The History of Venture Capital

First version 21 October 2026, last changed 24 October 2026 · Andreas Stockter

Venture capital is money that someone gives to a young company whose failure is more likely than its success, in exchange for a share in it rather than for interest, and without collateral. In most cases the investor loses everything and in a few gains many times the stake; the business lives on the few paying for the many. Almost every company that appears in the other chronicles on this site, Intel, Apple, Genentech, Cisco, Netscape, Google, Facebook, Tesla, OpenAI, came about this way. In 2025, 425 billion dollars flowed into young companies worldwide in this manner, almost half of it into artificial intelligence.¹ The history of this money is the history of the question of who pays for the uncertain, and it is younger than it seems: as a profession and an institution, venture capital has only existed since 1946.

Like the history of AI, the history of software development, the history of the Internet and the history of open source, this text follows a chain: each era resolves a bottleneck the previous one left open, and creates the next one in doing so. Here the bottleneck always has the same shape: where does money come from for something that does not yet exist, and how does it get out again? First the institution was missing, then the legal form, then the capital, then discipline, then the exit, then patience, and lastly scale. Every technical term is set in bold at the point where it is explained, and is only used after that. A glossary lists all terms alphabetically.

Contents: 1 Prehistory · 2 Doriot and the first fund · 3 Semiconductors and the limited partnership · 4 Computers, biotech, pension funds · 5 Networks, software, first bubble · 6 Internet and the bubble · 7 Platforms, cloud, unicorns · 8 Megafunds, crypto, AI · Chain · Timeline · Further reading · Sources

1. Prehistory: money without collateral (1840s–1945)

The problem. A bank lends money against collateral and receives interest; whoever does not pay loses their house or their machines. An invention has no collateral. It has no revenue, often not even a product, and its inventor nothing a bank could seize. A stock exchange admits companies that already are something, not those that want to become something. Whoever wanted to finance the uncertain therefore had to do a different business: not lend, but take part, that is, take a share in the future profit and lose the whole stake if there is none. This business is old. The whalers of New Bedford financed every voyage in the nineteenth century through agents who collected money from merchants, paid captain and crew with shares in the catch instead of wages, and kept a part of the profit themselves. Most voyages brought little, some nothing, a few a fortune, and the agents with the best eye for captains became rich. The economic historian Tom Nicholas has shown that this form, intermediaries, participation, a share of profit and a distribution in which a few hits decide everything, was already the pattern of later venture capital.²

Inventors and bankers. The inventions of the nineteenth century found their money with individuals. Alexander Graham Bell got the money for his experiments in 1875 from two men, Gardiner Hubbard and Thomas Sanders, in exchange for shares in the patents that came of them; from the contract of the three came the Bell Telephone Company. Thomas Edison had the development of the light bulb paid for in 1878 by J. P. Morgan and the Vanderbilts, who received the Edison Electric Light Company for it, a year before the first lamp burned. Henry Ford founded his company in 1903 with 28,000 dollars from twelve backers, merchants and clerks from Detroit, who got a hundred times their money back by the time Ford bought them out in 1919. The pattern was the same each time, and each time it was chance: it depended on whether the inventor knew a rich person who believed him.³

The rich families. In the first decades of the twentieth century chance became a sideline of great fortunes. The Phipps, partners of Carnegie, founded Bessemer Securities in 1911 to put their money into new companies too; Laurance Rockefeller financed Eastern Air Lines in 1938 and the aircraft maker McDonnell in 1939; the Whitneys, the Mellons and the du Ponts did similar things. These family offices, administrations of a family’s wealth, had money, patience and lawyers, but no procedure: every investment was a decision of the patriarch, and nobody outside the family could take part. The words “venture capital” arose in these years in the banking circles of the East Coast, as a name for money that ventures something.⁴

The state and the war. The largest backer of the new was produced by the Second World War. The Office of Scientific Research and Development under Vannevar Bush paid for radar, calculating machines, penicillin and the bomb, and it paid for them at universities and companies, not in laboratories of its own. In 1945 Bush demanded in “Science, the Endless Frontier” that the state continue this funding in peacetime, and from the demand came the National Science Foundation.⁵ On the West Coast, Frederick Terman, professor of electrical engineering at Stanford, had already in 1939 set two students, William Hewlett and David Packard, up in a garage in Palo Alto with a small loan and a first customer; after the war he brought defence contracts to the university and in 1951 founded the Stanford Industrial Park, where companies of his graduates settled.⁶ In Britain the Bank of England and the clearing banks founded the Industrial and Commercial Finance Corporation in 1945 to give small companies capital that the banks would not lend them; it later became 3i, the largest European investor in private companies.⁷

The bottleneck. In 1945 there was money for the uncertain: with families, with the state, with individual professors and bankers. But there was no institution whose profession it was to hand it out. Whoever wanted to found a company needed a rich acquaintance, and whoever wanted to venture their money needed a founder they knew personally. The engineers coming out of the war laboratories had neither.

2. Doriot and the first fund (1946–1957)

ARD. In June 1946 three men in Boston founded the American Research and Development Corporation, ARD for short: Ralph Flanders, president of the Federal Reserve Bank of Boston, Karl Compton, president of MIT, and Georges Doriot, a French-born professor at Harvard Business School who had led the army’s research into equipment as a general during the war. Their idea was to make venturing a profession: a company that collects money from many, puts it into young companies from the laboratories of MIT and accompanies them over years. ARD raised 3.5 million dollars, partly from insurers and universities, partly from private people through the stock exchange, and was thus the first venture capital firm: an enterprise whose sole purpose is to venture other people’s money in young companies and to look after them.⁸ Doriot called it “creative capital” and taught his students that one does not give a company money but time: a fund must be able to wait for years and stand by the founder instead of pushing him.

