← Home
Odd Lots · June 8, 2026 · 50m

How CoreWeave Sees the Market for Compute Right Now

Brannin McBee, co-founder and CDO of CoreWeave, returns to discuss the evolving cloud compute market three years after the company's previous appearance. CoreWeave has since IPOed and expanded funding through bond markets and Nvidia partnerships. The conversation explores persistent bottlenecks in AI infrastructure—chip supply scarcity, transformer availability, and electricity constraints—plus shifts in lease markets, Nvidia's new Vera Rubin systems, and the balance between training and inference demand.

This summary was generated from show notes and public descriptions, not from a full transcript review. Details may contain inaccuracies.

Highlights

GPU Supply Remains the Core Bottleneck in AI Infrastructure
Three years after CoreWeave's previous appearance, chip supply scarcity persists as a primary constraint on AI infrastructure growth, despite massive capital inflows into the sector.
Transformers and Electricity Are Co-Equal Bottlenecks with Chip Supply
Beyond GPUs, the market faces simultaneous scarcity in power infrastructure (transformers) and electricity, creating a three-way constraint on growth.
Nvidia Partnership Expansion and New Vera Rubin Systems Shape Market
CoreWeave's expanding relationship with Nvidia, including integration of new Vera Rubin systems, positions the company at the center of next-generation GPU infrastructure.
Training vs. Inference Demand Shapes Infrastructure Investment Differently
The market is bifurcating between compute for model training (massive, one-time investments) and inference (ongoing, lower-power), requiring different infrastructure strategies.

Editorial

Bond Market Funding Signals Maturation of AI Infrastructure as Asset Class
CoreWeave's ability to raise capital in bond markets, alongside equity and traditional venture financing, indicates that AI infrastructure companies are now treated as mature, credit-worthy entities.

Misc

CoreWeave IPOed since last appearance on Odd Lots
GPU sourcing was the original constraint; now chip supply remains scarce but competes with transformer and electricity bottlenecks
Nvidia partnerships expanding with new Vera Rubin systems
Bond market funding marks maturation of AI infrastructure financing
Was this useful?