China-based AI startup Moonshot AI is in early-stage negotiations with Microsoft, Amazon and Google over revenue-sharing agreements that would allow the U.S. cloud giants to host its Kimi K3 artificial intelligence model, according to three people familiar with the talks. The discussions could lead to the first major revenue-sharing arrangement between a Chinese AI company and a major U.S. cloud provider, creating an unusual commercial link between China’s rapidly advancing AI sector and America’s biggest cloud platforms.
Moonshot is reportedly seeking up to a 30% share of revenue generated from K3-related services on Microsoft Azure, Amazon Web Services (AWS) and Google Cloud. The negotiations are still at an early stage and may not result in agreements. Issues including the revenue split, access to data and mechanisms for auditing token usage remain unresolved. The talks are taking place despite heightened scrutiny of Chinese AI companies in Washington over technology, national-security and semiconductor concerns.
Moonshot AI Seeks Revenue Share From U.S. Cloud Platforms
The proposed arrangements would allow Microsoft, Amazon and Google to offer Kimi K3 through their cloud infrastructure, giving enterprise customers access to the model without having to build and operate the necessary computing infrastructure themselves.
Moonshot is seeking as much as 30% of revenue generated by K3-related services hosted through the three cloud platforms, according to Reuters’ sources. The terms are reportedly similar to those Moonshot has outlined for major customers using the model.
Kimi K3 Revenue-Sharing Talks
| Particular | Details |
|---|---|
| AI company | Moonshot AI |
| Model | Kimi K3 |
| Potential cloud partners | Microsoft, Amazon, Google |
| Platforms | Azure, AWS, Google Cloud |
| Moonshot’s reported revenue share request | Up to 30% |
| Stage of negotiations | Early |
| Final agreement | Not certain |
| Key unresolved issues | Revenue split, data access, token auditing |
The commercial structure would potentially give Moonshot access to the customer bases of three of the world’s largest cloud companies while giving the cloud providers another competitive AI model to offer enterprise customers.
Why Kimi K3 Needs Cloud Infrastructure
Kimi K3 is an open-weight model with approximately 2.8 trillion parameters, according to Reuters. While open-weight models can be downloaded and modified, running a model of this scale requires substantial computing resources.
That makes cloud infrastructure particularly important.
Instead of customers purchasing and operating their own large-scale AI infrastructure, cloud providers can host the model and charge customers for computing and AI services.
Kimi K3 Commercial Model
Moonshot AI
↓
Kimi K3 Model
↓
Microsoft Azure / AWS / Google Cloud
↓
Enterprise Customers
↓
AI Inference + Cloud Usage
↓
Revenue Generated
↓
Revenue Share Between Moonshot + Cloud ProviderFor Moonshot, this could provide a faster path to enterprise distribution than building its own global cloud infrastructure.
For cloud providers, hosting a high-performing third-party model could increase demand for their computing, storage and networking services.
A Potential First For Chinese AI And U.S. Cloud
If completed, the agreements could represent a significant milestone in the international commercialization of Chinese AI models.
Reuters said any deal could mark the first major revenue-sharing pact between a Chinese AI firm and a major U.S. cloud company.
That would be notable because the U.S. and China remain locked in a broader technology competition, particularly around advanced semiconductors and AI computing.
Why The Potential Deals Matter
| Area | Potential Impact |
|---|---|
| Moonshot AI | Access to global enterprise distribution |
| Microsoft | Additional AI model for Azure customers |
| Amazon | More AI offerings through AWS |
| Expanded model choice on Google Cloud | |
| Enterprise customers | Access to another advanced model |
| AI industry | Greater competition among model providers |
| U.S.-China technology ties | New commercial interaction despite tensions |
The talks demonstrate that commercial demand for AI models can create relationships that operate alongside broader geopolitical tensions.
Kimi K3 Is Gaining Attention For Its Performance
The potential cloud partnerships come after Kimi K3 attracted attention for its performance on several third-party evaluations.
According to Reuters, Arena.ai ranked K3 first on a benchmark measuring web-interface-building capabilities, while Artificial Analysis found its performance comparable with OpenAI’s GPT-5.5 and Anthropic’s Claude Opus 4.8 on tests involving complex, multi-step tasks.
These comparisons are significant because they suggest that Chinese AI companies are increasingly capable of producing models that compete with leading U.S. systems on selected benchmarks.
