Rack track · Presentation
Scaling Hyperscale Power Delivery: Managing Aggregated and Cascaded Power Shelf Architectures for High-Density GPU Systems
- When
- Thursday, February 4, 20272:40 PM to 3:00 PM PT
- Where
- Room ASan Jose McEnery Convention Center
- Track
- RackDesigning the 1MW Rack. Programmed by OCP Tech Day
- Format
- PresentationTopic: In-Rack Power Distribution
About this session
As hyperscale AI systems evolve toward large GPU clusters, rack power delivery is transitioning from single-shelf designs to aggregated multi-shelf architectures. While OCP standards enable modular power shelves with parallel scaling, emerging high-density systems introduce new challenges when multiple shelves are combined or cascaded across a rack. This session explores system-level considerations for managing aggregated and daisy-chain-like power shelf topologies, focusing on power balancing, coordination, and efficiency optimization. It highlights challenges in distribution imbalance, control-plane scaling, fault isolation, and dynamic load behavior across shelves. The discussion also covers opportunities to improve power utilization and reduce energy losses through coordinated shelf management, enabling more efficient and scalable power delivery for next-generation hyperscale GPU deployments.
General rate, until Feb 2, 2027
Register for Digital Infrastructure World 2027
One pass covers this session and every other session, plus OCP Tech Day, the keynotes and the exhibit hall, on all three days.
- All Access$799
- Open Compute Project members$399
- Teams of 3 or more, each$599.25