Summary
- Siemens has announced a strategic investment and exclusive global data centre partnership with Skeleton Technologies.
- The companies plan to combine supercapacitors with electrical infrastructure to manage rapid changes in AI processor power demand.
- The proposed offering targets existing AC installations and future 800V DC systems, with completion subject to customary formalities.
Siemens is investing in Skeleton Technologies and establishing an exclusive partnership to supply high-power energy storage systems for AI data centres, addressing rapid fluctuations in electricity demand that become more difficult to manage as computing installations grow.
The agreement will combine Skeleton’s supercapacitor technology with Siemens’ existing electrical infrastructure portfolio, extending its offering from utility connections and distribution equipment towards power management closer to the computing racks.
Large groups of AI processors can change their power consumption almost simultaneously as workloads begin, pause or move between computing stages. Siemens says these changes can occur within milliseconds and accumulate into fluctuations measured in megawatts over a few seconds.
Skeleton’s technology is designed to absorb or release energy quickly enough to moderate these disturbances before they place additional pressure on other electrical equipment. The companies also intend to support higher computing densities and provide retrofit options for existing facilities.
Siemens has worked with Skeleton as a technology partner since 2022 and has been an investor since 2023. The newly announced agreement establishes an exclusive global sales and scaling arrangement for Skeleton’s products in the data centre market, although completion of the transactions remains subject to customary formalities.
The financial terms of Siemens’ latest investment have not been disclosed. Skeleton separately reports that the investment values the company at more than $1 billion and will support manufacturing expansion in the United States.
AI changes how power systems must respond
Data centre electrical infrastructure must maintain stable power while supplying equipment whose demands can change substantially during normal operation. AI training and inference clusters place additional pressure on that task because large numbers of processors may operate in synchronised patterns.
An electrical installation must accommodate those changes without allowing unacceptable disturbances to affect other equipment or the utility connection. The engineering requirements extend across transformers, distribution systems, protective devices, power conversion equipment and controls.
Supercapacitors offer one approach to managing short-duration fluctuations. Unlike conventional batteries designed to store substantial amounts of energy for delivery over longer periods, supercapacitors can charge and discharge rapidly, making them suitable for applications requiring repeated bursts of high power.
When incorporated into an appropriate electrical system, a supercapacitor installation can absorb energy during a sudden reduction in demand and release it when consumption rises. Its role is to moderate short-term changes rather than provide the continuous electricity required to operate the entire facility.
That distinction limits the claims that can reasonably be made about the partnership. Supercapacitors cannot create additional long-term electricity supply, remove the need for a suitable grid connection or replace every function of conventional backup power systems.
They may, however, provide greater flexibility in handling transient demand, potentially reducing the extent to which particular electrical components must be designed around short-lived peaks.
Whether those improvements translate into additional usable computing capacity depends on the original facility design, the characteristics of the workloads and the configuration of the new equipment.
Retrofitting the existing data centre estate
Siemens says the planned offering will support alternating-current installations and retrofit projects, while providing a pathway towards future 800V direct-current architectures.
Many existing data centres were designed before the current generation of high-density AI systems became commercially significant. Operators seeking to install more powerful processors may encounter limits in their electrical distribution equipment even where sufficient physical space exists.
Replacing major electrical infrastructure can involve lengthy engineering work, equipment procurement and interruptions to operations. A storage system capable of moderating abrupt changes in demand could provide another option when upgrading an existing installation, provided the remaining equipment has sufficient capacity.
Such deployments still require careful electrical engineering. Protection settings, control software, switching arrangements, cooling and equipment maintenance must be coordinated so that the storage system behaves predictably during both routine operation and abnormal conditions.
The economics will vary between facilities. Operators will need to compare the price and operating requirements of supercapacitors with alternative measures, including changes to electrical distribution equipment, other forms of energy storage and adjustments to computing operations.
Shorter response times do not necessarily imply lower total operating costs. Storage devices, power electronics and control systems introduce their own capital expenditure and maintenance obligations.
Measurements from deployed systems will be needed to establish whether operators achieve lower disturbances, greater usable capacity or reduced expenditure compared with their existing infrastructure.
Manufacturing capacity follows AI investment
Skeleton’s expansion plans include increased manufacturing activity in the United States and additional research and development in Silicon Valley. Demand from hyperscale data centre operators is helping drive that investment, according to the company.
For Siemens, the partnership adds specialist high-power storage technology to a portfolio that already covers electrical distribution, infrastructure controls and related data centre systems.
The exclusive commercial arrangement could allow Skeleton to reach customers through Siemens’ established relationships and engineering capabilities, although future revenue will depend on products being delivered, qualified and adopted.
Manufacturing at greater volume will introduce the usual industrial challenges of consistent quality, dependable component supply and effective support arrangements. Those requirements are especially important for equipment installed within critical electrical infrastructure.
The proposed systems also sit within a wider market for power management technology, where data centre operators are examining batteries, power electronics, software controls and alternative electrical architectures.
Competition between these approaches will depend on their ability to meet specific technical requirements at an acceptable cost, rather than on a single claim of improved efficiency or computing density.
Siemens and Skeleton are addressing the speed at which AI infrastructure demands electricity, which is distinct from the total amount of electricity the industry requires.
The partnership offers an additional technical option for managing that operational problem. Its value will become clearer as the companies complete the investment arrangements, bring products to customers and publish evidence of performance in functioning data centres.












