Skeleton Technologies, which specializes in high-power systems for AI infrastructure, in October became one of Estonia’s latest unicorns thanks to a strategic investment led by Siemens, a key player in the architecture of AI data centers. The deal values Skeleton at more than $1 billion.
Siemens brings scale and customer access. Skeleton brings a fix for one of the biggest constraints on data center expansion: the sudden, large swings in electricity demand caused by AI workloads. Together they plan to develop products for next-generation AI data centers and sell them to hyperscalers, with Skeleton adding manufacturing in the U.S. to be closer to its largest customers, Arnaud Castaignet, Skeleton’s senior vice president of strategic affairs, said in an interview. Siemens will have exclusive global rights to sell Skeleton’s data center products.
At the Tallinn Digital Summit, which took place November 5 and 6 in Estonia, Ekaterina Zaharieva, European Commissioner for Startups, Research and Innovation, pointed to Skeleton as an example of a company that could help Europe in its ambition to build energy resilience and AI infrastructure.
The supercapacitors Skeleton makes don’t store bulk energy the way batteries do. They absorb and release power almost instantly and can do so over many more charge cycles, which makes them good at smoothing spikes rather than running things for hours. Skeleton’s flagship data center product, GrapheneGPU, pairs supercapacitor cells made with the company’s proprietary Curved Graphene material — which is engineered for extremely fast power delivery, high efficiency, and long cell life — with power electronics and control software.
The problem it solves stems from how AI chips behave. When thousands of GPUs start or stop a task at once, a cluster’s power draw can jump or collapse within milliseconds, sending spikes and electrical distortion, known as harmonics, back through the system. Skeleton’s equipment sits next to the servers and absorbs those swings before they reach the rest of the facility. The company says this cuts energy use by up to 40% and boosts GPU performance by double digits.
“AI is scaling faster than the power infrastructure behind it,” Taavi Madiberk, co-founder and CEO of Skeleton Technologies, said in a statement. “The constraint is not how many GPUs you can buy, but how much compute you can extract from every GPU and every megawatt available. In AI infrastructure, supercapacitors are becoming the new semiconductors. They will determine how much performance hyperscalers can deliver from their GPUs.”
Demand for data centers that can make more efficient use of electricity is expected to rise. The International Energy Agency (IEA) projects that data centers worldwide will use about 950 TWh of electricity in 2030, roughly twice the 485 TWh they consumed in 2025. Consumption by data centers built specifically for AI is expected to grow even faster, tripling over the same period. In the U.S., data centers are on track to account for nearly half of all growth in electricity demand through 2030, according to an earlier IEA report.
The IEA expects 20–25 GW of battery storage in data centers by 2030 to manage AI load swings. Why choose supercapacitors over batteries? “We don’t see lithium-ion batteries as a competitive threat because they address different applications,” Castaignet said.
Lithium-ion batteries have an important role in AI data centers, particularly in battery backup units and large-scale energy storage systems, he said. However, they are not suited to managing the extremely rapid, high-frequency power fluctuations generated by AI GPUs at rack level, inside the data center’s white space. “Tesla Megapack, for instance, operates at a completely different level of the electrical infrastructure and is not designed for microsecond-scale GPU power management,” Castaignet said.
“This is where Skeleton’s GrapheneGPU product has a decisive advantage,” he said. “Skeleton’s GrapheneGPU systems respond in microseconds (which is impossible for batteries), absorb and deliver very high power almost instantaneously, and withstand millions of charge-discharge cycles. This enables us to smooth GPU power demand and increase computing performance.
The architecture is equally important, he said. “By placing supercapacitor-based solutions on the DC side, close to the GPUs, we can address power fluctuations directly at their source, minimizing conversion losses and avoiding the additional response delays associated with managing these fluctuations through upstream AC-side systems. This provides a more efficient solution than relying on conventional battery storage elsewhere in the data center.”
