Founder Factories, by Spawner Rate

Where do the next unicorns actually come from? Not which school the founder went to. Which company they walked out of the week before.

So we find every founder. Find where they came from. Rank the source companies. And voilà.

We scanned 9 billion records to get there: companies, people, investments, employment histories, funding events, patents, and research output. 10.6 million alumni, across 364 private companies, tracked for who went on to found a venture-backed startup. Together, the companies they built have raised over $616 billion in venture capital. Read on for our take, or explore the data for yourself.

Universities and accelerators are excluded. This is about companies that spin out other companies. Google produces the most founders by volume, but the highest spawner rate belongs to its own research lab in King's Cross: Google DeepMind.

How We Built the Rankings

Every week we process billions of signals in our data platform: investment events, company records, employment data, founder profiles, social activity, patents, and scholarly publications. To build this study, we assembled the dataset in two passes, run separately for the global and European leaderboards, across all-time, five-year, and three-year windows:

We added Pass 2 because the household names kept turning up conspicuously absent from Pass 1. Microsoft, Tesla, Brex, Plaid, Google DeepMind, Palantir: none of these compete on rate alone, but stacking them side by side by volume instead is where it gets interesting. The differences that fall out are less about who is “better” and more about maturity and scale: a 200,000-person company and a 2,000-person research lab are never going to be measured fairly by the same yardstick, so we use both.

We evaluate companies across three dimensions:

Each metric tells you something different. A company with 1,000 funded founders sounds impressive until you realise it had 300,000 alumni. The rate is modest. A company with a high rate but only two founders might be noise. You need all three.

We also track what we call “Founder DNA”: the sectors that a company's alumni disproportionately choose to build in. Tesla alumni start energy and climate companies at 8.5x the market rate. 21% of everyone who left Tesla to found a company stayed in energy. That tells you something about what working there does to people.

We analyse multiple time windows because the companies producing founders today are not always the same ones that did so a decade ago. Netflix is a good example: nearly all of its spawning has happened since 2021, and the pace is still picking up. A company can be a late starter and a rising spawner at the same time: the all-time leaderboard alone would miss it. And we separate European and global rankings because the picture differs between the two.

A few caveats. This analysis measures venture-backed founders only. It does not capture bootstrapped companies, agencies, or profitable businesses that never raised. Capital raised is a signal, but it is not the same as value created. And we cannot fully separate selection effects from company effects: a high spawner rate may reflect what a company teaches its people, or just the kind of people it attracts in the first place. One more, in the spirit of openness: AI helped us build the pipeline, run the analysis, and edit this piece. It would have been odd not to let it.

The Top of the Table

We expected Google, Microsoft, Apple, Amazon, and Facebook at the top.

To our surprise, the number one company by spawner rate was Google DeepMind, the spin-off, not Google proper. The London lab based in King's Cross. With roughly 2,000 alumni and 58 founders, DeepMind has a spawner rate of 2.9%. Roughly one in every 34 people who have worked there has gone on to found a venture-backed company.

Palantir sits at number two, with a spawner rate of ~2.5% and $18.7 billion in capital spawned from 117 funded founders. High on both density and value.

OpenAI is at number three, with a spawner rate of ~2.1%. Its most notable spawn is Anthropic, which accounts for most of the $143 billion in capital spawned by OpenAI alumni. On capital spawned per alumnus, OpenAI is the clear outlier.

X, the moonshot factory is at number four. With 1,291 alumni and 24 funded founders, its spawner rate of ~1.9% puts it above all of traditional big tech. These are the people who worked on self-driving cars, balloon internet, and fusion energy. They left and started companies like Pacific Fusion.

ConsenSys, the Ethereum infrastructure company, rounds out the top five with a 1.7% spawner rate from 25 funded founders. Gnosis, a crypto trading and prediction-market platform, is its biggest spawn.

The top five by spawner rate: two Alphabet entities, the company that built GPT, the company that built software for the CIA, and an Ethereum infrastructure firm. The common thread is not size or sector, but intensity: places where the work is hard and the bar is high.

The Alphabet Constellation

Alphabet dominates this leaderboard.

