ABB is partnering with NVIDIA to speed up the event of next-generation, gigawatt-scale knowledge facilities designed to energy the explosive progress of synthetic intelligence workloads. The collaboration focuses on creating high-efficiency, scalable energy options that may assist NVIDIA’s deliberate 800-volt direct present (VDC) structure – a key enabler for the large 1-megawatt server racks anticipated in future AI services.
As knowledge heart energy consumption continues to rise globally, with whole demand forecasted to develop from 80 gigawatts in 2024 to about 220 gigawatts by 2030, the necessity for extra environment friendly power supply techniques has change into essential. In line with Dell’Oro Group, AI workloads will characterize round 70% of that enlargement, driving knowledge facilities towards basically new approaches to energy distribution and system design.
The ABB-NVIDIA collaboration will deal with creating energy architectures that mix medium-voltage uninterruptible energy provides (UPS) with direct present energy distribution on the server degree. This configuration, utilizing superior solid-state electronics, guarantees to ship larger effectivity and improved reliability for more and more power-hungry AI techniques. The aim is to streamline the power path from grid to chip, lowering conversion losses and enabling higher-density computing environments.
“ABB is main the event of key energy distribution applied sciences that may outline the following technology of information facilities,” mentioned Giampiero Frisio, President of ABB Electrification. “We’ve got been early traders in cutting-edge UPS, DC, and solid-state electronics that may assist knowledge facilities meet AI’s unprecedented energy calls for. Our collaboration with NVIDIA to advance 800 VDC architectures is a part of our broader dedication to driving innovation on this dynamic sector.”
Stable-State Medium-Voltage UPS: HiPerGuard
NVIDIA, whose GPUs type the spine of recent AI infrastructure, can be pushing for a shift towards higher-voltage energy techniques. “As AI calls for proceed to develop worldwide, knowledge facilities require new approaches to energy distribution that enhance effectivity and simplify design,” mentioned Dion Harris, Senior Director for HPC, Cloud, and AI Infrastructure at NVIDIA. “By means of this collaboration, NVIDIA and ABB are serving to the trade transfer towards 800-volt architectures that may assist the high-density AI infrastructure fueling the following wave of innovation.”
ABB’s intensive knowledge heart portfolio consists of clever energy administration, digital monitoring, and backup techniques that guarantee uptime and optimize power use. Almost 40% of ABB’s electrification R&D is devoted to improvements in knowledge heart applied sciences, together with DC distribution, safety gadgets, and superior cooling techniques.
Amongst ABB’s current improvements are the HiPerGuard – a solid-state medium-voltage UPS – which will increase power effectivity and energy density in a smaller footprint, and the SACE Infinitus, the primary IEC-certified solid-state circuit breaker designed to make DC distribution each sensible and scalable.
By combining ABB’s power engineering expertise with NVIDIA’s management in AI {hardware} and knowledge infrastructure, the partnership goals to ship {the electrical} basis for the following technology of AI-driven knowledge facilities – techniques able to supporting ever-larger fashions whereas minimizing their environmental and operational footprint.
