China Open-Source, Compute Arms Race, Reordering Global Trade | BG2 w/ Bill Gurley and Brad Gerstner

Dòng thời gian tập

1. 0:00

   Opening discussion of Trump's tariff deals with Europe and Japan and the resulting market optimism.

   • The EU deal imposes 15% tariffs on European goods entering the US while US goods face 0% tariffs going to Europe, plus commitments of $750 billion in energy purchases.
   • The Japan deal similarly has Japan paying tariffs to the US with no tariffs imposed on the US, and $550 billion of investment directed by the president.
   • Brad argues the president deserves credit because predictions of retaliation and trade wars have not materialized, only deals, and compares it to a CEO halfway through a high-risk strategy who is in line for a bonus.
2. 3:32

   Introduction of guest Sunny Madra, COO of Groq, and the AI action plan from the White House.

   • Sunny Madra joins from Saudi Arabia and is rumored to be raising $600 million at a $6 billion valuation for Groq, which hosts open-source models in inference clouds worldwide.
   • The AI summit and crypto summit under David Sacks produced the Genius Act for stablecoins, the Clarity Act for market structures, and a multi-pronged American AI strategic plan.
   • The AI action plan aims to maintain global AI leadership over China by extending government investment and accelerating distribution of the American AI stack globally.
3. 8:10

   China's dominance in open-source AI models and the compounding effect of open weights.

   • Qwen from Alibaba has passed 400 million downloads under the Apache 2.0 license, and six to seven high-quality Chinese open-source model providers are now among the fastest growing in the world.
   • Sunny explains that open weights allow Chinese labs to compound on each other's work, distilling and generating synthetic data, so they may pass the best US proprietary models by Q4 this year.
   • A 30-billion-parameter Qwen model released that day performs as well as GPT-4o, illustrating how quickly leading-edge models distill down into smaller turbo models.

Khái niệm chính

• China open-source models— Chinese open-weight models like Qwen and DeepSeek are dominating global downloads and rapidly closing the intelligence gap.
• compute arms race— Massive buildouts of gigawatt-scale AI data centers by xAI, OpenAI, and others reflect an unprecedented demand for compute.
• tariff deals— The Trump administration's tariff strategy has yielded deals with the EU and Japan, bringing trillions in investment and revenue.

Trích dẫn nổi bật

• I think if we look at Q4 this year or Q1 next year, I'd be willing to say, you know, the a top three worldwide model will be a US-based open source model.

  🔥

  — A bold prediction that US open-source models will catch up to Chinese leaders within months, despite current momentum favoring China.

  Nhảy đến 16:34

• Google went from 5 trillion tokens a month to 480 trillion tokens a month. And they had just put some press out that they crossed like, you know, 800 trillion. And I saw something today. They crossed a quadrillion. And I had to look that up. That's a,000 trillion. And so in a in a course of a year and a few months, they've gone from 5 trillion to a,000 trillion. So that's 200x right there.

  🤯

  — The staggering 200x growth in token consumption reveals the explosive, almost incomprehensible scale of AI inference demand.

  Nhảy đến 28:23

Hành động cụ thể