Z.AI shares slide 20% as Chinese AI developers mull licensing fees

Chinese AI developers may begin charging commercial licensing fees to cloud platforms that host their open-weight models, a shift that could reshape the economics of the $200 billion-plus AI inference market, according to Goldman Sachs.

"Domestic growth in adoption is very fast, and we think there's considerable adoption among small and medium enterprises in the global market, and even larger companies are starting to consider using Chinese models," Ronald Keung, Hong Kong-based head of Asia internet research at Goldman Sachs, said in an interview.

Chinese models such as Moonshot AI's Kimi K3 and Z.AI's GLM-5.2 have reached performance levels only slightly behind leading US competitors, according to benchmark data. Kimi K3 generated more than 930,000 downloads in its first week after a July release, a 200% increase from the prior week, Sensor Tower data shows. Z.AI reported revenue surged 132% to 724 million yuan ($107 million) last year, though its net loss widened 60% to 4.7 billion yuan ($694 million).

The shift from permissive open-source licensing to commercial terms would allow Chinese model builders to capture more revenue from surging global usage. Most Chinese models currently distribute under licenses such as MIT, allowing third parties to host and serve the models for free. Requiring commercial licenses for inference services could boost margins for developers while raising costs for cloud platforms that have built services around these free models.

The Open-Source Revenue Gap

Chinese AI developers have historically released models under open-source terms to drive adoption, but that strategy has created a revenue gap. While platforms like OpenRouter and cloud providers can freely download and host the core "weights" — the underlying parameters encoding a model's intelligence — the developers themselves capture little direct revenue from this usage. The top five most popular models on OpenRouter over the past month were all Chinese, according to platform data, highlighting the scale of unpaid consumption.

Keung said Chinese model makers could generate more revenue by requiring third-party vendors to purchase commercial licenses to provide inference services on their own infrastructure — a model he described as "paid weights" licensing. This approach would mirror how enterprise software vendors charge for runtime licenses rather than just distribution rights.

Competitive Pressure and the US Response

The rise of Chinese AI models has frustrated some US tech giants, but short of an outright ban, the cheaper alternatives are likely to continue gaining traction. Mozilla Chief Technology Officer Raffi Krikorian said he switched to Kimi K3 for many daily tasks within days of its launch, calling it "snappier" than Anthropic's Claude Fable. Cryptocurrency exchange Coinbase has also adopted Chinese models to help trim costs, according to people familiar with the matter.

US restrictions on advanced AI chip exports have not prevented Chinese developers from achieving competitive performance. The Trump administration this week accused Moonshot of using "covert" methods to build K3 based on Anthropic's Fable, a claim Beijing rejects as groundless. Meanwhile, US export controls that briefly kept Anthropic's Fable and Mythos models offline in June created an opening for Z.AI's GLM-5.2, which launched shortly after.

For investors, the licensing shift introduces a new variable into the Chinese AI market. Z.AI shares fell 19.9% on Tuesday, though the move appeared driven as part of broader sector rotation rather than company-specific news. The stock trades at a discount to US AI peers, reflecting ongoing concerns about profitability. If Chinese developers successfully transition to a licensing model, it could meaningfully improve unit economics — Z.AI's gross margin on inference services could expand by several hundred basis points, according to estimates from Goldman Sachs. Moonshot, which temporarily suspended new K3 subscriptions after overwhelming demand, is among the developers best positioned to benefit from the shift.

This article is for informational purposes only and does not constitute investment advice.