Google sits at #121 by spawner rate. On the other two measures it blows the table away: #1 in the world on funded founders, with 672 from 166,000 alumni, and #2 on capital spawned at $54.2 billion, behind only OpenAI, whose figure is mostly Anthropic.

And Google is only a third of it, with DeepMind at #1 and X at #4: three entities from one parent company, two of them in the top four, each the leader of its class. Across the three, Alphabet has produced over 750 funded founders and over $70 billion in spawned capital, $14.7 billion of it from DeepMind alone.

DeepMind is the sharpest version of the story: with roughly one eighty-third of Google's headcount, it runs at 7 times the rate. Its alumni have founded Reflection AI ($4.6B raised), Mistral AI ($4.0B), Physical Intelligence ($1.1B), and Reka AI, and its Founder DNA is AI and machine learning at 3.0x the market rate. Over half of everything its alumni build is AI.

DeepMind was originally funded in Europe and later acquired, and it still produces founders at the highest rate in the world from inside a trillion-dollar parent. Alphabet didn't water it down by absorbing it.

Something at Google causes people to walk out and build fantastic companies. Either it is the way Google hires, or it is what people learn once they are inside. Probably both.

Big Tech Is Less Impressive Than You Think (When Measured by Spawner Rate Alone)

Most of the companies you'd expect to dominate this leaderboard don't even crack the top 200 by spawner rate.

We rank 240 companies globally. Tesla, Facebook, Microsoft, Apple, and Amazon sit at #172, #179, #186, #194, and #195 respectively. Google, at #121, is the only traditional big tech company that makes it into the top half of the table. Netflix, the N in the old FAANG acronym, does not qualify for the all-time leaderboard despite 17,755 alumni. That's a recency story, not an absence one, and it gets its own section below.

By spawner rate, big tech is fairly unimpressive. But the companies their alumni do build are enormous. Tesla spawned Physical Intelligence ($1.1B). Microsoft produced xAI ($5.0B). Apple produced Verkor ($5.3B) and Together AI ($1.5B). Facebook produced Perplexity AI ($2.2B). The founders are fewer, but they swing bigger.

Count founders and big tech looks dominant. Use rates and they almost disappear. The detail that matters is what those founders go on to build.

You Are What You Ship

In the Founder DNA data, the capability that won a company its market keeps resurfacing in what its alumni build. Not for every company, though. The dataset splits into generalists and specialists.

The generalists. Old, broad, multi-business companies produce founders with flat DNA: no sector disproportionately represented, because the company itself doesn't do one thing.

Facebook is the flattest read in the dataset: nothing in its Founder DNA clears 1.5x, so there is no single sector its 316 funded founders disproportionately pick. Its biggest spawn by capital is Skild AI ($2.2B), an Enterprise Software company building general-purpose robotics AI, founded by a former Facebook AI Research manager. Its alumni leave with a way of building, not a product category.

Apple alumni are generalists too. The biggest share go into AI/ML (26.3%) and Enterprise Software (18.0%), roughly matching the market. Hardware & Semiconductors at 1.7x the market rate is the clearest signal that actually reflects what Apple does, and even that is modest.

Amazon is the third. Its highest-ratio signals look specific at first glance: construction at 4.2x the market rate, logistics and supply chain at 2.3x. But construction is 0.3% of its funded founders, essentially noise, and logistics is only 3.0%. The biggest real share of Amazon's 386 funded founders, 28.1%, goes into AI/ML, and that is just 1.5x, close to market rate. Amazon is a retailer, a cloud platform, a streaming service, and an advertising business all at once, and its founders, in aggregate, look like the market. Three of the biggest, oldest companies in the dataset, three of the flattest founder DNA profiles. When the company does everything, the alumni specialise in nothing.

The specialists. Companies built around one mission imprint it onto their founders. The narrower and more technical the core problem, the sharper the signal.

Tesla is the clearest case. You might expect Tesla spawns to be car companies. They are not. Look at the timeline below: Northvolt ($13.8B, bankrupt 2025), Form Energy ($1.8B), Electric Hydrogen ($798M), Lunar Energy, SPAN, Heron Power. It is almost entirely energy and climate companies. The DNA data confirms it: Energy & Climate Tech at ~8.5x the market rate, 21% of all its funded founders. It's not the cars. It's the batteries.