The proof: DEC. For ten years the results were middling. Then in 1957 two engineers from MIT’s Lincoln Laboratory, Ken Olsen and Harlan Anderson, came with the plan to build computers that would be smaller and cheaper than IBM’s. ARD gave them 70,000 dollars for 70 per cent of the company and a loan of 30,000, and demanded that the company first sell components, because “computer” sounded too risky to investors. The Digital Equipment Corporation built the PDP-1 in 1960 and the PDP-8 in 1965, the first minicomputer, went public in 1966, and in 1971 ARD’s stake was worth 355 million dollars, five thousand times the stake.⁹ This one investment brought more than all the others of ARD together. It was the first demonstration of the law that has governed every fund since: not the average counts, but the one outlier.

The limits of the form. Precisely the success showed that the form was wrong. ARD was an investment company traded on the stock exchange and subject to the Investment Company Act of 1940. The securities regulator forbade ARD’s employees to hold shares in the financed companies; whoever had found and looked after DEC received a salary and no share in the gain, and the best people left. ARD’s stock traded below the value of its holdings because investors distrusted the company, and every new issue of capital needed the regulator’s approval. Doriot led ARD until 1972 and then sold it to the conglomerate Textron.¹⁰ The lesson was that an institution for venturing needs two things: a share for the managers in the gain, so that the best stay, and a legal form that frees them from the rules for public companies.

The state as backer: SBIC. The Sputnik of 1957 frightened Congress, and part of the answer was money for small companies. The Small Business Investment Act of 1958 created the Small Business Investment Companies, private companies with a state licence that received twice, later four times, their own capital as a cheap loan from the state and had to put it into small companies. By 1962 there were almost 600 of them. The form had a flaw: whoever must service state loans needs running income, and so most SBICs lent their money for interest instead of venturing it; many went under in the 1960s, some in scandals. Two things remained. The programme paid for the apprenticeship of a whole generation of venture capitalists who later founded firms of their own, and it showed that the state can release money without handing it out itself.¹¹ Josh Lerner later described the programme as a model of what governments do right and wrong in promoting venture capital: they can create the incentive, but not the selection.¹²

The families become firms. In the same year as ARD, John Hay Whitney founded J. H. Whitney & Co. with ten million dollars, whose first success was an orange juice concentrate called Minute Maid, and the Rockefeller brothers pooled their holdings in a company of their own, which became Venrock in 1969. From the families’ sideline came a profession with offices, employees and procedures, at first in New York and Boston.⁴

The bottleneck. By 1957 the institution that had been missing in 1945 existed, and with DEC it had delivered its proof. But it sat on the East Coast, in a legal form that drove away its own people, or it lived on state loans that turned it into a bank. The engineers who wanted to venture something were moving in those years to California, to a component that would change everything, and there was nobody there yet to give them money.

Try it yourself. What DEC meant for ARD can be recalculated: 70,000 dollars became 355 million in 14 years, five thousand times. What annual rate of return that is, the 14th root tells: in Python 5000 ** (1/14), in a spreadsheet =5000^(1/14), result 1.84, that is 84 per cent a year, for fourteen years. This figure, the internal rate of return (IRR), is to this day the measure with which funds state their results; whoever hears it about a company should ask for the second figure too, the multiple of the stake, because 84 per cent a year is a fortune over 14 years and nothing over 14 months.

3. Semiconductors and the limited partnership (1957–1972)

The traitorous eight. William Shockley, co-inventor of the transistor, had founded a laboratory in Mountain View in 1956 and brought in the best young physicists and engineers in the country, and he ran it in such a way that eight of them wanted to leave after a year, among them Robert Noyce and Gordon Moore. One of them, Eugene Kleiner, wrote to the New York bank Hayden, Stone & Co. asking whether it knew anyone who would hire the group as a whole. The letter landed with a young analyst named Arthur Rock. Rock flew to California, listened to the eight and proposed to them what none of them had thought of: not employment, but a company of their own. He asked 35 corporations whether they would finance one; all declined. The 36th was Sherman Fairchild, inventor, heir and largest individual shareholder of IBM, whose company Fairchild Camera and Instrument put up 1.5 million dollars in exchange for the right to buy the new company for three million. The eight received shares and a company, Fairchild Semiconductor, which in 1959 founded with the planar process and the integrated circuit the semiconductor technology that holds to this day.¹³ Shockley called the eight traitors; the name stuck to them, as a title of honour.

The Fairchildren. In 1959 Fairchild Camera exercised its purchase right. The founders received around 250,000 dollars each, the parent company the firm and the profit, and from this the eight and their staff drew a lesson: whoever bears the venture should also keep the fortune. In the following ten years dozens of them left Fairchild and founded companies of their own, Amelco (1961), Signetics (1961), National Semiconductor (1967 under Charles Sporck), Intel (1968), Advanced Micro Devices (1969), and each of these companies needed money from someone who understood the business. In 1971 a newspaper called the area Silicon Valley for the first time.¹⁴ The family tree of these “Fairchildren” is at the same time the family tree of the firms that financed them.

The limited partnership. The legal form that ARD had lacked came to Palo Alto in 1959. Draper, Gaither & Anderson, financed with Rockefeller money, organised itself as a limited partnership, a form that oil and gas drillers had used for decades: the backers are limited partners, liable only with their contribution, do not interfere and tie up their money for about ten years; the managers are general partners, make all decisions and receive two things for it, an annual management fee of two per cent of the fund for salaries and office, and the carried interest, usually twenty per cent of the profit, after the limited partners have their money back. Such a fund is not a company but a contract for a term; it is not subject to the Investment Company Act because it does not collect its money on the stock exchange but from a few rich people and institutions, and its holdings, the portfolio, are usually twenty to thirty companies, of which the general partners know that most will fail.¹⁵ That the form works was proved by Arthur Rock and Thomas Davis: their partnership Davis & Rock invested three million dollars from 1961, among others in Scientific Data Systems and Teledyne, and returned a hundred million in 1968.¹⁶

Rock and Intel. In 1968 Noyce and Moore left Fairchild. Rock needed two days to raise 2.5 million dollars by telephone from fifteen backers, among them the endowment of Grinnell College, on whose board Noyce sat; there was no business plan, there were Noyce and Moore. Intel brought out its first memory chip in 1969, the first microprocessor in 1971, and went public in October 1971.¹³ Rock’s rule that he invested in people and not in ideas became an article of faith of the Valley, and his role, the intermediary who examines a plan, raises money and sits on the board, became the job description of the venture capitalist.

