Grace Shao on What the World Should Know About Chinese AI

June 22, 2026

AI Summary

5 min read

Grace Shao on What the World Should Know About Chinese AI

When DeepSeek V3 launched around Trump's inauguration in early 2025, it wasn't just another model release. The lab had quietly delayed its V4 rollout by three to four months to re-engineer inference onto Huawei hardware—a first serious effort to run frontier AI on a Chinese chip stack. "They kind of did one for the team," says independent AI researcher Grace Shao, who writes the Substack AIPROM. "They became a shared foundation layer for China's model ecosystem." Because everything is open source and open weight, other labs could study what DeepSeek did and start inferencing on Huawei silicon themselves.

Why Chinese AI Went Open Source (and What That Actually Means)

The conventional story about Chinese AI being open source versus America's closed frontier models has less to do with philosophy than pragmatism. "It was a very pragmatic business reason to start with," Shao explains. Chinese labs needed Western developers to trust them, and open-sourcing was a branding decision to build credibility. That said, the founder of DeepSeek has also openly stated he wants to share frontier research to propel the whole industry forward.

Continue reading the full summary in the app — free to try.

Read Full Summary →

Free • No credit card required

Listen to Audio Summary Open in App

Never miss an episode of Odd Lots

Get every new episode summarized in your inbox — free, ~5 minutes to read.

No spam. Unsubscribe anytime.

What you'll learn

1 (06:40) **Why Chinese AI Models Are Open Source** - The pragmatic business reasons behind open-source dominance, including trust-building for Western developers and a more collegial R&D ecosystem
2 (08:09) **How DeepSeek Changed the Ecosystem** - DeepSeek's V3/V4 releases transformed Chinese AI from a slump into a globally respected force, with a key hardware pivot
3 (10:56) **The Four Frontier Labs and Their Specializations** - DeepSeek, Moonshot, Zhipu AI (GLM), and Minimax each focus on a different vertical out of necessity
4 (12:00) **The Capital Stack: Lower Valuations, Real Revenue** - Chinese AI startups went public at $6-8 billion vs. US $20B+ paper valuations, but they are actually making money
5 (17:06) **The Two Major Constraints: Capital and Chips** - Export controls and limited VC funding force Chinese labs to optimize differently than US labs
6 (18:49) **How Chinese Labs Get Around the Data Bottleneck** - They wait for exclusivity windows on proprietary datasets to expire, then pay a fraction of the price
7 (19:20) **Energy Is NOT a Bottleneck in China** - Unlike the US, China has abundant, cheap, newly built energy infrastructure

+ Full timestamped outline available in the app

Show Notes

China's AI industry has changed a lot since DeepSeek released its cheap frontier model last year, and briefly sent US tech stocks falling. After being locked out of the most advanced chips, Chinese companies are now allowed to buy some Nvidia H200s. In fact, many of the big Chinese tech companies — like Baidu — are making a push to become full-stack players, with their own chips, models, and cloud infrastructure. Today's guest is Grace Shao, an independent AI researcher and the author of the AI Proem Substack. She's a bit of an insider when it comes to China's AI industry, and when we were in Hong Kong we spoke with her about the latest in open-source models, the competition among Chinese frontier labs, DeepSeek's place in an increasingly crowded Chinese AI market, China's manufacturing edge, where bottlenecks exist right now (spoiler: it isn't data centers), if Chinese grandmas are actually using OpenClaw, and finally, of course, AI psychosis.

Only http://Bloomberg.com subscribers can get the Odd Lots newsletter in their inbox each week, plus unlimited access to the site and app. Subscribe at bloomberg.com/subscriptions/oddlots

Subscribe to the Odd Lots Newsletter