Kimi K3 Positioning
| Factor | Kimi K3 |
|---|---|
| Parameters | ~2.8 trillion |
| Model type | Open-weight |
| Web-interface benchmark | Ranked first by Arena.ai |
| Complex-task performance | Comparable with leading Western models on some tests |
| Target market | Developers and enterprises |
| Distribution opportunity | Major cloud platforms |
The model’s open-weight structure can also make it attractive to organizations seeking greater control over deployment and customization.
Cost Is Another Advantage For Chinese AI Models
The negotiations also reflect a broader trend in the AI market: Chinese models are increasingly competing on cost as well as performance.
Reuters noted that leading Chinese AI models are often offered at significantly lower prices than comparable Western offerings. This can make them attractive to businesses looking for cheaper alternatives for inference and other AI workloads.
For cloud providers, hosting lower-cost models can also create additional consumption of cloud computing resources.
The economics therefore extend beyond the model itself.
AI Cloud Economics
Lower Model Cost
↓
More Enterprise Adoption
↓
Higher AI Inference Volume
↓
Greater Cloud Computing Consumption
↓
Higher Cloud Revenue
↓
Revenue Share With Model DeveloperThis model could make third-party AI models commercially valuable to cloud providers even when the cloud companies did not develop the underlying model.
Revenue Split Is Still A Major Sticking Point
The reported 30% revenue-share request is not yet an agreed commercial term.
According to Reuters, the parties are still discussing how revenue should be divided, along with data access and auditing of token usage.
Token measurement is particularly important because many AI services are priced according to the amount of text processed by a model.
If the number of tokens consumed determines the customer’s bill, the parties need a reliable method to determine usage and calculate the corresponding revenue share.
Key Negotiation Issues
| Issue | Why It Matters |
|---|---|
| Revenue split | Determines economics for Moonshot and cloud providers |
| Token measurement | Establishes billable AI usage |
| Data access | Determines what usage information Moonshot can see |
| Usage auditing | Ensures accurate revenue calculations |
| Pricing | Affects customer adoption |
| Cloud costs | Determines final profitability |
| Data governance | Important for enterprise customers |
The resolution of these issues will determine whether the talks can become commercial agreements.
Moonshot Already Has Smaller Cloud Partnerships
Moonshot is not starting from zero.
Reuters reported that the company has already signed similar revenue-sharing agreements with smaller cloud platforms, although details of the arrangements have not been disclosed.
In July, Chinese IT services company Chinasoft International disclosed that it had entered into a revenue-sharing agreement with Moonshot. The company did not reveal the percentage split.
These arrangements could provide Moonshot with experience in structuring commercial relationships around an open-weight model.
The proposed agreements with Microsoft, AWS and Google would represent a much larger step because of the scale and international reach of those platforms.
Cloud Giants Have Strong Incentives To Host More AI Models
Microsoft, Amazon and Google have each invested heavily in AI infrastructure and are competing to become the primary cloud platform for enterprise AI.
Offering multiple models can strengthen their positions because customers increasingly want flexibility rather than being locked into one AI model.
Why Hyperscalers May Want K3
- Adds another high-performing model
- Broadens AI choice for customers
- Generates inference workloads
- Drives demand for cloud infrastructure
- Strengthens AI marketplace offerings
- Creates additional enterprise engagement
- Reduces dependence on internally developed models
A third-party model can therefore complement rather than compete directly with a cloud company’s own AI systems.
U.S. Security Concerns Create A Major Risk
The potential partnerships face an unusual geopolitical complication.
Moonshot has come under criticism from U.S. officials. Treasury Secretary Scott Bessent said last month that he might add the company to a trade blacklist, according to Reuters. U.S. officials have also accused Moonshot of stealing from Anthropic’s advanced model Fable and illegally obtaining Nvidia chips.
Moonshot has rejected suggestions that Kimi K3’s performance was achieved through model distillation. The company told China’s National Business Daily that improvements came from changes to its underlying architecture.
These allegations remain disputed, and the negotiations do not mean that any U.S. government approval has been granted.
Geopolitical Risk Matrix
| Risk | Potential Effect |
|---|---|
| Trade restrictions | Could limit U.S. cloud availability |
| Export controls | Could constrain computing access |
| IP allegations | Could increase legal scrutiny |
| Data-security concerns | Could deter enterprise customers |
| Regulatory uncertainty | Could delay agreements |
| U.S.-China tensions | Could affect long-term partnerships |
The uncertainty could become one of the biggest obstacles to finalizing the proposed agreements.
