Moonshot AI is trying to put its Kimi K3 model in front of more enterprise and developer users through the biggest US cloud platforms. According to Reuters, the Chinese AI startup is discussing revenue-sharing agreements with Microsoft, Amazon, and Alphabet's Google.
If completed, the arrangements would allow Kimi K3 to be hosted on Azure, AWS, and Google Cloud. The talks are still early, and the source article makes clear that a final deal is far from certain.
What Moonshot AI is asking for
The central business point is revenue sharing. Moonshot is asking for up to 30 percent of revenue from K3 services, according to three people familiar with the discussions cited by Reuters.
That structure would give Moonshot a direct financial stake in how Kimi K3 performs on the cloud platforms. Instead of simply making the model available, the company would take a cut tied to customer use of the service.
The potential partners are the largest names in US cloud infrastructure: Microsoft, Amazon, and Alphabet's Google. The source article identifies their relevant platforms as Azure, AWS, and Google Cloud.
For Moonshot AI, access to those platforms would matter because cloud marketplaces are where many companies already manage computing, AI services, and developer tools. For the US providers, hosting Kimi K3 would expand the range of AI models available to customers, if the talks result in actual agreements.
Why these talks stand out
The source article says no Chinese AI company has previously reached this kind of deal with a major US cloud provider. That makes the negotiations notable even before any agreement is signed.
The reported talks sit at the intersection of model distribution, cloud computing, and AI revenue models. A model hosted on Azure, AWS, or Google Cloud can be packaged as a service, measured through usage, and sold through existing cloud relationships.
That also makes the details important. The talks are not only about whether Kimi K3 appears on a platform. They are about how money is divided, how usage is counted, and what access each side has to data.
Those details can shape whether a deal works commercially and operationally. A revenue split affects incentives. Token usage tracking affects billing. Data access affects how each party can run, monitor, and support the service.
The unresolved questions
Reuters reports that several open questions remain. The main issues named in the source are the revenue split, data access, and how token usage gets tracked.
- Revenue split: Moonshot is seeking up to 30 percent of revenue from K3 services, but the final terms are not settled.
- Data access: The talks still need to address what data can be accessed and under what terms.
- Token usage tracking: The parties need a way to measure usage accurately for services built around Kimi K3.
These are not minor implementation details. In AI cloud services, usage measurement and billing are closely linked. If token tracking is unclear, the commercial model becomes harder to operate.
Data access is also a central issue because hosted AI services depend on how requests, outputs, and service information are handled. The source article does not give the terms being discussed, so the only clear point is that this remains unresolved.
Political and legal pressure around Moonshot
The talks are taking place while Moonshot AI faces pressure in the United States. US Treasury Secretary Scott Bessent recently threatened Moonshot with a trade ban, according to the source article.
US authorities have also accused the company of stealing Anthropic's model Fable. Moonshot denies the allegations.
Those issues add uncertainty around any potential cloud hosting agreement. The source article does not say how the allegations or the threatened trade ban are affecting the talks, only that they are part of the current context.
Microsoft, Google, and AWS declined to comment to Reuters. That leaves the discussions in a preliminary stage, with the main facts coming from people familiar with the matter.
What happens next
The immediate question is whether the early-stage negotiations can become formal agreements. For now, there is no completed deal, no confirmed revenue split, and no public comment from the US cloud providers.
If Moonshot AI reaches terms with Microsoft, Amazon, or Google, Kimi K3 could become available through one or more major US cloud platforms. If the talks fail, the reported negotiations will still show how AI model companies are exploring new ways to reach customers through cloud infrastructure.
The important point is that the outcome remains open. Moonshot AI is seeking a path onto Azure, AWS, and Google Cloud, but the commercial, technical, and regulatory questions have not yet been resolved.