SpaceX makes Tesla look diversified: Aerospace & Defence at 32x the market rate, 27% of its funded founders. More than a quarter of everyone who left SpaceX to found a company stayed in rockets and defence hardware. The spawns read like a SpaceX supplier list: Long Wall ($509M), Castelion ($474M), Xona ($331M), Starcloud ($216M). Building rockets is a narrow, specific, brutally technical skill, and the people who can do it mostly keep doing it.

Both, of course, are Musk companies. You do not have to be a fan to read the data: something in those cultures produces founders of hard technical companies at a rate almost nobody else matches.

Palantir is the most concentrated of all by volume: AI/ML (28.8%) and Enterprise Software (29.8%) together account for 59% of its funded founders, almost exactly what Palantir itself is, an AI-powered enterprise software company. Layered on top is a smaller but sharper signal, Aerospace & Defence at 3.4x the market rate: the Palantir-to-Anduril pipeline. These are not people who happened to start defence companies. They worked inside a defence-adjacent company with a particular way of thinking about hard problems, and they took that thinking with them.

Seismologists expect a massive earthquake to hit San Francisco at some stage; they cannot tell you when, only that the pressure is building. Seeing the external factors pushing on Palantir right now, I expect a massive wave of spawners out of Palantir in the next three years.

Across all of these, sector choice tracks where someone worked far more closely than where they studied.

Where the Moat Is Too Deep

What about the semiconductor companies? ARM, Qualcomm, Intel, NVIDIA. Where are they?

Mostly absent, or barely there. AMD and TSMC do not appear in our data at all. ARM and ASML surface only in the European board's recent windows, with a handful of founders each: ARM's 6,926 alumni have produced seven. Broadcom makes the global board at #183. Intel sits at #196 with 140,000 alumni and just 120 funded founders, a spawner rate of 0.09%. Qualcomm is at #175, conservatively 37,000 alumni (of which I'm proud to count myself). NVIDIA fares much better at #149 with 14,000 alumni and 45 funded founders.

The sector affinity is strong. Intel's alumni over-index in Hardware & Semiconductors at 8.2x the market rate. Qualcomm's top signal is Telecom at 9.9x, then Hardware at 8.3x. These people know chips. But knowing how to design a chip and being able to start a chip company are two very different things. The barriers to entry are enormous: capital requirements, fabrication costs, IP licensing, and legal complexity. You can be the best chip designer in the world, but you cannot just spin up your own fab.

And then there is the sheer scale. These are enormous workforces, and the vast majority of those people are workers, not spawners. Engineers doing excellent, highly specialised work inside a machine that requires that specialisation. The moat that protects these companies from competition also protects them from producing competitors.

The Grandparents: When Spawners Spawn Spawners

This is the part that surprised us.

Some companies on our leaderboard were themselves spawned by other spawners. We started calling these “grandparent” relationships.

Airtable is a good example. It raised $1.35 billion, has 1,199 alumni, and has already produced 6 funded founders. But Airtable itself was co-founded by Andrew Ofstad from Google, and Howie Liu from Salesforce. Salesforce sits at #180 globally, with 71,000 alumni, 100 funded founders, and $7.0 billion in capital spawned. Airtable is its second-biggest spawn.

Google → Airtable → Anrok. Three generations.

And the DNA carries through. Ofstad brought AI and machine learning expertise from Google. Liu brought enterprise software from Salesforce. Airtable itself is an enterprise software company with strong AI capabilities. Its spawns follow the same pattern: Anrok ($109M) is enterprise software, Salient ($65M) is AI and machine learning.

Nuro was founded by two Google alumni. Scale AI traces back to Quora. These chains keep forming: high-rate spawner produces a company, that company starts producing founders of its own.

Founders Leave in Waves: The Netflix Story

Companies have life cycles: inception, growth, maturity, saturation, and eventually decline.