NASDAQ and the tax. Venturing includes the exit: a fund with a ten-year term must turn its stakes into money, through the sale of the company or through the initial public offering (IPO), in which the company sells shares to the public and the funds can dispose of their stakes on the exchange afterwards. For small technology companies there was no suitable market until 1971; they were traded over the telephone, without visible prices. On 8 February 1971 the NASDAQ began operation, an electronic quotation system without a trading floor that admitted young companies with few requirements; Intel went public there in 1971, Apple in 1980, Microsoft in 1986, and the NASDAQ became the exchange of venture capital.¹⁷ At the same time venturing became more expensive. In 1969 Congress raised the tax on capital gains in stages up to 49 per cent, and the stock market slump of 1973 and 1974 closed the window for public offerings. The sums that new funds raised fell by 1977 to 68 million dollars a year, for the whole industry.¹⁸

Kleiner Perkins and Sequoia. In this bad year 1972 the two firms arose that were to shape the business for half a century, both on Sand Hill Road in Menlo Park, a street at the edge of Stanford whose name became a synonym for venture capital. Eugene Kleiner, one of the eight, and Tom Perkins, who had built up the computer division at Hewlett-Packard, founded Kleiner Perkins with eight million dollars; Don Valentine, marketing chief at Fairchild and National Semiconductor, founded Sequoia Capital. Their doctrines differed and complemented each other. Kleiner Perkins co-founded companies itself, incubated them, like Tandem Computers in 1974 and Genentech in 1976, and placed partners in them as managers. Valentine invested in markets, not in people: he looked for the market that would become so big that even a mediocre team could win in it, and replaced founders when they did not grow with it.¹⁹ Both houses exist to this day, and a large part of the companies in this chronicle took money from one of the two.

The bottleneck. In 1972 the Valley had engineers, models, the legal form and the first fortunes. What it did not have was capital. The pension funds, the country’s largest pools of money, were not allowed to venture anything under the pension law of 1974, the tax took half of every gain, and the stock exchange no longer admitted young companies. The business knew its rules and had no money.

Try it yourself. The arithmetic of a fund is simple and explains a lot. A fund of 100 million dollars over ten years: two per cent fee is two million a year, twenty million over the term, eighty remain for investing. If the fund triples the money to 300 million, the limited partners first get their 100 back; of the profit of 200 million the general partners receive twenty per cent, forty million, the limited partners 160. If it only doubles it, carried interest is 20 million; if it loses, the fees remain. Whoever works this through understands why general partners like big funds and why limited partners look at the multiple after costs, not at the return before costs.

4. Computers, biotech and the pension funds (1972–1983)

Two bets. Two companies from these lean years showed what was at stake. In April 1976 Robert Swanson, a 28-year-old former employee of Kleiner Perkins, and the biochemist Herbert Boyer, who in 1973 had invented with Stanley Cohen the method of transferring genes from one organism to another, founded the company Genentech. Kleiner Perkins first gave it 100,000 dollars and then put Perkins himself on the board; in 1978 Genentech produced human insulin in bacteria, and on 14 October 1980 it went public: the share rose in the first hour from 35 to 88 dollars, a record, although the company was not yet selling a product.²⁰ Biotechnology was born as a field of venture capital, Amgen (1980), Chiron (1981) and Genzyme followed, and it was to remain the only field outside information technology into which a large part of the money flows permanently. The second company was called Apple. Mike Markkula, rich and retired as an Intel manager at 32, gave Steve Jobs and Steve Wozniak 250,000 dollars in January 1977 for a third of the company, partly as capital, partly as a guarantee, and wrote the business plan; in 1978 Venrock, Sequoia and Arthur Rock joined. On 12 December 1980 Apple went public, valued at 1.8 billion dollars, the largest IPO since Ford in 1956, and made around 300 employees millionaires in one day.²¹

Atari and the games. That amusement is a market too, the Valley learned from Nolan Bushnell’s Atari, which began in 1972 with the arcade game Pong and in 1975 took the money of Sequoia, Valentine’s first investment; a year later Warner bought the company for 28 million dollars. From Atari came the founders of Activision (1979) and the model for Electronic Arts (1982), both with venture capital, and games remained a strand of the business of their own up to Zynga, Supercell and Roblox.²²

The rules change. What then changed the industry was not companies but laws, in three steps. In November 1978 Congress lowered the top tax on capital gains in the Revenue Act from 49 to 28 per cent, at the urging of a congressman named William Steiger, to whom the industry owed the fight for the law; in 1981 the rate went to 20 per cent. In June 1979 the Department of Labor declared that the prudent man rule of the pension law ERISA of 1974, under which a trustee may only invest as cautiously as a prudent person, applies to the whole portfolio and not to each individual investment: a pension fund could from now on venture a small part of its assets. In 1980 Congress exempted venture capital firms from the supervision of investment advisers.²³ The effect was immediately measurable: new funds raised 68 million dollars in 1977, almost a billion in 1978, over five billion in 1983, and the pension funds became the largest limited partners, followed by the endowments of the universities, whose pioneer Yale became under David Swensen from 1985.¹⁸ The capital that had been missing in 1972 came through two paragraphs in legal texts.

Bayh-Dole and the university. A third law made the universities the source. Until 1980 inventions from state-funded research belonged to the state, which hardly used them. The Bayh-Dole Act of December 1980 transferred the patents to the universities, which have licensed them to companies ever since, also in exchange for shares. Stanford licensed the method of Cohen and Boyer to Genentech and hundreds of other companies and in 1998 the search algorithm of two doctoral students to their company Google, for 1.8 million shares.²⁴ Since then the big universities have had licensing offices, and the path from the laboratory to the fund is part of a professor’s job description.

