Everything About AI
NextFin News - China's Moonshot AI is weighing an initial public offering in Hong Kong that could raise as much as $5 billion as soon as this year, according to people familiar with the matter, a move that would put the maker of the Kimi chatbot at the center of the most consequential valuation test yet for China's artificial-intelligence sector. The Beijing-based lab has filed confidentially for the listing and added Bank of America as an overall coordinator alongside CICC, Deutsche Bank and Goldman Sachs.
The reported $5 billion target would be nearly nine times the proceeds of Zhipu AI's January Hong Kong debut and would demand a public-market valuation above the $35 billion Moonshot reached in a July private round - asking investors to underwrite a company whose value has more than tripled since January.
The Deal, the Timetable, and the Valuation Ladder
The offering is still unconfirmed: no prospectus has been published, and the company has not announced a final timetable. But the scaffolding is visible. In July, Moonshot distributed a shareholder resolution seeking backing for a Hong Kong listing that could land within six months, and it has been dismantling its offshore VIE structure - a prerequisite for a smoother approval path under China's revised overseas-listing rules. The company had previously explored a waiver route but abandoned it once it became clear Beijing would not grant an exemption.
The valuation ladder behind the IPO is steep. Moonshot was valued at $2.5 billion after a February 2024 round led by Alibaba Group, which took roughly a 36 percent stake. An August 2024 round led by Tencent Holdings and Gaorong Capital lifted it to $3.3 billion. By the end of 2025 it stood near $4.3 billion, then a $700 million round in January 2026 priced it at $10 billion. In May it raised at a valuation of roughly $20 billion. And in late July the company closed a $3.5 billion financing at a $35 billion post-money valuation - more than triple its value in six months and nearly double the $1 billion to $2 billion it originally set out to raise.
The July round was anchored by China's National AI Industry Investment Fund, a state-backed vehicle established in January 2025 with approximately $8.8 billion of capital, alongside Alibaba, Tencent, Meituan, IDG Capital and China Mobile. The state anchor matters: it is the same fund reported to be negotiating a lead role in DeepSeek's first external round at up to $50 billion. In effect, Beijing is picking the pace at which China's frontier AI labs are capitalized - and Moonshot is the first in line to take that price public.
Now the company is in discussions for a follow-on private round at a $50 billion pre-money valuation, which would be its final raise before the listing. That creates an awkward sequencing problem. New private investors would be asked to accept a price set in July by an oversubscribed round, with the IPO as the exit that justifies it. Investors who pass on the private round can simply wait for the float - a different negotiating position than earlier backers faced.
The Business Behind the Multiple: ARR, K3, and the Open-Weight Bet
The valuation story rests on two pillars: revenue growth and model credibility. On revenue, Moonshot's annual recurring revenue reportedly climbed from $100 million in March 2026 to $200 million in April and $300 million in June - a tripling in three months driven by paid chatbot subscriptions and enterprise API usage. Even so, that revenue remains smaller than Zhipu AI's, which is on track for $1 billion in annual sales.
On the model side, the catalyst was Kimi K3, released July 16, 2026. The company describes it in its own technical blog as its most capable model to date:
"Today, we are introducing Kimi K3 — our most capable model. Kimi K3 is a 2.8T-parameter model built on our Kimi Delta Attention and Attention Residuals, with native vision capabilities and a 1-million-token context window. It is the world's first open 3T-class model."
The model is priced for API access at roughly $3 per million input tokens and $15 per million output tokens - in the range of Anthropic's Sonnet-tier pricing. On Artificial Analysis's Intelligence Index, K3 scored 57 versus 60 for Claude Fable 5, and it beat Fable 5 outright on the Frontend Code Arena leaderboard with a 1,679 Elo score, becoming the first Chinese open-weight model to reach that mark. The company itself notes the gap that remains: "While its overall performance still trails the most powerful proprietary models, Claude Fable 5 and GPT 5.6 Sol, Kimi K3 demonstrated frontier-level performance across our evaluation suite."
