this model have materialized. That’s about to change.
On Wednesday, Aligned Data Centers announced it would pay for a new 31-megawatt/62-megawatt-hour battery alongside a forthcoming data center in the Pacific Northwest. The battery, developed by energy-storage specialist Calibrant Energy in partnership with the local utility, is now entering the construction phase and should be operating sometime next year. The kicker is, this deal will let Aligned get up and running “years earlier than would be possible with traditional utility upgrades,” per the companies.
If the plan works, would-be AI leaders will be jumping all over this battery-first strategy. In fact, many already are, they just haven’t publicly acknowledged it yet.
“There’s so much chatter right now about the potential to use energy storage in this manner to facilitate the connection that large power users want from the grid. But there hadn’t really been evidence of that theory being reality,” said Phil Martin, CEO at Calibrant, which is owned by Macquarie Asset Management. “It is possible, and it is being done — not as a proof of concept in a lab somewhere, but really a commercial project.”
Batteries aren’t, at first glance, a tool well matched to the needs of AI computing.
Lithium-ion chemistries have become quite competitive for short-form activities: First, it was managing second-by-second frequency fluctuations on the grid; now, in places like Australia, California, and Texas, batteries are shifting solar generation to compete with gas plants in the evening when demand rises.
Data centers, though, use energy around the clock — not literally at full blast 24/7, but a lot closer to it than current batteries can keep up with. Data-center developers have chased new gas, hydropower, and an exotic array of nuclear power plants in hopes of feeding the beast. But those options will take several years to come online, if they ever get built. The headlong rush into AI demands nearer-term solutions.
As a lot of exceedingly well-funded firms contemplated this conundrum, some thinkers started focusing on grid flexibility as a way to accelerate the computing-infrastructure buildout. Earlier this year, Duke University researcher Tyler Norris made waves in the AI-energy world with research that found today’s grid could handle quite a lot more data centers if the facilities could simply dial back their consumption for a couple hours at a time during moments of maximum demand.
The Aligned battery offers a concrete example of that kind of research. The utility studied just how big the battery would need to be to compensate for challenges imposed on the local grid by the data center. Aligned and Calibrant had their own calculations, Martin said, “but the validation of that, and the actual specification of that, came out of the interconnection study done on the utility side.”
Due to the local nature of the power constraint, the battery had to be built close to Aligned’s facility; the company ultimately provided the land to host the grid storage installation. In other cases, where a proposed data center runs up against a system-wide capacity constraint, a battery solution could be further away.
Another glimpse of the battery-enabled future came this summer when Redwood Materials, a richly funded battery-recycling startup, unveiled a new business line that repackages old EV batteries to serve data-center demand. The first installation, at Redwood’s campus near Reno, Nevada, fully powered a very small, modular data center using a solar array and a field of former EV battery packs laid out on the desert floor.
Redwood just got its own vote of confidence in that concept: On Thursday, it raised another $350 million from investors including AI-chip leader Nvidia.
Aligned’s commitment to paying for the battery itself could serve as a model of socially responsible AI-infrastructure development.
Some utilities around the country are jumping to build new power plants to support the projected data-center buildout, and charging their regular customers for the investment, hoping the AI titans eventually become paying customers. But this approach risks saddling consumers with unnecessary costs if the AI hubs don’t materialize.
Because Aligned is footing the bill, the utility’s other customers won’t be forced to pay for the data-center firm’s growth ambitions. But, though this one large customer will provide the land and funding, the battery will sit on the utility side of the meter. That means the utility can leverage the tech for other grid uses, like frequency management and capacity, when it’s not maintaining the flow of power to the data center during otherwise scarce hours.
In this case, Martin said, the permitting and buildout could move faster with the battery connecting to the utility grid instead of directly to the data center. In other situations, bigger batteries on the customer side of the meter might make more sense. Calibrant is already working on more and even larger batteries for the AI sector, he added.
“Whereas right now, we think this is unique, I think over a relatively short time horizon it’s going to be much more common,” Martin said. “It’ll start to look surprising if we don’t see projects like this at the largest loads as they connect [to the grid].”
A clarification was made on Oct. 25, 2025: This story originally stated that the local utility studied how many times per year the local grid could run out of electricity if the data center got built. The piece has been updated to clarify that the utility studied how big the battery would need to be to compensate for challenges imposed on the local grid by the data center.
.svg)