The PC boom. With the money came the first big market of the new era, the personal computer (the history of software development tells of it). Compaq, founded in February 1982 by three managers from Texas Instruments and financed by Ben Rosen and L. J. Sevin, built the first IBM-compatible portable PC and reached 111 million dollars in revenue in its first year, more than any company before; Lotus (1982) wrote the spreadsheet 1-2-3 and went public after a year and a half; Sun Microsystems (1982, Kleiner Perkins) built workstations, Silicon Graphics (1982, Mayfield) graphics computers, Adobe (1982) the printer language PostScript; Microsoft took a million dollars from Technology Venture Investors in 1981 for five per cent and never needed money again.²⁵ At the same time it became clear what too much money does. Between 1977 and 1984 the funds financed 43 makers of hard disks with 400 million dollars, each with a plan that promised a tenth of the market; in 1983 prices collapsed, and most disappeared. William Sahlman and Howard Stevenson described the case in 1985 as “capital market myopia”: every fund had acted reasonably for itself, all together had financed a market that had room only for a few.²⁶

The bottleneck. The capital was there, and with it the problem of capital. The funds grew bigger, the number of good founders did not, prices rose, and the vintages 1983 to 1985 brought their backers little. The industry learned that it is not money that is scarce but the companies that deserve it, and it needed a discipline that steers the money there.

Try it yourself. The stakes of those days can be read up. Apple’s IPO prospectus of December 1980 is digitised at the Computer History Museum (computerhistory.org) and names the holdings of Jobs, Markkula and Venrock. For every IPO since 1996 the prospectus, Form S-1, is freely available at sec.gov under EDGAR: Google’s (2004) contains the founders’ letter, Facebook’s (2012) the stakes of the funds that had come in in 2005. Whoever wants to understand the ownership of a company reads the section “Principal Stockholders”.

5. Networks, software and the first bubble (1983–1994)

The hangover. Abundance was followed by sobering. The funds of the vintages 1983 to 1985 had paid too much for too many, their results stayed below those of a government bond, and the pension funds, which had only just arrived, held their money back; the sums that new funds raised sank by 1991 to a little over a billion dollars a year. At the same time another business drew the attention: the debt-financed purchase of whole conglomerates, whose largest, the takeover of RJR Nabisco by KKR for 25 billion dollars, dominated the headlines in 1988. This private equity in the narrow sense, the purchase of mature companies with borrowed money, shares with venture capital the legal form and the fees, nothing else: it lives on debt and cost cutting, venture capital on growth and the one hit.¹⁸ For the industry the hangover was salutary. Whoever stayed learned that a company does not become more valuable because more money flows into it, and the vintages after 1986 were among the best of its history.

Network technology. The market of these years was the connecting of computers (the history of the Internet tells how the protocols for it came about). Bob Metcalfe, inventor of Ethernet, founded 3Com in 1979; and in December 1984 two employees of Stanford’s computing centres, Leonard Bosack and Sandy Lerner, founded Cisco Systems to sell the routers they had built for the university. By 1987, 76 backers had declined when Don Valentine gave 2.5 million dollars, for around a third of the company and the right to appoint the management. He brought in John Morgridge as chief executive, took Cisco public in February 1990 and had the founders thrown out the same year when they fell out with the management. In March 2000 Cisco was for a few days the most valuable company in the world.²⁷ The case became the object lesson of the Valentine model: what matters is the market, founders are replaceable, and the investor leads when it has to.

Software. The second field was software, and it changed the arithmetic of the business. A chip maker needs factories, a software house needs programmers and a copying machine; whoever occupies a market with little capital brings their fund many times the stake. Lotus went public in 1983 after a year and a half, Microsoft and Oracle in March 1986, Adobe in August 1986, Electronic Arts in 1989; Oracle had never taken venture capital, Microsoft only the one million of 1981.²⁸ Biotechnology ran on in parallel: Amgen, founded in 1980 and public since 1983, brought in 1989 with erythropoietin the first drug that turned a venture-backed company into a billion-dollar enterprise.

The contracts. In these years the contract that governs an investment took its present form, and William Sahlman described it in 1990 in a paper that is the basis of teaching to this day.²⁹ A fund does not buy ordinary shares but preferred stock with a liquidation preference: if the company is sold, it gets its money back before the founders see anything; only above that is the rest divided by shares. It does not give the money all at once but in financing rounds, named after the letters of the shares issued, Series A, B, C, each for new shares and only when the company has reached the goals of the previous one; with each round the founders’ share shrinks, the dilution, and, if things go well, its value grows. The founders’ own shares are subject to vesting: they are earned over four years, whoever leaves earlier leaves them behind. Before every round stands due diligence, the examination of technology, market, contracts and people, and at the beginning the term sheet, the non-binding summary of the conditions over which the negotiation takes place. The fund takes a seat on the board. The industry had organised itself in 1973 in the National Venture Capital Association, and the corporations began to play along with their own money: Intel founded Intel Capital in 1991, the forerunner of corporate venture capital, with which companies invest in startups to see markets and technology before they are finished.²⁹

Europe and Germany. Outside the USA the copy did not succeed. In Germany 29 banks founded in 1975, at the urging of the federal government, the Deutsche Wagnisfinanzierungs-Gesellschaft, the WFG, with a state guarantee for three quarters of the losses. It took stakes in around a hundred companies, lost almost everywhere and ceased work in the mid-1980s. Ralf Becker and Thomas Hellmann have described the case as an object lesson: the investments were chosen by bankers used to loans, the guarantee took away their reason to be strict, and a market on which one could have exited did not exist.³⁰ Britain had in 3i the largest investor in Europe, which however financed medium-sized firms, not founders; France, Japan and Germany tried state programmes, in Germany from 1989 with stakes by the Deutsche Ausgleichsbank. The same thing was missing everywhere: founders who had become rich and put the money back in, and a stock exchange that admitted young companies.

