Moonshot AI Thinks Open Models Can Become Big Business

Open-weight AI has always had an obvious business-model question: if anyone can download the model, where does the money come from?

Moonshot AI thinks the answer could still be measured in billions.

The Chinese AI company behind the Kimi family of models is reportedly targeting $2 billion in annualized revenue by the end of 2026, roughly double its reported August run rate. The aggressive target follows growing adoption of Moonshot's K3 models, which have become one of the more prominent open-weight alternatives to systems from OpenAI and Anthropic.

The number matters because the AI industry has increasingly divided into two camps.

One sells access to proprietary models.

The other gives model weights away and tries to build businesses around usage, APIs, enterprise services and infrastructure.

Moonshot is attempting to prove the second camp can still generate enormous revenue.

Kimi's usage is already operating at huge scale

According to data cited by TechCrunch, Moonshot's K3 models have recently generated as many as 300 billion tokens per day through OpenRouter, despite usage moderating somewhat after the model's initial launch surge.

That's a useful reminder that the open-model market isn't small.

Developers increasingly want models they can run themselves, customize or deploy through multiple infrastructure providers without becoming permanently tied to one AI company.

For enterprises, that flexibility can be valuable.

For the model maker, however, flexibility creates a monetization challenge.

A closed model provider controls the API.

An open-weight provider doesn't.

Users can potentially take the model elsewhere.

Open AI can make money — just differently

Moonshot's projected revenue remains well below the levels reportedly being generated by OpenAI and Anthropic.

TechCrunch notes another structural difference: because Moonshot distributes model weights openly, its economics can be less attractive than those of companies controlling access to every inference request.

But open-weight AI may not need to replicate the exact economics of closed platforms.

The business can emerge around hosted inference, premium APIs, enterprise deployments, customized models, tooling and developer ecosystems.

That begins to look less like selling one piece of software and more like building infrastructure around a standard.

Linux itself is free.

Entire companies were still built around making it useful to businesses.

AI could follow a similar pattern.

There is also a competitive controversy

Moonshot's growth comes as Anthropic has accused the company of using large-scale model distillation against Claude.

Anthropic alleged that requests associated with Kimi were routed to Claude models as part of efforts to collect outputs for training. The allegations are part of a broader fight over whether AI developers can use another company's model responses to improve competing systems. Moonshot's practices remain a contested issue rather than a settled legal finding.

That dispute matters beyond Moonshot.

As frontier models become increasingly expensive to train, companies have strong incentives to extract capabilities from stronger systems rather than reproduce every breakthrough independently.

The industry is therefore entering an awkward phase where companies advocate open ecosystems while simultaneously trying to protect the knowledge embedded in their own models.

China's AI companies are competing on economics too

Much of the geopolitical AI debate focuses on benchmark performance.

But the commercial race matters just as much.

A model doesn't need to be universally considered the world's best to become an important platform.

It needs developers.

Distribution.

Reliable infrastructure.

Competitive pricing.

And customers willing to build businesses around it.

Moonshot's $2 billion target suggests Chinese AI companies increasingly see global model competition as a revenue race, not merely a research race.

What happens next?

The biggest thing to watch is whether Kimi's heavy usage translates into sustained paid adoption.

If Moonshot reaches anything close to its target, it would challenge the assumption that the most valuable AI businesses must keep their strongest models closed.

Open-weight AI may generate lower margins.

But if the ecosystem becomes large enough, lower margins on enormous usage can still create a very large company.

The next AI battle may therefore be less about open versus closed and more about which model strategy can build the strongest economy around itself.