To reduce high AI operating costs, Microsoft is currently evaluating the introduction of the Kimi K3 model developed by the Chinese AI company Moonshot AI into its Copilot service. Previously, this service mainly relied on technologies from OpenAI and Anthropic, and the addition of the new model is expected to reshape its underlying architecture.

Outstanding performance with cost-saving advantages, potentially saving up to $600 million annually

As the largest open-source model in terms of parameters globally, Kimi K3 has demonstrated excellent performance comparable to top-tier models in multiple authoritative evaluations. Internal estimates from Microsoft show that switching to this model could reduce the cost per token for inference by up to 60%, saving the company approximately $602 million annually.

Substantial profits collide with political risks, policy red lines become the biggest concern

However, this highly commercially attractive decision is facing serious external political considerations and policy risks. The current U.S. government is planning to further restrict domestic companies from using Chinese AI models, and may even implement targeted bans.

If Microsoft deeply integrates Chinese open-source technology into its flagship product, it could face strong dissatisfaction and scrutiny from the White House. Faced with significant economic benefits and a complex political situation, Microsoft is still in a cautious evaluation and decision-making phase.