Moonshot AI Sets $2 Billion Revenue Goal as Kimi Usage Surges and Scrutiny Grows

Moonshot AI is aiming to generate $2 billion in annualized revenue by the end of 2026, a bold target that highlights how quickly demand for its Kimi model has translated into commercial traction. The Chinese AI lab is pursuing the milestone even as it faces fresh allegations from Anthropic over model distillation and the ethics of how its system was trained.

The revenue goal, first reported by Bloomberg, would mark a sharp step up from Moonshot’s estimated August run rate and underscores the growing market for open-weight AI systems. But the company’s rise is unfolding alongside escalating concerns about whether some of the performance behind Kimi came from using rival model outputs in ways that may have crossed legal and competitive lines.

Why Moonshot’s revenue target matters

Moonshot’s revenue ambition is significant because it suggests that open-weight AI models can support a real business at scale, even if they do not command the same margins as closed systems from OpenAI or Anthropic.

For much of the AI boom, the industry narrative has centered on giant proprietary models sold through premium subscriptions and enterprise contracts. Moonshot’s trajectory points to a different path: releasing model weights more broadly while still capturing usage, services, and distribution value from one of the world’s most talked-about AI assistants.

That matters not only for Moonshot but for the broader AI market. If the company can approach $2 billion in annualized revenue, it would offer one of the clearest examples yet that open-weight models can become meaningful revenue engines in their own right.

How does Moonshot compare with OpenAI and Anthropic?

Moonshot is still far smaller than the leading U.S. AI companies, but its growth rate is eye-catching. Recent reports have pegged OpenAI’s annual revenue at around $40 billion and Anthropic’s at roughly $65 billion, figures that reflect the enormous commercial appetite for closed-weight frontier models.

Moonshot’s model is structurally different. Because Kimi’s weights are available more freely than those of proprietary rivals, the company likely earns less per user and faces more limited monetization power. Even so, the company’s targets suggest there is room for strong revenues outside the dominant closed model playbook.

What is driving Kimi’s growth?

Kimi’s recent popularity is the main reason Moonshot believes its revenue can climb so quickly. The company’s K3 model, released over the summer, appears to have expanded adoption across users and developers.

Although usage has eased slightly in recent months, Kimi remains highly active on model hubs and routing platforms. OpenRouter data shows that K3 variants are still producing as many as 300 billion tokens per day, a huge level of activity that implies substantial demand even if the pace has moderated from its peak.

Token generation is not the same as revenue, but it is an important proxy for how heavily a model is being used. High usage typically gives a provider more opportunities to convert attention into paying plans, API access, enterprise contracts, or infrastructure-linked services.

Why are open-weight models still attractive to businesses?

Open-weight models remain attractive because they can spread quickly, be integrated by developers more easily, and build ecosystem momentum faster than tightly controlled closed systems. For companies and startups, that openness can lower adoption barriers and speed experimentation.

At the same time, open-weight deployment can make monetization harder. When users can access model weights directly, the provider often has less pricing power and less control over where the model is hosted or how it is packaged. That trade-off is one reason Moonshot’s revenue ambitions stand out.

• Open-weight models can drive broad usage faster.
• They often create larger developer ecosystems.
• They usually deliver lower margins than closed systems.
• They can still support major revenue if adoption is strong enough.

What are Anthropic’s allegations against Moonshot?

Anthropic has accused Moonshot of using a prolonged distillation process that allegedly routed Kimi user requests to Claude Opus, effectively substituting Anthropic’s model output in place of Moonshot’s own responses.

According to the allegations, the activity involved nearly 300,000 requests and produced more than 23 million responses from Anthropic models that were then used in Moonshot’s training. If accurate, that would amount to a serious competitive and legal dispute over whether one lab extracted value from another company’s proprietary model without permission.

Moonshot has not publicly resolved the allegations, and Anthropic’s complaint comes at a delicate moment for the Chinese startup, which is trying to convince the market that its technical progress can translate into durable revenue.

Anthropic says the disputed activity involved extensive use of Claude Opus outputs and that the material was incorporated into Moonshot’s training pipeline, raising questions about model sourcing and consent.

Why distillation is such a sensitive issue

Distillation is a common AI technique in which a smaller model learns from the outputs of a larger, stronger model. In legitimate contexts, it is used to compress capabilities, reduce costs, and improve deployment efficiency.

The dispute arises when one company is accused of using another company’s outputs without authorization or in a way that breaches contractual or technical limits. That can turn a routine machine-learning method into a legal and competitive flashpoint.

For frontier labs, the stakes are high. Their models are often the product of huge capital investment, scarce talent, and proprietary data pipelines. If competitors can replicate performance by harvesting outputs, the economics of model development become much harder to defend.

How big is Moonshot’s opportunity in China and beyond?

Moonshot’s opportunity is substantial because it sits at the intersection of Chinese AI demand, developer interest, and the global appetite for capable lower-cost models. Kimi has emerged as one of the better-known AI products from China’s laboratory ecosystem, and its visibility gives Moonshot an opening to expand revenue in multiple markets.

In China, AI competition has intensified as startups and larger internet companies race to offer assistants, coding tools, search products, and enterprise copilots. A highly used assistant can become a platform for subscriptions, workplace tools, and developer APIs.

Outside China, open-weight models have another advantage: they can be adopted by organizations that want to customize models locally, avoid vendor lock-in, or keep more control over deployment. That makes Moonshot’s model commercially relevant even beyond its home market.

What does the token data tell us?

The OpenRouter figures suggest Kimi remains a model with serious real-world demand, even if day-to-day activity has cooled from earlier highs. A daily throughput of 300 billion tokens is a large-scale signal that users, developers, or third-party applications are still routing substantial work through the model family.

Still, token volume alone cannot prove profitability. Revenue depends on pricing structure, hosting costs, enterprise conversion, and the mix between free and paid access. Moonshot’s challenge is to turn visible usage into recurring income without undermining the accessibility that helped Kimi spread in the first place.

Why this could reshape the open-weight AI market

Moonshot’s goal is important because it tests a central assumption in AI: that the biggest profits belong only to companies that tightly control their models. If Moonshot can scale revenue materially while keeping Kimi open-weight, it would strengthen the case for a more open ecosystem.

That could encourage other labs to pursue similar strategies, blending openness with aggressive commercialization. It could also push investors and customers to rethink how they judge AI leadership, especially if usage and revenue are no longer exclusive to the most secretive frontier models.

Yet the company’s progress is inseparable from the controversy surrounding its training methods. A strong revenue story could be undercut if the allegations gain traction or if legal findings constrain how Moonshot built or improved its models.

What happens next?

The near-term story will hinge on two questions: whether Moonshot can keep translating Kimi usage into paid revenue, and whether the Anthropic dispute escalates into a more formal legal or regulatory confrontation.

If the company approaches its $2 billion target, it will show that open-weight AI can become a major commercial category. If the allegations worsen, however, Moonshot could find itself fighting to preserve both its technical reputation and the business momentum it has built around Kimi.

For now, the company stands as a reminder that the AI market is broadening. The race is no longer only about who builds the smartest model. It is also about who can turn model usage into sustainable revenue without crossing lines that competitors, regulators, or courts may later scrutinize.

Timeline of key developments

Moonshot AI’s rise captures two sides of the current AI era: the commercial promise of fast-growing models and the deepening disputes over how those models are built. Its next phase will help show whether open-weight AI can be both widely used and highly profitable, or whether the economics of the sector still favor the most guarded systems.