- Dummy loads burn up to 45% of energy just to keep power levels stable
- Skeleton’s GrapheneGPU delivers up to 40% more computing with the same hardware
- AI workloads are killing energy efficiency, GrapheneGPU smooths power draw in real time
As artificial intelligence systems grow more demanding, many data centers have found themselves consuming nearly twice the energy they technically need.
This overuse isn’t due to system flaws or outdated hardware, it’s rooted in how GPUs behave, as their power demand can swing drastically within seconds, from full throttle to idle.
To cope, operators often deploy dummy loads, deliberate energy wasters, to maintain a stable power draw – but these data centers deliberately slow the performance of tens of thousands of GPUs to prevent power outages
Dummy loads mean massive wastage of energy
While this avoids damage and blackouts, it means up to 45% of energy is lost as heat, performing no useful computation.
Skeleton Technologies now claims it may have a more efficient alternative, one that allows GPUs to run at full capacity without overwhelming the grid.
The Estonian company developed GrapheneGPU, a peak-shaving system using proprietary Curved Graphene supercapacitors.
Unlike lithium-based systems, these capacitors can respond in just 10 microseconds, absorbing energy during idle periods and discharging it instantly when GPU loads spike.
The result, according to Skeleton, is the ability to maintain consistent GPU performance without stressing the grid or resorting to throttling.
Their tests suggest the system can deliver up to 40% more FLOPS – floating point operations per second – using the same GPUs, simply by removing the performance penalties associated with thermal de-rating and power instability.
“GrapheneGPU delivers up to 40% more computing with the same energy footprint, while cutting both capital and operating costs by reducing grid upgrade needs, energy waste, and cooling,” said Taavi Madiberk, CEO of Skeleton Technologies.
“Powered by our patented Curved Graphene, this is a fundamental shift in how AI infrastructure can scale – sustainably and economically”.
The company also reports up to a 44% reduction in the power capacity that data centers must reserve from the grid.
The core unit, the GrapheneGPU PCS 50, delivers up to 80 kW of peak power in a standard 1OU form factor, compatible with existing infrastructure and cooled by air or liquid.
Importantly, it avoids lithium entirely, using Skeleton’s patented graphene-based material instead.
According to Skeleton, this technology has been tested under rigorous hyperscaler-grade GPU workloads with positive results. However, it has not been independently tested for real-world performance and durability.
The first shipment of this technology will commence in Germany by June 2025. The company also has a U.S. production site planned for early 2026.
