Semi Doped Quick Takes: Nvidia Keynote at GTC
15 snips
Mar 17, 2026 They break down Nvidia's GTC keynote highlights and the shift in AI into training, inference, and agentic eras. They cover new hardware moves like Groq's LPX and Vera Rubin systems, plus rising CPU demand and novel interconnects. They dig into scale-up architectures, AI-native storage concepts, and the tradeoffs between building agents and buying SaaS.
AI Snips
Chapters
Transcript
Episode notes
Debrief Recorded On The Road After Flight Issues
- Austin Lyons recorded the debrief from a hotel room in San Jose after flight delays and an analyst dinner.
- He streamed the keynote from a plane and mentioned meeting Nvidia staff like Ian Buck and Gilad at the analyst relations dinner.
Hosts React To DLSS Gaming Demo
- Vikram Shekhar watched Jensen's keynote early and skipped many cinematic demo videos but enjoyed the DLSS gaming demo.
- He said the DLSS demo turned blocky game faces into photorealistic characters, making him want to upgrade a GPU.
Grok LPX Paired With GPUs For Ultra Low Latency Decode
- Nvidia integrated Groq's deterministic decode chips (Grok LPX) as an ultra-low-latency decode tier alongside HBM GPUs.
- The Rubin system pre-fills KV cache and attention on HBM GPUs then sends activations to Grok racks for sequential SRAM-based decoding.