Chinese artificial intelligence (AI) corporation Moonshot AI unveiled its gigantic open-weight (weights-disclosed) model "Kimi K3" while mandating paid licenses for corporations above a certain size. It is a strategy to grow the open-source ecosystem while securing revenue. The New York Times (NYT) said, "Chinese AI models are advancing rapidly, but the corporations that developed them have yet to find a clear strategy to turn their technical success into stable revenue," it said.
According to the information technology (IT) industry on the 28th, Moonshot drew industry attention in early this month by unveiling the gigantic open-weight model "Kimi K3," which has 2.8 trillion parameters. Kimi K3 ranked third globally in major AI performance evaluations, following Anthropic's "Claude Fabulous 5" and OpenAI's "GPT-5.6 Sol," earning assessments that it stands shoulder to shoulder with the most advanced U.S. AI models.
The day before, Moonshot released the weights and programming code so anyone could download and use the model. Developers can modify the model based on this or apply it to their own services. However, in reality, most users find it difficult to run the model themselves. That is because Kimi K3 is so large and resource-intensive that it is difficult to operate on a regular PC.
The NYT noted that the model is too large to download and use on a general computer, and that while Kimi K3 is technically open source, most users gain access to the model by using Moonshot's subscription service or through other corporations that have license agreements with Moonshot. Kevin Xu, founder of AI investment-focused hedge fund Interconnected Capital, said it is meaningful that Moonshot is the first case to explicitly present such commercial use conditions.
Behind the technical achievements of Chinese AI corporations lies a weakness of structurally worsening profitability. The NYT pointed out, "Even China's AI powerhouses cannot figure out how to generate revenue with AI." AI models require enormous computing power to run. In China, service fees are set low, leading to a phenomenon in which deficits grow as the number of users increases.
While U.S. big techs generate relatively stable revenue based on powerful business-to-business (B2B) cloud ecosystems, China relies on free or low-cost consumer (B2C) platforms and thus is not making money from AI technology. The Chinese government, while hoping its corporations can compete with U.S. corporations, is also grappling with how to keep strategic and economic benefits within China.
Wei Sun, senior AI analyst at Counterpoint Research in Beijing, China, said, "Open models are strong as a distribution strategy, but they are not a complete business model," adding, "You can raise funds for the next training cycle through an initial public offering (IPO), but that alone cannot create a sustainable revenue structure."
Chinese AI corporations facing an existential crisis are seeking breakthroughs by revising their open-source strategies. Moonshot's licensing policy announced this time can be seen as part of that effort. Moonshot opened the model free of charge to general developers to aim for ecosystem expansion, while requiring giant service corporations to obligatorily obtain a "commercial license." It is a tightrope strategy to capture the benefits of open source while preventing big techs from free-riding and securing fair compensation.
Sam Sachs, senior fellow at the School of Advanced International Studies (SAIS) who attended a Shanghai conference, said, "The crux of the debate in China is how to reconcile the conflicting signals about the strategic advantages of Chinese technology used around the world," adding, "Chinese authorities want industry champions, but they are also considering how not to kill the goose that lays the golden eggs."
China's AI industry has struggled for years under U.S. trade sanctions that restrict purchases of advanced chips. It also has far less capacity to purchase computing power compared to U.S. corporations with deep pockets. A case in point is Moonshot's announcement two days after unveiling Kimi K3 that it would not accept new subscribers because it had not secured enough chips needed for the service.
Jeffrey Ding, a professor at George Washington University who studies U.S.-China AI competition, pointed out that China's shortage of AI systems is a serious problem. He said, "The reason Chinese corporations chose an open-source strategy by releasing their models was not because open source itself was the goal, but because it was a strategy to secure a competitive edge as challengers," adding, "For Chinese corporations, the ultimate goal is to generate revenue."