DeepSeek Panic, US vs China, OpenAI $40B?, and Doge Delivers with Travis Kalanick and David Sacks

FriedbergCommentary

Friedberg questions whether OpenAI's massive fundraise and data-center moat thesis still holds after DeepSeek, suggesting model commoditization is happening faster than expected and the value will accrue elsewhere in the stack.

“it's a window for OpenAI to say, we're going to go up against Meta. This is it. We're going to be the players... it becomes harder to believe in the context of what happened in the last week.”▶ Listen · 30:40 ⚑

ChamathCommentary

Chamath is skeptical of OpenAI's $40B raise at $340B valuation, arguing the underlying thesis — that ChatGPT is the killer consumer app — puts them on a collision course with Meta's billions of DAUs, making the valuation hard to justify.

“if OpenAI raises $40 billion at $340 billion, that just hit the wire. The underwriting logic at $340 billion, exactly what you just said, Friedberg. It is the wrapper, meaning ChatGPT is the next killer app... which puts them on a”▶ Listen · 29:49 ⚑

ChamathCommentary

Chamath argues Meta is best positioned to respond to DeepSeek because it must embrace and extend the innovations (PTX, GRPO), and it already has billions of DAUs — making it the only company that can realistically bring AI to a billion-plus people.

“I think all the pressure right now I think is on Meta because I think Meta has to show up with the next iteration of LLaMA that beats and exceeds Gemini, that exceeds R1. And I think that that is going to be crucial for us to have a”▶ Listen · 40:47 ⚑

ChamathCommentarytracking against the call

Chamath argues DeepSeek's use of PTX (bypassing CUDA) and GRPO (bypassing PPO) signals an 'emperor has no clothes' moment for NVIDIA's moat, pointing toward a more heterogeneous chip environment that reduces NVIDIA's lock-in advantage.

“apologies to the NVIDIA bulls, but it's going to create a more heterogeneous environment. And the reason is because there's too much money and risk on the line to go through a single point of failure. A chip, a high-level framework to get”▶ Listen · 41:47 ⚑

Travis Kalanick argues Tesla has a significant manufacturing advantage in the autonomous vehicle race, and cheap AI is already turbocharging FSD performance, with 10x improvement in miles-per-intervention in just 3 months.

“cheap AI makes cheap autonomy. Okay? So if, as cheap AI gets out there and proliferates and gets broadly distributed, we should expect autonomy gets easier and easier and easier. And you see some of the stuff that's happening with Tesla and”▶ Listen · 1:27:05 ⚑

Travis Kalanick is bullish on Waymo's technology maturity, noting it has normalized the robotaxi experience to the point where riders don't think twice, though he flags power grid constraints as the key bottleneck for scaling.

“You get in a Waymo today and it's like, you're not even thinking twice. You're just like, it's all good. You just get in, you get out... The technology works.”▶ Listen · 1:26:07 ⚑

JasonCommentary

Jason argues the application layer is the best place to invest as AI models commoditize rapidly, drawing the analogy that you want to be YouTube not storage, or Uber not GPS — the real value accrues at the application layer.

“The answer to your question is the application layer because this is all going to become storage... Do you want to be in the storage business or you want to be in the YouTube business? Do you want to be in the Uber business or do you want”▶ Listen · 43:42 ⚑