The era of entirely free open-source artificial intelligence models in China appears to be drawing to a close. Alibaba Group is preparing to implement a new commercial licensing structure for its upcoming Qwen3.8-Max model, signaling a strategic shift in how the country’s tech giants plan to monetize their massive investments in foundational AI research.
According to a report by Reuters, the Chinese technology conglomerate will ask major commercial users of the open-weight model for a share of the revenue they generate using the system. The impending policy change marks a departure from Alibaba’s previous approach to open-source distribution.
Historically, the company has charged developers for utilizing its models when hosted directly on its proprietary cloud computing platform. Still, it has largely permitted its open-source offerings to be downloaded and run in customers’ own data centers without imposing licensing fees. The decision to introduce a revenue-sharing requirement for Qwen3.8-Max, which is expected to be released during the week of August 10, suggests that Alibaba is now seeking to capture a portion of the downstream value created by its most advanced open-weight models.
Following the Moonshot Precedent
Alibaba’s new licensing strategy closely mirrors a controversial but increasingly common playbook pioneered by other Chinese AI developers. Last month, Beijing-based startup Moonshot AI released its highly anticipated Kimi K3 open-weight model with a specific provision tucked into its licensing terms. The Kimi K3 license explicitly requires any organization offering the model as a commercial service and generating more than $20 million in annual sales to negotiate a formal commercial agreement with Moonshot.
The specifics of Alibaba’s revenue-sharing demands remain fluid. According to sources cited by Reuters, Moonshot has been demanding up to a 30 percent revenue share from its largest commercial partners. Alibaba is reportedly planning to ask for a similar revenue-sharing arrangement, though the exact percentage is still being negotiated internally. The fact that both a nimble startup like Moonshot and a massive incumbent like Alibaba are adopting similar licensing frameworks indicates that the Chinese AI industry is converging on a unified business model to monetize open-source development.
(Related: Alibaba Prices Qwen3.8-Max at $2 Per Million Tokens, Undercutting OpenAI)
The Open-Source Freemium Strategy
The transition toward paid commercial licenses reflects a broader maturation of the open-source AI ecosystem. Developers are increasingly adopting a “freemium” approach: offering the core software weights at zero cost to encourage widespread adoption and experimentation, while imposing steep fees on heavy commercial use or enterprise-grade deployments. This strategy allows companies to build massive developer communities while still establishing a path to profitability.
“You pay for collaboration with these open-weight model labs to make sure that you’re optimizing your deployment. You pay for getting early access for the next revision of the model,” Paddy Srinivasan, CEO of cloud computing firm DigitalOcean Holdings, told Reuters. Srinivasan confirmed that his company, which hosts Kimi K3 and other Chinese models for U.S. developers, has already established a formal commercial agreement with Moonshot. Dan Fu, vice president of kernels at Together AI, noted that the real value for commercial partners lies at the application layer, where companies optimize how models utilize tokens to perform specific, useful tasks.
(Related: Moonshot AI Targets $50 Billion Valuation in Final Pre-IPO Round as Hong Kong Listing Nears)
Implications for the Global AI Race
The shift toward revenue-sharing agreements comes as Chinese AI firms aggressively push to capture global market share from U.S. rivals like OpenAI and Anthropic. Open-source models have been a critical weapon in this competition, allowing Chinese developers to bypass export controls and distribute their technology directly to international developers. By offering models that are nearly as capable as American closed-source systems, but at a fraction of the cost, companies like Alibaba and Moonshot have rapidly built a global user base.
However, the introduction of mandatory revenue-sharing agreements could alter the economic calculus for international developers considering adopting Chinese models. While Kimi K3 currently costs about one-third of Anthropic’s Fable model based on input and output token pricing, a steep revenue-sharing requirement could erode that cost advantage for highly successful applications. As U.S. startups like Thinking Machines Lab, founded by former OpenAI executive Mira Murati, begin releasing their own powerful open-source models, Alibaba and its domestic peers will have to carefully balance their desire for immediate monetization against the risk of driving developers toward American alternatives.
Timing and the Broader Monetization Shift
The timing of Alibaba’s move is significant. Qwen3.8-Max is expected to be released during the week of August 10, meaning the revenue-sharing terms will be in place from the model’s first day of public availability. This contrasts with Kimi K3, where the licensing terms were embedded in the release documentation after the model had already gained substantial traction. By announcing the commercial terms before or at launch, Alibaba is establishing clear expectations for enterprise users from the outset, reducing the risk of a backlash from developers who feel the rules were changed after they had already built applications on the model.
The broader context is one of an industry-wide reckoning with the economics of open-source AI development. Training a frontier model with 2.4 trillion parameters requires enormous capital investment, and the traditional open-source model of releasing software for free and monetizing through cloud services has proven insufficient to fund the next generation of development.
By introducing revenue-sharing requirements, Alibaba and Moonshot are effectively arguing that the most commercially successful applications built on their models should contribute to funding the research that made those applications possible. Whether the global developer community accepts this argument will determine how quickly the Chinese AI industry can sustain its current pace of model development.