The bottleneck. In 1994 the industry was disciplined, small and profitable, with about three billion dollars a year, and its markets were mature: the PC belonged to Microsoft and Intel, the network to Cisco. The next step needed a market bigger than all previous ones and an exchange that would take a company before it made a profit. Both came in the same year.

Try it yourself. The model contracts under which almost every round is closed today lie open at the National Venture Capital Association (nvca.org, Model Legal Documents): the term sheet, the purchase agreement for the preferred stock, the shareholders’ agreement. Whoever reads the term sheet finds the terms of this section again on a few pages, liquidation preference, anti-dilution, vesting, board composition, and sees that the negotiation over a financing is less about the valuation than about these clauses.

6. Internet, e-commerce and the bubble (1994–2002)

Netscape. In April 1994 Jim Clark, who had built up Silicon Graphics, and the 22-year-old Marc Andreessen, who had co-written the browser Mosaic at the University of Illinois, founded a company to build a better browser; John Doerr of Kleiner Perkins gave it five million dollars. On 9 August 1995, sixteen months after the founding and without ever having made a profit, Netscape went public. The share had been issued at 28 dollars, opened at 71, and by the evening the company was worth almost three billion dollars.³¹ The day changed the rules. Until then a company went public when it had profits, usually after six to eight years; now it went when it occupied a market, and the exchange paid for the future. For the funds this meant that the exit no longer lay years away but months (the history of the Internet tells the years from the net’s point of view).

Get big fast. What followed was the biggest experiment venture capital has ever made. Jeff Bezos’s Amazon took eight million dollars from Kleiner Perkins in June 1996, at a valuation of 60 million, and went public in May 1997; its motto “Get big fast” became the doctrine of the era: whoever is big first in a new market wins it, so losses in the first years are not a mistake but the price. Sequoia came into Yahoo in 1995, which went public in April 1996. Benchmark, a firm founded in 1995 with only five partners, gave 6.7 million dollars in June 1997 for 22 per cent of eBay; at the IPO in September 1998 and in the months after, the stake became worth five billion, the best deal a fund had made until then.³² Draper Fisher Jurvetson put 300,000 dollars into the email service Hotmail, which Microsoft bought in 1997 for around 400 million. And in August 1998 Andy Bechtolsheim, co-founder of Sun, wrote two doctoral students at Stanford a cheque for 100,000 dollars for a company that was not yet registered; in June 1999 Kleiner Perkins and Sequoia together gave Google 25 million dollars, each for about a tenth of the company, stakes that were worth several billion at the IPO in 2004.³³ In between stood PayPal: founded in 1998 as Confinity, financed in 1999 with three million from Nokia Ventures, merged in 2000 with Elon Musk’s X.com, sold in 2002 to eBay for 1.5 billion. Its founders and first employees, Peter Thiel, Elon Musk, Reid Hoffman, Max Levchin, David Sacks, Roelof Botha and others, founded and financed in the following years Tesla, SpaceX, LinkedIn, YouTube, Yelp, Palantir and the Founders Fund; the press called them the PayPal mafia.³⁴ They showed the law that has driven the Valley since the Fairchildren: every exit produces founders and backers for the next wave.

Open source goes public. The movement that gave software away also received venture capital and stock prices in these years. Red Hat took money in 1998 from Intel, Netscape, Benchmark and Greylock and went public on 11 August 1999, VA Linux on 9 December 1999 with the largest first-day gain ever measured; the history of open source tells what became of the two. For the funds it was the proof that money could be made with software anyone may copy, through support, subscriptions and services.

The numbers. Venture capital in the USA rose from eight billion dollars in 1995 to around a hundred billion in 2000, more than in the preceding twenty years together. In 1999, 476 companies went public, and their shares rose on the first day by 71 per cent on average; the median age of a technology company at its IPO was four years. Every company with “.com” in its name found money, and funds that had raised a hundred million in 1997 raised a billion in 2000.³⁵ The grocer Webvan took 375 million in venture capital and as much again at its IPO in November 1999, built warehouses for a billion, and was insolvent in July 2001; Pets.com went public in February 2000 and closed in November. On 10 March 2000 the NASDAQ reached 5,048 points; in October 2002 it stood at 1,114, 78 per cent lower. Venture capital fell to 20 billion a year, the funds of the vintages 1999 and 2000 returned less to their backers than they had received, and for the pension funds and endowments a lost decade began.³⁶

Yozma, Neuer Markt and China. The bubble was worldwide, and the rest of the world learned different things from it. Israel had put a hundred million dollars into ten funds in 1993 with the programme Yozma, each of which needed a foreign partner, and withdrew again in 1997; the funds remained, and with them arose the densest venture capital scene outside the USA, with Check Point (1993, public in 1996) as the first proof.³⁷ Germany opened the Neuer Markt on 10 March 1997, a stock market segment for growth companies on the model of the NASDAQ, on which more than 300 companies listed within three years, Intershop, EM.TV, Mobilcom; it brought the country founders and funds for the first time, then accounting frauds such as Comroad’s, a price loss of 96 per cent and on 5 June 2003 its closure. The Neuer Markt had created the exit before the discipline existed, and left behind a distrust that closed the German stock market to young companies for years.³⁸ In China in the same period fell the two decisions that are among the best in the history of the business: Masayoshi Son of SoftBank gave Jack Ma’s Alibaba twenty million dollars in January 2000, and the South African publisher Naspers bought almost half of Tencent in 2001 for 32 million. At Alibaba’s IPO in 2014, at 25 billion the largest of all time, SoftBank’s stake was worth around sixty billion.³⁹

The bottleneck. After the bubble the window through which the money had gone out closed. The Sarbanes-Oxley Act of July 2002, Congress’s answer to the accounting scandals of Enron and WorldCom, required audited control systems of every listed company and made the IPO too expensive for small companies; instead of several hundred there were now a few dozen a year.⁴⁰ The funds that had raised billions in 2000 found no companies that needed them, and the backers withdrew. What the industry needed was patience: a way to let companies grow privately for ten years, and money that could wait that long.

