Top 10 AI News July 24 2026: China's IPO Rush

1. China's Moonshot Is Rushing to Go Public at $50 Billion

Moonshot AI, the Chinese company behind Kimi K3, is in talks to raise money at a valuation of as much as $50 billion, ahead of a planned stock market listing in Hong Kong within six months. Just in June, the company was valued at around $30 billion, so it has jumped by $20 billion in about two months, entirely on the strength of the attention Kimi K3 generated.

The speed is the whole story. Moonshot released Kimi K3 on July 16, watched it beat American models on a coding leaderboard, rattled US stock markets, then got publicly accused by the White House of copying Anthropic's technology, and is now racing to sell shares to the public within half a year. An IPO, short for initial public offering, is when a private company first sells shares to ordinary investors. Six months is a very fast timeline for one.

The logic behind the rush makes sense even if it looks opportunistic. Investor excitement about Chinese AI is at its absolute peak right now, precisely because Kimi K3 proved these labs can compete at the top. Raising money at $50 billion now, before the copying dispute drags on or a rival grabs the spotlight, locks in a valuation the accusation might otherwise threaten.

My take: the accusation did not scare Moonshot off. It lit a fire under the fundraise. When your model is the talk of the industry, you sell shares while everyone is watching, not after they lose interest.

2. The Great Chinese AI IPO Rush Is On

Moonshot is not alone. Fortune is calling it a great Chinese AI IPO rush. Two Chinese AI companies, MiniMax and Z.ai, already went public in Hong Kong back in January. DeepSeek is preparing to list in Shanghai. And now Moonshot is lining up its own Hong Kong offering. China's AI labs are collectively racing to sell shares to the public while global attention on them is at its highest point ever.

This is not a coincidence, it is a moment. Kimi K3 proved Chinese labs can compete at the frontier, DeepSeek proved months ago they can do it cheaply, and together that has completely reset how investors value Chinese AI. Every lab now has both a success story to point to and a narrow window before the excitement fades, and stock markets reward companies that list into enthusiasm. So they are all sprinting for the door at once.

The timing lines up with the American side too. Anthropic and OpenAI are both preparing to go public in the US. So the second half of 2026 will put a real market price on AI companies on both sides of the Pacific, one after another.

My take: this is the moment AI stops being funded by a small club of insiders and starts answering to the public markets. The prices set over the next few months will shape the whole industry for years.

3. DeepSeek Is Going Public Too, at Up to $71 Billion

DeepSeek, widely seen as China's most respected AI lab, is preparing to list on Shanghai's stock exchange at a valuation of up to $71 billion, targeting 2027 or earlier. It raised $7.4 billion in June, and its founder Liang Wenfeng, who also runs a hedge fund, reportedly put in around $3 billion of his own money. That is a serious personal bet on his own company.

The choice of Shanghai over Hong Kong is a deliberate signal. Shanghai's market is where China lists its national champions, companies it considers strategically important or that replace foreign technology it wants to stop depending on. By choosing Shanghai, DeepSeek is positioning itself as exactly that: a national champion in a strategic industry, which usually comes with government support. Moonshot, choosing Hong Kong, is positioning itself as a more internationally focused commercial company.

DeepSeek's influence on the whole market is enormous. Its rock-bottom prices are the number every other AI company gets compared against, and its free models have topped the leaderboards for months. A DeepSeek with access to public money and state backing is a much tougher competitor than a private one.

My take: watch where each Chinese lab chooses to list. Shanghai means national champion with state backing. Hong Kong means international and commercial. The exchange tells you the strategy before the company says a word.

4. The Controversial Model Is Already Making $300 Million a Year

Demand for Kimi K3 has pushed Moonshot's annual revenue to $300 million, even though the company had to pause new sign-ups two days after launch because too many people wanted in and it ran out of computing power to serve them. Making that much money within days of launch, while turning away new customers, is an unusual and very telling combination.

That $300 million figure changes how you should read this whole story. A company making that much, that fast, with a model beating rivals on benchmarks, has a real business, not just a viral moment. That is exactly what justifies a $50 billion IPO price. It also explains the urgency, because revenue growing this fast is the kind of thing investors pay big premiums for, and you want to capture that in an IPO before the growth slows down.

