Taming Data Gravity: Sustainable Archiving for HPC and AI at Scale

HPC and AI environments are producing data at a rate that outpaces traditional storage models. Simulation outputs, model checkpoints, experimental data pipelines and massive training datasets continue to accumulate, creating data gravity - large, cold datasets that weigh down high performance infrastructure, increase energy consumption and drive up costs.

As archives grow into the petabyte and exabyte range, relying on performance tier storage for long term retention becomes impractical. Power and cooling budgets are strained, HDD fleets grow inefficient and NVMe capacity gets consumed by data that rarely changes.
This session explores how organisations in the HPC space are redesigning their archival strategy to turn data gravity into an operational advantage rather than a burden. Attendees will learn how next generation, tiered long term storage solutions can:

• Reduce archive power and cooling requirements by up to 75%
• Free up NVMe and HDD capacity for active compute workloads
• Improve durability and resilience by integrating modern tape and intelligent hierarchical storage
• Optimise workflows end to end, from data generation to long term retention and future reuse
• Support institutional sustainability and cost efficiency goals without compromising accessibility

Whether you are expanding your AI pipeline, scaling simulation output, or facing rising energy constraints, this session will provide a practical, strategic framework for building a sustainable, high efficiency archival architecture.