Report 041 · Energy Storage
A data center needs power in months. The grid can take five years to give it any. So a new kind of buyer is showing up for grid-scale batteries, and their reason has almost nothing to do with storing energy. They are buying a way past the line. It works, and it is worth understanding exactly what that battery does and, just as important, what it cannot do.
Most of what I write about batteries here is about the cells themselves: why LFP won the grid, why a zinc-bromide cell can't burn, what actually wears a pack out. Those are chemistry stories. This one isn't. This is about a reason to buy a battery that a chemist would never think of, and it is quietly becoming one of the biggest drivers of grid storage in the country.
The reason is a traffic jam. To connect a large new load to the transmission grid, you get in a queue, and the queue is enormous. According to Lawrence Berkeley National Laboratory's "Queued Up" analysis, more than 2,000 gigawatts of generation and storage were sitting in U.S. interconnection queues, and storage is now the single largest category in that pile. The wait is the problem: the report finds the timeline to reach commercial operations exceeds five years in the regions with data to measure it, and the median wait just to get an interconnection agreement signed runs well over three years.
Now put a data center next to that. An AI data center wants to be drawing power in a year, maybe two, because the economics of the build assume it. Five years in a queue is not a delay to that plan, it is the death of it. So the developers went looking for a way around the line, and one of the answers turned out to be a battery.
Here is the mechanism, because it is not obvious. The battery is not there to power the data center. A grid-scale battery of the size being deployed holds a couple of hours of energy at most, and a hyperscale facility would drink that in a blink. The battery is there to make the load look friendlier to the grid.
A data center that can lean on an on-site battery during the grid's worst moments, and that can throttle its own draw, is a "flexible" load rather than a rigid one. A flexible load asks less of the wires. It can promise not to pull its full demand at the exact hours the system is strained, because the battery covers the gap. That promise is what lets a utility connect it years sooner, without waiting for new transmission to be built, because the connection no longer has to be sized for the worst case every second.
The clearest early example is a deal from Aligned Data Centers, which paired a data center in the Pacific Northwest with a 31-megawatt, 62-megawatt-hour battery, planned to be operational in 2026. The reporting on it is blunt about the point: the battery was "sized to allow it to get interconnected years earlier than would be possible with traditional utility upgrades." As Matt Barnes of Calibrant Energy described the shift, demand has "grown exponentially" as data centers recognize "the critical role that energy flexibility, and specifically battery storage, can play" to "accelerate interconnection." The battery is bought for time, not for kilowatt-hours.
This is not a fringe workaround anymore. It is now the thing the federal government is actively trying to formalize. On June 18, 2026, the Federal Energy Regulatory Commission issued Section 206 "show cause" orders to all six of the regional grid operators it oversees, giving them 60 days to justify or reform how they connect data centers and other large loads. The orders followed an October 23, 2025 advance notice from the Department of Energy asking FERC to enable faster, non-discriminatory interconnection of large loads.
Read the list of what FERC told the operators to address and the battery is right there in it. Among the five issues are co-location and behind-the-meter generation, and new transmission service specifically for "flexible" large loads. The point is to let a load that can co-locate its own resources or agree to limit its grid usage avoid triggering transmission upgrades it would otherwise have to wait for. Commissioner Judy Chang said plainly that this contemplates running the system "tighter than we have done," with "more loads on the system served by co-located or behind-the-meter generation, and potentially more use of batteries, load control systems, and backup resources." When a FERC commissioner names batteries as part of the plan to unclog the queue, the workaround has become policy.
This is the part I want to be careful about, because it is the part the excitement tends to skip, and this publication is where I get to be exact about my own field.
A two-hour battery is a bridge, not a generator. It does not create energy, it moves energy in time, and it can only cover a short gap before it is empty. It buys a data center flexibility and a faster connection, but the actual electricity to run the servers, day in and day out, still has to come from somewhere: the grid, a gas plant, a solar farm, something. The battery reshapes the demand curve. It does not feed the load. Anyone who tells you a battery lets a data center skip the grid entirely is confusing a shock absorber with an engine.
And there is a systems cost that comes with using storage this way. A battery bought to smooth a data center's grid interface gets cycled on a hard, business-driven schedule, and how you cycle a pack is a large part of what decides how long it lasts. That is the whole argument of an earlier report here: the duty cycle is part of the battery's design life, not a decision you make after buying it. A pack run aggressively to keep a data center connected is a pack aging faster than its datasheet cycle count implies, and that wear is a real cost hiding behind the fast connection.
This intersection, storage sized and scheduled for what a facility actually needs, is the work I do on the energy side, and I'll be exact about my role: I help design the AI battery-cycling systems for a veteran-owned (HUBZone) energy-storage integrator. I don't own that company and earn nothing from this link; I flag it because it's a field I build in, not just write about. Full policy here.
The story being sold is "batteries are powering the AI boom," and that framing quietly implies the battery is a source of energy. It usually isn't. The battery is buying a place at the front of a five-year line, by making a hungry, rigid load behave like a polite, flexible one. That is a genuinely useful thing and it is why storage demand keeps climbing, but it is a grid-timing tool, not a power plant. If someone pitches you a battery as the answer to a data center's energy problem, ask the two questions that separate the tool from the hype: how many hours will it actually run the load, and where does the energy come from when it's empty. The honest answers are "not many" and "the grid you were trying to skip."
U.S. Army combat veteran (Counter-IED / Electronic Warfare), peer-reviewed researcher in microwave spectroscopy, and founder & CEO of Shroombiosis. Consults on laboratory operations, AI, and supplement formulation.