The Underwater World of China's 1.5 Market (5 Years—Issue 183—5,032 Transaction Leads)

Summary

• China’s five-year panorama of the 1.5 market: across Issues 1–183 of 36Kr’s “Insider Message Board,” 5,032 transaction leads covered 410 targets in 35 industries; 63% were sellers, 29% buyers, and just 8% M&A—“at a high level, these messages are about selling.” Transaction volume was 318 in 2021, peaked at 1,436 in 2024, fell to 750 in 2025, and stands at 205 so far in 2026, moving inversely to the IPO window: U.S. listing restrictions and a closed IPO approval channel pushed activity underwater, while Hong Kong’s reopening pulled liquidity back into the secondary market.
• The market’s most informative signal is the buyer share: it was just 12.4% before 2022, reached 56.6% in 2026, and for the first time in five years exceeded the seller share. Capital is actively hunting hard-tech names such as newly listed Unitree and BrainCo, as well as already listed Changxin Technology, while consumer and internet names crowd the sell list—“capital has effectively voted with its feet across every market”; secondary-share transactions are a leading indicator of sentiment and the public market.
• ByteDance topped the rankings in all four eras, appearing 562 times, or more than 11% of all leads; SpaceX followed with 260 and SHEIN with 224. 明浩’s view: “The world before Didi blew up in June 2021 was the old world… If ByteDance and Xiaohongshu aren’t among your investment targets over the next five years, you can take it as a total wipeout.”
• The supply-demand ratio is the clearest read on market recognition of private companies: Xiaohongshu is the only net-buy name on the list, with 129 buy requests versus 63 sell offers—the exact inverse of the marketwide 1:2 ratio—and its valuation may recently have reached roughly $50B. Animoca, dubbed “the Tencent of Web3,” had 135 sell offers and zero buy requests. SHEIN’s 1:4.5 ratio maps to a valuation decline from $100B to roughly $40B; Discord is at 1:10 and DJI at 1:7—the most direct test is whether anyone actually accepts its primary-market valuation.
• U.S. primary-plus trading hit a record $240B in 2025, with Pre-IPO at $106B, or more than 40%; Q1 2026 alone reached $233B, and 明浩’s off-the-cuff estimate puts full-year growth potentially above 100%. Charles Schwab bought the trading platform Forge for $660M in March 2026, bringing the participation threshold down to $5K; Forge’s “real Mag7” index—SpaceX, OpenAI, xAI, Databricks, Anthropic, Stripe and defense company Anduril—rose 94.7% in 2025, versus 17.7% for the S&P 500.
• The other side of underwater trading is loss of control: SPVs have “reportedly gone six or seven layers deep,” and by the seventh or eighth layer, trades may already be beyond OpenAI’s visibility, while each transaction prices the underlying asset differently. That is why OpenAI and xAI issued near-simultaneous notices this June warning that unapproved SPVs “may be disavowed at any time”—明浩 believes the warnings were tied to intense trading activity, IPO preparations and efforts to trace share ownership.
• The entire Chinese data pipeline—from collection and cleaning to dashboards and the final deck—was handled by Qwen Work Assistant. 明浩’s working takeaway is that humans are responsible for defining task boundaries, structuring complex tasks and dynamically reviewing outputs. The conclusion is blunt: “There is no need for excessive fear; these Agents are already extremely capable today, so start using them now.”

Deep dive

1. What Is the 1.5 Market? A Transitional Trading Layer Built Around Three Real-World Problems

• 明浩 opened with three questions: how can a startup’s third- or fourth-largest shareholders cash out at a “right price”; how do early employees sell options that are “just scraps of paper until they can be monetized”; and when a VC fund’s “seven plus two” life reaches year nine and it “has to settle the books,” who takes the stake in a portfolio company that has not IPO’d but still has value? Then there are the Chinese guests who claim to be “SpaceX investors”: “SpaceX has never raised money in China, so how did their investors get in? Through what channel? At what price?”
• The definition is straightforward: a transitional equity market between the primary and secondary markets, providing a transfer channel for shares that are not yet publicly tradable. The U.S. has mature platforms, variously referred to as S-transactions or the private-equity secondary market; the concepts are somewhat distinct, but they are probably variations on the same form.
• Three forces created the market: the urgent need to monetize equity before an IPO; a “pricing anchor” formed without a new financing round—especially important for companies approaching an IPO and seeking a stable stock price, with SpaceX conducting many such transactions before listing; and the dense rollout of U.S. regulations in recent years alongside a group of “正规” formal trading platforms.

