Mon, Jul 20, 2026|Sean Michael Cummings

China Slams on the AI Accelerator

A Chinese startup just made fools of the White House...

It all started on June 9, when AI company Anthropic rolled out its most advanced models yet.

Claude Mythos 5 and Claude Fable 5 were both on the cutting edge. In fact, Mythos was too powerful for the "Everyman" – it was available to only a handful of top companies. Fable was almost as capable, but it had built-in security measures to make it safe for the public.

On June 12, the White House flagged both models as threats to national security, slapping export controls on them. It claimed they would be too dangerous in the wrong hands.

The controls kept Anthropic from releasing these models abroad without a special license. The U.S. also forbade foreign nationals from using them... either inside or outside the country.

But only a day later, a Chinese startup called Z.ai released a new AI model that was nearly as good. Known as GLM-5.2, it came close to Fable on many key benchmarks for large language models... And it scored better on long-horizon coding than OpenAI's GPT-5.5.

Plus, unlike its U.S. rivals, GLM-5.2 is open-weight. That means anyone can download and run it, no matter where they are. Users can even tweak the code themselves.

The release made a huge splash. And it spooked the U.S. government...

At the end of June, the U.S. Department of Commerce dropped the export restrictions on Mythos and Fable.

Tech leaders in the U.S. had warned that regulation could make American AI fall behind... And China quickly proved their case.

And last week, a Beijing-based company proved again that China is catching up in the AI race...

Chinese startup Moonshot just announced another major AI breakthrough... Kimi K3, the world's largest open-weight AI system to date.

What's more, Kimi K3 surpasses Claude Fable when it comes to front-end coding (building the visual, interactive components of websites and apps), according to a benchmark test by AI evaluator Arena.ai.

Its front-end test has various models code live apps or websites. Then, human judges blindly vote on which products work the best in a series of one-on-one comparisons. The votes are aggregated into a score, with a higher score showing more coding proficiency.

Take a look at how Kimi K3 blew up the leaderboard...

A New AI Contender Emerges

Of course, this is a big deal for China. With GLM-5.2 and Kimi K3, the country is proving that its budget offerings can rival the West's flagship AI models.

Whether these new competitors are disruptive enough to kill the AI boom remains to be seen. But they do form a dark cloud over the bull market thesis...

If U.S. businesses adopt these open-weight models instead of OpenAI's and Anthropic's, Western hyperscalers could turn off the capital-expenditure spigot. If they did, demand for chips would plummet... devastating the stock market.

Today the Philadelphia Semiconductor Index (SOX), which tracks the American chip sector, is about 19% off its June 22 peak – on the brink of a bear market. Take a look...

I recommend keeping a watchful eye on your exposure to U.S. chips in the coming weeks. This sector is in a downtrend.

And if corporate America starts choosing budget Chinese options over the West's best-in-class AI offerings, this crash could deepen further.

Watch SOX closely from here. A decisive break below its recent lows would tell us the market is finally taking China's AI threat seriously.

Good investing,

Sean Michael Cummings

China Slams on the AI Accelerator

A New AI Contender Emerges

Further Reading