Who Feeds the GPUs? Inside AI's Hidden $30B Layer | Renen Hallak, VAST Data

Name: Who Feeds the GPUs? Inside AI's Hidden $30B Layer | Renen Hallak, VAST Data
Uploaded: 2026-09-26T00:00:00.000Z
Duration: 4222 s
Channel: The MAD Podcast with Matt Turck
Description: VAST Data CEO Renen Hallak explains the storage tier feeding xAI-class clusters: one AI cloud's 500PB request grew by 2 exabytes in a week. He covers shared-everything architecture, why inference raises storage demands again, and data enclaves that expose private data without surrendering it.

TMThe MAD Podcast with Matt Turck

Full transcript

English

Summary：VAST Data CEO Renen Hallak explains the storage tier feeding xAI-class clusters: one AI cloud's 500PB request grew by 2 exabytes in a week. He covers shared-everything architecture, why inference raises storage demands again, and data enclaves that expose private data without surrendering it.

Watch on YouTube

Core points (3)

1Storage demand from AI clouds is exploding: 500PB asks became 2.5EB within a single week.
2Inference needs fast private data, not just big archives — the next storage bottleneck.
3Data enclaves let models compute on enterprise data without the data leaving custody.