Oluwadi
Investigating the Overlooked
Region
What
Who For
Analysis
Oluwadi
Investigating the Overlooked
Moonshot AI, the Beijing startup behind the Kimi chatbot, confidentially filed an A1 application with the Hong Kong Stock Exchange on September 3, 2026, formally starting the IPO process it has been building toward all year.[1][2] Seven days later, on September 10, Anthropic published a threat intelligence report naming Moonshot as one of seven China-based AI labs it accuses of running large-scale campaigns to distill Claude's outputs into their own models -- specifically alleging that Moonshot routed nearly 300,000 Kimi customer requests to Claude over a single 10-day window, through more than 5,380 fraudulent accounts, without telling the customers whose questions it was answering.[10][11] Both things are true at the same time, about the same company, in the same week.
Moonshot closed its Series C at a $4.3 billion valuation at the end of 2025.[5] From there the number moved fast: a $700 million raise lifted it to $10 billion in January 2026; a $2 billion round pushed it to $20 billion by early May; and a $3.5 billion round -- closing in the weeks after Moonshot released its Kimi K3 model on July 16 -- valued the company at $35 billion in late July.[3][5][6] The round now in motion, disclosed alongside the HKEX filing, targets a pre-money valuation of up to $50 billion -- Moonshot's likely final private round before listing.[1][4] That is roughly an elevenfold increase in nine months from the Series C, not the eightfold jump from a $6.25 billion baseline sometimes attached to this story -- no financial-press source found in reporting this piece supports a $6.25 billion figure for Moonshot at any point in the timeline, and the real multiple is larger, not smaller.
An A1 filing is a confidential first step under Hong Kong Stock Exchange rules -- it starts regulatory review without immediately disclosing financials publicly, and it is a step, not an approval. Moonshot is reportedly seeking to raise about $3 billion in the listing itself, with Goldman Sachs, CICC, and Deutsche Bank advising, and is targeting a listing as early as the first quarter of 2027.[1][2] As a prerequisite, Moonshot unwound its offshore "red-chip" holding structure and redomiciled onshore in mainland China -- a structural step Chinese regulators have increasingly required of companies chasing a Hong Kong listing rather than a U.S. one.[1] The company itself has offered no on-record comment on any of this; every figure above is sourced to people described as familiar with the matter, not to a Moonshot disclosure.[1]
The valuation run tracks a real product cycle, not just capital chasing itself. Kimi K3 launched July 16, 2026 as a 2.8-trillion-parameter mixture-of-experts model with a 1-million-token context window, with open weights released July 27 under a license that requires companies above $20 million in annual revenue to negotiate a commercial contract before reselling it.[6] It debuted third on the Artificial Analysis leaderboard, behind Anthropic's and OpenAI's flagship models.[6] Moonshot's annualized recurring revenue moved with it: about $100 million in March 2026, $200 million in April, $300 million in June -- and, according to figures Moonshot reportedly gave investors, past $1 billion in August, with a $2 billion year-end target.[7]
Alongside the fundraising, Moonshot is restructuring how it sells. Reporting from Caixin Global on September 11 describes Moonshot pairing with five Chinese IT system integrators -- including Teamsun and Kingsoft Cloud -- to build joint forward-deployed engineering teams that embed at client sites, splitting token-usage and implementation revenue by contract.[8] Separate reporting the next day named the model explicitly: Moonshot is following the approach Palantir built its own business on in the U.S., trading a pure per-token consumer or API product for on-site engineers who wire the model into a client's actual business processes.[9] That is a real business-model shift, not a rebrand -- forward-deployed engineering is a materially different cost structure and sales motion than shipping a chatbot, and it is the kind of move a company makes when it is trying to show durable, non-consumer revenue to IPO investors, not just growth.
The distillation claim against Moonshot comes from one source: Anthropic's own September 2026 threat intelligence report, which this outlet already covered in full for its other findings -- a Russian drone-targeting case, five biological-misuse cases, and the same distillation section's claims against six other Chinese labs.[10] On Moonshot specifically, Anthropic's claim is precise: that Moonshot's Kimi product relayed customer requests to Claude's API, returned Claude's answers to those customers as if Kimi had generated them, and did this nearly 300,000 times over one 10-day window through more than 5,380 accounts Anthropic says were fraudulently created to evade its access controls.[10][11] That is a specific, falsifiable factual claim -- not a vague accusation -- and it has not been tested in court, audited by an independent party, or admitted to by Moonshot, which has not commented on the report any more than it has commented on the IPO filing.[11] It is also being made by a direct competitor with its own commercial interest in the claim landing publicly, in a report this outlet has already noted is self-investigated, self-judged, and self-published, with no outside regulator or auditor involved.[10]
Why does this matter? Neither fact in this piece cancels the other out, and resolving them into a single verdict would misrepresent both. The valuation trajectory is real and heavily corroborated across independent financial press -- Moonshot has closed successive rounds at $10 billion, $20 billion, and $35 billion in nine months, and is now filing for a Hong Kong listing while raising at a $50 billion pre-money valuation, on the strength of a model, K3, that genuinely ranks near the top of independent benchmarks and a revenue base that grew roughly tenfold in five months. None of that requires the distillation allegation to be true or false. Separately, Anthropic's claim against Moonshot is specific and detailed enough to be a real allegation rather than an insinuation, but it comes from one interested party, has not been independently verified, and Moonshot has not answered it. A company can be building a genuinely fast-growing product on real revenue and also be credibly accused of having built part of that product by quietly routing customer traffic through a rival's model -- investors and readers evaluating Moonshot's IPO story get to hold both of those facts at once, without either one being allowed to erase the other.
Companion piece on this outlet: "A Russian Freelance Team Used Claude to Build a Drone Swarm That Picks Its Own Human Targets and Detonates With No One in the Loop...", which first documented Anthropic's distillation allegation against Moonshot and six other Chinese labs in full.