Loading...
Why Moonshot AI is in talks with Amazon, Google, Microsoft
Moonshot AI is preparing for an IPO

Why Moonshot AI is in talks with Amazon, Google, Microsoft

Aug 26, 2026
04:37 pm

What's the story

China's Moonshot AI is in talks with tech giants Microsoft, Amazon, and Google's parent company Alphabet for revenue-sharing deals. The agreements would allow these US cloud companies to host Moonshot's popular Kimi K3 model. If successful, this would be the first major revenue-sharing agreement between a Chinese artificial intelligence (AI) firm and a leading US cloud provider.

Revenue negotiations

Moonshot seeks up to 30% revenue share

Moonshot AI, which is preparing for an IPO, is seeking up to a 30% share of the revenue generated from K3-related services on Microsoft's Azure, Amazon Web Services (AWS), and Google Cloud.

The talks are still in their early stages, and there's no guarantee they will lead to agreements.

This comes despite US national security concerns that have banned AI chip exports to China.

Controversies and accusations

US officials accuse Moonshot of stealing from rival AI firm

Moonshot AI has been criticized by US Treasury Secretary Scott Bessent, who hinted at potentially adding the company to a trade blacklist.

US officials have accused the Beijing-based firm of stealing from Anthropic's advanced model Fable to develop Kimi K3 and illegally obtaining NVIDIA chips.

However, Moonshot denies these allegations, claiming that improvements in Kimi K3's performance are due to original changes in its underlying architecture.

ADVERTISEMENT

Model evaluation

Kimi K3 has performed well in 3rd-party evaluations

Kimi K3 has performed well in third-party evaluations.

It topped a benchmark by Arena.ai for web interface-building capabilities and is said to perform on par with OpenAI's GPT-5.5 and Anthropic's Claude Opus 4.8, especially in complex tasks.

Despite being an open-weight model that can be downloaded and modified, analysts believe few customers would run such a system with 2.8 trillion parameters due to high computing costs involved.

ADVERTISEMENT