Kimi-maker Moonshot AI targets $2B in annual revenue

Moonshot AI, the Chinese artificial intelligence company behind Kimi, is pursuing a sharp expansion in commercial income after the summer release of its K3 model. Bloomberg reported Friday that the company aims to reach an annualized revenue rate of $2 billion by the close of the year. The objective would put one of China’s most visible AI laboratories under pressure to translate model attention into durable customer spending in a matter of months.

The target is twice the revenue pace Moonshot AI was reported to have recorded in August. Annualized revenue is a measure that projects a current level of sales over a full year; it is not the same as revenue already booked for the year. That distinction matters because the reported goal depends on the company substantially increasing its run rate before December rather than simply maintaining the level it had reached late in the summer.

At the center of that calculation is K3, an open-weight model that has become a significant part of Moonshot’s commercial pitch. Open-weight releases can be adopted, adapted, hosted or evaluated by outside developers and businesses more easily than fully closed systems, depending on the accompanying terms. For Moonshot, the model’s reception has created an opportunity to turn technical reach into services, usage and other forms of business demand.

The company’s ambition follows a period in which K3 generated considerable interest after arriving this summer. Its early momentum is a key reason the revenue goal is being viewed as plausible enough to pursue, even if it remains demanding. A fast-moving model launch can raise a laboratory’s profile and broaden access to potential customers, but an initial surge does not by itself establish a recurring sales base.

Usage data from OpenRouter illustrates both the scale of K3’s reach and the uncertainty surrounding it. The platform has recently recorded up to 300 billion tokens per day produced by K3 models. Tokens are the small units of text and other data processed by language models, so the figure points to a very large volume of activity. It does not, however, directly reveal how much of that activity becomes revenue for Moonshot.

There is also a counterweight to the headline number: K3’s usage has edged down in recent months. The decline appears modest, but it is relevant when a company is seeking to double its reported revenue pace in a short window. High token totals can demonstrate developer interest and infrastructure demand, while changes in the trend can signal that a model must keep proving its appeal as users compare alternatives or shift workloads.

Moonshot’s plan highlights a broader challenge for AI developers: popularity and monetization are related but not interchangeable. An open-weight system may circulate widely among researchers, startups and software builders, yet the model’s creator still needs a clear path from adoption to paid business. The coming months will test whether the attention around K3 can support the kind of repeatable commercial activity implied by a $2 billion annualized target.

The report arrives as AI companies compete not only on model performance but also on distribution, developer usage and the ability to make products economically useful. That competition was also reflected in promotion for Disrupt 2026, an event set to feature OpenAI, Anthropic, Replit and other companies across six industry-focused stages, with tickets advertised at a 25% discount. The breadth of that gathering underscores how many firms are seeking to convert AI interest into business traction.

For Moonshot, the next evidence will be whether K3’s activity stabilizes or expands and whether the company reports a faster revenue run rate as year-end approaches. The announced benchmark is ambitious precisely because it would require a major acceleration from August. If achieved, it would mark a significant commercial step for the Kimi maker; if not, the gap between substantial model use and sustained sales will remain the central question.

Source: TechCrunch