The company’s GrapheneBBU (based on its SuperBattery, which is designed to fill the technological gap between supercapacitors and traditional batteries) competes directly with lithium-ion battery backup solutions. The company says it requires up to 67% less space than conventional lithium-ion solutions, recharges in under 90 seconds, and provides up to 90 seconds of backup power. Less space dedicated to backup systems means more room for computing infrastructure, Castaignet said. “Depending on the customers’ needs and applications, we can either use supercapacitors or SuperBattery in our systems.”
The Tallinn-based company, which was founded in 2009 by Madiberk and Oliver Ahlberg, started in transport, supplying supercapacitor products for rail, trams, buses, and heavy equipment. It has since moved into defense. Its products are used in air defense systems, specialized guided missiles called interceptors, and counter-drone systems. “Defense platforms are becoming in some ways like data centers,” Castaignet said. “Complex military systems, whether they be interceptors, jammers, or radars, all operate using small peaks of power, which need to be orchestrated inside the military platform, and therefore you need power architecture that can react very fast.”
Skeleton works with about a third of the world’s top 20 defense companies, Castaignet said.
In June, at the Eurosatory defense show in Paris, Skeleton announced a partnership with KNDS France for air defense, mobile, and turret systems, and signed a strategic memorandum of understanding with GDELS-Santa Bárbara Sistemas, the Spanish unit of General Dynamics European Land Systems, centered on advanced counter-unmanned aircraft systems and interceptor missile technologies aimed at bolstering Estonia’s defense capabilities. GDELS-Santa Bárbara will integrate Skeleton’s high-power energy storage systems into its military mobility platforms. Skeleton also signed an agreement with Portugal’s Tekever at Eurosatory.
Today, AI data centers account for about 50% of Skeleton’s business. By 2027 that share is expected to become “the majority of our revenue,” rising to 80% to 90% by the end of the decade, Castaignet said.
The relationship with Siemens will play a major role in Skeleton’s go-to-market strategy. The October investment by Siemens — which Skeleton described as a nine-figure sum, made in return for a minority stake and exclusive distribution rights — is a deepening of a long-standing relationship. Siemens Financial Services joined Skeleton’s 2023 €108 million round, and Siemens Digital Industries designed and automated Skeleton’s SuperFactory near Leipzig, Germany.
Competitive Landscape
Skeleton competes on supercapacitor cells with companies such as Japanese energy storage manufacturer Musashi Energy Solutions, and on power systems with companies such as U.S.-based Vertiv, which offers power systems for data centers but doesn’t make its own cells.
“Our first differentiator is that we are fully integrated: We do the cells and the systems,” Castaignet said. Skeleton supplies not just energy storage but also the power conversion, control, and system architecture needed to optimize an AI data center from the GPU rack to the electrical grid. “Secondly, our cells have the highest power density on the market. Power density means smaller systems, smaller modules, much more power, and much faster reaction time, which is absolutely critical for AI data centers.”
Another plus: Skeleton says it is the only company in its sector that does not do any of its manufacturing in Asia. Castaignet noted that Skeleton is the first Estonian unicorn focused on hardware — a recognition, he said, of hardware’s growing importance to technology sovereignty.
The company already operates a full European value chain, producing cells at its Leipzig SuperFactory and power systems in Finland.
The Siemens money will pay for a U.S. build out. Skeleton plans to spend more than $200 million on scaling up operations and production and on research and development in San Jose, California, close to the hyperscalers it works with. The Estonian scale-up, which is using a contract manufacturer’s facility, also plans to make cells in the U.S. by the end of 2027 at the same facility. Once that is in place, Castaignet said, Skeleton will be the only company in its segment that makes everything from cells to finished systems on American soil.
“Data centers, especially in the U.S., are considered as strategic infrastructure, so they want to reduce their dependency on foreign supply chains,” Castaignet said. “That is why manufacturing in the U.S. will be a key differentiator.”
