This Week in Startups

CoreWeave’s Brannin McBee on the future of AI infrastructure, GPU economics, & data centers | E1925

36 snips
Apr 4, 2024
Guest
Brannin McBee
Brannin McBee, Co-founder and CDO of CoreWeave, shares insights from transitioning his company from cryptocurrency to AI infrastructure. He discusses the soaring energy demands of GPUs, emphasizing innovative cooling solutions and the significant costs tied to GPU operations. The conversation highlights CoreWeave's expansion with new data centers to tackle surging AI demands, and the competition in AI hardware, particularly against NVIDIA's dominance. McBee also explores the evolving landscape of AI applications and their implications for industries like advertising and e-commerce.
Ask episode
AI Snips
Chapters
Transcript
Episode notes
INSIGHT

GPU Economics and Scale

  • Renting a single GPU costs around $4/hour, but CoreWeave specializes in large-scale deployments.
  • Scaling creates a more scarce resource, as building large, contiguous GPU fabrics is engineering-intensive.
INSIGHT

Power Bottleneck and GPU Efficiency

  • A major bottleneck in the market is not just data center space, but also sufficient power supply.
  • GPUs consume a lot of power, about 10% of CoreWeave's infrastructure cost, but are more efficient than CPUs for parallel workloads.
INSIGHT

Data Center Cooling

  • Data centers require significant cooling, currently achieved through forced air, but liquid cooling is the future.
  • Liquid cooling will improve energy efficiency by directly cooling chips, lowering the energy overhead.
Get the Snipd Podcast app to discover more snips from this episode
Get the app