NextFin News - Estonia's Skeleton Technologies has crossed the $1 billion valuation mark, becoming the Baltic nation's newest tech unicorn after Siemens AG invested more than $100 million in October to fund U.S. manufacturing of energy-storage systems for AI data centers. The deal crowns a 17-year pivot for the Tallinn-based supercapacitor maker — from space and automotive components to the front line of the AI power buildout — and puts a capital-intensive hardware manufacturer, rather than a software company, at the center of Europe's latest venture milestone.
The investment, confirmed by the company and Chief Executive Officer Taavi Madiberk, values Skeleton above the unicorn threshold without disclosing the precise figure. It is the second major capital event for the company this year: in May it closed a €33 million first tranche of a pre-IPO round that lifted total venture funding to €392 million, and it has targeted a U.S. public listing in 2027. Siemens Financial Services, the industrial group's financing arm, now joins existing backers including Marubeni, FirstFloor Capital, EIT InnoEnergy, Axon Partners Group, SmartCap and Taiwan's Taiwania Capital.
Why Siemens Is Buying Power, Not Just a Stake
The transaction is best read as vertical integration by a company already riding the data-center equipment supercycle. Siemens Energy, the power-technology spinoff, reported record third-quarter sales of €11.45 billion in August, up 18.5%, with profit before special items more than tripling to €1.62 billion. Gas-turbine orders from U.S. AI data centers and Middle East power projects accounted for about half of the quarter's turbine orders, and the company's shares have risen nearly sevenfold over the past two years alongside U.S. rival GE Vernova. In February, Siemens Energy announced a $1 billion expansion of U.S. manufacturing for gas turbines and grid equipment that will create more than 1,500 jobs.
Siemens Financial Services' stake extends that logic downstream, from turbines and transmission hardware into the millisecond-scale power conditioning that AI racks actually require. A data center can secure a grid connection and install turbines and still lose uptime to voltage sags that last thousandths of a second — the kind of transient that traditional battery-based uninterruptible power supplies, with response times measured in tens of milliseconds, are comparatively slow to arrest. Supercapacitors respond in milliseconds and can endure hundreds of thousands of charge cycles with minimal degradation, making them a complement to, rather than a replacement for, lithium-ion backup.
"As AI infrastructure continues to scale, reliable, high-performance power solutions have become critical," Madiberk said. "This funding round brings additional strategic partners to the table and strengthens our ability to deliver the next generation of power solutions that cut AI data center energy consumption by 40 percent, increase computing power by 40 percent, and enable faster power grid connection."
Those performance claims are company figures and have not been independently verified. The company's own Leipzig factory opening materials cite a 44 percent reduction in total electricity consumption and 40 percent more computing power from the same GPU investment. But the direction of the problem is not in dispute. The International Energy Agency projects global data-center electricity demand will more than double to roughly 945 terawatt-hours by 2030, driven almost entirely by AI, while Goldman Sachs Research forecasts a 165 percent increase in data-center power demand over the same horizon. Power, not chips, is increasingly the binding constraint on AI expansion.
The Technology: What a Supercapacitor Actually Does
Skeleton's product sits in a narrow band of the energy-storage spectrum. Batteries store large amounts of energy chemically and release it over minutes to hours. Electronic capacitors discharge almost instantly but hold little energy. Supercapacitors occupy the middle: they store charge electrostatically at the electrode surface, which removes the kinetic barriers that slow ion insertion in batteries, delivering bursts of power far faster than chemical storage at the cost of lower energy density.
The company's differentiation rests on its patented Curved Graphene material, protected through more than 70 patent families, and on manufacturing scale. Its €220 million SuperFactory near Leipzig, opened in November 2025, is designed for annual output of up to 12 million cells and is already supplying Siemens, General Electric and Hitachi Energy for European grid applications as well as major U.S. hyperscalers for AI infrastructure. The plant uses a fully European value chain and no lithium, cobalt or manganese. Skeleton opened a U.S. plant in Houston earlier in 2026, having acquired a battery factory in Finland in 2024 to scale SuperBattery manufacturing. Its first customer was the European Space Agency, and its cells are used in BMW's i7 and M-series vehicles — a pedigree in mission-critical and automotive applications that predates the AI data-center narrative.
The AI use case is specific: smoothing the sharp load fluctuations created when GPU clusters ramp up and down, reducing the need to oversize UPS systems, extending battery life by limiting shallow cycling, and helping facilities connect to the grid faster. In June 2026 the company launched GrapheneUPS, a high-density uninterruptible power supply for AI data centers that it says delivers 50 percent lower volume than competing systems for the same performance. The economic pitch to operators is capital efficiency — lower cost per kilowatt for short power bursts — rather than long-duration backup, where lithium-ion remains the clear choice.
The addressable market is expanding fast enough to support several winners. The global supercapacitor market was valued at $8.95 billion in 2025 and is projected to reach $39.01 billion by 2032, a 23.4 percent compound annual growth rate, according to one research estimate; a second forecast puts the market at $6.8 billion in 2025 growing to $18.4 billion by 2034 at an 11.7 percent annual clip. The gap between the two forecasts is itself informative: nobody disputes double-digit growth, but the size of the prize depends on how quickly data-center and grid applications convert from pilots to specifications.
