Sold Out of GPUs, Selling Shares

Moonshot AI ran out of GPUs for paying customers on July 19 — then told investors it wants a $30 billion Hong Kong listing within six months. The same company whose open-weight Kimi K3 helped crash global chip stocks is now pitching scarcity to shareholders while turning away subscribers.

What's new: On July 19, Moonshot posted on X that Kimi K3 demand had pushed its compute cluster "close to the limits of our current capacity" and suspended new consumer subscriptions with immediate effect. That same weekend, Bloomberg reported Moonshot had distributed a shareholder resolution seeking approval for a Hong Kong IPO as soon as six months out, while wrapping a funding round that could value the three-year-old startup above $30 billion.

Why it matters: The IPO is not a liquidity rescue. Moonshot reported $300 million in annual recurring revenue in June — up from $200 million in April — and CEO Yang Zhilin has said the firm holds more than 10 billion yuan in cash reserves. This is an opportunistic land grab to set the public-market valuation benchmark for Chinese AI before DeepSeek's planned 2027 listing. Goldman Sachs and China International Capital Corporation are in talks to underwrite the deal.

The irony: Kimi K3 helped trigger a global semiconductor selloff by proving frontier AI can match U.S. leaders at a fraction of the cost. Days later, Moonshot could not serve the customers who wanted to pay for it. Scarcity did not vanish from the AI economy. It relocated from Wall Street's narrative to Moonshot's own data center.

The numbers:

• $30 billion+ — target valuation in Moonshot's current funding round, per Bloomberg and The Next Web
• $300 million — annual recurring revenue in June, up from $200 million in April and roughly $100 million earlier in 2026
• ~100x — implied revenue multiple at a $30 billion valuation on $300 million ARR
• 2.8 trillion — parameters in Kimi K3, the model whose demand overwhelmed Moonshot's cluster
• July 27 — scheduled release of K3's full open weights under a Modified MIT license
• Six months — target window for the Hong Kong listing after the July 19 shareholder resolution
• 70%+ — share of Moonshot revenue now coming from enterprise API contracts, per BigGo Finance citing people familiar with the business

The quotes:

• Moonshot AI, X post, July 19: "Kimi K3 has received far more love than we expected, and our GPUs are feeling it. Over the past 48 hours, demand has pushed close to the limits of our current capacity."
• Moonshot, same post: "To protect the experience of existing subscribers, we're temporarily pausing new subscriptions and prioritizing compute for current members."
• Lian Jye Su, chief analyst at Omdia, to ABC News, July 20: "This does show Moonshot AI does not have sufficient compute chips to serve the current surge in demand."
• Su, same interview: K3 is "very demanding" in compute requirements, making allocation "challenging and expensive."
• Morgan Stanley analyst Gary Yu, cited by Decrypt, July 18: "K3 has received positive feedback globally, signaling an all-round catch-up of Chinese LLMs with U.S. leaders in model size, performance, and pricing."

The IPO mechanics: Moonshot is dismantling its offshore VIE structure — a prerequisite for a smoother Hong Kong listing under China's revised CSRC rules, per The Next Web. The shareholder resolution is the formal start of a process that typically precedes a listing by about six months. Backers include Alibaba, Tencent, Meituan, and China Mobile. Yang Zhilin, a former Tsinghua professor who worked at Meta and Google, co-founded the company in 2023.

The race: Moonshot is not the only Chinese lab eyeing public markets. DeepSeek is targeting a 2027 IPO. Z.AI is reportedly on track for $1 billion in annual sales — three times Moonshot's current ARR. Moonshot's edge is timing: K3 gave it a live demonstration that no roadshow slide could replicate, even if the demonstration also exposed the compute gap.

What changes July 27: The full K3 weight drop will let any lab run the model locally without API fees. That is the event Wall Street feared when chip stocks sold off — and the event Moonshot is racing to price before independent evaluators and open-source adopters settle the benchmark debate. Alibaba previewed its own 2.4-trillion-parameter Qwen3.8-Max on July 19 at WAIC Shanghai, promising open weights "soon" with no date set.

What we cannot verify: The 70% API revenue share comes from a single outlet citing unnamed sources. Moonshot has not published audited financials. The $30 billion valuation may close higher or slip toward the $20 billion figure some reporting cites from its May Meituan-led round. Whether CSRC and Hong Kong regulators clear the listing before K3's open-weight release stress-tests the commercial moat remains an open regulatory question.

Convina's view: Moonshot just proved the bull and bear cases for Chinese AI in the same weekend. The bull case: frontier models that rival Claude and GPT, $300 million ARR tripling in months, and a Hong Kong IPO that could reprice the entire sector. The bear case: a company so compute-constrained it cannot take your money, pitching a 100x multiple into a market that just entered a semiconductor bear market because it believes AI is getting cheaper. The subscription pause is not a crisis — it is the most honest slide in the prospectus. Investors are not buying a model. They are buying the rack bill for a model that ships free on July 27.