Skeleton Technologies

39 GW of new AI data center capacity will run on 800 VDC by 2030, according to SemiAnalysis. NVIDIA has already made it the power architecture for its next-gen racks. Skeleton is now partnering with DG Matrix to build it.   We're combining our supercapacitor-based power systems with DG Matrix's Interport solid-state transformer platform. Interport supports 800 VDC natively; Skeleton's GrapheneCBU800 and GrapheneBBU800 are built for the same voltage class.   Together they solve a core AI data center problem: GPU pulse loads that hit faster than the grid, or any conventional UPS, can respond. Microsecond-speed absorption, paired with real-time power routing from grid to rack.   Full announcement in the comments.

The Nordics have become a hotspot for AI infrastructure tech, reports Site Selection Magazine, highlighting Skeleton’s leading role globally. The article examines how our footprint is expanding across Europe and into the US, helping solve the power bottleneck across AI data centers and grids with power systems built on trusted supply chains. We’ve grown rapidly in Finland with R&D and systems integration centered on Varkaus and Lappeenranta, each home to a skilled industrial heritage and thriving research hubs. This has enabled Skeleton to become a full provider of power systems for AI infrastructure.

Power is the ultimate bottleneck for AI data centers. But one layer down, you'll find a second critical one: heat. And it determines how well you solve the first.   Grid capacity is scarce, so data centers stretch it with buffering. Supercapacitor-based power systems are designed to support that efficiently and smoothly, absorbing the millisecond load swings of GPU clusters so more grid power gets turned into compute power.   How hard a buffer can work is set by heat. More specifically, how fast it can escape from cells. That’s measured as thermal resistance.   It caps how much current the system can absorb continuously without cooking itself. Underestimate it, and the buffer either has to be derated, oversized, or accepted as a fast-ageing component. In each case, the amount of grid connection you can actually reclaim shrinks.    Solve thermal resistance and every megawatt works harder.   Our new whitepaper, authored by Dr Florian Türke and Markko Püü, explains how we we measure and design for thermal resistance at Skeleton. Link in the comments.   Here's the issue. The values given on datasheets for thermal resistance of cells rarely reflect their long-term real-world usage inside data centers.   The measured value is heavily affected by the test setup, while the actual value is heavily affected by the design of the power system into which the cells are integrated and also the surrounding data center cooling design.   In high-current testing, the power cables alone can carry away 30–50% of the heat.   That’s why, at Skeleton, we separate the cell from its test environment by combining physical measurement with calibrated 3D thermal simulation. It enables us to report three distinct values for thermal resistance:   → Directly measured. This is the one that usually gets the focus in datasheets. It’s the raw lab result, environment included.   → Corrected. This enables more meaningful comparisons across different cells. It’s based on standardized natural-convection conditions.   → Pure cell. This is the one that engineers can properly work with by adding the thermal resistance of their own cooling setup. It measures purely internal hot spot to surface, independent of any cooling system. No lab assumptions baked in.   That last value matters because thermal resistance doesn't stop at the cell wall. The total heat path runs from the hot spot inside the cell, through the module, into the cooling system around it.   Most suppliers only control the first step.   Skeleton builds full power systems for AI infrastructure: cells, modules, and the power systems they operate in. We're not handing customers a cell value and wishing them luck with the rest. We engineer both halves of the heat path together, optimized as one product, so every megawatt of grid connection works as hard as it can to unlock the full potential of AI data centers. Read the full (ungated) whitepaper on the link in the comments.

Great insights here from Dr Mario Rapisarda. The research community is buzzing, as supercapacitor-based power systems increasingly move from the lab into real-world applications solving the world’s most critical power bottlenecks.

