Moonshot AI, the Beijing-based startup behind the acclaimed Kimi K3 language model, is in advanced discussions with Microsoft, Amazon and Google about revenue-sharing arrangements that could see the American tech titans host the Chinese firm’s flagship product on their respective cloud platforms.
The talks, confirmed by three sources with direct knowledge of the matter, would represent the first significant revenue-sharing agreement between a leading Chinese artificial intelligence company and major U.S. cloud providers. Under the proposed terms, Moonshot would receive up to 30 per cent of all revenue generated through K3-related services delivered via Azure, Amazon Web Services and Google Cloud.
The negotiations underscore a remarkable shift in the global AI landscape. Chinese models, frequently available at a fraction of the cost of Western alternatives, are increasingly attractive to American enterprises despite ongoing national security concerns in Washington that have resulted in sweeping export restrictions on advanced AI chips destined for China.
Discussions at Critical Stage
According to those familiar with the proceedings, the current negotiations remain at an early phase with no certainty of reaching final agreements. Key sticking points include how revenue would be divided between the parties, arrangements for data access, and mechanisms for auditing token usage.
Tokens serve as the fundamental units of text processed by AI systems, and tracking their consumption is central to calculating invoices under usage-based billing models that have become standard across the industry.
Moonshot has pursued comparable arrangements with several smaller cloud platforms, the sources indicated, though specific details of those deals were not disclosed. In July, Chinese IT services provider Chinasoft International announced a revenue-sharing partnership with Moonshot, declining to reveal financial terms.
The Kimi K3 model, which operates with 2.8 trillion parameters, has demonstrated impressive capabilities in independent assessments. Arena.ai ranked it first in benchmarks evaluating web interface-building proficiency, while analysis from Artificial Analysis suggests performance levels comparable to OpenAI’s GPT-5.5 and Anthropic’s Claude Opus 4.8, particularly on tests measuring complex, multi-step reasoning tasks.
Enterprise adoption through major cloud providers remains the primary pathway for AI model developers seeking to scale their customer base. Despite K3 being an open-weight model that organisations can download and modify, analysts note that few businesses possess the infrastructure necessary to operate such a computationally demanding system independently.
Political Headwinds
The timing of these negotiations proves complicated given sharp criticism from senior U.S. officials. Treasury Secretary Scott Bessent suggested last month that Moonshot could face addition to a trade blacklist, a prospect that would severely restrict the company’s ability to conduct business with American entities.
U.S. authorities have accused the Beijing-headquartered firm of improperly obtaining technology from Anthropic’s sophisticated Fable model to accelerate development of Kimi K3. Authorities have also alleged that Moonshot illegally secured Nvidia chips in violation of export controls.
Moonshot has forcefully rejected claims that its model achieved performance improvements through distillation techniques, telling China’s National Business Daily that gains stemmed from original modifications to underlying architecture.
IPO Ambitions
Founded in 2023 by Carnegie Mellon-educated AI researcher Yang Zhilin, Moonshot counts Alibaba among its principal backers. The startup secured more than $2 billion in funding during May and is preparing for a potential listing on the Hong Kong Stock Exchange, according to sources familiar with the matter.
Alibaba itself is separately pursuing revenue-sharing agreements with major users of its newly released open-source AI model, those sources added.
Microsoft, Google and AWS declined to comment on the discussions. Moonshot did not respond to requests for clarification regarding the potential partnerships.
Why it Matters
The potential revenue-sharing agreements between Moonshot AI and America’s dominant cloud providers would mark a pivotal moment in the escalating competition between Chinese and Western artificial intelligence developers. If concluded, such arrangements would demonstrate that despite mounting political tensions and export controls designed to limit China’s AI advancement, commercial incentives remain powerful enough to bridge the divide. For American enterprises, access to cutting-edge Chinese AI capabilities at competitive price points could accelerate adoption across sectors ranging from software development to financial services. Simultaneously, the deals would provide Moonshot with vital distribution infrastructure and revenue streams as it navigates the path toward a public listing, while raising fresh questions about the effectiveness of Washington’s strategy to contain China’s AI ambitions through technological restrictions.