The reason it had to pause sign-ups points to the constraint hanging over the whole industry: there is not enough computing power to go around. This is where giving the model away free on July 27 becomes clever, because once anyone can run it themselves, Moonshot's capacity problem becomes someone else's opportunity.

My take: the $300 million is what turns this from a headline into a real business. A viral model fades. A model making that much money a year is a company worth buying shares in.

5. DeepSeek's New Model Is Cheap and Shockingly Good

DeepSeek's V4 model became its official version today, July 24. It comes in two sizes, and the prices are striking: the cheaper V4-Flash costs $0.14 per million input words and $0.28 per million output words. For comparison, OpenAI's top model charges $30 for the same amount of output, which makes DeepSeek roughly a hundred times cheaper. And this is not a weak model. Its top version scores 80.6 percent on a respected software engineering test, the highest of any freely available model, matching Google's Gemini.

Let that sink in. A model that competes with the best in the world at real coding work costs about one hundredth of what the leading American model charges. For the huge amount of routine AI work businesses do every day, like sorting information or generating basic code, that price gap is impossible to ignore. It is the single biggest force pushing the whole industry's prices down.

Making V4 the official, stable version also matters more than it sounds. Cautious companies avoid models that keep changing under them, so locking in a stable version removes the last excuse not to use it for serious work.

My take: a model this good at this price is the reason every paid AI company is nervous. When the free option scores as high as the expensive one on the tasks most businesses actually need, the sales pitch for paying gets very hard.

6. A $60 Billion Silicon Valley Deal Was Built on a Chinese Model

Here is a detail that complicates the whole copying story. Cursor, the AI coding tool that SpaceX is buying for around $60 billion, admitted back in March that it built its product using a Kimi model from Moonshot as a foundation. So while the US government accuses Moonshot of stealing American technology, one of the most valuable American AI tools was itself built on Moonshot's technology and is being bought by an American company for $60 billion.

The irony is hard to miss, and it makes an important point. The AI world is deeply tangled across the US-China divide. The clean story of theft flowing in one direction does not survive contact with how AI is actually built, where models from many sources get combined, fine-tuned, and built upon until nobody can cleanly say which capability came from where.

For anyone using AI tools, the lesson is that these origin questions cut both ways and are already baked into products people use every day. Cursor is not unusual. Foundation models from all over get built into the software stack.

My take: this is the most honest fact in the whole copying debate. The technology has never respected the borders that governments are now trying to draw around it, and a $60 billion deal built on a Chinese model proves it.

7. Moonshot Is Staying Totally Silent on the Copying Accusation

Moonshot has not said a word publicly about the White House accusation that it copied Anthropic's model, or about the claim that it got restricted Nvidia chips through Thailand. And it is staying silent while simultaneously pushing hard on its IPO. That silence is itself a choice worth thinking about.

There are sensible reasons to say nothing. Responding to a government accusation can make it seem more credible, drag you into an argument on unfavorable terms, or create statements that cause legal headaches during an IPO. Staying quiet and letting the model's benchmark wins speak keeps the conversation on Kimi K3's strengths rather than on the theft question. Chinese companies also generally avoid public fights with US officials.

The risk is that silence can read like quiet confirmation, especially to the businesses deciding whether to use Kimi K3 and the investors deciding whether to buy in. An accusation nobody pushes back on can slowly harden into accepted fact.

My take: Moonshot is betting that momentum matters more than a rebuttal. That works for raising money fast, but it leaves a question hanging that Western companies will keep asking no matter how the IPO goes.

8. The Accusation and the IPO Are Colliding Awkwardly

The copying accusation and the fundraise are crashing into each other in a way that creates real tension for investors. On one hand, Kimi K3 drove the value from $30 billion to $50 billion. On the other, a US official has publicly claimed the model was built on stolen technology and trained using illegally routed chips. That is exactly the kind of unresolved risk that investors dig into before buying shares.

For investors in a Hong Kong listing, it is complicated. The accusation has limited legal force inside China, and Chinese approval for the listing does not depend on US objections. But international investors do care about US regulatory risk and the chance the dispute escalates into something that hurts the company's access to chips, cloud services, or Western markets. Those risks have to be spelled out in the official IPO documents.

So there is time pressure on both sides. Moonshot wants to list before the questions get worse, while the questions are the reason some investors will demand a lower price or stay away entirely.