Try it yourself. The numbers of this section can be checked. Jay Ritter of the University of Florida has kept tables of all American IPOs since 1980 with number, first-day gain and age of the companies (site.warrington.ufl.edu/ritter); there stand the 476 of 1999 and the 21 of 2008. The IPO prospectus of Pets.com of February 2000 lies in EDGAR and explains on twenty pages of “Risk Factors” exactly what happened nine months later; it is a lesson about what is in prospectuses and who reads them.

7. Platforms, cloud and the unicorns (2003–2015)

Google opens the window. On 19 August 2004 Google went public, at 85 dollars a share in an auction that bypassed the banks, valued at 23 billion dollars, and with two classes of shares that left the founders the majority of the votes although they held a minority of the shares. This dual-class structure, until then customary with media houses, became the model for Facebook, LinkedIn, Snap and almost every big technology company after: the founders keep control, the backers get shares, and the exchange accepts it.⁴¹ Salesforce had gone public two months earlier and showed that software could be sold as a subscription over the net, software as a service, a model that made revenue predictable and gave the funds a new favourite field. But IPOs remained rare. The usual exit became the sale to a corporation: Skype to eBay in 2005 for 2.6 billion dollars, YouTube, which Sequoia had financed with 11.5 million, to Google in 2006 for 1.65 billion.

Facebook and the social network. Peter Thiel gave Mark Zuckerberg 500,000 dollars in the summer of 2004 for a tenth of Facebook, Accel 12.7 million in May 2005 at a valuation of just under a hundred million; it became the best investment in Accel’s history. Then came money of a new kind. Microsoft paid 240 million in 2007 for one and a half per cent, valued at 15 billion, and in 2009 Yuri Milner’s Digital Sky Technologies gave 200 million, without a board seat, and at the same time bought employees’ shares from them. This growth equity, money from hedge funds, sovereign funds and corporations for companies that have already delivered the proof, lifted the limit that had hitherto driven companies to the stock exchange: whoever gets billions privately does not have to go public. Facebook went only in May 2012, valued at 104 billion dollars, and was eight years old by then.⁴² Twitter, LinkedIn and Zynga followed the pattern; the funds that were in early, Benchmark, Greylock, Sequoia, Accel, Union Square Ventures, became the addresses of the decade.

Founding more cheaply. At the same time the price of the beginning fell. From 2006 Amazon rented out computers by the hour over the net, and a company that had needed five million dollars for servers in 2000 needed five thousand in 2008 (the history of the Internet tells of the rise of the cloud). Paul Graham founded Y Combinator in 2005, the first accelerator: twice a year it takes in a few dozen teams, gives each a few thousand, later a hundred thousand dollars for six to seven per cent, three months of mentoring and at the end a demo day before investors. From its batches came Dropbox (2007), Airbnb (2009), Stripe (2009) and Coinbase (2012).⁴³ Eric Ries described the method of the new companies in 2011 as the lean startup: build a minimal product early, measure, learn, instead of planning for years. The small sums at the beginning got a stage of their own, the seed round, and backers of their own, the business angels, founders who have become rich and give their own money and advice, networked through platforms such as AngelList (2010). Marc Andreessen and Ben Horowitz founded in 2009 with 300 million dollars a firm that understood the fund as a service provider to founders, with departments for recruiting, press and sales; Andreessen’s essay “Why Software Is Eating the World” of 2011 became the programme of the decade.⁴⁴

The phone and the platforms. The iPhone of 2007 and the App Store of 2008 created the market that “get big fast” would have needed. Uber, founded in 2009, took eleven million from Benchmark in 2011 and 3.5 billion from the Saudi sovereign fund in 2016, the largest financing of a private company until then, and went public in 2019 valued at 82 billion. Airbnb took 600,000 dollars from Sequoia in 2009. Instagram had thirteen employees when Facebook bought it in April 2012 for a billion; WhatsApp had 55 when Facebook paid 19 billion in February 2014, of which around three went to Sequoia, which had invested 60 million.⁴⁵ The platform, a service that brings providers and users together and becomes more valuable for all with every participant, became the business model of the era, and its economics, losses until market leadership, then profits without marginal costs, was the justification for sums no fund had ventured before. In Europe arose Spotify (Stockholm, 2006) and Berlin’s Rocket Internet (2007), which rebuilt business models from the USA and had its greatest success with Zalando in 2008.

Cleantech, Tesla and SpaceX. Not every wave held. Between 2006 and 2011 the funds put 25 billion dollars into solar cells, biofuels and batteries, Kleiner Perkins set up a fund of its own with Al Gore in 2008, and the solar maker Solyndra received in addition a guarantee from the Department of Energy of 535 million; it went under in 2011, and a study by MIT calculated in 2016 that the funds had lost more than half of the 25 billion, because factories and chemistry do not grow by the rules of software.⁴⁶ Two exceptions justified the field in hindsight. Tesla, founded in 2003, took its first round in 2004, 7.5 million dollars, mostly from Elon Musk, later money from Draper Fisher Jurvetson and a loan from the Department of Energy, and went public on 29 June 2010, the first IPO of an American carmaker since Ford in 1956; ten years later it was the most valuable carmaker in the world. SpaceX began in 2002 with a hundred million from Musk’s PayPal proceeds, took twenty million from the Founders Fund in 2008, a billion from Google and Fidelity in 2015, and in 2024 at 350 billion dollars was the most valuable private company in the world.⁴⁷

Fintech and crypto. Money itself became a market: Square (2009) let anyone accept card payments, Stripe (2010) let anyone collect money on the net, financed by Sequoia, Andreessen Horowitz and Thiel, valued at 95 billion in 2021. Bitcoin had arisen in 2009 without any venture capital, as the software of an unknown; the business around it received it from 2013, when Andreessen Horowitz put 25 million into the trading platform Coinbase, which came to the exchange in 2021 at 86 billion.⁴⁸

