AI agents · OpenClaw · self-hosting · automation

Quick Answer

DeepSeek vs Moonshot vs Z.ai vs MiniMax: China's AI IPO Race

Published:

The short answer

DeepSeekMoonshot AI (Kimi)Z.ai (Zhipu)MiniMax
Status (Sep 13, 2026)Pre-IPO: CITIC Securities hired for Shanghai STAR MarketConfidential Hong Kong filing; exploring HK + Shanghai dual listingListed in Hong Kong, 2026Listed in Hong Kong, 2026
Valuation500B yuan ($75B) in current private round; >$50B post-money in June≥$50B target; $4.3B at start of 2026~$63B market cap (early Sep)~$16B market cap (early Sep)
RaiseNot set$3–5B
Recent private capital~$7.4B round (June 2026)$500M Series C (Jan), $3.5B (July)Listed out of needListed out of need
Revenue signalNot disclosed>$300M ARR~$136M H1 2026 revenue~$117M H1 2026 revenue
Stock since listing>7x YTD, but down >50% since late June~+5% YTD, ~75% below March high
Flagship modelsV4 Pro, V4.1 Flash (Sep 10)Kimi K3 (open weights Jul 27)GLM 5.3MiniMax M3, Speech 2.8

DeepSeek is the biggest and least transparent; Moonshot is the best-capitalised newcomer; Z.ai and MiniMax are the cautionary tales about what happens after the bell rings.

DeepSeek: Shanghai, CITIC and a $75 billion private mark

Reuters reported on September 9, 2026 that Hangzhou-based DeepSeek has tapped CITIC Securities to prepare a potential listing on Shanghai’s technology-focused STAR Market, with the IPO process expected to begin in 2026. Nothing else is fixed: not the timing, the raise, or the IPO valuation. What exists is a private financing that could value the company at about 500 billion yuan (~$74–75 billion) — a step-up of at least ~48% from the >$50 billion post-money of its ~$7.4 billion June 2026 round.

The choice of Shanghai over Hong Kong or New York keeps DeepSeek inside mainland capital-market and regulatory structures, which fits a company increasingly tied to national technology policy. It also imposes mainland pre-listing tutoring, disclosure and regulatory review — the first time DeepSeek’s finances will be visible. The commercial pitch is cheap inference at scale: the V4.1 Flash launch on September 10 ($0.15/$0.60 per MTok off-peak, $0.003 cache hits) landed the day after the IPO news.

The risk overhang is reputational. Anthropic’s September 10 threat report alleges DeepSeek relayed some customer requests to Claude without disclosure and harvested reasoning traces for training. DeepSeek had not responded in detail at publication; a STAR prospectus will have to address data-handling practices.

Moonshot: Hong Kong’s largest 2026 listing, if it prices

Beijing-based Moonshot AI, maker of Kimi, confidentially filed for a Hong Kong IPO in early September 2026 (Reuters, September 3; WSJ), targeting a valuation of at least $50 billion and a raise of $3–5 billion. On September 10 SCMP reported it is exploring a dual Hong Kong–Shanghai listing, with an offering possibly by October.

Moonshot’s distinguishing feature is that it did not need to list. It raised a $500 million Series C in January 2026, then $3.5 billion in July, taking total 2026 funding past $8 billion from a ~$4.3 billion valuation at the turn of the year. Its models — Kimi K3 open-weighted on July 27 at $3/$15 per MTok — have brand recognition outside China that Z.ai and MiniMax lacked at listing. The open question is revenue: Moonshot recently crossed $300 million ARR, not radically above Z.ai’s and MiniMax’s half-year figures, yet it is asking for roughly the same valuation as Z.ai commands after a 7x run.

Moonshot also appears in Anthropic’s report, accused of silently forwarding ~300,000 Kimi requests to Claude over ten days. Expect underwriters to be asked about it.

Z.ai and MiniMax: what listing actually looked like

Z.ai (formerly Zhipu) and MiniMax became China’s first public frontier-model labs when they listed in Hong Kong earlier in 2026 — partly, per Asia Tech Review, because their cash was dwindling. Their first earnings reports set the reference points:

  • Z.ai: ~$136 million H1 2026 revenue; market cap just over $63 billion; shares up more than 7x year-to-date but down more than half since late June. It has pivoted toward cloud-based services aimed at a global audience, and its GLM models earned respect abroad.
  • MiniMax: ~$117 million H1 2026 revenue; market cap about $16 billion; up ~5% in 2026 and roughly three-quarters below its March 2026 high.

The lesson for DeepSeek and Moonshot is that Hong Kong retail enthusiasm inflates and then reprices. Both incumbents now trade on revenue trajectory, not narrative.

Why the race is happening now

  1. Compute costs are permanent. Training, inference capacity and GPU access need capital on a scale private rounds cannot repeatedly supply; public equity also funds acquisitions and retention packages.
  2. The US comparison is coming. Anthropic and OpenAI are both moving toward listings; Chinese labs want valuations set before investors can benchmark them against American peers with larger revenues.
  3. Windows close. STAR and Hong Kong investors are receptive to strategic-technology issuers today; the Z.ai drawdown shows how fast sentiment turns.
  4. Regulatory alignment. Shanghai and Hong Kong listings keep data, disclosure and ownership inside frameworks Beijing controls.

What to watch

  • DeepSeek’s first filing — the first look at its revenue, margins and compute spend, and its answer on the Claude-relay allegations.
  • Moonshot’s pricing and venue — whether the dual listing happens and whether $50 billion holds against $300 million ARR.
  • Z.ai’s next quarter — the stock’s post-June slide is the market’s live verdict on Chinese AI unit economics.
  • US-listing calendar — an Anthropic or OpenAI S-1 would reset comparables for every Chinese lab.

Sources