Moonshot AI's open-weight model Kimi K3, released Monday (local time). [Reuters]
Moonshot AI's open-weight model Kimi K3, released Monday (local time). [Reuters]

Chinese AI company Moonshot AI on Monday (local time) released the weights and related materials for its open-weight model Kimi K3. Along with the release, Moonshot specified that companies operating a platform-as-a-service (PaaS) business using the Kimi K3 model and generating more than $20 million in total sales over any consecutive 12-month period must enter into a separate commercial license agreement with Moonshot AI.

The New York Times reported that while Chinese AI models are advancing rapidly, the companies behind them have no clear strategy for turning that success into profit.

Moonshot's technical showcase — a second DeepSeek shock?

The Kimi logo and Kimi K3 model on display at the Moonshot AI booth during the World Artificial Intelligence Conference (WAIC) in Shanghai, China, on July 18. [AFP]
The Kimi logo and Kimi K3 model on display at the Moonshot AI booth during the World Artificial Intelligence Conference (WAIC) in Shanghai, China, on July 18. [AFP]

Moonshot drew widespread attention earlier this month when it unveiled Kimi K3, a massive open-weight model with 2.8 trillion parameters. Kimi K3 ranked third globally in performance benchmarks, trailing only Anthropic's Claude Fabulous 5 and OpenAI's GPT-5.6 Sol — putting it on par with America's most advanced models.

On Monday, Moonshot released the model's programming code and other technical information, making it available for anyone to download and use in theory. Users can also adapt it to their own needs.

Moonshot also announced paid licensing terms, acknowledging that most users cannot run the model properly due to its size.

The New York Times noted that Kimi K3 is technically open-source but that the model is too large to download and run on a standard computer. Most users will access it either through Moonshot's subscription service or through other companies that have signed a license agreement with Moonshot.

Kevin Xu, founder of Interconnected Capital, a hedge fund that invests in AI technology, said Moonshot deserves attention for being the first company to spell out such commercial requirements explicitly.

Even the middle ground can't escape the monetization trap — 'more users, more losses'

The Kimi application by Moonshot AI installed on a mobile phone screen in Beijing, China, on July 17 (local time). [AFP]
The Kimi application by Moonshot AI installed on a mobile phone screen in Beijing, China, on July 17 (local time). [AFP]

Behind the technical achievements of Chinese AI companies, however, lies a structural weakness: deteriorating profitability. The New York Times said "even China's AI powerhouses are struggling to figure out how to make money from AI."

AI models demand enormous computing power to run. In China, service fees are set extremely low, creating a situation where losses grow as the user base expands.

Major US tech companies generate relatively stable revenue backed by powerful business-to-business cloud ecosystems, while Chinese firms rely on free or low-cost consumer-facing platforms and are failing to monetize their AI technology. The Chinese government, meanwhile, wants its companies to compete with American rivals but is wrestling with how to keep the strategic and economic benefits within China.

Wei Shen, a senior AI analyst at Counterpoint Research in Beijing, said "an open model is a powerful distribution strategy, but it is not a complete business model." He added that while a company "can raise funds for the next training cycle through an initial public offering, that alone cannot create a sustainable revenue structure."

Moonshot revises its open-source strategy — walking a tightrope between sharing and profit

Visitors at the Kimi K3 booth of Chinese AI startup Moonshot at the World AI Conference in Shanghai, China, on July 17. [AP]
Visitors at the Kimi K3 booth of Chinese AI startup Moonshot at the World AI Conference in Shanghai, China, on July 17. [AP]

Facing a survival crisis, Chinese AI companies are revising their open-source strategies in search of a way forward. Moonshot's newly announced licensing policy is part of that effort.

Moonshot is opening the model to ordinary developers free of charge to expand its ecosystem, while requiring large commercial service providers to obtain a paid license. It is a balancing act — capturing the benefits of open-source while blocking big tech companies from free-riding and ensuring fair compensation.

Sam Sachs, a senior researcher at the Johns Hopkins School of Advanced International Studies (SAIS) who attended the Shanghai conference, said "the core of the debate in China is how to reconcile the conflicting signals about the strategic benefits of Chinese technology being used around the world." He added that "Chinese authorities want industry champions, but they are also thinking about how not to kill the goose that lays the golden eggs."

China's AI industry has struggled for years, hampered by US trade sanctions that restrict purchases of advanced chips. Chinese companies also have far less capacity than their well-funded American counterparts to buy computing power.

A case in point: just two days after announcing Kimi K3, Moonshot said it would stop accepting new subscribers because it had not secured enough chips to support the service.

Jeffrey Ding, a professor at George Washington University who studies the US-China AI competition, said China's shortage of AI infrastructure is a serious problem. He said Chinese companies adopted an open-source strategy not because open-source was an end in itself, but as a competitive tactic for challengers trying to gain an edge, and that "the ultimate goal for Chinese companies is also to generate profit."


yckim6452@heraldcorp.com