XNO® for AI Data Centres: High-power batteries for the demands of AI
AI training clusters can spike or drop power draw by hundreds of megawatts in under a second and, without effective buffering, risk operational disruption and grid destabilisation. The batteries serving uninterruptible power supply (UPS) and battery backup units (BBUs) have the potential to meet this challenge and smooth power fluctuations, but most aren’t currently up to the task.
Echion Technologies’ new whitepaper sets out how, and what changes when its niobium-based anode technology, XNO®, is integrated into AI data centre operations.
Download the full whitepaper now
AI computing has created battery demands that didn’t exist at scale before. To maintain centre operations and stabilise the grid, UPS and BBUs require batteries that deliver continuous, symmetric, high-rate cycling, thousands of times per day across a lifetime of multiple years.
Incumbent battery technologies were not built to meet these demands. Under required fast charge conditions, incumbent batteries can suffer lithium plating, forcing manufacturers and operators to cap charge rates well below their discharge rates and overbuild their installations. This requires more raw materials and centre floor space than necessary and ultimately drives up economic and environmental costs. The duty cycle then compounds this problem further, as continuous high-rate cycling degrades incumbent technologies faster, shortening their lifecycle and requiring regular mid-life replacements.
XNO®, Echion’s niobium-based anode technology, removes these limitations altogether, delivering:
· 50,000+ cycles to 80% capacity (10–90% DoD)
· 30C maximum symmetric charge/discharge rate
· 1 MW continuous charge and discharge, 1.5 MW peak
· 10x the life of the best incumbent high-power rack
· 85% lower battery capital cost vs incumbent lithium-ion systems
· 80% lower 10-year cost of ownership compared to incumbent lithium-ion systems
· Zero mid-life replacements at this duty
In a worked 270 MW AI facility example, XNO® meets the full smoothing duty at a fraction of the cost and floor space of incumbent alternatives. For operators, this means more compute power per rack, increased floor space that can be translated into GPU-driven revenue, and a facility that grid managers are more likely to connect.
The whitepaper explores:
· The AI data centre power challenge and the emerging regulatory case for grid smoothing
· Where batteries sit in an AI data centre’s architecture
· Why incumbent chemistries fail under high-rate symmetric cycling
· What XNO® does differently and how it works
· Inside the XNO® UPS rack design
· A full 270 MW worked example quantifying the impact at real data centre scale
· Market and validation