What is venture capital

US startups raised a record $515.8 billion of venture capital in nine months of 2026, but the cash paid back to fund investors arrives years later.
Venture capital is money professional investors put into young private companies for a share of ownership, betting that a few winners will repay all the losses. In the first nine months of 2026, US startups raised $515.8 billion of it, 44% more than in the whole of 2021, the previous record year. That money sets the price of most new technology companies long before the public can buy their shares.
Beginners tend to read that number as a measure of how well venture capital is doing. It measures money going in. What matters to the investors behind venture funds is the cash that comes back out, and it arrives years later, in steps, from very few companies. Between those two flows sits the fact that defines the asset class in 2026: AI companies took 82.7% of US venture dollars invested through September, while net cash flow to the investors in venture funds has been negative by $202 billion since 2022.
Both facts describe one machine running at two speeds. Money goes into a fund over four or five years, comes out over the following five to ten, and a handful of outcomes inside a portfolio of dozens decides the return. Without that timetable, a record year of spending looks like a record year of returns.
The arithmetic is unforgiving: a $100 million fund has to turn each invested dollar into about $4.40 for its investors to triple their money.
How venture capital funds work
A venture capital fund is a partnership with two kinds of partners: a general partner, the firm that picks the companies and runs the fund, and limited partners, the investors who supply almost all of the money. Limited partners range from pension funds and endowments to individuals who meet the legal tests for private investments. The general partner puts in a small slice of its own: the median manager of a US venture fund commits 1.7% of the fund.
Limited partners do not hand over their money on day one. They sign a commitment, and the manager draws it down through capital calls as it finds companies to back. Model fund agreements give the manager five years to make new investments and ten years in all, plus two one-year extensions. In practice the median US fund raised between 2017 and 2020 had called 10% of its money after two quarters, half after eight quarters and 90% after fifteen quarters, close to four years.
The money goes into a few dozen companies, with a reserve held back for later rounds in those that do well.
What sets venture capital apart from every other asset class is the shape of its returns. Across the hundreds of venture funds that one long-standing fund investor backed from 1985 to the mid-2010s, about 6% of the investments, taking 4.5% of the dollars, produced about 60% of all returns. The best funds lost money on more of their deals than merely good funds did. They won because their biggest winners returned close to 70 times the money, against about 20 times at good funds. That pattern, the power law, drives every venture decision.
How venture capital firms make money
Venture capital firms make money in two ways: an annual management fee on the fund, and a share of the profits called carried interest, or carry. The standard arrangement is known as 2 and 20. For every vintage from 2018 to 2025, the median US venture fund charged a 2% annual fee during its investment period, and the middle half of funds took exactly 20% of profits. Most funds lower the fee in later years: 81.9% have at least one step-down. A hurdle, a minimum return before carry starts, is rare: about one fund in eight of $100 million or more has one.
The arithmetic is easiest to see in round numbers. Take a $100 million fund with a 2% fee on commitments for ten years and 20% carry with no hurdle.
- Fees: 2% × $100 million × 10 years = $20 million. That leaves $80 million to invest in startups.
- Target: the limited partners want three times their money after fees and carry, or $300 million.
- Carry: the manager takes 20% of everything above the $100 million paid back. If the investments return P, investors get P − 0.2 × (P − $100 million), which equals 0.8P + $20 million.
- Solving 0.8P + $20 million = $300 million gives P = $350 million. Carry is 20% × $250 million = $50 million, and $350 million − $50 million leaves $300 million for investors.
- Gross multiple: $350 million from $80 million invested is 4.375 times, about 4.4x.
A manager therefore has to multiply each invested dollar by about 4.4 to give investors 3x. Fees explain the difference: one committed dollar in five never reaches a startup. Step-downs trim the fee, but legal and administrative costs add to it; in the median fund raised in 2017, fees and expenses together have consumed 19% of commitments.
Why venture funds need a few huge winners
Venture funds need a few huge winners because the power law decides where the $350 million comes from. If one or two companies have to supply about 60% of it, as the power law implies, those winners must return about $210 million. In a portfolio of 25 companies with $3.2 million in each, a single winner that carries most of that load returns roughly $200 million, about 60 times its check, by our calculation. That is why managers pass on good companies that cannot become very large.
Time matters as much as the multiple: three times the money over ten years is 11.6% a year if every dollar went in on day one and came out at the end (3 to the power of one-tenth, minus 1). Over the ten years to 30 June 2026, US venture funds as a group returned 17.57% a year after all fees and carry, against 13.86% for the same cash flows put into world stocks. A dollar compounding at those rates for ten years grows to $5.04 against $3.66, by our calculation (1.1757 and 1.1386, each to the tenth power).
Venture capital funding in 2026
Venture capital funding in 2026 is at a record, and concentrated in AI companies and in the largest managers. The $515.8 billion invested in US startups through September already exceeds the $319.7 billion of all of 2025 by 61%, by our calculation. The number of deals grew far more slowly, to an estimated 14,199 in nine months against 15,866 in all of 2025, so the average deal rose from about $20 million to about $36 million (each period's total divided by its deal count).
AI accounts for the size. AI companies took 88.4% of deal value in the first quarter, 84.1% in the second and 65.9% in the third. Deal value fell to $98.4 billion in the third quarter, down roughly 40%, as fewer giant rounds closed, while an estimated 5,012 deals made it the second-busiest quarter on record.
The same concentration runs through fundraising. US venture firms raised $108.5 billion across 699 funds in the first nine months of 2026, up 38.6% on all of 2025. Funds of $500 million or more took 78.1% of that money while making up 6% of the funds closed, and firms on their fourth fund or later took 88.2%, the highest share on record. Andreessen Horowitz alone closed funds totaling $23.8 billion, more than a fifth of all US venture capital raised this year, part of the shift we tracked as record venture money went to fewer hands.