The open-weight strategy is the company's distinguishing bet. By releasing model weights rather than walling off access behind an API, Moonshot is betting that developer adoption and ecosystem lock-in will matter more than closed-model scarcity - the same logic that made Meta's Llama family a distribution force. K3's reception, which triggered a selloff in global tech stocks and drew a "concerning" reaction from former White House AI czar David Sacks, gave the IPO thesis a demonstration no pitch deck could replicate.
The company, founded in early 2023, is led by co-founder Yang Zhilin, a former Tsinghua University professor who worked at Meta and Google and contributed to Transformer-XL, rotary position embeddings and group normalization. Moonshot is named after a Pink Floyd album.
Why Hong Kong, and Why Now
The timing is not accidental. Hong Kong has become the default venue for China's frontier AI labs after U.S. listings closed off and domestic A-share rules proved ill-suited to loss-making, variable-interest-entity structures. Zhipu AI became the first Chinese foundation-model developer to list, pricing its January 8, 2026 offering at HK$116.20 a share and raising about $558 million at a market value near $7.1 billion. A day later, MiniMax listed at HK$106.7 billion, about $13.7 billion.
Those two listings set the anchor, and MiniMax supplies the cleanest comparable. MiniMax's annual recurring revenue was about $150 million in February 2026 and approached $300 million by the end of May; at a market value near $18.6 billion on that roughly $300 million run-rate, it has traded at a price-to-revenue multiple of about 60 times. Moonshot's own $300 million of ARR, measured on the same metric, implies an $18 billion private-market reference point before the K3 premium. The question Moonshot's bankers must answer is whether K3's benchmark performance and open-weight distribution justify a two-to-three-fold premium over MiniMax.
The answer depends on what public investors are actually buying. A $5 billion raise at a $50 billion pre-money valuation would value Moonshot at roughly 167 times annualized revenue run-rate - a multiple that prices not just current growth but a decade of dominance. For comparison, the most valuable U.S. AI companies trade at high but not infinite multiples of revenue, and they carry audited financials, transparent cap tables, and established enterprise contracts.
The Cyclical Wave Riding a Structural Shift
This is the core judgment: Moonshot's valuation surge is a cyclical wave superimposed on a structural shift, and the two are being confused in the deal math. The structural shift is real and durable. China's AI stack is consolidating around a handful of well-capitalized labs, the state has made frontier model development a national priority, and the open-weight distribution model is rewiring how foundation models reach customers. None of that reverts on its own.
The cyclical wave is the capital rush, and three historical-cycle comparisons make the pattern clear. First, the 2020-2021 technology IPO and SPAC boom: a study of that class found global revenue multiples fell from 17.9 times in March 2022 to 7.4 times in March 2023, a 59 percent decline, and by March 2023 only 5 percent of the class was trading above its listing price - with not a single SPAC above water. Second, China's 2020-2021 electric-vehicle SPAC frenzy, where Lucid, Nio and their peers commanded multiples that priced in decades of adoption before delivery growth normalized and valuations compressed. Third, the 2023-2024 generative-AI private-round escalation, where valuations doubled and tripled on model-release momentum before revenue caught up.
The mean-reversion pattern is consistent: when capital is abundant and a technology narrative is hot, private valuations front-run revenue by years; when liquidity tightens or growth normalizes, multiples compress toward the revenue line. Moonshot's three-and-a-half-fold valuation increase in six months is a cyclical signature, not a structural one. Revenue tripled in three months; valuation more than quadrupled in roughly the same window. The gap between the two is the wave.
But the structural leg is what keeps the floor from falling out entirely. Open-weight distribution creates genuine switching costs for developers, and China's state-backed capital stack means these labs will not face the funding cliff that broke many 2021 SPACs. The right read is not "bubble" or "breakthrough" - it is a durable industry built at a cyclical price.
The Second-Order Question Nobody Is Asking
The first-order read is simple: a $5 billion IPO validates China's AI sector and rewards early backers. The second-order question is harder: what happens to the other Chinese labs once Moonshot sets the public benchmark?
