Stenobird

Episode

Inside the $41B AI Cloud Challenging Big Tech | CoreWeave SVP

Podcast
Gradient Dissent: Conversations on AI
Published
Jan 6, 2026
Duration seconds
3199
Processing state
processed
Canonical source
https://wandb.ai/site/resources/podcast
Audio
https://episodes.captivate.fm/episode/167db7d2-6e2f-406e-b26b-19e329f838c8.mp3
JSON
/v1/public/podcasts/gradient-dissent/episodes/inside-the-41b-ai-cloud-challenging-big-tech-coreweave-svp
Markdown
/podcast/gradient-dissent/inside-the-41b-ai-cloud-challenging-big-tech-coreweave-svp.md

Actions

  • POST https://stenobird.com/v1/public/podcasts/gradient-dissent/episodes/inside-the-41b-ai-cloud-challenging-big-tech-coreweave-svp/transcription-requests
    Idempotently request low-priority transcript generation for this episode.
  • GET https://stenobird.com/podcast/gradient-dissent/inside-the-41b-ai-cloud-challenging-big-tech-coreweave-svp.md
    Read the agent-friendly Markdown representation of this episode resource.

Summary

General-purpose clouds are struggling to meet the specialized demands of large-scale AI training. CoreWeave's approach focuses on maximizing GPU throughput and observability to solve the unique bottlenecks of modern workloads.

Topics

  • AI Infrastructure
  • GPU Orchestration
  • Cloud Computing
  • Machine Learning Operations
  • CoreWeave
  • Data Throughput
  • Cloud Strategy
  • Kubernetes

Highlights

  • Main idea: The shift toward AI workloads requires a 'Neo Cloud' model that prioritates GPU throughput over API consistency
  • Practical takeaway: Maximizing 'goodput' requires specialized hardware considerations like liquid cooling and optimized data paths
  • Failure mode: General-purpose clouds fail when they cannot make the specific hardware assumptions necessary for massive AI training jobs
  • Main idea: Customer success in AI infrastructure depends on deep observability and automated orchestration like Slurm on K8s
  • Practical takeaway: High-quality product feedback is best gathered by presenting functional, high-fidelity prototypes rather than abstract ideas

Chapters

  1. 1:00 Introduction: Lukas Biewald introduces Corey Sanders, SVP of Product at CoreWeave, discussing his transition from Azure to the specialized AI cloud space.
  2. 9:00 The Evolution of AI Workloads: A discussion on how the AI revolution has created a new tier of business-critical workloads that mirror the previous analytics wave.
  3. 13:05 CoreWeave's Technological Innovations: Exploring how specialized assumptions regarding hardware, such as liquid cooling and optimized data movement, provide a competitive edge.
  4. 25:05 Customer Engagement and Future Prospects: How CoreWeave uses deep observability and technical customer enablement to drive product adoption and user satisfaction.
  5. 29:05 Comparing Cloud Approaches: Contrasting the 'Neo Cloud' model with traditional public cloud strategies regarding customer service and infrastructure flexibility.
  6. 37:00 Balancing Executive Roles and Hands-On Projects: Corey discusses the importance of staying close to the technical implementation and the challenges of GPU availability.
  7. 41:20 Product Development and Customer Feedback: The value of using high-fidelity mockups and functional prototypes to elicit actionable feedback from engineers.