Moonshot Is Preparing For A Potential Hong Kong IPO
The company is also approaching the public markets.
Moonshot was founded in 2023 by Yang Zhilin, a Carnegie Mellon-trained AI researcher, and is backed by several Chinese technology companies, including Alibaba.
The startup raised more than $2 billion in May 2026 and is preparing for a potential Hong Kong listing, according to sources cited by Reuters.
The ability to demonstrate international commercial adoption could be valuable as Moonshot prepares for a possible IPO.
Moonshot AI At A Glance
| Metric | Detail |
|---|---|
| Founded | 2023 |
| Founder | Yang Zhilin |
| Headquarters | Beijing, China |
| Key model | Kimi K3 |
| K3 parameters | ~2.8 trillion |
| Funding raised in May 2026 | More than $2 Bn |
| Strategic backer | Alibaba among others |
| IPO | Potential Hong Kong listing |
| U.S. cloud discussions | Microsoft, Amazon, Google |
A major cloud partnership could potentially strengthen Moonshot’s commercial credentials, although the talks are far from finalized.
Open-Weight Models Are Changing AI Distribution
The potential deal also highlights how open-weight AI models are changing the traditional model-distribution structure.
A closed AI model typically requires users to access it through the developer’s API or an approved platform. Open-weight models can instead be downloaded and deployed by customers.
But the largest models still require substantial infrastructure.
That creates a hybrid model in which the weights are openly available while cloud providers become the preferred route for enterprise deployment.
Open-Weight Model Distribution
Open-Weight Model
↓
Available For Download
↓
Most Enterprises Face High Compute Costs
↓
Cloud Hosting Becomes Attractive
↓
Hyperscaler Marketplace
↓
Enterprise DeploymentThis could become an important commercial model for future AI companies.
What The Talks Mean For Microsoft, Amazon And Google
If an agreement is reached, the three cloud providers could gain access to an AI model that is attracting attention for its performance and cost profile.
However, they would also need to evaluate regulatory and reputational risks associated with hosting a Chinese AI model.
For Moonshot, the trade-off is different.
The company would potentially surrender a portion of its revenue in exchange for distribution, infrastructure and enterprise access.
Potential Benefits And Risks
| Party | Potential Benefit | Main Risk |
|---|---|---|
| Moonshot | Global distribution | Revenue-share dilution |
| Microsoft | More Azure AI workloads | Regulatory exposure |
| Amazon | More AWS AI usage | Geopolitical risk |
| More Cloud AI offerings | Compliance concerns | |
| Enterprise users | Model choice | Data/security questions |
The commercial opportunity is therefore substantial, but the political and regulatory risks cannot be separated from the business case.
The Bigger Picture
Moonshot AI’s negotiations with Microsoft, Amazon and Google illustrate how quickly the global AI market is becoming interconnected despite U.S.-China technology tensions. A successful agreement would potentially give Kimi K3 access to three of the world’s largest cloud platforms while creating a new revenue stream for Moonshot. The company’s reported request for up to a 30% share also shows that model developers are increasingly seeking to monetize their intellectual property through cloud distribution rather than relying exclusively on direct customer relationships.
The bigger issue is whether commercial demand can overcome geopolitical and regulatory barriers. Kimi K3’s reported performance and low-cost positioning are creating interest, but U.S. officials have raised allegations concerning intellectual property and semiconductor procurement, while Washington continues to scrutinize Chinese AI companies. If Moonshot can secure agreements with major U.S. cloud providers despite those concerns, it would represent a significant milestone for China’s AI industry and demonstrate the growing importance of cloud platforms as the global distribution layer for AI models.
Looking Ahead
The immediate focus will be on whether Moonshot and the three cloud providers can resolve the commercial mechanics of the proposed partnerships. Revenue allocation, token-use auditing and data access remain key outstanding issues, while the companies will also need to assess regulatory and security considerations. Reuters reported that the negotiations are still preliminary and there is no guarantee that any agreement will be reached.
For Moonshot, access to Azure, AWS and Google Cloud could accelerate enterprise adoption of Kimi K3 and strengthen its position ahead of a potential Hong Kong IPO. For Microsoft, Amazon and Google, hosting the model could increase cloud consumption and expand their AI offerings, but the potential commercial gains would have to be weighed against geopolitical and compliance risks. The outcome could provide an early test of whether Chinese AI models can build significant commercial distribution through U.S. cloud infrastructure
Get the day’s top stories in your inbox
One concise email. No spam, unsubscribe anytime.
