The Next Wave - AI and The Future of Technology The Hidden Industry Powering Every AI Company
21 snips
Nov 12, 2025 Evan Conrad, CEO of SF Compute, is an expert in AI infrastructure and data centers. He discusses the transformation of GPU clusters into tradable commodities to combat credit risk and volatility in AI finance. The conversation covers the challenges of financing GPU deals, U.S.-China competitiveness in power capacity, and regulatory hurdles. Evan also shares insights on the future of compute infrastructure, predicting groundbreaking advancements by 2050, including self-driving vehicles and innovative delivery systems.
AI Snips
Chapters
Books
Transcript
Episode notes
Compute Needs Long-Term Offtake
- GPU clusters require long-term offtake contracts to be financeable because lenders underwrite against customer credit risk.
- Mispriced credit risk in short-term startup contracts can cascade into broader market instability.
Contract Duration Mismatch Creates Systemic Risk
- Short-term customer contracts sold by inference providers create mismatch with long-term GPU cloud obligations.
- A VC pullback could expose this mismatch and trigger widespread defaults in the compute stack.
GPUs Behave Differently Than CPUs
- GPU pricing power is much lower than CPUs because AI apps are price-sensitive and margins are thin.
- That forces GPU providers to rely on longer contracts and different capital structures than traditional cloud providers.
