Every Market Begins in the Street

If you work at a company people want a piece of, you have already received the message. It arrives on LinkedIn, from someone you have never met, with a title like Director of Private Markets at a firm you have never heard of. It says they are working with buyers for shares of your company. It quotes a price, usually a premium to the last round and sometimes a number your own finance team has never seen, and asks whether you would be open to a confidential, no-obligation conversation about liquidity for your vested equity. Your colleagues got the same message, at a slightly different price. Nobody at your company sent it. Nobody at your company can tell you whether the number is real. And if you ask, the answer is that any transfer requires board approval, which answers a question about permission and says nothing about price.

That message is the market forming. It has a bid, a spread and a sales force; the bid is usually an inquiry about supply dressed up as an offer, to be shopped to buyers or wrapped into a vehicle once the seller says yes. The only party not in the room is the company. Founders usually find out when an employee forwards the message, or when the number in it becomes the price everyone quotes. The most closely watched private company in the world found out from the news.

On a Monday at the end of June 2025, at a presentation in Cannes, Robinhood announced a giveaway of “OpenAI tokens” and “SpaceX tokens” to retail investors across the European Union, the opening move in a plan to sell tokenized private companies to the public. OpenAI learned of it when everyone else did. Its reply, posted two days later from the company's newsroom account, is as close as a corporate statement gets to a shout: “These ‘OpenAI tokens’ are not OpenAI equity. We did not partner with Robinhood, were not involved in this, and do not endorse it.” Robinhood's chief executive did not dispute a word of it. The tokens, he agreed, “aren't technically equity,” but they “effectively give retail exposure to these private assets.” Then he added the detail that deserved more attention than the dispute: since the announcement, “we've been hearing from many private companies that are eager to join us.”

Both were right, and that is the problem. OpenAI had acquired a public price, a number that moved by the hour, on a venue the company had not chosen, for an instrument it had not issued. Until then its value had been set once every eighteen months or so, in a conference room, by people it had chosen. The market for OpenAI had formed. OpenAI was the last to know.

It did not stop there. By May of 2026, OpenAI and Anthropichad both rewritten their transfer policies to declare any unauthorized sale of their shares void, whether direct, through a special-purpose vehicle, tokenized, or by forward contract. Void in the legal sense: a sale the company will treat as never having happened. “The sale will not be recognized,” OpenAI's policy now reads, “and carry no economic value to you.” Anthropic went further and published a list of eight venues whose transactions it would not honor, among them some of the largest names in the industry. A token tracking Anthropic fell from $1,400 to $900 within a day. A month later SpaceX went public, and investors who had bought in through vehicles stacked four and five layers deep learned that they would not know how many shares they owned, or whether they owned any, for months, and in the bottom layers most of a year.

This is the state of the market for the most valuable companies of our era. A prediction market will quote you odds on when OpenAI lists and what it will be worth at the close of its first day, refreshed by the minute; nobody, including OpenAI, can quote you a price for OpenAI today. On a single day in February 2025, the same share of Cerebraswas quoted at $13 on one venue and $57 on another. A bond trader from 2001, the last time a major market worked like this, would recognize every part of it: the phone calls, the spread, the middlemen whose entire business is knowing something you don't.

It was not always so. For most of the twentieth century a company that mattered had a price, and the price was a fact, printed in the morning paper, the same for everyone. Coca-Cola, General Motors and Standard Oil were quoted daily on a New York sidewalk before the Stock Exchange would have them, and on the Big Board after. A company grew up in public, and its price, set by thousands of strangers who disagreed with one another, was the most honest thing about it. The exchange is one of the great inventions of civilization. It turned ownership into something that could be measured, divided and moved, and in doing so it built the modern economy. And over the last thirty years we have quietly stopped using it for the companies that matter most. In 1996 there were more than 8,000 listed companies in the United States; today there are about 4,700. The average unicorn now spends nine years private before it lists, if it lists at all; more than 1,200 of them are waiting. The entire growth phase of this era's defining companies, the years in which nearly all of the value is created, now takes place inside a legal structure that was never designed to host a market, and that is hosting one anyway.

The brokers, the platforms and the vehicle sponsors who work inside that structure are doing difficult, mostly honest work in a market with no rails, and much of what they charge is the price of that work: finding a seller, finding a buyer, papering a transfer, waiting out a right of first refusal. But look at where the margin comes from. It is the spread: the distance between what the seller received and what the buyer paid, and the layers in between. A market with one price and one layer would erase it. History is unambiguous about what intermediaries do with that incentive. When they have built venues, they have built them to protect the spread: the Buttonwood Agreement of 1792, the founding document of the New York Stock Exchange, was a promise among twenty-four brokers to fix their commissions and deal only with each other, and the fixed commission survived until the government abolished it in 1975, over the objections of an exchange whose chairman warned that competition would “freeze the public out of the securities markets.” When regulators began publishing corporate-bond trades in 2002, the dealers warned that transparency would come at the cost of liquidity. Investors' trading costs fell by roughly half; the dealers' margins, headcount and market share fell with them. The public exchanges will not build this either; they were built for the disclosure regime of a mature company, and a company at its Series B cannot live under it. A market formed with the company has no infrastructure yet. It has to be built, and it will not be built by the people the spread pays.