Unicorns. On 2 November 2013 Aileen Lee, founder of a seed fund, counted in an essay 39 American software companies founded since 2003 and valued at more than a billion dollars, and called them unicorns because they were so rare. Two years later there were over 140, and the word meant the opposite.⁴⁹ Behind it stood three shifts. Growth equity let companies stay private: technology companies were on average four years old at their IPO in 1999, twelve in 2018. The JOBS Act of April 2012 raised the number of shareholders from which a company must disclose its figures from 500 to 2,000, the limit that had forced Google and Facebook to the exchange.⁵⁰ And the valuations themselves became a matter of negotiation: a billion valuation in a round in which the new shares carry preferential rights says little about the value of the old shares, but the number stood in the newspaper. What stayed the same was the law behind everything: in the data of Horsley Bridge, a fund of funds that has financed hundreds of funds since 1985, six per cent of all investments brought sixty per cent of all gains. That is the power law of the business: the distribution of results has no average that explains anything, but a tail in which everything lies, and a fund without the one hit is a bad fund, however good its other companies.⁵¹

The bottleneck. By 2015 founding was cheap, money plentiful and staying private possible. What was missing was oversight. Founders with multiple voting rights, backers without board seats, valuations without a market: Theranos, founded in 2003, had raised 700 million dollars from families and corporations, hardly any from funds, was valued at nine billion and had, as the Wall Street Journal began to report in October 2015, no working product. The scale of the money had run away from scrutiny, and the largest backer in history was still to come.⁵²

Try it yourself. The simplest instrument of the seed round is the SAFE (Simple Agreement for Future Equity), which Y Combinator published in 2013: five pages, free at ycombinator.com/documents, a promise of shares at the next round instead of a valuation today. The arithmetic: a SAFE of 500,000 dollars with a valuation cap of five million becomes ten per cent of the company at the next round, however highly the round is valued. Whoever wants to trace a company’s rounds finds them at crunchbase.com; for Uber there are more than twenty from the Series A of 2011 to the IPO of 2019.

8. Megafunds, crypto and AI (since 2016)

The Vision Fund. In May 2017 Masayoshi Son, who had made the industry’s best deal with Alibaba, closed the SoftBank Vision Fund at 93 billion dollars, 45 of them from the Saudi and 15 from Abu Dhabi’s sovereign fund: more than all American venture funds of the previous year had raised together. The fund wrote cheques of a billion and more, to Uber, Didi, DoorDash, Grab and WeWork, and its business model was to give companies so much money that the competition gave up. WeWork, a lessor of office space, was valued at 47 billion in January 2019; its IPO prospectus in August 2019 disclosed losses of two billion a year and the founder’s dealings with his own company, the IPO was called off, SoftBank took over the company in October at a sixth of the price, and the Vision Fund lost over thirty billion dollars in its worst year.⁵³ Theranos ended in court: Elizabeth Holmes was found guilty of fraud in January 2022 and sentenced to eleven years in prison.⁵² Both cases became the lesson that money does not replace scrutiny, and both were forgotten when interest rates fell.

Zero interest and record. In 2020 and 2021 money was almost free, the pandemic shifted commerce, work and entertainment onto the net, and venture capital reached sums that had never existed: worldwide over 600 billion dollars in 2021, in the USA over 300. Companies went public by merging with an empty, already listed shell, a SPAC; 613 such shells went public themselves in 2021. Cryptocurrencies had raised billions in 2017 by selling their own tokens, digital units without rights, and now became a field of the funds: Andreessen Horowitz set up a crypto fund of 4.5 billion in 2022, and the exchange FTX was worth 32 billion in January 2022, with Sequoia, Temasek and the Ontario Teachers’ Pension Plan as backers. In November 2022 FTX collapsed within a week because its founder had shifted customer money into his own trading fund; Sequoia wrote off 214 million, and Sam Bankman-Fried was convicted in 2023.⁵⁴ By then the central banks had already raised interest rates, venture capital fell by a third in 2022 and by another in 2023, valuations sank in down rounds, and on 10 March 2023 Silicon Valley Bank collapsed, where half of American startups kept their money, after a run that funds themselves had triggered when they advised their companies to withdraw. The state guaranteed all deposits.⁵⁵

The AI rounds. Then came the technology that shifted every scale (the history of AI tells it). OpenAI had been founded in 2015 as a non-profit organisation with a billion dollars in pledges and had given itself in 2019 a subsidiary with capped profit, into which Microsoft put a billion. After ChatGPT, Microsoft gave around ten billion in January 2023; in October 2024 followed 6.6 billion at a valuation of 157 billion, in March 2025 40 billion, 30 of them from SoftBank, at 300 billion, the largest private financing there had ever been, and on 27 February 2026 110 billion from Amazon, Nvidia and SoftBank at a valuation of 730 billion before the money; by the end of March the round grew to 122 billion and the valuation to 852.⁵⁶ Anthropic, founded in 2021 by former OpenAI researchers, took billions from Amazon and Google, in September 2025 13 billion at 183 and in February 2026 30 billion at a valuation of 380 billion, led by Singapore’s sovereign fund and the hedge fund Coatue, and in May 2026 a further 65 billion at 965 billion, led by Altimeter, Dragoneer, Greenoaks and Sequoia, 15 of it money already pledged by the cloud corporations; Elon Musk’s xAI took 20 billion in January 2026, with Nvidia and Cisco among the backers.⁵⁷ In 2025, 425 billion dollars of venture capital flowed worldwide, by Crunchbase’s count almost half and by the OECD’s 61 per cent of it into AI companies, a fifth into just five, and in the first quarter of 2026 alone around 300 billion, more than ever in a single quarter.¹ The form had changed: the backers were no longer funds on Sand Hill Road but the corporations that sold the chips and the data centres, Nvidia, Microsoft, Amazon, Oracle, the sovereign funds from the Gulf and Singapore, and banks that financed data centres like power plants; in January 2025 OpenAI announced with SoftBank and Oracle that it would put 500 billion into data centres.⁵⁸ The sums paid not for ideas but for infrastructure, and the question of whether this was still venture capital or the financing of power plants arose for the first time.

