Barron: Moonshot AI's Latest Model Dings Tech Stocks.

In early 2025, the release of DeepSeeks R1 artificial-intelligence model sent tech stocks sliding. This week investors met Kimi, the latest AI challenger from China, though the reaction was less drastic.

Moonshot AI, the Chinese lab backed by Alibaba Group 9988 3.68%.., on Thursday released its newest model, dubbed Kimi-K3. The model outperformed its most advanced peers from Anthropic and OpenAl across an array of coding tasks, according to model evaluator Arena

AL.

The iShares Semiconductor exchange-traded fund fell

1.6% on Friday, extending its losing streak to three days.

The tech-heavy Nasdaq Composite, meanwhile, dropped

1.4%.

For markets, the headline is simple: Investors are starting to question whether the U.S. maintains the same commanding lead in AI that many had assumed over the last 18 months, wrote Mizuho Securities managingdirector Daniel ORegan in a note Friday.

Unlike Anthropics Claude or OpenAIs GPT model families, Kimi-K3 is open-weight, meaning enterprises can host it on their own servers or private cloud without paying Moonshot directly. Moonshot will release the full model weights on July 27.

Enterprises have increasingly used cheap open-weight models for simple tasks to save on token prices. But if Kimi-K3 can mimic advanced U.S. models at a fraction of the cost per token, it begs questions about the sustainability of the entire AI investment boom-from Anthropic and OpenAl to the cloud providers and chip makers who depend on them.

Kimi-K3 is essentially the spark hitting a room already filled with gas, said Harrison Rolfes, a private markets analyst at PitchBook. If intelligence gets cheaper faster than expected, the justification for hyperscalers pouring hundreds of billions into data centers loses its floor.

Chinese leader Xi Jinping fanned the flames in a news conference Friday, endorsing the continued creation of open-source AI models and criticizing the

monopolization of AI development by any one country.

The market isnt panicking just yet. Recall that when DeepSeek released R1 on Jan. 27, 2025, the Semiconductor ETF plunged 7.8% and the Nasdaq tanked 3.1%. Tech stocks recovered in short order and capexestimates only ballooned further.

This parallels the first time [DeepSeek] came to global attention, and for a week we thought OpenAI was toast, said Giuseppe Sette, co-founder of Reflexivity, an investment analysis platform. Better and cheaper tech simply means more adoption.

Wall Street is also savvier now about how AI economics work. Kimi-K3 looks cheap on a per-token basis, said Willy Lee of Neostellar Capital NSLR 4.58%..

J, a firm that

has invested in OpenAI. However, it is still a massive model that could strain memory and networking hardware and guzzle tokens.

Dean W. Ball, head of Strategic Futures at OpenAI, echoed that opinion in a social media post Friday.

Its a very good model! Ball conceded. In my fairly limited use, it also seemed very token hungry. Its not obvious to me that this model is actually that cheap to run.

After last years DeepSeek Moment, investors will hope history repeats itself. If so, AI adoption will grow, the closed-source U.S. models will retain market share, and money will keep trickling down throughout the AI ecosystem. Kimi-K3 at least poses another risk to thatbest-case scenario.