A supercapacitor company from Estonia just became Europe’s newest unicorn, and the reason has less to do with capacitors than with the awkward fact that artificial intelligence is starving for electricity. Skeleton Technologies has closed a multi-hundred-million-dollar round to expand its manufacturing and energy storage operations into the United States, a move that lands right as utilities, chipmakers and hyperscalers scramble to figure out where the power for the next generation of data centers will actually come from.
Supercapacitors are not batteries, and that distinction matters. Where lithium-ion cells store energy chemically and release it slowly, supercapacitors hold charge electrostatically and discharge it almost instantly. They charge in seconds, survive hundreds of thousands of cycles, and do not degrade the way batteries do. For decades they lived in niche corners of the market: regenerative braking on buses, backup power for wind turbine pitch systems, voltage stabilization on rail networks. Solid, unglamorous engineering. Then AI arrived and made the electric grid’s weak spots impossible to ignore.
The problem is not simply that AI data centers consume enormous amounts of power. It is that they consume it unevenly. A training run can push a facility toward its ceiling for hours, then drop sharply. Inference traffic spikes and dips minute to minute. Traditional grid infrastructure was never designed for that kind of whiplash, and grid operators have grown increasingly vocal about the strain. Supercapacitors offer something batteries cannot: near-instantaneous response to sudden demand swings, smoothing the load before it reaches the broader grid.
That is the thesis Skeleton is selling, and investors appear to be buying it. The company has spent years building out its curved graphene electrode technology, which it claims delivers higher energy density than conventional activated carbon designs. Whether that edge holds up at industrial scale is the open question. Manufacturing supercapacitors at the volume AI infrastructure demands is a different animal from producing them for a few thousand buses, and the company’s US expansion will test whether it can execute on that transition.
The timing is not accidental. Washington has been pushing hard to onshore energy storage and semiconductor supply chains, and supercapacitor production sits at the intersection of both. Skeleton already operates a plant in Germany and has been weighing US sites for years. The funding round gives it the capital to stop weighing and start building, though the company has not disclosed final site selections or production timelines in detail.
There is a broader pattern worth noticing here. For the past two years, the AI investment story has been dominated by chips, models and data center construction. Power has been treated as an afterthought, a line item that utilities would somehow sort out. That assumption is cracking. Constellation Energy’s deal to restart Three Mile Island, the rush toward small modular reactors, and now a supercapacitor unicorn all point to the same realization: compute is cheap relative to the electricity needed to run it.
Whether supercapacitors become a meaningful part of the answer or remain a useful but limited tool is genuinely unclear. They store far less total energy than batteries, which limits them to short-duration applications measured in seconds or minutes. Longer outages still require lithium-ion, flow batteries or gas peakers. Skeleton’s pitch is not replacement but complement, a fast-twitch muscle alongside the slow-twitch endurance of conventional storage.
That is a reasonable argument, but it also means the company’s fortunes are tied to a market that is still forming. Utilities are experimenting with hybrid storage configurations, and no standard architecture has emerged. If grid operators settle on battery-only designs, Skeleton’s addressable market shrinks. If they conclude that fast-response smoothing is essential, it grows considerably.
The US expansion also puts Skeleton in direct conversation with a policy environment that has grown more favorable to domestic energy manufacturing but less predictable overall. Tariffs, permitting timelines and federal incentives all shape whether a new plant pencils out. The company will need more than a good technology story to navigate that.
What the raise really signals is a shift in where AI money is flowing. The first wave went to the companies building models. The second went to the companies building chips. The third, apparently, is going to the companies keeping the lights on. Skeleton Technologies just raised a lot of money betting that the least glamorous layer of the stack is about to become one of the most contested.