The Baltic Unicorn Club, and Why This One Is Different
Estonia, a country of about 1.3 million people, has produced roughly 10 unicorns, including ride-hailing group Bolt, payments company Wise (which relocated to London), identity-verification firm Veriff, and software businesses Pipedrive, Katana MRP and Tuum. The list is overwhelmingly software and consumer-internet. Skeleton is different: it is a hardware manufacturer with factories, inventory, working capital needs and industrial customers.
That distinction matters for the 2027 IPO plan. Public-market investors price software margins and recurring revenue very differently from factory economics. Skeleton's path to a U.S. listing will be judged on order books, gross margins, and the ability to convert Siemens' strategic endorsement into contracted revenue — not on user growth. The May pre-IPO round, at €33 million for a first close, suggested a larger round still being assembled; the October Siemens injection both fills that gap and de-risks the story by attaching a blue-chip anchor investor whose parent company is already selling into the same customers.
The Counter-Case: A Cyclical Capex Wave, Not a Structural Moat
The strongest argument against treating Skeleton as a durable winner is that the AI power buildout is a cyclical capital-expenditure wave, not a structural shift in Skeleton's favor. Every large industrial and electrical supplier is chasing the same demand: Siemens Energy and GE Vernova in turbines and grid equipment, Hitachi Energy with its own $1 billion U.S. manufacturing expansion announced in September 2025, and a field of battery and capacitor specialists that includes Eaton, Panasonic, KEMET, Nippon Chemi-Con and Maxwell Technologies, now part of Tesla. If data-center construction slows — because grid interconnection queues stall, permitting tightens, or hyperscaler capex guidance rolls over — the order surge reverses with it. Supercapacitors are a component, not a platform, and component suppliers in a gold rush often see margins compress as capacity catches up with demand.
There is also execution risk in the technology claims. The 44 percent energy-saving and 40 percent compute-gain figures are internal performance assertions; until they are validated across multiple hyperscaler deployments and published in operator case studies, they should be treated as marketing targets rather than contracted specifications. And while supercapacitors win on response speed and cycle life, lithium-ion capacitors and hybrid battery-supercapacitor architectures offer higher energy density, meaning the winning design for any given facility may not be Skeleton's.
The answer to that counter-case is equally concrete. Even if AI data-center construction proves cyclical, the underlying power-quality problem is structural: AI workloads create millisecond-scale load swings that traditional infrastructure was not designed to smooth, and every additional gigawatt of compute adds to that volatility regardless of the construction cycle. Siemens' investment is not merely financial — it is a channel. A financing arm does not write a nine-figure check into a component supplier unless the parent company's sales teams can bundle that component into broader power deals. That bundling, more than any single performance claim, is what could convert Skeleton from a niche supplier into a standard specification.
The falsifying signal is specific: if Siemens Energy's own data-center order growth stalls for two consecutive quarters while AI capex guidance from the largest hyperscalers remains intact, the thesis that power quality is the binding constraint — and that Skeleton sits at the choke point — is wrong. A simpler test: if Skeleton's 2027 IPO prices below its last private valuation, the market is saying the hardware multiple, not the AI narrative, sets the value.
The Bottleneck That Makes the Trade
There is a second-order reason to take the power-quality thesis seriously: the grid itself cannot keep up. The U.S. interconnection queue has swelled to roughly 2,600 gigawatts of backlog, with the median wait for a project to reach commercial operation approaching five years and some data-center connections facing delays of up to 12 years. Nearly half of the U.S. data-center capacity planned for 2026 — about 7 gigawatts of 12 gigawatts announced — has been canceled or delayed, according to industry analysis, with transformer shortages and interconnection waits cited as the primary causes. Before 2020 a high-voltage transformer took 24 to 30 months to deliver; today the wait is about five years. Wood Mackenzie estimates U.S. power-transformer demand has risen 119 percent since 2019 while manufacturing capacity has not kept pace.
That bottleneck changes the value of Skeleton's pitch. If a data center cannot get a grid connection, anything that lets it draw less peak power or connect faster moves from a nice-to-have efficiency feature to a permit-enabling necessity. The company's claim that its systems enable faster grid connection is not just an operating-cost argument; it is a schedule argument, and in a market where a finished data center can sit idle for years waiting for power, schedule is the most expensive line item on the balance sheet.
What Comes Next
In the short term, the story is sentiment and validation: more named investors in the pre-IPO round, announced customer wins at U.S. data centers, and the first independent case studies behind the 44 percent energy claim. In the medium term, it is revenue conversion — whether the Siemens relationship produces contracted orders that show up in audited financials ahead of the 2027 listing. In the long term, it is whether supercapacitor-based power conditioning becomes a standard line item in AI data-center design, the way turbines and switchgear are today.
The base case is that Skeleton reaches its 2027 U.S. IPO as a valued niche leader, carried by the AI power wave and Siemens' endorsement. The upside case is that power quality becomes a contractually mandated requirement for AI facilities or that interconnection rules begin to reward fast-response power conditioning, turning a performance advantage into a specification lock-in. The downside case is that the data-center capex cycle peaks before the IPO, leaving a factory-heavy balance sheet priced for growth that no longer arrives.
Skeleton's unicorn status is less a verdict on supercapacitors than on the industrialization of AI: the market is no longer paying only for the model, it is paying for the megawatt. The companies that win the next leg of the AI buildout will be the ones that remove the bottlenecks between a chip and a working grid connection — and Skeleton just got paid to stand at that bottleneck.
Explore more exclusive insights at nextfin.ai.