Germany and Europe. The largest German venture of these years managed without funds. BioNTech was founded on 1 October 2008 in Mainz and received 150 million euros from the twins Andreas and Thomas Strüngmann, who had become rich with the generics maker Hexal, and from the Munich fund MIG; for twelve years the company wrote losses, went to the NASDAQ almost unnoticed in 2019, and in December 2020 delivered the first approved vaccine against COVID-19, which brought it nineteen billion euros in revenue in 2021. It was Doriot’s lesson, patience instead of money, realised by a family office.⁵⁹ Beside it a German scene grew: Celonis from Munich became in 2021 the first German company valued at more than ten billion, DeepL from Cologne, Personio, N26; the federal government set up the High-Tech Gründerfonds in 2005, which became the country’s most active seed investor, in 2021 the Zukunftsfonds with ten billion euros and in 2024 the WIN initiative, in which insurers and corporations pledged twelve billion by 2030. The numbers nevertheless stayed small: 17 billion euros of venture capital in the record year 2021, 8.4 billion in 2025, in all of Europe around 60 billion against 320 in the USA.⁶⁰ The reasons are the old ones: no pension funds allowed to venture, no stock exchange that takes young companies, since the Neuer Markt, and too few who have become rich and put the money back in. France’s Mistral, founded in 2023, raised more than two billion euros in two years, most recently from ASML, and became the hope for a European AI model; Israel delivered with Wiz, which Google agreed to buy in 2025 for 32 billion dollars and took over in March 2026, the largest exit outside the USA.⁶¹ China, which in 2018 had almost drawn level with the USA, lost its venture capital to politics after 2021: the government stopped Ant’s IPO, forced Didi off the New York exchange and put the platforms under supervision, the USA banned investments in Chinese AI and chip companies in 2023, and the state itself became the largest backer with its guidance funds.⁶² DeepSeek, whose model frightened the industry in January 2025, was financed by a hedge fund, without a dollar of venture capital.

Where the chain stands. Eighty years after Doriot’s 3.5 million, a single financing round amounts to 110 billion dollars. Each era resolved its bottleneck: the institution in 1946, the legal form in 1959, the exit in 1971, the capital in 1979, discipline in the 1980s, the market in 1995, patience after 2002, the price of the beginning in 2005. What has remained is the law of DEC: a few hits pay for everything, and whoever finds them becomes rich and finances the next. What has shifted is the scale. A ten-year fund with a two per cent fee cannot lead rounds of a hundred billion; only corporations, states and debt can, that is, exactly the backers the business was meant to bypass in 1946. What is open is whether venture capital finances AI or merely witnesses it, and whether Europe, which has failed at the same gap for fifty years, finds its two paragraphs in the law.

Try it yourself. The rounds of the AI companies are public: OpenAI, Anthropic and xAI announce them in their own statements, crunchbase.com collects them. The arithmetic from section 2 can be applied to them: whoever had come into OpenAI in January 2023 at a valuation of 29 billion dollars held 29 times as much in April 2026 at 852 billion, in a little over three years. And whoever wants to know how a German company was financed finds every capital increase of a GmbH or AG in the commercial register (handelsregister.de), free of charge, together with the names of the shareholders.

The chain in brief

1. Until 1945 individuals and rich families finance the uncertain by acquaintance and chance; the institution is missing.
2. ARD (1946) makes venturing a profession and proves it with DEC; the legal form drives away its own people, and the state (SBIC) breeds banks instead of venture capitalists.
3. Semiconductors and the limited partnership (1957–1972) give the Valley founders, models and the form; the capital is missing.
4. Tax cut, pension funds and Bayh-Dole (1978–1983) open the sources; the abundance finances 43 hard-disk makers.
5. The 1980s (1983–1994) bring discipline, network technology, software and the contracts of the business; the exit remains narrow.
6. Netscape and the Internet (1994–2002) create the market and the bubble; Sarbanes-Oxley closes the IPO.
7. Cloud, accelerators and growth equity (2003–2015) make founding cheap and staying private possible; the unicorns grow without oversight.
8. Megafunds, zero interest, crypto and AI (since 2016) burst the scale; what is open is whether the form still holds and who keeps up outside the Valley.

Timeline

Further reading

• Tom Nicholas: VC. An American History. Harvard University Press, Cambridge, MA 2019. The history of the business from whaling to the unicorns, by an economic historian.
• Sebastian Mallaby: The Power Law. Venture Capital and the Making of the New Future. Penguin Press, New York 2022. The narrative from Arthur Rock to SoftBank, after conversations with those involved.
• Spencer E. Ante: Creative Capital. Georges Doriot and the Birth of Venture Capital. Harvard Business Press, Boston 2008. The biography of the man who made venturing a profession.
• Leslie Berlin: The Man Behind the Microchip. Robert Noyce and the Invention of Silicon Valley. Oxford University Press, New York 2005. Fairchild, Intel and the making of the Valley.
• Randall E. Stross: eBoys. The First Inside Account of Venture Capitalists at Work. Crown Business, New York 2000. Two years at Benchmark, with eBay as the main character.
• Paul Gompers, Josh Lerner: The Venture Capital Cycle. 2nd edition, MIT Press, Cambridge, MA 2004. The scholarly account of funds, contracts and exits.
• John Carreyrou: Bad Blood. Secrets and Lies in a Silicon Valley Startup. Knopf, New York 2018. The Theranos case, by the reporter who uncovered it.
• Christoph Keese: Silicon Valley. Was aus dem mächtigsten Tal der Welt auf uns zukommt. Knaus, Munich 2014. A German publishing manager’s view of the Valley and its backers (in German).

Sources

The chronicle rests on the original papers, laws, IPO prospectuses, press releases and contemporary reports named in the footnotes.