VCs use funds from the secondary market to realize their dreams in the primary market

The dislocation between the primary and secondary markets is completely tearing down the traditional boundaries of investment.

On one hand, most investors haven't seen carry (performance share) for a long time, so they divert part or even most of their energy to the secondary market.

Due to complex institutional terms, you may be in a position that is not eligible for carry, your project has not exited, the institution's DPI has not reached the standard, or you left your job before dawn breaks. Even if you stay long enough, your project exits, and the DPI meets the requirements, the carry may still be paid to you in installments over several years.

Most domestic investors have never seen carry.

In the widely discussed criticism of Moonshot AI by Zhu Xiaohu at the end of 2024, one focus of the debate was whether the carry from the investment in Xiaohongshu made by Zhang Yutong, former partner of GSR Ventures and current president of Moonshot AI, when he was still at GSR Ventures, could still be honored.

Without performance share, these investors, many of whom come from top-tier backgrounds and have access to cutting-edge information and sharp trading acumen, have to pin another important source of income on the secondary market.

I once heard a VC investor say that after a decade of focusing on the tech sector, the most profitable sum of money he made came from the NVIDIA stocks he bought.

On the other hand, the secondary market is accounting for an increasingly large proportion of institutional returns.

More and more funds are engaging in transactions across both the primary and secondary markets, investing in early-stage projects as well as secondary market targets, riding the favorable secondary market trends of recent years.

Meanwhile, even if you invest in an early-stage startup, it may achieve faster growth after its IPO. The paper floating profit (asset appreciation) spans both the primary and secondary markets; but if you exit at this point, the real cash you ultimately earn, by legal and tax definition, belongs to the secondary market.

For example, Zhipu AI currently has a market value of nearly HK$900 billion. At its IPO in January this year, its valuation was just over HK$50 billion, representing an increase of nearly 20 times.

To achieve the same return in the primary market, you would have had to invest in Zhipu AI when its valuation was below 2.5 billion yuan. Unfortunately, Zhipu AI was founded in 2021, and by early 2023, when people realized it could become China's representative foundational large model company, its valuation had already reached 4.3 billion yuan. In other words, only funds that could foresee OpenAI would launch GPT-3.5 and recognize Zhipu AI's large model capabilities before the 2022 venture capital downturn, or those that were extremely lucky, could earn the same 20x return in the primary market.

In short, it seems that investing in the secondary market is much easier than in the primary market now.

As a result, a group of VC investors, who manage hundreds of millions or billions of US dollars and are looking for the next world-changing genius in a garage or a residential apartment, have been devoting more effort to studying the secondary market in recent years.

Transactions Can Happen Even Without an IPO

SpaceX went public last month, with a current market value of nearly $2 trillion. If SpaceX were a country, its market value would be close to South Korea's annual GDP.

Anthropic has also submitted its prospectus, and OpenAI is seeking listing opportunities.

But many people have overlooked one thing: before these companies went public, there had already been frequent instances of employees selling old shares, institutions transferring their stakes, and SPVs being everywhere — the trading activity was even higher than that of many listed companies.

They have long been more "public" than public companies.

This is not just a phenomenon overseas. You may have seen messages in some groups seeking to buy old shares of Moonshot AI, China's remaining unlisted large model company still in the game, even though there is no official trading channel. ByteDance, the "universe factory" that has been recognized by the market for years but never gone public, has long been the dream target for people looking to acquire old shares.

More and more institutions are starting to deploy in both the primary and secondary markets; more and more startups already have market attention and pricing power comparable to listed companies even before their IPOs.

There's no time to wait around — good projects are appreciating too fast, and time waits for no one. The money in the primary market is not investing in the future, but in projects that are the future right now.

VCs are acting like PEs, eager to exit; the primary market is resembling the secondary, chasing consensus; the secondary market is starting to look like the primary, where prices rise before value is fully realized.

Amid the AI boom, a chaotic convergence is taking place.

The Primary Market Resembles the Secondary Market

For a long time in the past, the capital market had a clear division of labor: the primary market, represented by VCs, was in the business of "belief" — placing bets before a company was proven; the secondary market was in the business of "validation" — waiting for the company to go public, release data, and deliver growth, before deciding how much it was worth.

So the former bets on what will happen in five years, while the latter bets on what will happen in three months.

But in this AI cycle, many things are starting to look different: the primary market is increasingly resembling the secondary market, prioritizing certainty.

AI has taken most of the money from VCs. According to PitchBook, in 2025, the AI and large model sectors captured 65.6% of US VC deal value — that is, taking up more than two-thirds of the money in the market, without the number of target companies increasing by two-thirds.

Johnny, the initiator of Hongyue Hub whom I once interviewed, told me about the differences he observed between the past and the present: "Before Alibaba went public in 2012 and 2013, fundraising in the primary market was not as smooth as people thought — everyone thought the valuation was too high. When ByteDance reached a valuation of $70-80 billion, there wasn't the same stampede of investors as we see today; people thought its ceiling was clearly visible." But today, the concentration of top AI companies is extremely high, and everyone is flocking to them, even scrambling to get a stake. "This level of concentration is something we have never seen before."

AI does require massive amounts of capital, otherwise startups will run out of resources before they can reach the IPO finish line.

Their costs remain high: training models, expanding business, and maintaining service capabilities all require computing power, which in turn relies on data centers, electricity, and infrastructure.

Falling behind in computing power and power generation facilities means hitting a production capacity bottleneck. Even with a leading model, it is difficult to translate technological advantages into revenue growth.

