NeuroAI NEUROAINEUROAI.SITE
ESC

China's AI Exit Crunch, and the 34.6 Billion Yuan Secondhand Fix

With IPOs scarce, Chinese AI venture capital is turning to S funds and fund-stake transfers; the Shanghai platform has now moved 34.6 billion yuan of secondhand stakes.

2026-11-02 · 1171 words · NeuroAI
China's AI Exit Crunch, and the 34.6 Billion Yuan Secondhand Fix

A fund raised in 2021 with a seven-year term reaches 2026 holding AI companies that are alive, growing, and nowhere near an IPO window. The fund has already been extended once. Its limited partners (LP) want their money back, and the general partner (GP) has one asset left to sell: the stake itself.

That is the deal China's private equity market has spent two years learning to do. It is called a secondary, or an S fund (S基金) transaction, and it has become the key plumbing in Chinese AI financing.

Key takeaways

  • Shanghai's fund-stake transfer platform had completed 169 transactions totalling about 34.6 billion yuan (≈ US$4.9B) as of 30 April 2026 — the largest such market in China.
  • In the first four months of 2026, completed transactions there reached 3.80 billion yuan (≈ US$535M), up about 1,084%.
  • Fund-stake pledges are a parallel channel: 89 pledges raised 22.49 billion yuan (≈ US$3.17B).
  • More than 90% of Chinese secondary deals are still simple single-stake transfers, according to Coller Capital, versus roughly half of global volume being GP-led restructurings.
  • Policymakers responded in 2026: Beijing wrote S funds into its five-year plan, and Shanghai published rules allowing market-based pricing for state-owned capital (国资) stake transfers.

Why the exit door narrowed

Chinese venture capital has historically exited one way. A senior manager at a local state-owned investment company, quoted by the Shanghai Securities News, put it plainly: over 90% of venture exits used to come via IPO; now they need M&A and equity transfers.

The IPO route has not closed, but it has become selective. Shanghai's July 2026 direct-financing measures push to extend the STAR Market (科创板) fifth listing standard (第五套上市标准) to AI companies and to guide large model (大模型), embodied AI (具身智能), quantum computing and brain-computer interface firms toward listing — an admission that the gate is too narrow for the 2020–2023 pipeline.

M&A has picked up some slack. PwC's mid-2026 review counted 7,316 China M&A deals in the first half — up nearly 30% year on year, worth US$241.6 billion, up 56%, both three-year highs — with ten high-tech mega-deals in AI, storage and semiconductors.

But M&A only helps companies someone will buy whole. For a 2021-vintage fund holding minority stakes in forty AI startups, neither an IPO nor a trade sale is available. The stake itself has to become tradeable.

What the Shanghai platform actually moved

Shanghai's fund-stake transfer platform, run by the Shanghai Equity Exchange, is the deepest in the country. Its disclosed numbers give a rare look at scale:

  • As of 30 April 2026: 198 stakes listed cumulatively, 169 transactions completed, and total completed value of roughly 34.6 billion yuan (≈ US$4.9B).
  • Pledge business: 89 completed stake pledges, raising 22.49 billion yuan (≈ US$3.17B).
  • 2025 full year: transaction financing through the platform exceeded 10 billion yuan (≈ US$1.4B).
  • January to April 2026: new transaction financing volume of 11.01 billion yuan (≈ US$1.55B), of which completed value was 3.80 billion yuan (≈ US$535M), up 1,084.42% year on year.
  • GP-led transactions (接续基金) on the platform totalled 1.5 billion yuan (≈ US$21M) in 2025.

Two things stand out. The growth rate is enormous off a very small base — 3.8 billion yuan of completed deals in four months is small next to the trillions of yuan committed to Chinese funds. And pledging has raised more money than selling: an LP who pledges a stake is borrowing against an illiquid asset rather than exiting it, which tells you the discount on an outright sale is still painful.

Who is buying, and why pricing is the hard part

The buyer base is broadening. Shanghai's July 2026 measures call for studying an S master fund and for attracting insurers, financial asset investment companies (AIC) and asset management companies (AMC) to set up or participate in S funds.

The seller base is distinctive: much of the supply comes from state-owned capital (国资) and government-guided funds. That creates a specific problem — transferring a state-owned stake below cost or book net asset value (NAV) can look like a loss of state assets, so managers are reluctant to sign.

Shanghai's April 2026 guidance from the municipal state-assets regulator addressed this. Its 16 measures let a transferring entity rely on a third-party valuation report — based on project conditions, comparable market cases and asset liquidity — to set a reasonable price adjustment, and relax valuation constraints and simplify approval.

Practitioners say the gaps are national. A deputy general manager at an equity exchange, quoted by the Shanghai Securities News, noted that rules for state-owned fund exits, transfer procedures and discounted deals remain unclear, leaving many state-backed funds unwilling to transfer.

According to Li An, chief scientist at BrainNet (脑机网), secondary markets mature in a fixed order — simple stake sales first, then GP-led restructurings, then reliable pricing — and China is still early in that sequence.

The structural gap with global markets

Yang Zhan, investment partner at Coller Capital, calls China's S fund market still in its infancy, though trending up. In a May 2026 interview he noted that over 90% of domestic transactions are single-stake deals — one buyer, one stake, one manager — whereas GP-led structures made up close to half of global secondary volume last year.

Shanghai Equity Exchange chairman Xu Jun lists the same four frictions: valuation difficulty, information asymmetry, complex structures and insufficient liquidity. The exchange has responded with infrastructure, not subsidies — a stake valuation system, an industry valuation guideline, quotation services and matching built on large models.

The direction of travel is clear even if volumes are small: Shanghai's measures also encourage continuation-fund (接续基金) and restructuring deals, and propose extending the platform nationwide.

Honest limitations

  • Data comes from official Shanghai disclosures, the Shanghai Securities News, Securities Times and PwC. China publishes no comparable national totals, so these figures cannot be extrapolated.
  • Almost no disclosed data is AI-specific. The link between S funds and AI money is structural — AI and hard-tech funds raised in 2020–2023 have long hold periods — not measured by any dataset found here.
  • Discounts to NAV, the key price in a secondary deal, are not publicly disclosed in China. Statements about pricing pain are inferences from the pledge-versus-sale split and practitioner commentary, not transaction data.
  • Policy documents describe intentions. Whether state-owned sellers transact in volume depends on rules not yet published; this article assesses no listed security and recommends no institution.

What readers can do now

  • If you are an LP in a Chinese fund nearing its term, ask your GP now: is a continuation fund contemplated, is the stake listable on a licensed platform, and what valuation basis applies? Answers are cheaper before the extension vote.
  • If you are a founder raising from Chinese funds, ask about fund vintage and remaining life. A lead investor with two years left behaves differently from one with eight.
  • Watch the implementing rules for state-owned stake transfers, not the headline policy. Whether selling below book value can be done without liability is what unlocks or freezes the market.

Related coverage

Cite this article

NeuroAI (脑机网), "China's AI Exit Crunch, and the 34.6 Billion Yuan Secondhand Fix," 2 November 2026, https://neuroai.site/a/na-capital-ai-fund-secondary-exit

Machine-readable: BibTeX · RIS · JSON · Markdown

More in “Capital & Markets” → · Back to home · Markdown version