The Circuit EP: 157 GTC! OFC! Micron Earnings (margin city)
35 snips
Mar 23, 2026 They unpack Nvidia's GTC buzz, product launches, and how vertical stacks reshape data center infrastructure. They explore optical networking trends, co-packaged optics reliability, and manufacturing challenges. They react to Micron's surprising earnings and debate whether memory margins signal a lasting industry shift.
AI Snips
Chapters
Transcript
Episode notes
Competition Now Means Matching The Full Stack
- Competing with NVIDIA now means matching multiple adjacent products, not just a single GPU rack, and many vendors lack the full portfolio.
- Ben noted NVIDIA can't supply everyone, but occupying the full stack raises the bar for competitors like AMD.
Attach Rates For Vera Rubin And Grok Are Key Unknowns
- Attached-rate questions matter: buyers will mix Vera Rubin CPU racks and Grok inference racks with GPU pods, but exact ratios are unknown.
- Ben and Jay emphasized rack-level economics (8–10 rack pods) determine revenue upside for CPUs and Grok units.
Liquid Cooling And Co-Optimization Favor NVIDIA Ecosystem
- NVIDIA's new racks are liquid-cooled and tightly integrated, reducing interoperability with third-party components.
- Ben said the stack's design choices (liquid cooling, shared networking) reinforce a mostly NVIDIA-only deployment model.