The MAD Podcast: How AI Gets Built — with Matt Turck Who Feeds the GPUs? Inside AI's Hidden $30B Layer | Renen Hallak, VAST Data
41 snips
Sep 24, 2026 
Renen Hallak, founder and CEO of VAST Data, builds infrastructure for massive AI workloads. He explores what an AI factory needs, why enterprises may control their own models and agents, and how agent memory, vector databases, and confidential computing are reshaping the stack. He also discusses exploding data demand, sovereign AI, neocloud survival, NVIDIA and xAI partnerships, and the future of intelligence infrastructure.
AI Snips
Chapters
Transcript
Episode notes
Agents Turn Experience Into Organizational IP
- Agents can accumulate organizational IP by learning from daily experiences, fine-tuning at night, and sharing knowledge across teams.
- Hallak envisions an intelligence marketplace where specialized model capabilities can be leased without exposing their weights.
Hallak Chased The Seven Million Dollar Proof
- As a student, Renen Hallak spent six months trying to solve P versus NP before realizing the problem exceeded his abilities.
- He connected that experience to AI's potential to accelerate difficult problems humans may not solve alone.
VAST Bet On Fast Access To Massive Data
- VAST began with the premise that neural networks needed both enormous data volumes and extremely fast access, not merely better algorithms.
- Its DASE architecture puts shared storage across a fast network so thousands of nodes can access information without constant node-to-node coordination.