The exchange has always worked this way, with the company. A company arrives on the New York Stock Exchange by applying: it files a registration statement, signs a listing agreement, and discloses on the exchange's schedule, and in return its shares trade in one place, at one price, in front of everyone. The venue brings the rules and the price; the company brings its consent and the facts. The private market of 2026 has exactly half of that arrangement. The company's consent has become decisive, as the void clauses and the collapse of the Anthropic token showed, and there is still no venue for the company to bring it to.

One platform will not be enough for this. Every company of consequence will eventually list somewhere long before it goes public, on terms it helped set, and that will require standards, infrastructure, and a generation of engineers who might otherwise spend their careers shaving microseconds off the public market's plumbing. When the over-the-counter market got its first screen in February 1971, the system was built by Bunker Ramo, a manufacturer of military electronics, because nobody in finance could build it. The tools are better now, and the problem is larger.

Some will say founders want it this way. Opacity has its uses: it protects a weak last round, it keeps employees from watching the price, it lets a story run ahead of the numbers. That was true for as long as opacity was on offer. It no longer is. The choice left to a founder in 2026 is between a price they helped form and a price a stranger formed for them, and the void-transfer policies of this spring are what it looks like when founders discover that at scale. Prohibition buys time, and only time. The New York Stock Exchange tried to cut the bucket shops off from its ticker in 1889 and abandoned the embargo within days, because demand does not disappear when you stop serving it. It goes somewhere worse.

Some will say regulation makes this impossible. Regulation made Nasdaq possible. Congress mandated a national market system in 1975, and the JOBS Act of 2012 rewrote the rules of staying private in a single bill. Every rail a private-company market needs already exists in some form: the tender offer, the transfer agent, the regulated venue, exemptions written for exactly this purpose. All of it exists in pieces, waiting for someone to assemble it.

And some will say ordinary investors have no business in private companies. They are already there, through the worst door available: tokens the company disavows, vehicles that take two-and-twenty at every layer, brokers charging as much as eighteen percent for access. The public is in. The only open question is whether there will be rails under the next wave.

None of this is new; it is the oldest story in finance. Every market that has ever mattered began exactly this way, in the street, at a discount, without permission, and the ones that became engines of prosperity did so for reasons that are specific, repeatable, and almost entirely forgotten. To see what happens next, and why the outcome is not yet decided, it helps to start at the beginning, which comes long before 2012 or 1971. The beginning is Amsterdam, 1602.

Every Market Begins in the Street

The venue came centuries before the share. From about 1285 the Van der Buerse family kept an inn in Bruges, and by the fourteenth century the square in front of it had become the city's commercial center, where merchants dealt in bills of exchange, goods and, above all, information. The family's coat of arms showed three purses, beurzen, and the square took their name; every bourse, borsa, bolsa and Börse in the world is named after that square. Exchange rates were published there regularly from 1370. When Bruges declined, the merchants moved to Antwerp, which put up the first building designed as an exchange in 1531. The order of invention matters. The meeting place came first, then the published price, and only then, centuries later, the tradable share.

In August 1602 the subscription books of the Dutch East India Company closed. In Amsterdam alone, 1,143 people had put in 3,674,945 guilders, and across the company's six chambers the total came to almost 6.5 million. The charter that created the company contained a sentence that had never appeared in one before: “Conveyance or transfer may be done through the bookkeeper of this chamber.” A share could be sold. To sell one, the seller presented himself at the East India House, two directors approved the transaction, and the bookkeeper entered it in a special register. The company had created the first tradable share, and with it the first secondary market, and it had placed its own register at the center of that market from the first day.

The trading itself happened wherever traders could stand. Deals were struck on the Nieuwe Brug, the northernmost bridge over the Damrak by the harbor, and in St. Olaf's Chapel nearby when it rained; the city had let merchants use the chapel in bad weather since 1586. Amsterdam did not open an exchange building until August 1611, nine years after the shares existed. For nine years the world's first stock market was a bridge, a chapel and a ledger. By 1688 it had grown sophisticated enough that Joseph de la Vega, a merchant who traded there, wrote the first book about a stock exchange ever published. He called it Confusión de Confusiones, and he gave its shareholder character a piece of advice that has aged well: whoever wishes to win in this game “must have patience and money, since the values are so little constant and the rumors so little founded on truth.”

