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

Open-weight AI models can now achieve significant revenue, even with lower margins than proprietary models.

Moonshot AI, the company behind the Kimi model, is reportedly targeting $2 billion in annualized revenue by the end of 2024. This ambitious goal represents a doubling of its previously reported revenue run rate from August. The target is fueled by the success of its K3 model, which, despite a slight recent dip in usage, continues to process approximately 300 billion tokens daily according to OpenRouter data. Although its projected revenue is significantly lower than giants like OpenAI ($40 billion) and Anthropic ($65 billion), Moonshot AI's open-weight model strategy demonstrates potential profitability in the open-source AI space. However, the company faces controversy, with Anthropic alleging that Moonshot AI engaged in model distillation by routing nearly 300,000 user requests to Anthropic's Claude Opus model and collecting over 23 million responses for its own training data.

Moonshot AI's aggressive revenue target signals a growing market for open-weight AI models, challenging the dominance of closed-weight frontier models. Even with a business model that allows for lower margins due to freely available model weights, the company's projections indicate substantial financial viability. This success suggests that developers and businesses can find lucrative opportunities by building on or contributing to open-source AI, potentially fostering wider adoption and innovation in the field. It also puts pressure on larger, closed-model providers to justify their premium pricing and integrated ecosystems.

The controversy surrounding Anthropic's allegations of model distillation raises significant questions about the ethical boundaries and legality of training AI models in the open-weight ecosystem. If proven, Moonshot AI's practices could lead to increased scrutiny from competitors and regulators, potentially impacting how open-weight models are developed and commercialized. This situation highlights the ongoing tension between rapid innovation and intellectual property protection in the AI landscape, and future developments will determine the acceptable norms for data acquisition and model training in this rapidly evolving sector.