AI

Moonshot AI Seeks 30% Revenue Share From Microsoft, Amazon, Google

Chinese startup negotiates unprecedented cloud distribution deals for its Kimi K3 model amid U.S. trade tensions.

Omega Editorial· August 26, 2026· 3 min read

Chinese AI Startup Pursues Major Cloud Partnerships

Moonshot AI is negotiating with Microsoft, Amazon, and Google to distribute its Kimi K3 large language model through their cloud platforms, according to Reuters, which cited three sources familiar with the discussions. The Chinese startup is requesting up to 30% of revenue generated from services tied to K3 on Azure, Amazon Web Services, and Google Cloud.

The talks remain at an early stage and may not produce final agreements, Reuters reported. Moonshot did not respond to requests for comment, while Microsoft, Google, and AWS all declined to comment.

Why It Matters

If completed, these agreements would represent the first major revenue-sharing arrangement between a Chinese AI company and leading U.S. cloud providers. The deals would give Moonshot access to enterprise customers globally while offering cloud platforms a competitive alternative to Western models. However, the negotiations unfold against a backdrop of escalating technology restrictions and national security concerns that could complicate or derail any partnership.

Outstanding Deal Terms

Several technical and commercial issues remain unresolved in the negotiations. The parties are still working through the precise revenue split, data access protocols, and mechanisms for tracking token consumption. Tokens represent discrete text units that models process, and tracking their usage forms the foundation of consumption-based pricing models.

Moonshot has already established similar arrangements with smaller platforms. Chinasoft International, a Chinese IT services company, announced a revenue-sharing agreement with Moonshot in July, though financial terms were not disclosed.

Model Performance and Scale

Kimi K3 has attracted attention in independent testing. Arena.ai ranked it first on a leaderboard for web interface construction, while Artificial Analysis found its performance comparable to OpenAI's GPT-5.5 and Anthropic's Claude Opus 4.8 on complex, multi-step reasoning tasks.

The model contains 2.8 trillion parameters, making it the largest publicly available open-weight model. While the weights are freely downloadable and modifiable, the computing resources required to operate a system of this scale make cloud distribution the practical path to enterprise adoption for most organizations.

Moonshot released Kimi K3's weights for unrestricted public download last month. The launch drove the company's daily revenue to at least six times its previous level and pushed shares of Chinese AI competitors lower. The company reported $300 million in annual recurring revenue in June, up from $200 million in April, and is pursuing new funding at a $50 billion valuation ahead of a potential Hong Kong IPO.

Regulatory Headwinds

The negotiations proceed amid heightened U.S. scrutiny of Moonshot. Treasury Secretary Scott Bessent indicated last month he might add the company to a trade blacklist. U.S. officials have accused Moonshot of training Kimi K3 using banned Nvidia chips and distilling capabilities from Anthropic's Fable model.

Moonshot has disputed the distillation allegations, telling China's National Business Daily that the model's capabilities stem from novel architectural innovations rather than copying from other systems.

Reuters first reported the details of these negotiations.

#moonshot ai#kimi k3#cloud computing#revenue sharing#china ai#enterprise ai

This is an original analysis by the Omega editorial team. Source reporting: AI Watch.

Want systems like this working for your business?

Book a Call

More in AI

AI· 3 min read

Nvidia Q2 Earnings Test Whether AI Spending Extends Beyond Hyperscalers

Customer concentration and vendor-backed financing deals complicate the picture as the chipmaker reports results after market close.

Via AI Watch · Aug 26, 2026
AI· 4 min read

Nvidia Earnings Will Test Whether AI Boom Is Broadening or Fragile

Four metrics in the August 26 report will reveal demand health, customer concentration, and financing risks across the AI infrastructure buildout.

Via AI Watch · Aug 25, 2026
AI· 4 min read

MIT Calls for Structural Overhaul of Higher Education for AI Era

A new report from the Institute's Ad Hoc Committee recommends course-specific AI policies, redesigned assessments, and stronger emphasis on residential learning.

Via AI Watch · Aug 25, 2026