Uploaded January 2026 | Updated September 2026, 2 weeks ago
Earlier this year, Weights & Biases joined forces with CoreWeave to build an end-to-end AI cloud platform.
In this episode, we look under the hood at what that actually means for the stack. Lukas Biewald sits down with Corey Sanders to discuss why general-purpose infrastructure is throttling AI innovation.
After two decades building traditional cloud infrastructure, Corey explains why the future of AI training belongs to a specialized “Neo Cloud.”
• Specialization: AI workloads break general-purpose clouds. Neo Clouds exist because they’re built for one thing, and do it better.
• GPUs: In AI, idle GPUs are wasted money. Everything, from storage to networking, should exist to keep them fed.
• Infrastructure: Liquid cooling, caching, and scheduling aren’t optimizations, they’re requirements for next-gen AI training.
• Shift: As AI becomes business-critical, customers will leave “best-in-suite” clouds for best-in-class platforms.
Watch the full episode to see how the new AI tech stack is being built from the metal up.
Earlier this year, Weights & Biases joined forces with CoreWeave to build an end-to-end AI cloud platform.
In this episode, we look under the hood at what that actually means for the stack. Lukas Biewald sits down with Corey Sanders to discuss why general-purpose infrastructure is throttling AI innovation.
After two decades building traditional cloud infrastructure, Corey explains why the future of AI training belongs to a specialized “Neo Cloud.”
• Specialization: AI workloads break general-purpose clouds. Neo Clouds exist because they’re built for one thing, and do it better.
• GPUs: In AI, idle GPUs are wasted money. Everything, from storage to networking, should exist to keep them fed.
• Infrastructure: Liquid cooling, caching, and scheduling aren’t optimizations, they’re requirements for next-gen AI training.
• Shift: As AI becomes business-critical, customers will leave “best-in-suite” clouds for best-in-class platforms.
Watch the full episode to see how the new AI tech stack is being built from the metal up.
