Qanot / Unicorn formation
Where billion-dollar companies come from, and how fast they stop being one.
The word is imprecise and the dataset is self-reported, which is exactly why it is worth reading carefully. What follows is the shape of private billion-dollar company formation over the last decade — and the part of the distribution that gets charted far less often, which is how many of them do not stay there.
On these figures. Unicorn counts are compiled from commercial trackers that disagree with each other by 10–20% and revise retroactively. Everything below is directionally reliable and precisely wrong. Treat the shape as real and any individual number as an estimate — and before this page goes live, replace each dataset with a dated, cited pull.
What the word actually means, and why it misleads.
A unicorn is a private company that has been assigned a valuation of $1B or more in a primary funding round. Note what that sentence does not say. It does not say the company is worth $1B. It does not say it is profitable, or growing, or that anybody has sold a share at that price.
A valuation is the output of a negotiation over a small slice of preferred stock, usually with liquidation preferences attached. A company can be assigned $1B by selling 5% with a 1× preference and a ratchet — and the common stock held by founders and employees can be worth a fraction of what the headline implies.
Why we care anyway: not because the label means much, but because the rate at which it is awarded is the single best public proxy for how much late-stage capital is in the system — and late-stage capital is what eventually prices the companies we back at formation.
2021 was not a trend. It was an interest rate.
Between 2013 and 2018 the world produced somewhere between twenty and a hundred new unicorns a year. In 2021 it produced more than five hundred. Nothing about human ingenuity changed in that window. The cost of capital did.
When risk-free rates sit near zero, the discount applied to distant cash flows collapses, and crossover funds that normally buy public equities move down into private rounds looking for growth. The result is a repricing of the same companies, not the appearance of better ones.
The correction is visible in the same chart: 2023 formations fell roughly 86% from the peak. Companies that were unicorns in 2021 did not become worse businesses in 2023. The buyer left the room.
Two countries hold two thirds of them.
The United States accounts for roughly half of all unicorns ever formed; China for another seventh. Add India and the United Kingdom and you have accounted for about three quarters of the global total.
This concentration is usually read as a statement about talent. It is better read as a statement about capital density. Silicon Valley does not produce more capable engineers than Shenzhen, Bangalore, Warsaw or Tashkent. It produces vastly more people whose full-time job is writing a first cheque to one.
Where formation-stage capital is dense, companies that would otherwise die at the prototype stage survive long enough to compound. Where it is absent, the same companies are stillborn — and the region concludes it lacks founders, when what it lacks is the first $150,000.
| Market | Share | Approx. count | Seed funds active | Read |
|---|---|---|---|---|
| United States | 52% | ~730 | High | Capital density, not talent density |
| China | 14% | ~200 | High | State-directed capital plus scale |
| India | 6% | ~85 | Rising | The clearest precedent for our thesis |
| United Kingdom | 4% | ~55 | High | Fintech-weighted |
| Germany | 2.4% | ~34 | Moderate | Enterprise-weighted |
| Israel | 1.7% | ~24 | Very high per capita | Density beats population |
| Central Asia | 0.14% | 2 | Near zero | The gap this fund exists to close |
Fintech, software, and now models.
Fintech has been the largest single category for most of the last decade, for a structural reason: financial services is the one industry where the product is information, the incumbents are slow, and the regulatory moat that protects them also caps how fast they can respond.
The AI slice is the one genuinely moving. It was negligible in 2019 and now accounts for a substantial share of new formations — and an outsized share of the capital, since model companies raise larger rounds earlier than any category before them.
Our read: the interesting position is not building a frontier model. It is being the company that owns the proprietary data and distribution in a language and market the frontier labs will not prioritise for another decade. Uzbek is spoken by roughly 35 million people. No frontier lab is optimising for it.
Seven years, not two.
The median company that reaches a billion-dollar valuation takes about seven years to get there. The press covers the eighteen-month cases because they are astonishing; they are also roughly one in twelve.
This matters more than it sounds. A seven-year median means a fund writing first cheques today is underwriting outcomes that land in the early 2030s — past the end of a standard ten-year fund life if you start late. It is an argument for writing the first cheque rather than the third.
It is also the argument against the accelerator model of forcing a company to raise a Series A in twelve weeks. Twelve weeks is enough to establish slope. It is not enough to establish a business, and pretending otherwise produces companies that raise well and die quietly.
The chart nobody publishes.
Unicorn counts are almost always reported cumulatively, as though the status were permanent. It is not. Companies get marked down, recapitalised, acquired below their last round, or quietly wound up — and most trackers do not remove them.
The stranded middle is the most instructive group. These are companies that raised at a billion-dollar price in a hot window, grew into perhaps a third of it, and now cannot raise a flat round without triggering preferences that would wipe out the common stock.
They are not failures in an operating sense. Many have real revenue and real customers. They are failures of price discipline at the formation stage — a valuation taken early that the business then spent five years trying to grow into.
This is the single strongest argument for the terms we publish. We do not press for a high price at the first cheque, and we do not want you to take one from anybody else. A company that raises $150K at a sensible price and reaches default-alive has every option. A company that raises $3M at $30M pre before it has customers has one.
Two data points, and what they imply.
Kaspi.kz — Kazakhstan
A payments, marketplace and banking super-app serving a country of 20 million. Listed in London in 2020 and on Nasdaq in 2024, with a market capitalisation that has traded in the tens of billions of dollars. It is the proof that a Central Asian company can be priced by global public markets rather than by regional discount.
Public market · verify current figures before launch
Uzum — Uzbekistan
An e-commerce, fintech and marketplace group built for the Uzbek domestic market, reported to have crossed a $1B valuation in a 2023 round — the first company in the country to do so. Built for 37 million people who were, a decade ago, considered too small and too cash-based a market to serve.
Reported private valuation · verify before launch
Two companies is not a pattern and we will not pretend otherwise. But consider what produced them: no local seed ecosystem, no domestic venture funds at formation stage, no institutional angel base, and founders who had to be their own first investor for years.
That is the interesting part. These outcomes happened in spite of the capital environment, not because of it. The question a fund should ask is not "will Central Asia produce unicorns" — it demonstrably has — but "what is the base rate when the formation-stage capital gap is closed, given that the rate without any capital at all is already non-zero?"
India is the nearest precedent. In 2013 it had a handful of billion-dollar companies and almost no domestic seed infrastructure. A decade of formation-stage capital later it has roughly eighty-five. We are not forecasting that. We are pointing out that the mechanism is known.
What we changed because of this data.
- Small first cheque
- Because the median company needs seven years, and a large early round compresses the time available to find product-market fit before the preference stack becomes unmanageable.
- High position count
- Because the return distribution has a fat right tail and a thick zero. Forty positions is the minimum to have a reasonable chance of intersecting the tail at all.
- No pressure on price
- Because the stranded middle is full of good companies that took a hot-market valuation and spent five years underwater on it.
- Default-alive over growth
- Because we cannot forecast the rate environment at your Series A, and default-alive is the only state that does not depend on forecasting it.
- Domestic-first markets
- Because the sector data says fintech and vertical software win, and both are won by being undeniable in one market before being adequate in ten.
- Language and data moats
- Because the AI slice is growing fastest, and the defensible position for a regional company is the data and distribution the frontier labs will not prioritise.