Moonshot AI, the Chinese lab behind the open-weight Kimi model family, is targeting $2 billion in annualized revenue by the end of 2026 — twice its August run rate and a dramatic increase fueled by the summer launch of its K3 model.
Since K3’s debut, usage metrics have softened a bit, but the OpenRouter platform still records as many as 300 billion tokens generated daily by K3 models. Analysts estimate K3 delivered over $1 billion in annual recurring revenue in August alone, up sharply from roughly $300 million in June.
Open-Weight vs Closed Ecosystems
Moonshot operates in the open-weight model space, releasing its weights publicly. That grants it lower profit margins than rivals like OpenAI and Anthropic, which keep their models proprietary. Still, Moonshot’s rising revenue suggests open-weight models, though traditionally less lucrative, can scale at commercial levels.
To put this into context: OpenAI is estimated to be pulling in around $40 billion annually, and Anthropic closer to $65 billion. These figures dwarf Moonshot’s goals — but Moonshot is carving out room in a lower-margin, higher-distribution model that could make open LLMs more competitive.
Distillation Controversy and Allegations
Moonshot is under fire from Anthropic over alleged model distillation activity. Anthropic claims Moonshot routed nearly 300,000 customer requests to its own Claude models during a 10-day span, which were then displayed to users pretending to be Kimi responses. Approximately 23 million exchanges from May through July are tied to this campaign.
An illicit distillation move involves using outputs from a more capable model (here, Claude), gathering chain-of-thought transcripts, and using that content to train a different model, often without disclosure or authorization.
This puts Moonshot in a contentious position: gaining traction and revenue while being accused of risky or unauthorized training practices.
What This Tells Us
Moonshot’s ambition to double its already high run rate is a bold indicator that open-weight models are being taken seriously by the market. There’s growing demand for LLMs that offer transparency and reuse of weights — but this model also exposes such companies to threats around intellectual property and distillation ethics. If Moonshot succeeds in reaching $2 billion, it could reshape how AI labs approach open-source release strategies.
On the flip side, the distillation allegations highlight growing tension in the AI ecosystem over training data, ownership, and ethical practices. Labs operating with open models are being scrutinized for how they benefit from adversarial techniques that enable capability transfer from proprietary models without licensing or oversight.
It’s worth watching how regulations evolve around model training disclosures, data pipelines, and cross-company accountability. Moonshot’s next moves will not just be about revenue — success will depend heavily on trust, transparency, and whether the broader AI community and regulators accept its methods.
