™The Programmable Data Center — Evan Conrad, SF Compute
A 37-minute interview recorded at AI Engineer New York: SF Compute CEO Evan Conrad on GPU compute trading — physical settlement, order-book liquidity, interrupting and moving workloads, and spot versus forward pricing.
Key points Tap a time to jump
Why NVIDIA can't build a compute market itself: neutrality as a precondition for trading
Physical settlement: the buyer gets usable compute, not just cash settlement against an index
Liquidity and the GPU order book
Interrupting and moving workloads: whether a workload can be interrupted and how fast it recovers from storage
Batch versus live inference leases
Physical versus cash settlement
The emerging compute market ecosystem
Sovereign compute
Comments
More to watch
All →- 43:33AI in NYC: How Tenex Builds an AI-Native Engineering Team — Alex Lieberman & CJ Hess
- 40:32Dylan Field (Figma CEO) on AI and design, the blocked $20B Adobe deal, and the SaaS debate
- 1:15:20High Taste Testers Ep. 1: OpenAI Shipped Everything. What Actually Works?
- 1:34:05All-In Podcast E293: CoreWeave Crash, AI Trade Cracks, Anthropic IPO Red Flags, Bill Gates Doomer
Latest news
All →- Oct 22Legora Raises $550M Series E at $8.8B Valuation
- Oct 21ChatGPT Adds Locked Chats Behind a PIN or Face ID
- Oct 21PHP2Shell: Two Critical Flaws Threaten ~70% of Websites, Discovered with AI Help
- Oct 21OpenAI Expands ChatGPT Ads to 89 Countries
- Oct 21OpenAI's Frontier Models Come to Google Cloud
- Oct 21Pigment Raises $250M Series E at $2.3B Valuation
Today
Oct 22 Thursday- Brief
- GitHubGitHub Weekly Vol.2: 6 Projects Worth Your Time
Tested or read — weeklyAll →
- Quiz
- Call