French Investors Value China’s Moonshot AI, Maker of Kimi K3, at $30 Billion

With Moonshot AI heading for a stock market listing, a group of mostly French investors has bought its way into the Chinese company behind the Kimi model family. The money was pooled in a special purpose vehicle (SPV) that put roughly €10 million into the last financing round before the IPO. One of the participants is Pyratz Corp. (MLPTZ), an investment firm listed on Euronext Access Paris, whose CEO Bilal El Alamy explains to Trending Topics how the structure works and why the money went to Beijing rather than to Mistral.

The company said the are investing at an $30 billion pre-money valuation of Moonshot AI. Chinese competitors that are already listed on exchanges are valued at $46.6B (Z.ai) and $13B (Minimax). Chinese AI lab DeepSeek recently has closed its first-ever external funding round, raising roughly $7.4 billion. The valuation exceeded $50 billion, sources said.

What makes the bet legible is how far Moonshot has closed the gap to the American frontier. Kimi K3, the first open model in the three-trillion-parameter class, ranks level with Gemini 3.1 Pro on Artificial Analysis’s Intelligence Index and ahead of Claude Opus 4.8, at roughly a thirtieth of the API price of Western frontier models. Among open-weight models it sits at the top of the field, the first to reach the frontier tier rather than trailing it by the usual six to twelve months. Frontier-level quality that companies can download and host themselves is what European buyers are reacting to.

A Listed Investor Buying Off Its Own Balance Sheet

Pyratz is not a conventional fund. “We are an investment firm, built a little bit on the model of American companies like Pershing Square or Berkshire Hathaway,” El Alamy says. “We’re listed ourselves, we invest off our own balance sheet, and we invest across all stages, but only in technology companies.” Since 2021, the firm says it has built a portfolio of 29 companies that have collectively raised more than €150 million, including one billion-dollar company, with six exits so far.

The path to Moonshot ran through Pyratz’s own product work. “We do a lot of research and development internally, and we work with the technologies of the companies we invest in,” El Alamy says. “We’ve been using the models of Moonshot quite a lot, we tested deploying them, self-hosting them basically, and they’re really performing. So when we had the opportunity, we made the investment.”

€100,000 at a $30 Billion Valuation

The official Pyratz announcement discloses no figure. In conversation, El Alamy is more specific: “We invested around €100,000. We’re an independent investment firm and it’s our own money, so we just contributed as we could.” The round, which closed over the summer, valued Moonshot AI at $30 billion pre-money.

He justifies that price by pointing to the European comparison: “Honestly, for the performance of their models and the revenue they do, they do twice as much revenue as Mistral, for example, it was a good investment.”

How the SPV Is Put Together

A single €100,000 ticket does not fit into a round of that size, so the investors banded together. “We syndicated ourselves with a bunch of other investors to buy this,” El Alamy says.

The vehicle is structured and managed by Techmind, a French investment intermediary that operates as an investment studio, packaging deals for angel investors, family offices and corporates. The SPV runs under the name Techmind Moonshot AI. “The pool amounts to about 10 million,” El Alamy says. “Obviously, when you invest at 30 billion, 10 million is quite negligible. The point is to recognize the performance of those Asian models.”

He declines to name the other participants in detail, since it is a private SPV. What he will say: “It’s mostly French investors, mostly founders of French tech companies that participated.”

The shares themselves were bought through Cathay Capital, a French-Chinese investment firm with an established name in France.

Pyratz is in no hurry to sell in any case. “We are pretty much long-term investors,” El Alamy says. “The fact that we’re listed basically allows us to never have to sell the shares if we don’t want to. We’re not a fund where we have to return the money at some point. People who hold shares in our company are liquid anyway. So our strategy is to keep the most performing investments.”

The IPO Is Around the Corner

The timing of the investment is deliberate. “Moonshot plans to list on the Hong Kong Stock Exchange during the first semester. That’s why I say it’s a pre-IPO investment: the next fundraising is supposed to be the IPO,” El Alamy says.

According to reporting by the South China Morning Post, the company has already filed confidentially for a Hong Kong listing and is targeting a debut as early as the first quarter of 2027. A final financing round is running in parallel that could value Moonshot at up to $50 billion. It also emerged recently that the company is weighing a dual listing, adding a quotation on the Shanghai Stock Exchange’s Star Market, where other Chinese tech heavyweights have gone public lately. Trending Topics reported over the summer that Moonshot was aiming to list within six months after the Kimi K3 coup.

Why Moonshot and Not Mistral

For a French investor, the obvious question is why the money did not go to the domestic AI champion. El Alamy does not dodge it: “From a pure model perspective, yes, I’m not going to lie. I think nine people out of ten would say the same thing if you asked them.”

He now sees the strengths of the two companies in different places: “What we’ve been seeing with Mistral recently is that since they’re not really able to compete on the large models, they started diversifying towards becoming a cloud provider. They deploy Chinese models on their own infrastructure, at the expense of their own model. What Mistral is going to be good at is the hosting part, the data centers. They’re building a bunch of them, and different models will be deployed there.”

He resists framing this as a contest: “The playbook is never to use just one model, it’s to use multiple models. So it’s not against Mistral, it’s pretty much complementary. We chose to deploy their model and Chinese models, not American ones.”

Self-Hosting Is the Real Driver

The strategic backdrop is the shift toward open weights. “Sixty-three percent of AI usage now goes through open-weight models, and the trend is definitely accelerating,” El Alamy says, pointing to regulated industries as the driver: “Even though providers like Anthropic and OpenAI say they don’t suck up everything you do, they pretty much do. So for all the regulated industries, finance, healthcare, insurance, people are not really using those models.”

He gives short shrift to the argument that Chinese models are a security risk: “Some people might say: be careful, it’s Chinese, it’s not sovereign. For me, that’s a little bit bullshit. A model is formalized, it’s a matrix of coefficients. There’s no backdoor in it. Models can be biased, but all models are biased differently. That’s exactly why it’s important to use various models, compare the answers and challenge them against each other. From a cost-efficiency perspective, the Chinese models are really, really good.”

Pyratz could run the openly available Kimi models without holding a stake, and El Alamy concedes as much. The deal is meant to open the distribution side: “One of the particularities of our structure is that when we invest in a company, we try to help it deploy its solution better in France. We’re building a joint offering where people can self-deploy the models.”

“An open-weight strategy can compete at the frontier”

Beyond that, Pyratz wants to combine Moonshot’s open weights with technology from portfolio company Zama, which works on fully homomorphic encryption. It allows computations to be performed on encrypted data without exposing the underlying information. In the official announcement, El Alamy puts it this way: “Moonshot AI has shown that a focused team, strong architecture and an open-weight strategy can compete at the frontier without matching the spending of the largest US laboratories. This investment also supports a broader objective: exploring sovereign AI infrastructure in which powerful open-weight models can work with confidential-computing technologies.”

Moonshot AI was founded in 2023, employs around 300 people and has raised roughly $5 billion since inception, far less than the leading US labs. Its latest flagship model, Kimi K3, carries 2.8 trillion parameters and a one-million-token context window. About 70 percent of revenue comes from enterprise API usage, according to Pyratz, with the remainder from paid consumer subscriptions.