Moonshot Files for Hong Kong: Inside the Kimi Maker's $50 Billion Bid to Become China's Defining AI IPO

Three years after a Tsinghua-trained researcher founded a chatbot company in Beijing, Moonshot AI has quietly taken the step that could define how the world prices Chinese artificial intelligence: the company behind the Kimi assistant has confidentially submitted an IPO application to the Hong Kong stock exchange.

Reuters and the Wall Street Journal confirmed the filing on September 3, after Chinese tech publication LatePost first broke the news. According to sources familiar with the plan, Moonshot is looking to raise around $3 billion — and Bloomberg reports the offering could stretch to as much as $5 billion, with a listing possible before the end of 2026. The startup’s most recent private funding round valued it at roughly $50 billion, meaning the float would be one of the largest technology IPOs anywhere this year, and unambiguously the biggest public-market event of China’s AI cycle so far.

What we know about the filing

A few things separate this from the usual rumor-cycle listing story.

First, the confidential route. Hong Kong’s exchange allows sensitive applicants to file behind closed doors while regulators review, shielding financial details from competitors until the company is ready to road-show. Moonshot’s use of it suggests management wants the option to move fast — the company told investors as early as July that it was preparing to list within roughly six months of its Kimi K3 breakthrough.

Second, the banks. Moonshot is working with advisers including Goldman Sachs, according to the Straits Times — a signal that this is being built as a global-markets deal, not a domestic-only float.

Third, the money already on the table. Moonshot has raised more than $5.5 billion to date from a cap table that reads like a map of Chinese tech power: Alibaba, which led its $1 billion round back in 2024 and remains the largest outside shareholder, alongside Tencent and a long list of Chinese state-adjacent and venture funds. Its July 2026 round — $3.5 billion at a $35 billion valuation — reportedly passed its funding goal, and the pre-IPO round that closed in August pushed the paper value toward $50 billion.

The valuation ride: $2.5B to $50B in thirty months

The numbers worth holding in your head:

• February 2024: valued around $2.5 billion in an Alibaba-led financing.
• May 2026: roughly $2 billion raised at a $20 billion valuation.
• July 2026: $3.5 billion raised at $35 billion, after passing its funding target.
• August 2026: pre-IPO round values the company near $50 billion.
• IPO (target): raising $3–5 billion, possibly by year-end.

That is a 20x markup in about thirty months — a trajectory that makes even the headiest US AI rounds look gradual. The fuel behind it is Kimi K3, the 2.8-trillion-parameter open-weight mixture-of-experts model Moonshot released at the World Artificial Intelligence Conference in Shanghai in July. K3 landed near the top of global capability leaderboards, beat several leading US systems on coding and web-development benchmarks, and — by one widely cited estimate — contributed to a week in which semiconductor market value swung by trillions of dollars.

Revenue is growing fast but remains modest relative to the valuation: reporting through mid-2026 put annualized recurring revenue around $300 million, up from roughly $200 million earlier in the year. At $50 billion, the market is being asked to pay well over 150x current ARR — a multiple that only makes sense if you believe open-weight model distribution becomes a platform business rather than a product one.

The open-weight business model — and the revenue-sharing gambit

That platform thesis is where Moonshot’s pre-IPO moves get genuinely interesting. On August 26, Reuters reported that Moonshot is negotiating with Microsoft, Amazon, and Google over hosting Kimi K3 on Azure, AWS, and Google Cloud — and seeking up to a 30% share of revenue generated from K3-related services on those clouds.

If those terms stick, they would create something new in the industry: an open-weight model maker collecting a toll from the very US hyperscalers that sell compute. It is the logical endpoint of the open-weights strategy — you cannot charge for access to weights anyone can download, but you can charge for the ecosystem that forms around them: fine-tuning pipelines, enterprise deployments, agentic tooling, and the brand trust that comes with being the reference implementation.

There is an irony here that investors in the IPO will have to underwrite. The same open weights that make Moonshot’s negotiating position possible also make its moat unusually permeable — its full 96-shard, 1.56-terabyte K3 release has been downloaded, mirrored, and redeployed far beyond anyone’s control. And the company’s rise has already been weaponized in Washington: US officials spent August scrutinizing how Moonshot obtained Nvidia compute through Alibaba’s cloud infrastructure, an arrangement that helped trigger the ongoing regulatory push around remote GPU rentals. None of that appears in the valuation yet.

Why Hong Kong, and why now

The venue tells its own story. A US listing was effectively off the table — Chinese AI companies sit squarely in the crosshairs of American investment restrictions and audit disputes — and Hong Kong is enjoying its strongest fundraising year in a decade, with IPO proceeds already above $41 billion through mid-August.

More importantly, Hong Kong gives Moonshot access to the southbound flow of mainland capital through Stock Connect, meaning Chinese retail and institutional investors can buy the AI story they have watched unfold domestically. For a company whose product is used by tens of millions of Chinese consumers and whose flagship model became a national-technical-pride moment, that is the natural buyer base.

The stakes: a price discovery event for Chinese AI

Every AI valuation to date — OpenAI’s, Anthropic’s, xAI’s, and Moonshot’s own private marks — exists in a market of a handful of insiders negotiating with a handful of funds. An IPO is different. It forces the open market to attach a number to a Chinese AI company’s future, in a currency everyone can trade, while the sector is still burning cash on training runs that grow more expensive each generation.

Moonshot’s pitch is essentially: we are the open-weights winner in the world’s second-largest AI market, we have a consumer franchise (Kimi), a developer franchise (K3), and a credible path to taxing global cloud distribution. The skeptics’ counter is a $50 billion price on $300 million of revenue, a geopolitically exposed supply chain, and open weights that competitors — including Alibaba’s own Qwen family, which out-downloads Kimi by more than 50 to one — can fork at will.

Whichever way the book builds, the listing will set a reference price for every Chinese AI company queuing behind it. DeepSeek, MiniMax, Zhipu, and the rest of the 2023 founding cohort are all watching. And so, less comfortably, are regulators in Washington, who have spent this summer trying to fence off exactly the kind of US-China AI entanglement — cloud deals, revenue shares, investor overlap — that this IPO will now formalize and price.

The six-month clock Moonshot told investors about in July is running. If the deal lands this year at anything near the target, “China’s AI boom” stops being a private-market abstraction and becomes a ticker.