The public owned the first venture fund. Then it was locked out.

Venture capital finances innovation, and nothing before it could do the job. A bank will not lend to a company with no collateral and no cash flow. A public market cannot price a technology that does not yet work. A venture fund can, because it concentrates people who understand the technology, writes cheques in stages as the risk falls, takes a board seat, and keeps a share of the upside large enough to pay for the losses that come with the job. The losses are the point. A venture portfolio depends on a power law: most of its companies fail, a few return the fund, and the best one can be worth more than all the rest combined, so the whole craft is finding the few and holding on to them. Since the 1950s that craft has financed the semiconductor, the personal computer, biotechnology, the internet, and most of what was built on them. It is the best machine anyone has built for turning savings into new industries.

Its shape was not inevitable. The first venture firm, American Research and Development, was founded in Boston in 1946 as a publicly traded fund, and anyone could buy its shares. In 1957 it put $70,000 into Digital Equipment Corporation for 70 percent of the company; by 1971 the stake was worth $355 million. The industry did not copy the model. Rules written for public funds made it hard to pay ARD's own people a share of the gains, so its best staff left to form limited partnerships, which could. From 1959 the limited partnership became the vehicle: a pool raised from a small number of wealthy investors, locked for ten years, run by general partners who took a fee on the assets and a fifth of the profits. When pension funds were allowed in after 1979, new commitments rose from about $200 million a year to more than $3 billion within five years, and venture capital became an asset class. The public had owned the first venture fund. The structure that replaced it was built for the manager and the regulator, and the public was left outside it.

For four decades the structure fit the job. A fund raised its money, spent five years finding companies, and had five more to see them to an exit. In 1999 the median technology company went public four years after it was founded, so a ten-year fund could back a company at birth and still hold it through its first years as a public stock. Carry aligned the partner with the investor, because the general partner got rich only if the limited partners did. And the economics were simple: venture money bought sales capacity, and when the company reached scale you sold it to an acquirer or to the public. The internet made the outcomes bigger, so the funds got bigger, and the logic held.

There was always a problem at the centre of it, and it is an old one. Whenever one person acts with another's money, their interests overlap but are not the same. Economists call this the principal-agent problem, and a venture fund is a hard case of it, because the investor hands over every decision for a decade and gets a quarterly report. The manager's fee rewards raising more, the deployment schedule rewards spending it, the carry rewards taking big swings with other people's money, and the next fund rewards marking the current one up. The investor's own agent, the officer who chose the fund, prefers marks that never show a loss. None of these people is dishonest. It is that nobody in the chain is paid to do what the owner of the money would do. The industry knows this, which is why the fund agreement, the advisory committee, the key-person clause and the ILPA principles exist: an apparatus for managing the gap. It manages it. It cannot close it, because the manager still holds every decision, the owner cannot intervene or leave, and everything the fund learned lives in partners who eventually walk out of the door.

For a long time the apparatus was enough, because the returns covered it. The model is not broken. It is out of date, because the world it was designed for has changed in three ways, and each makes the others worse.

The power law has steepened, and the clocks have come apart. The top one percent of American venture-backed exits produced 17 percent of exit value between 2005 and 2010, 34 percent between 2017 and 2022, and 80 percent since 2023, or 45 percent if you leave out SpaceX. The winners are bigger, and they compound longer in private: the median technology company now goes public twelve years after it was founded, longer than the life of the fund that first backed it. A fund built to hold for ten years is forced to exit just as its best position starts to compound, and its reserves, fixed at the start, run out when the winner needs more. The largest exits ever recorded are happening now, and the funds that found those companies were designed in a way that stops them holding on.

The funds have grown to match the winners, and the capital has concentrated. Venture is a game of winners, and owning a meaningful piece of a bigger winner takes a bigger cheque, so funds grow with the outcomes. Access compounds it. Nothing in this asset class is scarcer than access: the best companies choose their investors, and they choose the firms that backed the last winners, so the largest families and institutions allocate to the most prestigious firms, and those firms get bigger and more prestigious. In the first quarter of 2026 the five largest firms took 73 percent of all commitments, and Andreessen Horowitz alone raised $15 billion this year, close to a fifth of the total; it and Sequoia each manage around $90 billion. Fees make the loop sticky, as Manidis argues: a fund that must deploy $15 billion cannot wait, concentrate, buy at a discount or hand money back, because its investors expect capital to be called on schedule, and returning it ends the next fundraise. So the largest firms funnel enormous sums into the same consensus companies, and AI now absorbs more than 80 percent of American venture dollars.

And the exits have narrowed. The IPO window is open for a handful of names and shut for the median good company, and the acquirers who once bought the rest have pulled back. So funds mark their companies up each quarter and distribute little. Distributions have fallen to about 10 percent of net asset value from 25 percent in the mid-2010s, cash flow to limited partners has been negative in five of the last six years, private equity and venture returns have trailed public markets since 2020, and secondary sales of fund stakes are up more than 150 percent in four years as investors pay to get out early. Felipe Montealegre's reading is that we have lived through a rare period in which capital demanded no premium for being locked up, and that the premium is returning. If he is right, a ten-year claim on a manager is about to be worth less, and a liquid claim on the same companies worth more.

Investors and managers have noticed, and they are already unbundling the fund. Special purpose vehicles, one-deal funds that let an investor say yes or no to a specific company at a specific price, went from an edge case to a core strategy in 2025: formations on Carta are up 116 percent in five years, and in Odin's survey of 56 managers, 84 percent use them or plan to, most often to follow on into a winner after the fund's reserves are gone. Secondary markets let limited partners sell what they were never meant to sell. Continuation vehicles, NAV loans and evergreen funds let managers hold what the fund's term says they must sell. Each is a patch negotiated deal by deal with its own fee, and each is a signal. The market wants flexibility and control: to choose its deals, leave when it likes, and keep the winners.

The industry's own diagnosis says the same. Manidis's conclusion is that the winners of the next decade will not look like venture funds. They will be large, patient, permanent pools of unconflicted capital that can hold forever or sell tomorrow, funded by a new base of individual investors, because that is how private equity solved the same problem a decade ago; by 2023 Blackstone was raising more from individuals than from institutions. In his words, permanent capital with long durations and deep conviction is the only scarce product in the market.

He is right about the product and about where the capital will come from. We disagree about the vehicle. The most patient, long-term capital of the next decade will not be a new form of venture fund. It will be agentic capital vehicles. The industry can add patience, liquidity and a wider base of owners only by building around the manager, because the manager's judgment is the product it sells, and every patch adds a fee and a negotiation without changing who owns the fund, who decides, or who is allowed in. A structure that changes those three things has to be built from different parts.

It also has to reckon with a change the venture fund was never designed for. The fund was built around scarce judgment and expensive work, and the work is no longer expensive. What that does to the fund is where the engine comes in.