AI is the Pivot -- Data is the Power

In this RAISE Summit keynote, Paul Bloch (President & Co-founder, DDN) lays out why data infrastructure—not GPUs—is the rate limiter for AI at scale. Drawing on DDN’s 25-year HPC heritage, he shows how the company now powers many of the world’s largest AI training and inference clusters (from NVIDIA internal supercomputers to xAI) and why sovereign, exabyte-class storage is the backbone of Europe’s AI ambitions.

‍

Key discussion points include:

• HPC → AI factories: How sovereign HPC in France (CEA, EDF, IDRIS/CNRS, TotalEnergies) evolved into sovereign AI—complete with national “AI factories” and multi-billion-euro gigafactory plans.
• Where scale breaks: End-to-end data pipeline design—high-throughput ingest, prep/indexing, fast checkpoint/restore to protect accuracy and avoid wasted GPU epochs, and low-latency, multi-tenant inference.
• “Data is the asset”: Exploding real-time/video feeds push enterprises from hundreds of TB to multi-exabyte footprints; model sizes sprint toward the trillion-parameter era.
• Performance proof (RAG on cloud): Swapping native S3 for DDN Infinia delivered 22× faster time-to-result and ~40% fewer GPU hours on AWS—small wins per query, massive savings at millions of queries.

‍

Ecosystem & partnerships:

• NVIDIA training stacks (A100→H100→B/GB series) run atop DDN EXAScaler/Infinia.
• Google Cloud: new, fully managed DDN option to stand up AI storage workloads in GCP with a few clicks.
• Rising AI clouds and EU sovereignty: Mistral AI Cloud, Scaleway, OVHcloud, Fluidstack (plus partners like CoreWeave, Lambda).
• What’s next: Million-GPU clusters, “physical AI” (edge-to-core), strict data locality/ethics, and storage software that’s GPU-aware, cloud/on-prem portable, and ops-proven.

Building or modernizing AI data pipelines? Benchmark your checkpoint/restore and RAG paths with DDN Infinia (now on Google Cloud) and stress-test multi-tenant latency before you scale. Meet the DDN team from the floor to see EXAScaler/Infinia in action—and subscribe for more RAISE sessions on AI infrastructure and sovereign AI.