The building, when it comes, is put up by the intermediaries, and it is put up to protect them. On 17 May 1792, twenty-four New York brokers signed that single sentence under a buttonwood tree on Wall Street. It is the founding document of the New York Stock Exchange, and it is a price-fixing agreement. What made the Exchange something more than a cartel came later, and it came from the companies. In 1853 the Exchange began requiring listed companies to provide complete statements of their shares outstanding and capital; in 1899 it required regular financial statements from every listed company; in 1910 it abolished the Unlisted Department through which shares had traded before a company qualified. The venue set the terms, the company met them, and in return its shares traded in one place at one price. A listing has been a partnership between the venue and the company for a hundred and seventy years.

Chicago showed what a standard does. The city became the center of the American grain trade in the 1840s, and the Board of Trade was founded in 1848 into a market of “repeated gluts and shortages, and chaotic price fluctuations,” where every lot had to be inspected and priced on its own. The Board's contribution was to make grain interchangeable: graded, warehoused, and from 1865 traded in standardized futures contracts. Once a bushel of No. 2 spring wheat was the same bushel in every contract, a price for it could exist. Amsterdam's transferable share had done this for ownership; Chicago's grade did it for goods. Every market that has ever consolidated has needed a unit that means the same thing to both sides.

There has also always been a street outside the building. From the 1860s the stocks the New York Stock Exchange would not list traded on the Broad Street sidewalk, in a stretch traders called the gorge, among “curbstone brokers” who took their orders from clerks leaning out of the office windows above. Each firm had its own hand signals, and its boys wore distinctive caps so the clerks could find them in the crowd. The companies on the Curb were the ones the Big Board considered too young or too speculative, and the list of them reads like a history of the twentieth century: Coca-Cola, General Motors, Standard Oil, Shell and Philip Morris all traded on the sidewalk before they qualified for the floor. About 80 percent of the orders on the Curb came from New York Stock Exchange member firms, which is to say the incumbents fed the market they would not host. When the curbstone brokers finally organized, in 1908, they named their body the Curb Agency and declared that it had “NO ORGANIZATION WHATEVER,” to keep it out of anyone's reach. They moved indoors on 27 June 1921 and in time became the American Stock Exchange. The market for unlisted companies had existed for sixty years before it had a roof, and the best companies in it applied for a listing and moved on.

The street also had bucket shops. From the 1870s, storefronts across America let anyone with a dollar “buy” stocks on margins as high as 100 to 1, with no shares ever changing hands; the customer was betting against the house on the next number to come over the ticker, and the house, which could see all the orders, front-ran and manipulated at will. The Supreme Court in 1906 defined the bucket shop as a place “nominally for the transaction of a stock exchange business” but “really for the registration of bets.” For the first time ordinary people could take part in the market, and they did so through an instrument that tracked the price of a share without conferring any part of one. The New York Stock Exchange's first response, in 1889, was to cut the bucket shops off from its ticker. The embargo hurt the Exchange's own members more than the bucket shops and was abandoned within days. What ended them was New York's Martin Act in 1921, a fraud statute, together with a regulated market that, by then, was cheaper and safer to use than the bet.

Fragmentation also ran along the map. In the nineteenth century an American company's shares traded where the company was, because that was where the information was: Philadelphia, Boston, Baltimore, Charleston and New Orleans all had exchanges before the Civil War, and Pittsburgh, Chicago, San Francisco, Cincinnati, Cleveland, Los Angeles, St. Louis and Detroit added theirs by 1907. Detroit traded automobiles, Los Angeles oil, San Francisco mines. The regionals also listed the companies New York rejected, on lighter disclosure. In 1934, when exchanges first had to register with the new Securities and Exchange Commission, twenty-four did and nineteen more were temporarily exempted: forty-three venues for one country's stocks. The telegraph and the ticker had already made a New York price visible everywhere; the 1934 Act imposed uniform listing standards, which removed the regionals' main reason to exist; Chicago, Cleveland and St. Louis merged into the Midwest Stock Exchange in 1949, and San Francisco and Los Angeles merged in 1953. Each step was the same step. As soon as everyone could see the same information, the reason to trade in forty-three places disappeared.

The largest fragmented market of all was never on an exchange. Thousands of American stocks traded over the counter, priced by telephone between dealers whose quotes were collected in a daily booklet, the pink sheets, printed since 1913. A 1963 study by the Securities and Exchange Commission laid out the opacity of that market, and the dealers' association answered it with a screen. On 8 February 1971 the National Association of Securities Dealers Automated Quotations system went live, built under contract by Bunker Ramo, a maker of military electronics, and for the first time a dealer in Denver and a dealer in New York could see the same quote at the same moment. The over-the-counter market was given a screen and became Nasdaq, which today lists the largest companies on earth. Congress followed in 1975 with a mandate for a national market system, and the regional exchanges never recovered from the combination.