If Moonshot lists at $50 billion and the stock holds, every private Chinese AI valuation gets repriced upward - DeepSeek, MiniMax, Zhipu, and the dozens of smaller labs all benefit from a higher anchor. If it lists and breaks - if the stock trades down 30 or 40 percent in its first quarter - the entire private round at $50 billion becomes a marking-to-myth exercise, and follow-on funding for every other lab freezes. The IPO is not just Moonshot's exit; it is the pricing mechanism for China's entire AI capital stack.
That creates a perverse incentive for the banks and the company: the deal almost certainly gets done, because the state anchor and the strategic importance of the listing make failure politically costly. The real risk is not a cancelled IPO - it is a successful IPO at a price that the secondary market rejects. A $5 billion raise that prices well but trades poorly is worse for the sector than no IPO at all, because it locks in a false benchmark that takes years to unwind.
There is also a cross-border dimension. Moonshot's open-weight model is priced at Sonnet levels but performs within striking distance of U.S. frontier labs. If public investors reward that, capital flows into Chinese AI intensify despite export controls and geopolitical friction. If they do not, the open-weight strategy becomes a cautionary tale about giving away the crown jewels for distribution.
The Counter-Thesis: Why the Skeptics Could Be Right
The strongest case against Moonshot's valuation is straightforward: revenue does not support it, and the competitive moat is thinner than the benchmarks suggest. At $300 million of ARR and a $50 billion valuation, Moonshot trades at roughly 167 times revenue. Even assuming 100 percent annual growth for three years, revenue would reach $2.4 billion - still implying a 20 times multiple on a company with no proven profitability, facing brutal domestic competition and U.S. chip restrictions that cap its compute scaling.
The benchmark argument is also vulnerable. K3's 57-versus-60 Intelligence Index score is close but not superior, and benchmark leadership in foundation models has proven fleeting - the gap between releases is measured in months, not years. Open-weight distribution, while powerful for adoption, also means competitors can distill and replicate capability, eroding the pricing power that closed models enjoy. Zhipu's $1 billion revenue trajectory and MiniMax's public-market validation suggest Moonshot is not the uncontested leader the $50 billion tag implies.
Furthermore, Hong Kong's liquidity for large-cap technology has been inconsistent since the 2021 regulatory crackdown. A $5 billion offering in a market where daily turnover for even the largest names can thin out quickly risks a weak aftermarket - and a weak aftermarket is the signal that would prove the bullish thesis wrong.
What to Watch: The Falsifying Signals
The base case is that Moonshot lists in Hong Kong before year-end, raises between $3 billion and $5 billion, and prices at a valuation between $40 billion and $50 billion. The stock likely opens strong on scarcity value - there is no pure-play Chinese frontier AI comparable in public markets - then grinds lower as quarterly revenue disclosures fail to match the multiple.
The upside case: Kimi K3 adoption accelerates, ARR crosses $1 billion within 12 months, and the stock sustains a premium to MiniMax, validating the open-weight strategy and triggering a wave of higher-priced listings across China's AI sector.
The downside case: the IPO prices at the low end of the range, the stock falls more than 30 percent within its first quarter, and private valuations across Chinese AI compress toward public comparables. The specific falsifying signal is this: if Moonshot's first two quarterly revenue disclosures show ARR growth below 50 percent year-over-year while the stock trades below its IPO price, the structural-premium thesis is wrong - the market is pricing a cyclical wave, not a durable moat.
Watch three things. First, the final IPO size and price - anything below $3 billion or above $5 billion signals weak demand or overreach. Second, the first-quarter revenue disclosure - ARR trajectory is the only number that matters. Third, the aftermarket performance of MiniMax and Zhipu in the weeks around the pricing - they are the canary for Hong Kong's appetite for AI at these multiples.
The closing judgment: Moonshot's IPO is less a test of one company than a referendum on whether China can build public-market AI champions at American-style valuations. The structural shift in Chinese AI is real; the price being asked for it is cyclical. When the two collide in Hong Kong's order book, the market will decide which one wins.
Explore more exclusive insights at nextfin.ai.
