The key idea is that China's decentralized state-structure enables *political* competition among local governments, which creates strong incentives to back local firms and thus a systemic reluctance to allow inefficient or underperforming firms to exit the market.

My piece about DeepSeek (www.chinatalk.media/p/deepseeks-...) is the essay of the week on The Syllabus! How cool is that!

In case it wasn’t clear, the painting in my post is of DeepSeek eating Sam Altman

Nice piece (with a useful graphic) by @kyleichan.bsky.social on China’s overlapping industrial sectors and how advancements in one sector reinforces advancements in others. Useful for thinking about industrial policy strategy.

www.high-capacity.com/p/chinas-ove...

In 2017 Sam Altman wrote a blog post to rally the tech sector against Trump. He argued that tech CEOs must resist Trump's policies, even if there is "business risk" in doing so. He also suggested that tech workers "have a lot of leverage" and should "do something." blog.samaltman.com/time-to-take...

The US and China are each other's most important AI research partners. Nice infographic by @restofworld.org.

restofworld.org/2025/us-chin...

Microsoft is putting AI nationalism front and center for the incoming Trump administration. In Microsoft's President Brad Smith's new blog post, he makes explicit that the company is all in on using AI to reassert American dominance and help beat down China.

blogs.microsoft.com/on-the-issue...

When analyzing how green coalitions form (or whether they get opposed by counter-interests), I argue that we should pay attention to the country's:

- Growth Model (export- vs consumption-led)
- Development stage (early vs late developers)
- Political structure (pluralist vs corporatist)

Interesting: Microsoft is the biggest buyer of Nvidia's Hopper GPUs.

Equally interesting: Chinese-owned firms ByteDance and Tencent are the next largest buyers (they bought a modified version of the hopper chip bc of US export restrictions).

link: www.ft.com/content/e85e...

The key argument is summarized in this table highlighting how the US's hidden industrial policies differed from the East Asian model of industrial policy:

Racist stereotyping of Chinese students at #neurips (one of the top AI conferences) by a MIT Media Lab professor who runs a lab about making computers emotionally intelligent.

Normally, services aren't considered strong growth drivers because they are simultaneously produced and consumed, and service-related labor has intrinsic value. Baumol's classic example is that it took a string quartet exactly the same amount of time to play a piece in 1965 as it did in 1865. 2/

US growth has been the envy of the world. Business investment is a key driver of this.

Investment in tech alone (dominated by Google/Microsoft/Amazon/Apple) is bigger than the entirety of Chinese investment (or 4x Chinese tech investment). European investment in tech is barely on the map. 1/

Tesla is the only car company that makes 100% of the cars it sells in the US, domestically. Among US-assembled carmakers, Tesla’s parts are also third-most sourced from the US.

In other words- Trump’s proposed tariffs would make other carmakers much less competitive than Tesla in the US market.

China is both the solution and the problem in the green transition. It is by far the biggest renewable energy manufacturer and greenhouse gas emitter (the US is a distant second on both dimensions).

Two interesting charts that show how important U.S. institutions (especially firms) are for the Chinese AI ecosystem. The first is where Chinese AI firm founders graduated. The second shows which companies founders came from. Blue is US/foreign schools or firms.