Beijing-based AI company Moonshot is in preliminary talks with Microsoft, Amazon, and Alphabet's Google to host its Kimi K3 model on Azure, Amazon Web Services, and Google Cloud, according to three sources cited by Reuters. The Chinese startup is seeking up to 30% of revenue generated from K3-related services on the three platforms, terms similar to arrangements Moonshot has discussed with other major customers. If completed, the deal would become the first significant revenue-sharing agreement between a Chinese AI developer and a major US cloud provider.
The proposed revenue split would give Moonshot a substantial share of the economics while leaving cloud providers responsible for infrastructure and customer distribution. However, the companies haven't resolved critical operational details, including how revenue would be divided, how Moonshot would access data, and how token usage would be audited—details that matter because token consumption drives usage-based AI billing. Kimi K3 is an open-weight model with 2.8 trillion parameters, allowing customers to download and modify its weights, but operating a model of that scale at enterprise level requires substantial computing capacity, making major cloud platforms crucial for wider adoption. Moonshot has already signed revenue-sharing deals with smaller cloud providers, and Chinese IT services firm Chinasoft International disclosed a similar arrangement in July without revealing financial terms.
The negotiations face significant political obstacles. US Treasury Secretary Scott Bessent said last month he might add Moonshot to a trade blacklist, and US officials have accused the company of using outputs from Anthropic's Fable model to help develop K3 through distillation and of illegally acquiring Nvidia chips. Moonshot has rejected the distillation allegations, telling China's National Business Daily that K3's gains came from original changes to its underlying architecture. Independent evaluations have shown strong performance: Arena.ai ranked K3 first for 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 demanding, multi-step tasks.
For Moonshot, access to Azure, AWS, and Google Cloud would transform K3 from an open model that customers can download into a service that's far easier for businesses to deploy. The cloud companies face a difficult calculation: K3 could give their enterprise customers another capable and potentially lower-cost model, but a partnership with Moonshot could attract additional scrutiny from Washington. Data access, usage verification, and US policy could ultimately determine whether the agreements happen at all. Moonshot raised more than $2 billion in May and is preparing for a potential Hong Kong listing. A successful US cloud rollout could strengthen K3's position as a global commercial platform while giving channel partners another model to deploy and support, though partners would still need to weigh data governance, contract continuity, and the risk that future US restrictions could disrupt customer deployments. The talks highlight how geopolitical tensions increasingly shape the infrastructure choices available to enterprises, forcing cloud providers to balance commercial opportunity against regulatory exposure in ways that could reshape which AI capabilities reach which markets.