AI data centres
More usable compute from the same connection
Near-lossless conversion holds efficiency at peak as AI workloads fluctuate, turning constrained power capacity into more usable compute and less heat to cool.
AI data centres in detailA conventional DC/DC converter is tuned for one operating point. As AI workloads swing, its efficiency falls away and more power turns into heat.
Talk to our power engineersAccess to power has become the main bottleneck in AI data centres. The grid connection you have is the grid connection you will have for years.2
AI workloads fluctuate and EV charging is one of the most variable loads on the grid. Conventional converters are optimised for a narrow peak-efficiency point and lose ground everywhere else.2
Every watt lost in conversion turns into heat that has to be cooled. Cutting the loss frees capacity twice: once at the rack, once at the chiller.2
Software-Defined Power Conversion (SDPC) combines soft switching with digital control of the conversion process. The converter adapts in real time to changing loads and input voltages.
Zero Voltage Switching and Zero Current Switching remove the moment where voltage and current overlap in the switch. That overlap is where conventional converters burn energy. Higher switching frequencies then make the converter smaller and more powerful.
Advanced modelling lets SDPC use the full potential of ZVS and ZCS at every operating point. Where a conventional converter drops off at partial load, SDPC stays close to its peak. Diagnostics and updates run remotely, without interrupting operations.
Every Vensum product is built from the same validated blocks of hardware, firmware and control software. They are reused and reconfigured for new voltages, power levels and applications, so a new converter is developed significantly faster than an application-specific design.
How SDPC differs from an application-specific converter, from Vensum's own technology description.
| Property | Conventional converter | Vensum SDPC |
|---|---|---|
| Efficiency | Peaks at a narrow operating point | Near peak across very wide ranges |
| Fast load changes | Efficiency falls away | Adapts in real time |
| Switching | Hard switching losses | ZVS and ZCS soft switching |
| Design | Built from scratch per application | Reconfigured from validated blocks |
| Service | On-site | Remote diagnostics and updates |
Compatible with distributed and centralised power architectures, compact enough for new builds and retrofits.
More usable compute from the same connection
Near-lossless conversion holds efficiency at peak as AI workloads fluctuate, turning constrained power capacity into more usable compute and less heat to cool.
AI data centres in detailSmaller stations, lower operating cost
Compact converters fit new and existing infrastructure and reduce the footprint of charging stations. Higher efficiency means lower energy losses for owners and operators.
EV charging in detailMore usable energy from every cycle
Energy is converted every time it is stored and released. Near-lossless conversion means more usable energy per cycle, faster charging and discharging, and intelligent control of DC energy flows.
Storage and microgrids in detailBorn out of research in deep technology, with patented and patent-pending technologies. Series production starts in 2027.3
“The question in data centres is no longer only how powerful the chips are, but how efficiently the used electricity actually reaches them.”
Tell us the application and the power level. Our power electronics team replies with how SDPC would fit.