2. From Niche to Mainstream: $240B in U.S. Trading in 2025, Potentially Double in 2026

• U.S. primary-plus trading reached a record $240B in 2025, with Pre-IPO accounting for $106B, or more than 40%, and annualized growth above 50%. Q1 2026 alone reached an all-time high of $233B, according to FactSet. 明浩’s off-the-cuff estimate was that full-year growth of “more than 100% is entirely possible,” because SpaceX, Anthropic and OpenAI—companies valued above $1T and potentially even $2T—are moving through the Pre-IPO phase.
• Three factors are driving the market mainstream: traditional brokerages and investment banks are entering, as shown by Charles Schwab’s $660M acquisition of Forge in March 2026; the threshold has fallen to $5K—“you can say you are a SpaceX investor, and you genuinely are”; and this cohort of AI stars is collectively staying private, including OpenAI, Stripe, Ramp, Midjourney, Harvey, OpenRouter and Cohere. Their valuations have multiplied several times over, making supply and demand “extremely, extremely hot.”
• The logic fits in one sentence: “Liquidity is always the core metric in financial markets.” Hot names are rapidly revalued in the primary-plus market, attracting incremental capital and creating a virtuous cycle of expansion.

3. Forge’s “Real Mag7” Index: Outperforming Public Markets, but Increasingly a Game for a Few Giants

• The constituents are SpaceX, OpenAI, xAI, U.S. defense company Anduril—which 明浩 calls “America’s DJI”—Databricks, Anthropic and Stripe. The index rose 94.7% in 2025, versus 17.7% for the S&P 500 and 20.8% for QQQ. In Q1 2026, the S&P 500 fell 4.4% and QQQ fell 5.9%, while the index gained 16%. 明浩 also said it may have outperformed the S&P 500, QQQ and various benchmark indexes by more than 10x over the past two or three years.
• Sector rotation is visible in the sequence from AI leadership to chips taking over: AI-related names rose 191.3% in 2025; aerospace and defense led in Q4, followed by chips. Chips rose 354.1% in the first half of 2026 and 424.2% over the past 12 months; defense gained 200% over 12 months, while consumer names fell 11%. “This is not merely a secondary-market pattern; the primary-plus market is doing the same thing. The themes are rotating.”
• SpaceX is the case study: an internal buyback valued it at $200B in August 2024, $800B in December 2025, and $1.3T at listing. SPVs reportedly went “six or seven layers deep,” with countless economic entities participating—explaining why so many people in China say they are SpaceX investors. But 明浩 warned that many investors advised against buying around the IPO: “Its lock-up will unwind very quickly… Can the market absorb that much liquidity? That’s the question.”

4. The Other Side of the Underwater Market: Layered SPVs and the Companies’ Pushback

• The market is called “underwater” because the transactions are theoretically outside the company’s control. A fund has LPs, those LPs have another tier of LPs, and trades at the seventh or eighth layer seem to have lost any connection to OpenAI’s layer—OpenAI may not even know about them. “It has become possible to participate with just $5K. How could it know about everything?”
• But each transaction prices the underlying asset differently. That is why, on roughly the same day this June, OpenAI and xAI both posted notices on their websites warning that they could refuse to recognize unapproved SPVs and underwater transactions at any time. 明浩 inferred that the warnings were related to the two companies’ intense trading activity at the time, IPO preparations and efforts to trace the ownership chain.

5. Methodology and China’s Market Panorama: 5,032 Leads, 63% Sellers

• The China analysis is based on Issues 1–183 of 36Kr’s “Insider Message Board”: five years, 5,032 leads, 410 targets and 35 industries—the “bulletin board for China’s 1.5 market.” Qwen Work Assistant, the episode’s sponsor, handled the full workflow despite major formatting differences across the five-year dataset, from collection and structuring in Excel to cleaning, tagging, dynamic dashboards and the final PPT.
• The breakdown was 63% transfers, 29% buy requests and just 8% M&A requests—two sell offers for every buy request. That confirms monetization as the market’s primary reason for existing: “at a high level, these messages are about selling” (“大面上来看，这些消息是在卖”).
• The annual curve runs counter to the secondary market: 318 leads in 2021, 980 in 2022, 1,343 in 2023, a peak of 1,436 in 2024, 750 in 2025 and 205 so far in 2026. U.S. listing restrictions and blocked IPOs drove activity underwater; over the past two years, “listing in Hong Kong is no longer that difficult,” pulling liquidity back into the formal secondary market. M&A leads only emerged when IPO exit pressure spiked in 2024, then declined again as Hong Kong and A-share liquidity loosened—“mutually reinforcing, very interesting.”