The entrepreneurial spirit that creates worthwhile differences is characterized by a relentless pursuit of innovation, consistent and meticulous actions focused on the details, and an unwavering belief in the mission and impact of one’s business. After speaking with Taavi Madiberk, the CEO of Skeleton Technologies—an Estonian company with which Marubeni Corporation has had a capital and business partnership since 2019—I have once again found myself reflecting on these thoughts. With super-capacitors as its core business, Skeleton has become a leading next-generation company in Estonia, a country known for producing numerous unicorns (start-up companies valued at one billion dollars or more). Its technology, which can deliver high power instantaneously, is known to have caught the attention of Elon Musk during his student days. Through steady research and development, Skeleton has evolved and commercialized its technology, steadily growing its revenue through applications such as solutions for energy-efficient data center, stable power transmission line, and commercial vehicle starting systems. In particular, President Taavi explained to me that energy-efficient solutions for data centers lead to improved GPU performance and significant reductions in lifecycle costs, and that there is great potential for substantially reducing capital expenditure for data centers. He also noted that the steady and patient application development being carried out jointly with Marubeni is now translating into business opportunities for AI data centers. I am convinced that Skeleton’s belief in the difference super-capacitors can make will further expand its business, and I look forward to seeing its rapid growth moving forward.   「違い」を生み出す起業家精神とは、貪欲な革新性の追求、細部にこだわる地道なアクションの積み上げ、そして事業のミッション・インパクトを信じ続ける信念です。丸紅が2019年より資本業務提携をしているエストニアのSkeleton TechnologiesのTaavi社長との対話を経て、改めてこのことに思いを馳せています。 Skeleton社はスーパーキャパシタを軸とし、数々のユニコーンを生み出すエストニアを代表する新たな次世代企業となっています。瞬時に大電力を供給できるテクノロジーで、イーロン・マスク氏が学生時代に注目していたことも知られています。Skeleton社は地道な研究開発で技術を進化・商業化し、データセンター向け省電力運用、送電線安定運用、商用車のエンジン始動などで売上を確実に伸ばしつつあります。 特にデータセンター向け省電力運用はGPUのパーフォマンス向上・ライフサイクルコストの大幅削減に繋がり、データセンター向けCAPEXを大きく削減していく可能性が見出されていることをTaavi社長より今回伺いました。そして、丸紅と共同で進めている地道な忍耐を伴うアプリケーション開発が現在のAIデータセンター向けの商機に繋がっているとの言葉を頂いています。 スーパーキャパシタが生み出す違いを信じる信念は、Skeleton社の業容を更に大きくしていくものと確信しています。Skeleton社のここからの飛躍的な成長に期待をしています。

Europe must combine the scale of major defence groups with the speed of specialised tech companies, Skeleton's Arnaud Castaignet tells Defence24. Recommended reading. Interviewer Dr Aleksander Olech notes that military advantage no longer depends only on tanks, artillery and missile systems. Energy resilience, mobility, platform endurance, drones, counter-drone systems and the rapid integration of new technologies into existing weapons are becoming just as important. That's why the joint declaration between Skeleton Technologies and KNDS France reflects a significant, wider shift in Europe’s defence industry. We're working together to build a fully sovereign defence industrial base with fast iteration cycles - measured in months, even weeks - that make better use of our whole continent's diversity. Integrating specialist technologies from both defence and non-defence companies is vital. "French industry has enormous platform expertise, engineering depth and industrial scale," explains Skeleton Senior VP Arnaud Castaignet. "But innovation in areas such as power systems, autonomy, sensors, cyber, electronic warfare, drones, and counter-UAS systems can also come from Tallinn, Warsaw, Riga or Vilnius." European defence sovereignty is not only about final assembly or having European flags on platforms, he adds. "It also depends on control over critical components, supply chains, software, materials and enabling technologies. At Skeleton, building a non-China-dependent value chain has been a strategic priority from the outset, because security of supply is just as important as technical performance for critical defence capabilities." These capabilities all have different power requirements, often at the same moment. "The role of the electrical architecture is to orchestrate them so the entire platform performs as one integrated system. That orchestration is what ultimately delivers energy resilience: keeping the platform operational, responsive and effective throughout the mission." The full interview is on the link in the comments.

👇To scale AI, we must scale the systems powering it. Skeleton CEO Taavi Madiberk, CPO Tero Järveläinen, and VP of US Operations Andrew Worley returned to NVIDIA this week to discuss exactly that.

The supercapacitor community is gathering at Poznan University of Technology this week as we work together to solve the global power bottleneck. 👇 #ISEECap is the flagship gathering of supercapacitor researchers where Mario Rapisarda, PhD will discuss how Skeleton's supercapacitor-based high power systems serve as a complementary power layer to AI infrastructure, grids, and more. Thank you to everyone involved in bringing this event together and propelling the sector to new heights, including co-chairs Elzbieta Frackowiak and Krzysztof Fic, as well as supporters such as Jon Ajuria and François Béguin.

👇Coming soon: Our partners at Siemens stopped by the Skeleton SuperFactory in Leipzig to give you a good look around at how our teams build world leading power systems for AI infrastructure and more.

We’re proud to share that we’ve broken the record for highest peak current amongst UL-certified supercapacitors.   As data center power density demand rises, power systems like GrapheneGPU that can deliver extremely high peak current are critical.   In addition, the bar for safety is higher than ever as data centers move closer to residential areas. Backed by the UL 810A certification, our supercapacitors pose zero risk of thermal runaway.