Now Playing
Vera Rubin's Orchestration Moat, Moonshot Loophole & CME GPU Futures
Audio is available on Spreaker — see link below.
Nvidia's AI hardware moat has shifted from GPU supply to data orchestration — and regulators are scrambling to close a cloud-access loophole that let Moonshot train on 20,000 Hopper chips without a single export violation. Today's briefing covers Vera Rubin architecture, TSMC's chiplet scale-up, CME GPU futures, and China's persistent equipment gap.
Audio is available on Spreaker — see link below.
Nvidia's competitive advantage is no longer about who has the most GPUs. That framing is already out of date.
That said, the competitive threat hasn't disappeared. It's relocated.
Running underneath all of this is a structural manufacturing shift. Chiplets are becoming the default architecture across the industry, and TSMC is the infrastructure enabling that transition.
The policy story this week centers on a gap that took regulators longer than expected to notice. Chinese AI firm Moonshot trained its Kimi K3 model on twenty thousand Hopper chips accessed via Alibaba cloud.
Two more developments worth tracking. CME launches a GPU compute futures market on October fifth.
The two real watchpoints from here are Nvidia's orchestration moat durability and the BIS export control rule's legal path. On Nvidia: the question is whether hyperscalers build competing infrastructure layers fast enough to matter before Vera Rubin locks in customer dependency.
Chapter summary auto-generated from the verified script. Listen to the full episode for the complete content.