6. From Finding Buyers to Finding Assets: Buyer Share 12%→57%

• The biggest inflection point was the buyer share: before 2022, buy requests accounted for only 12.4%, less than one-eighth; in 2026, the figure rose to 56.6%, marking the first time buyers outnumbered sellers. Capital is actively seeking newly listed hard-tech names such as Unitree and BrainCo, along with already listed Changxin Technology, while high-valuation consumer and internet names from the previous era are flooding the sell side. “That is the story of the previous era.”
• The rotation path is clear: 2021–22 brought internet platforms and consumer names, including ByteDance, SHEIN, Xiaohongshu and HEYTEA; 2022–23 brought Web3 and the metaverse, including Animoca, Discord and Epic; 2023 brought the AI breakout, including OpenAI, xAI, Kimi and Anthropic; and from 2024, hard tech rose to the fore, including Unitree, AgiBot, Changxin and LandSpace. The timing is slightly offset from the primary market and even the A-share market, but the trend is unchanged.
• The conclusion is direct: “When robots replace consumer names as the No. 1 hot-search topic, capital has effectively voted with its feet across every market” (“当机器人取代消费成为热搜第一”). Secondary-share trading is a leading indicator of sentiment and the public market.

7. ByteDance, Clear Five-Year Leader: Remnants and Afterglow of the Old World

• ByteDance ranked first across all four eras, appearing 562 times, or more than 11% of all leads. SpaceX followed with 260 appearances, or 5%, then SHEIN with 224; Musk’s brain-computer-interface company Neuralink and Xiaohongshu came next. Neuralink’s presence alongside SpaceX has the same origin: when Musk was arranging SpaceX transactions, he also arranged various SPV trades in Neuralink, bringing large numbers of Chinese individuals and funds into them.
• SHEIN ranks third because its peak valuation reached $100B but has now fallen to roughly $40B. “For the past few years it has been constantly working through an IPO, constantly trying to determine whether it is a Guangzhou company, a Singapore company or something else.”
• The lower end of the 2025–26 rankings was reshuffled: AgiBot ranked third, Unitree fourth, Rokid—the glasses maker—fifth, and Volant Aerotech sixth, putting hard tech clearly in the ascendant. 明浩’s strongest call was: “The world before Didi blew up in June 2021 was the old world… If ByteDance and Xiaohongshu aren’t among your investment targets over the next five years, you can assume you’ve been wiped out. This tally seems to prove I wasn’t wrong.”
• Xiaohongshu’s valuation rose from more than $10B, to the $32B price Zhu Xiaohu once said it was being sold at, and perhaps roughly $50B most recently. SHEIN, by contrast, fell from $100B to roughly $40B.

8. Supply-Demand Ratios Best Capture Market Recognition—and Lessons from Working with Agents

• 明浩 called this the most telling cut of the data. Against the marketwide 1:2 buy-to-sell ratio, ByteDance was 226:336, or 1:1.5, and remained in demand; SpaceX was 101:159; Neuralink was 106:113, or roughly 1:1. SHEIN was 42:182, or 1:4.5—“there are plenty of sellers, but not nearly as many buyers.” Cainiao was 1:5, Discord 1:10 and DJI 1:7. DJI’s business is strong, but it has seen “no capital-market action” for a long time, leaving its investors impatient.
• The two extremes were Xiaohongshu, the only net-buy name on the entire list, with 129 buy requests versus 63 sell offers—an exact reversal of the marketwide 2:1 seller-to-buyer ratio—and Animoca, billed as “the Tencent of Web3,” with a peak valuation that should have exceeded $7B, after once listing and later delisting in Australia. Animoca had 135 sell offers and zero buy requests. “Supply and demand determine how coveted a name is—or whether anyone actually accepts its primary-market valuation.”
• The closing lesson from working with Agents was that humans handle “defining task boundaries and structuring complex tasks,” along with reviewing intermediate outputs and dynamically signing off on results; Jeff Dean’s evaluation-experiment loop is the model. General-purpose Agents evolved from coding agents are “too specialized,” creating a faint sense of fear, while work-assistant products do not. “There is no need for excessive fear; these Agents are already extremely capable today, so start using them now.”