The cleanest experiment came last, in a market that had resisted every previous consolidation. American corporate bonds traded over the counter, dealer to dealer, with no public record of prices, for the whole of the twentieth century. In July 2002 the regulator began requiring dealers to report their trades to a system called TRACE and disseminating the prices to the public. The dealers had argued that transparency would come at the cost of liquidity, and in the least liquid bonds trading activity did fall, by about 41 percent on one estimate. For the bonds whose prices were published, the cost of trading fell by roughly half, and execution costs fell by a fifth even for bonds that were not yet covered, because a published price for one bond tells you what its neighbors should cost. Dealer employment and compensation fell; the largest dealers lost market share. Nothing in the structure of the market had changed except that the price was public.

That is the whole history, and it has one shape. A market forms wherever value and demand meet, before anyone gives permission and long before anyone builds a venue: a bridge in Amsterdam, a coffee house in London, a sidewalk in New York, a telephone line between two dealers. In that state it is expensive. Prices differ from one corner to the next, the spread between them is the intermediary's income, information is the only real asset, and the people who pay most for the absence of rules are the smallest participants, the ones in the bucket shops. Then, in every case, the same three things arrive: a unit that means the same thing to both sides, a venue where that unit trades in one place, and a published price. When they arrive together the spread collapses, participation grows by an order of magnitude, and the market becomes an engine of prosperity for the companies in it. And in every case the transition was made with the companies, whose consent and whose disclosure were the price of admission: the VOC's bookkeeper, the Exchange's financial statements, the Curb's best companies applying for the floor. The shadow market is the signal that the real one is about to form.

The Retreat Into Private

On 15 May 1997 Amazon sold shares to the public at $18 each. It raised $54 million at a valuation of $438 million; it was three years old and had sold $16 million of books in the previous quarter. Everything that happened next, the climb from $438 million to more than two trillion dollars, happened in public, where anyone with a brokerage account could own a piece of it and the price was printed every morning. On 11 June 2026 SpaceX sold shares to the public at $135 each. It raised $75 billion, two and a half times the previous record, at a valuation of about $1.75 trillion; it was twenty-four years old. The public got SpaceX at four thousand times the size at which it got Amazon, and it got it at the end. OpenAI and Anthropic have filed to follow the same way.

Nothing in the nature of rockets or language models required this. What changed between 1997 and 2026 was a set of rules, each of them reasonable when written, that together moved the growth of the era's defining companies out of the public market and into a legal structure with no market in it. Three rules did most of the work, and one assumption held them together.

Five Things Followed

This is the private market in 2026. It is Amsterdam before the exchange building, the Curb before it moved indoors, the bond market before anyone published a trade. The rules written in 1982 to keep the public out of a short, closed phase now keep it out of the longest and most valuable phase in a company's life, and the public has come in anyway, through tokens and stacked vehicles, the way it once came in through the bucket shops. The 24 million households the rule admits mostly cannot find the door; the ones it excludes have found the wrong one. And the company, whose consent is now the only thing that makes any of it legitimate, has nowhere to bring it.

The venues, the vehicle sponsors and the banks are doing what a market with no rails rewards, and doing much of it well. Their margin is the spread and the layers, and they cannot be expected to build the thing that would remove it. Every venue in the private market is built around the company, and the market is still waiting for one built with it.

The last three years have settled four questions that used to be open.

Put the four together and the shape of the missing piece is plain. It is a venue on which a private company lists, in the sense the word has carried since 1853: the company applies, agrees to the venue's rules on who may participate and in what structure, discloses on the venue's schedule, and in return its shares trade in one place, in windows the company opens, at one price per window, for a fee of one or two percent of each trade. The participants hold one instrument in one layer. The company's operating cap table stays what it was. And the venue exists before the synthetic market does, so that the LinkedIn message and the token arrive to find the official route already open.

Every venue built so far routes around the company. The missing one starts from the company's consent. The company of 2026 has the consent power of a listing and no exchange to list on.

The debate about private companies assumes a choice between two conditions. A company can stay private, which means opaque, restricted and priced by negotiation, or it can go public, which means the disclosure regime of a mature corporation, quarterly scrutiny and the litigation that comes with it. The history in Section 2 shows the choice is false.

From 1602 to 1910 the world built a third condition, the listing, and then over thirty years the American company stopped using it during the phase of its life in which it creates most of its value. The new model restores the listing to that phase, in the form the private stage can use: a market of windows. It has five principles.

The largest companies already do this, at great effort, every 132 days. Investors oversubscribe nearly every window they open. The structure itself is four hundred years old. We are building the venue described here, and this document is the standard we intend to be held to.