Moonshot AI, a leading artificial intelligence lab in China, has set an ambitious target to reach $2 billion in annualized revenue by the end of this year. This goal represents a doubling of the company’s revenue run rate reported in August and highlights the commercial momentum behind its K3 open-weight AI model, which launched this summer. Although usage of the K3 model has slightly declined recently, data from OpenRouter indicates it still processes up to 300 billion tokens daily. Despite this growth, Moonshot’s revenue remains significantly lower than major competitors like OpenAI and Anthropic, whose revenues are estimated at $40 billion and $65 billion respectively. The difference is partly due to Moonshot’s open-weight model approach, which offers freely available model weights and results in lower profit margins compared to closed-weight models. However, Moonshot’s development methods have drawn criticism and legal scrutiny. Anthropic recently accused Moonshot of conducting a prolonged model distillation operation that funneled nearly 300,000 requests from Moonshot’s Kimi model directly to Anthropic’s Claude Opus model. Anthropic claims that over 23 million responses were harvested from its models without authorization to train Moonshot’s AI, raising questions about the legality and ethics of Moonshot’s training practices. This controversy underscores the challenges and competitive tensions in the AI industry, especially between open and closed model strategies.