Moonshot AI targets $2 billion annual revenue on the strength of open-weight Kimi models

Moonshot AI, the Chinese lab behind the Kimi model family, is aiming to reach $2 billion in annualized revenue by the end of the year, according to a Bloomberg report cited by TechCrunch. The goal would double the revenue run rate the company recorded in August, and it follows a strong stretch for the lab's K3 model since its release earlier this summer.

An aggressive goal for an open-weight lab

The target stands out because Moonshot's flagship models are open weight, meaning the underlying parameters can be downloaded and run by anyone. That openness has fueled adoption, but as TechCrunch notes, it also leaves the company with far slimmer margins than its closed-weight competitors, since it cannot gate access to the models themselves.

Adoption remains substantial even as it cools. K3's usage figures have slipped somewhat in recent months, but OpenRouter data currently shows K3 models generating as many as 300 billion tokens per day on the platform.

Still a fraction of US lab revenue

For perspective, recent reports put OpenAI's revenue at $40 billion and Anthropic's at $65 billion, leaving Moonshot's projection far behind the leading US labs. Even so, a company that gives away its weights doubling its run rate within a matter of months is a signal that open-weight AI can support a genuine business, even if it is less lucrative than closed frontier models.

Anthropic's distillation allegations

The revenue news lands in the middle of an escalating dispute. Earlier this week, Anthropic accused Moonshot of a sustained distillation operation in which nearly 300,000 requests from Kimi were routed directly to Claude Opus, so that users were in effect receiving Opus responses rather than output from Moonshot's own models. In total, Anthropic alleges that more than 23 million responses from its models were harvested for use in Moonshot's training data. TechCrunch is blunt in its assessment, describing the alleged practices as contentious and possibly unlawful.

Why it matters

Moonshot's $2 billion goal is one of the clearest data points yet that open-weight AI can be monetized at scale. The argument for freely available models has usually rested on ecosystem building rather than direct income; a lab doubling its run rate on the back of K3 suggests that API and enterprise demand around open models can translate into serious revenue, even at lower margins.

The Anthropic allegations, meanwhile, expose the fault lines beneath that growth. If output from closed frontier models is quietly feeding open-weight training pipelines, the relationship between open and closed labs becomes more contentious and legally murky. How this dispute resolves could shape what open-weight labs are permitted to train on, and how durable their momentum proves to be.

The comparison with OpenAI and Anthropic also puts the number in context. Two billion dollars would be a milestone for the open-weight camp, but it still leaves Moonshot well behind the US frontier labs it is most often measured against.