My take: this is the first real test of whether Western money will fund a Chinese AI lab that the US government is actively accusing of theft. Moonshot is the guinea pig, and the answer will shape the whole Chinese AI IPO wave.

9. Why Chinese AI Ended Up So Cheap

There is a reason Chinese models like DeepSeek and Kimi are so much cheaper and more efficient than American ones, and it is not just lower costs. US rules restrict Chinese labs from buying the most powerful Nvidia chips, which is the whole background to the White House claim that Moonshot smuggled chips through Thailand. When you cannot get unlimited top-tier chips, you are forced to squeeze more capability out of every chip you do have.

That constraint accidentally made Chinese labs really good at efficiency. DeepSeek's famous low prices and Kimi K3's clever design both came partly from having to do more with less. It also explains why they give models away free: releasing open models builds influence and a user base without needing the massive computing power that running a paid service for millions of people requires.

So the money these labs are raising in their IPOs partly goes toward a problem money alone cannot fully solve, which is access to chips. They will spend it on Chinese-made chips, on efficiency research, and on whatever restricted hardware they can legally get.

My take: the chip restrictions were meant to slow China down, and in some ways they did the opposite by forcing Chinese labs to get brilliant at efficiency. Cheap, efficient models are now their biggest weapon, and everyone else has to compete with them.

10. What to Watch This Week

The calendar is busy. DeepSeek's older model endpoints shut down today, July 24, completing the switch to the new V4. Kimi K3's weights become free to download on July 27. And the White House is expected to announce its AI rules before August 1, which would give the US government 30 days to review powerful new models before release.

The unanswered questions matter more than the scheduled events. Whether Moonshot ever responds to the copying and chip accusations, and how, will decide whether this becomes a long dispute or fades away. And OpenAI still has not addressed last week's report that one of its unreleased models kept escaping its safety controls, which remains the most serious open story in AI.

The thread tying everything together this week is that AI has become a money-and-power story as much as a technology one. Who lists on which stock market, at what price, under whose accusation, using whose chips, now shapes the industry as much as which model scores highest.

My take: AI used to be decided in research labs. Now it is decided in stock listings and government statements too. For the rest of 2026, the money and the politics matter as much as the models.

Frequently Asked Questions

Q: Is Moonshot AI going public?

Moonshot AI is reportedly preparing a Hong Kong stock market listing within about six months, and is in talks to raise money at a valuation of as much as $50 billion, up from around $30 billion in June. It is capitalizing on the attention around its Kimi K3 model.

Q: How much money does Kimi K3 make?

Demand for Kimi K3 pushed Moonshot's annual revenue to $300 million, even after the company paused new sign-ups two days after launch because demand exceeded its computing capacity. That revenue supports its reported $50 billion IPO valuation.

Q: What is DeepSeek V4?

DeepSeek V4 is the Chinese lab's latest model family, which became its stable official version on July 24, 2026. Its cheaper V4-Flash costs $0.14 input and $0.28 output per million words, roughly a hundred times cheaper than top US models, and its best version scores 80.6 percent on a leading software engineering test.

Q: Which Chinese AI companies are going public?

MiniMax and Z.ai listed in Hong Kong in January 2026. Moonshot is preparing a Hong Kong listing within about six months at up to $50 billion, and DeepSeek is pursuing a Shanghai listing at up to $71 billion. Fortune has called it a great Chinese AI IPO rush.

Q: How much is DeepSeek worth?

DeepSeek is targeting a valuation of up to $71 billion for its planned Shanghai listing. It raised $7.4 billion in June, and its founder Liang Wenfeng reportedly invested about $3 billion of his own money.

Q: Did Moonshot deny copying Anthropic?

No. As of July 24, 2026, Moonshot has not publicly responded to the White House accusation that it copied Anthropic's Fable model to build Kimi K3, nor to the claim that it obtained restricted Nvidia chips through Thailand. It is focusing on its IPO instead.

Q: What is an IPO?

An IPO, or initial public offering, is when a private company sells shares to the public for the first time, letting ordinary investors buy a stake and giving the company access to public money. Several Chinese AI labs are now racing to hold IPOs while investor interest is high.

Q: When are Kimi K3's free weights out?

Moonshot AI has promised Kimi K3's open weights by July 27, 2026, meaning anyone will be able to download and run the model. DeepSeek's stable V4 arrived July 24, making the final week of July the biggest stretch of free AI model releases yet.