Against this backdrop, securing more capital has almost become the common goal of top AI companies. Those who fall behind get eliminated.

Why are VCs willing to fund money-burning machines like Anthropic and OpenAI? This follows the PE logic of "certainty first".

Since receiving what is widely regarded as around $10 billion in strategic support from Microsoft in early 2023, OpenAI has quickly become one of the most sought-after assets in the global capital market.

In the past, the most top-tier primary projects were only available to a small number of funds, and many individuals wanted to participate but the opportunities were not public.

But in this round of financing in February this year, OpenAI raised $3 billion from individual investors. For the first time, individuals got the chance to participate in primary market projects.

Over the past two years, Anthropic has been the most coveted "room" that global capital wants to squeeze into. Amazon got in, Google got in — money from both the primary and secondary markets is pouring in.

Then Anthropic did a very interesting thing.

In mid-May, it released an announcement stating that any stock transfer not approved by the board of directors would be deemed invalid. No SPVs are allowed, and platforms like Forge and Hiive have no official authorization at all.

What does this mean?

When everyone is trying to squeeze through the same door, the person inside the door gets to decide who can enter and who cannot. The rules are written by the winners.

Then the question arises: when capital no longer places scattered bets, but concentrates its wagers on a handful of winners, can these winners really generate large enough returns to meet everyone's expectations?

Venture Capital No Longer Loves Risk

This level of concentration and consensus is unprecedented.

Over at least the past year, the global market has stopped discussing who the next OpenAI or Anthropic will be. Instead, the focus has shifted to "who can get a stake in Anthropic and OpenAI" and "who can get onto the shareholder list of the most popular projects".

When investors' discussions change from "is this worth it?" to "can I get in?", the process of value discovery has essentially ended. For many late-entering institutions, what they are facing is no longer a startup waiting to be validated, but a scarce asset widely recognized by the market.

And as more and more capital flows into the same few companies, new trading demands emerge. The market wants to acquire stakes, old shareholders want to realize their returns, and employees want to unlock their wealth early. Once consensus is formed, liquidity naturally follows.

For many investors, what they are waiting for is no longer the IPO itself. Because the market has already completed transactions long before the IPO.

And a crucial prerequisite is that, both in China and the US, the IPO market has been sluggish in the past two years, leaving a large amount of capital trapped in the primary market.

In the past, the biggest feature of venture capital was its willingness to wait. After investing in a company, investors would spend five, seven, or even ten years accompanying its growth, before exiting through an IPO.

But today, more and more capital has to focus on exit efficiency.

For many investors, the biggest risk is not misjudging a company, but misjudging the exit timing, Johnny said. "This in turn pushes investors to favor projects with a clearer listing path and more definite exit expectations."

Those who are truly passionate about primary market investment will continue to deploy capital in promising projects. Johnny believes that the change does not lie in the investment philosophy itself, but in the actual behavior of investors.

In theory, investors still want to invest in good early-stage projects, but the reality is that if you don't rush in, the valuation will double next month — how can you explain to your LPs that your judgment is still sound?

Pursuing certainty is also a way to be accountable to LPs.

The DPI crisis is a dilemma faced by both China and the US. Cambridge Associates, one of the world's largest institutional investment consulting firms, mentioned that for US venture capital funds, the median DPI of most vintages (fund establishment years) since 2015 is below 1x.

China does not have a unified public DPI database, but the figures that are circulating privately are roughly the same.

Given the exit pressure mentioned earlier and the current DPI pressure, investing in certain projects makes it easier to report to LPs. (Who knows if it's the LPs putting pressure on institutions to get into that specific project, or if it's the LPs themselves scrambling to get in?)

Venture Capital is no longer venturing, but instead pursuing Certainty and Consensus. We might as well rename it CC.

Angels Are Disappearing

The primary market's investment style is becoming like the secondary market, and the secondary market is also moving towards the primary market. Isn't this a kind of mutual convergence?

PitchBook data shows that in 2025, $194.7 billion in VC investment came from non-traditional investors, one of the highest levels in the past decade. These non-traditional investors include hedge funds, mutual funds, and crossover funds.

Typically, institutions such as Tiger Global, Coatue, Altimeter, and D1, which used to focus on trading listed tech companies, have started to frequently appear in the primary market, participating in financing for leading projects like OpenAI, Anthropic, Scale AI, and Databricks.

Among them, Coatue Management has heavily invested in public market stocks of NVIDIA and Microsoft, while placing bets on OpenAI and Scale AI. Tiger Global Management, D1 Capital Partners and other typical tech hedge funds are also appearing more and more frequently on the cap tables of top AI projects.

In addition, although this trend did not start after the AI boom, more and more institutions now have their VC and secondary market operations accounting for roughly half of their total business each.

In fact, the performance of many VCs is now being propped up by the secondary market. Many VC investors who have never received carry make their most profitable money from stocks of companies like NVIDIA and SK Hynix.

The primary market's pursuit of certainty, and the secondary market's follow-on investments in star projects — this logic may be rational, feasible, and return-generating for investors.

But what about the entrepreneurs in this era?

If you are not among the top 1% of superstars, if your track record does not have the glamorous AI label, if you cannot tell a story that excites the entire market within three months, you may not even get noticed.

We are now discussing the convergence and chaos of the primary and secondary markets, but looking back ten years from now, the golden age when a founder, a garage, and a bold world-changing idea alone could leverage capital and achieve meteoric growth seems to be over.

This article is from the WeChat official account "UnDefined Players", written by Wang Yutong, and published with authorization from 36Kr.