For a fund investor, that concentration cuts both ways. It puts the AI winners inside a short list of large funds, and it ties the 2026 vintage to one theme, the same build-out whose financing we set out in our analysis of the AI bubble. Fanaura backs AI-native companies, and how Fanaura invests starts from that timetable: the price paid in a record year is fixed today, while the cash it returns is counted a decade later.
The largest venture capital firms
The largest venture capital firms in the US each advise tens of billions of dollars, and the one common yardstick is the regulatory assets they report to the Securities and Exchange Commission on Form ADV. The figure counts the gross value of the funds a manager advises, including commitments not yet called.
| Manager | Regulatory assets, $ billion | Accounts | Latest Form ADV filing |
|---|---|---|---|
| Andreessen Horowitz | 106.5 | 119 | 30 Mar 2026 |
| Insight Venture Partners | 92.2 | 181 | 9 Sep 2026 |
| Sequoia Capital | 82.2 | 60 | 4 Nov 2026 |
| Thrive Capital | 50.5 | 54 | 6 Jul 2026 |
| Lightspeed Venture Partners | 50.0 | 51 | 6 May 2026 |
| General Catalyst | 45.5 | 49 | 5 Oct 2026 |
| New Enterprise Associates | 35.6 | 27 | 26 Aug 2026 |
| Bessemer Venture Partners | 20.2 | 36 | 13 May 2026 |
As of 15 November 2026, from each manager's latest Form ADV filing with the SEC, Item 5.F. Sequoia Capital is its US entity only. Firms that file as exempt reporting advisers report no regulatory assets and are left out, as are crossover managers that also run hedge funds.
The top three firms together advise more than the next five combined, by our calculation from the table. Because the figures include money still to be called, the table also maps who will write the next four years of AI checks.
When investors get their money back
Investors in a venture fund get their money back late, in steps, and mostly from the few companies that list or are sold. Among US venture funds kept on one large fund-administration platform, the median fund raised in 2017 had paid back 0.37 times its investors' money in cash by 30 June 2026, nearly a decade after it started. The median 2018 fund had paid back 0.15 times, and the median 2019 fund 0.04 times.
Call the distance between what a fund is worth on paper and what it has paid out the paper-cash gap. It widens when valuations rise faster than companies sell: on 30 September 2026 a record 992 US unicorns, private companies valued at $1 billion or more, were worth $5.7 trillion on paper. Fund investors now treat cash returned, not paper value, as the measure that counts.
Even a giant exit closes the gap slowly. SpaceX sold shares at $135 in June 2026 in the largest stock market listing ever, raising $86.25 billion, yet the funds that backed it could not sell on that day. Their shares unlock in steps over the following year, the pattern measured by the float multiple, and a fund then sells or hands out stock at whatever price each release date brings. The listing value is a headline; the cash arrives in installments.
The strongest objection is that the gap is already closing. US venture-backed companies produced a record $2,291.1 billion of exit value in the first nine months of 2026, more than eight times the $274.5 billion of 2025 by our calculation, and acquisitions reached $465.3 billion, a decade high. The answer lies in what the figure counts. Exit value is the price of a company at listing or sale, not cash paid to funds. Most of the 2026 total comes from SpaceX: its listing, and its $60 billion all-stock purchase of Cursor, which alone made up 53.1% of third-quarter exit value and was paid in shares. Only 18 venture-backed companies went public in the third quarter. The fact that would settle it is cash: if full-year 2026 figures show net cash flow to fund investors turning positive, the gap is closing faster than this primer assumes.
Mistakes new venture investors make
Four mistakes cost new venture investors the most, and each has a practitioner's answer.
- Reading paper as cash. A fund can report healthy paper gains and still, nine years in, have paid back about a third of the money, as the median 2017 fund shows. The answer is to judge a manager on the cash each earlier fund has returned.
- Committing everything in one year. Over the five years to 30 June 2026, a window that opens near the 2021 peak, US venture funds returned 8.30% a year after fees, against 12.69% for world stocks on the same cash flows. Over ten years venture won. The answer is pacing: similar commitments every year, so no single vintage decides the result.
- Treating a commitment as a one-off payment. The median fund calls half its money in two years and nearly all of it within six, so an investor needs cash ready for calls years after signing, and a missed call counts as a default under the fund agreement.
- Backing the average manager. When about 6% of deals produce about 60% of returns, a fund that misses its few winners returns little. Manager selection is the decision that matters most, and patience is the second: Fanaura invests its own capital and that of its limited partners, with no fund to raise and no fund clock, so a winner can be held until it pays rather than sold to meet a deadline.
What changes in venture capital next
Venture capital in 2027 will be more concentrated, not less. Our view, checkable in January 2027 when full-year figures for 2026 appear: AI companies will have taken more than three-quarters of US venture deal value for the year, and funds of $500 million or more will have raised more than 70% of the money committed to new US venture funds. A drop below either line would mean the concentration of 2026 was a one-year event rather than the new structure of the market.
The $515.8 billion invested in US startups this year will be judged in the 2030s, by the same measure that shows 0.37 times for the funds of 2017: cash returned to the people who committed the money.
Invest alongside Fanaura as a limited partner. →
More from Fanaura
Invest alongside Fanaura as a limited partner
Fanaura is a private family office. We invest our own capital and the capital of our limited partners (LPs), usually through a separate SPV for each deal, in AI-native companies, games, technology and real estate across the US, the EU and Latin America. Accredited investors can talk to us about joining as an LP.
Get the news by email
One short note each weekday. No marketing, and you can unsubscribe at any time.