China's Moonshot AI Says It Can Hit $2B in Revenue. Here's the Catch

The Kimi maker is targeting $2 billion in annualized revenue — proof open-weight models can make money, and a glimpse of the gray area that made it controversial.

Published: 2026-09-14 Category: Quick Take Sources: TechCrunch

The Claim

Moonshot AI, one of China's most prominent AI labs and the maker of the popular Kimi models, is telling Bloomberg it targets $2 billion in annualized revenue by year-end — double its reported August run rate. The bullishness tracks the success of its K3 open-weight model, released this summer. OpenRouter data reportedly shows K3 models generating up to 300 billion tokens a day on the platform, even as usage has dipped slightly in recent months.

The headline is arresting: an open-weight lab projecting serious money. But context matters. OpenAI's revenue is put at $40 billion and Anthropic's at $65 billion. Moonshot's $2 billion, while impressive for a lab whose weights are freely downloadable, is a rounding error next to the closed-weight frontier.

The Open-Weight Economics

This is the crux of the story. Because Moonshot's model weights are freely available, it has far lower margins than closed-weight competitors like OpenAI and Anthropic. Anyone can run K3 themselves — which is precisely what undermines pricing power. Moonshot is betting it can still monetize through hosted API, enterprise tools, and its consumer app, even when the core model is free.

If Moonshot hits $2 billion, it's a notable data point for the open-weight camp: yes, you can build a real business on top of freely distributed weights, particularly if you control the infrastructure, the data, and the distribution layer around the model. It undercuts the assumption that only closed frontier models are commercially viable.

The Catch: The Distillation Allegations

The uncomfortable counterweight is that Moonshot's road to a strong model is itself disputed. Earlier this week, Anthropic accused Moonshot of a long-running model distillation campaign — routing nearly 300,000 requests from Kimi directly to Claude Opus, effectively serving Opus in place of Kimi's own models, and collecting over 23 million responses from Anthropic models for training. Anthropic called it "distillation" that borders on illegal.

This matters for the revenue story in two ways. First, if Moonshot's model quality in part reflects its rivals' frontier models, then its commercial position is built partly on borrowing from someone else's expensive work — a fragile foundation. Second, it foreshadows a legal and licensing battleground over what "distillation" of a competitor's model means, which could redefine how open-weight labs train and what they owe the labs that trained the models they learn from.

The takeaway: Moonshot's $2 billion target is real news for open-weight economics, but it arrives wrapped in the very controversy — appropriation of rival labor — that could cap how far that model goes. Growth built on borrowed frontier capability is growth with a legal overhang.

Sourced and summarized from TechCrunch's reporting, Bloomberg's revenue figures, and Anthropic's distillation allegations.