3 min read
Two weeks ago the White House accused Moonshot AI of building its flagship model by covertly siphoning off a US rival’s work. This week Moonshot is reportedly chasing a $50 billion valuation ahead of a stock-market debut. In the AI gold rush, it turns out, an accusation of industrial-scale copying is not a scandal to survive. It is a launch announcement.
The money on the table
The numbers are frankly silly. Moonshot AI, maker of the Kimi models, is seeking pre-IPO funding at around $50 billion and lining up a Hong Kong listing within six months, with analysts floating a figure north of $100 billion by the time it actually lists. Not bad for a three-year-old startup running on roughly $300 million of annual revenue. DeepSeek, the lab that rattled Western tech stocks last year by being cheap and good, is preparing its own float on Shanghai’s Nasdaq-style STAR market. And behind the model-makers, a pack of Chinese chip designers, MetaX, Biren and Moore Threads, are all filing to raise the cash to build the domestic silicon that Washington’s export bans were supposed to deny them.
The delicious irony
Look at the timing for a second. The American argument all year has been that China’s AI is derivative, that it only keeps pace by copying, distilling and dodging chip controls. Maybe some of that is even true. But the market’s response to all this alleged thievery is to value these companies at tens of billions of dollars, because investors do not much care where the capability came from, only that it exists, it is cheap, and people are paying for it. The export controls meant to strangle Chinese AI have instead handed it a cracking founding myth: the scrappy underdog building world-class models on a shoestring while the empire tries to cut off its supplies. That story sells shares.
Revenue versus valuation
A word of caution for anyone tempted to get excited. A $50 billion valuation on $300 million of revenue is a price-to-sales ratio north of 150. That is not investing on fundamentals, it is investing on the belief that these firms will be gigantic later and you want in before the door shuts. It might pay off spectacularly. It might also be the kind of number people point at ruefully in a few years. Chinese listings come with their own political weather too: Beijing can giveth and taketh away with a single regulatory memo, as anyone who held Ant Group or Didi can bitterly confirm.
The bigger point
Whatever you make of the copying claims, this is the real story: AI is no longer a two-horse race with America comfortably in front. It is a genuinely global scramble in which the Chinese labs are good enough, cheap enough and now well-funded enough to matter, and in which “they stole it” is starting to sound less like a knockout blow and more like the losing corner’s excuse. The chequebooks have spoken, and they are not remotely embarrassed.
The AI tool stack actually worth paying for
One email a week. The tools, models and moves that matter, minus the hype and with the horseshit filter set to maximum. Free.
Did you know: Moonshot is not just listing, it is also releasing the full weights of its 2.8-trillion-parameter Kimi K3 model to the public, roughly 1.4 terabytes of file, making it the largest open-weight AI model ever released. Giving away the crown jewels while raising billions on the stock market is a very 2026 way to run a business, and a headache for anyone in Washington still hoping to keep this technology bottled up. (None of this is investment advice, and pre-IPO Chinese shares are about as far from a sure thing as markets get.)