While a company is growing, the builders stay. The product is still changing, the hardest problems are still open, the upside is still ahead. At maturity the work shifts from invention to optimisation: scaling infrastructure, improving margins, defending share, raising prices on customers who are already locked in. Some people enjoy that phase. The ones who joined to invent lose their reason to stay, and they tend to leave together.

The best big tech companies renew themselves: new revenue streams, new product markets, new industries to disrupt. Meta, Google, Microsoft, Amazon. We all know that story, and they cycle too, but with so many businesses running at once the waves blur together. At a company that does one thing, the cycle is single and sharp.

“Products mature. Companies mature. Builders often don't.”

The cleanest cycle in our data is Netflix. Its defining periods were existential bets: DVD-by-mail to streaming, then licensed distribution to original content. Bets like that need builders. By 2021 the streaming war had ended with Netflix on top, and the work became optimising a dominant platform: retention, pricing, advertising, and content going moneyball, with spend flowing to what moves the dial rather than to big swings. And while Netflix worked out how to stop you sharing your account with your sister, its best builders were walking out the door in droves. Before 2021, Netflix alumni had founded four venture-backed companies in total. Since 2021: thirteen, ten of them in the first three years alone. You can see the wave break in the timeline below.

Some would call these people the builders. They are not content to sit around and keep the lights on. Maybe they are only happy when they are building something, or moving toward something bigger and better. A saturated product has nothing left to teach them, so they leave and start the next thing.

None of this is a failure story. Netflix did not stop succeeding; it succeeded so completely that the inventing was done. The startup ecosystem is not separate from corporate success. It is a direct consequence of it.

The waves connect across industries too. The momentum moved from social media into cloud, from cloud into big data, from big data into fintech, and from fintech into AI, and it tends to be the same kind of person each time: whoever was early to one wave shows up again at the start of the next. That is what this dataset really tracks: momentum, moving from company to company, so that when the next wave breaks we already know where the founders will come from.

Switch from the global view to Europe and the picture changes.

The global leaderboard shows a rich spread of sectors: enterprise software, AI, energy, transport, aerospace, and hardware. Switch to Europe and the palette narrows. The top 30 European spawners are dominated by fintech: Revolut, Klarna, N26, Monzo, Checkout.com, Adyen, Wise, Qonto, Funding Circle, SumUp, 4finance, and PayU. That is twelve fintech companies in the top thirty. Health tech, e-commerce, gaming, and professional services fill most of the rest.

None of this is a knock on fintech. Adyen and Wise are serious engineering companies, and payments is a fight Europe actually won. The worry is the next wave. Google DeepMind is the only AI/ML parent company in the entire European leaderboard of 82; the global leaderboard of 240 has 14.

That gap is a challenge to be met, not a verdict. We believe the biggest companies of the next two decades will be built in AI, hardware, energy, robotics, and, for the first time in a generation, medicine, where AI has the research moving at a pace we have not seen before. The most capital-intensive of those industries tend to be dominated by America, but not all of them are: Europe holds its own in medicine, Novo Nordisk being the obvious example, with a university research base that deserves a lot of the credit. The upside is that none of this is settled: the models, the hosting, even the social platforms Europe currently rents from America can be built here, and the dependence can shift back. Whether Europe faces up to that challenge, or a change of administration in Washington takes the urgency out of it, is a question this leaderboard will start to answer over the next few years. As a European fund, we are not neutral about which way it goes.

The pressure runs both ways, too. Europe's existing strengths are not guaranteed: look at what China is shipping and you can see the car industry, luxury included, coming under strain. There is catching up to do on one front and defending to do on the other.

The children, at least, tell a different story. European spawners have produced 180 AI companies, 91 of them founded since 2022. DeepMind is doing most of the heavy lifting (Reflection AI at $4.6B, Mistral AI at $4.0B, Ineffable Intelligence at $1.1B, Recursive at $650M, H Company at $230M), but the momentum is spreading. Spotify spawned Agentio. Darktrace spawned Geordie AI. Monzo spawned Gradient Labs. These are signals that the pipeline is forming.

The question is whether Europe can turn these children into spawners themselves, whether Mistral AI, founded in Paris by a DeepMind alumnus, will one day sit on this leaderboard producing its own funded founders. If it does, Europe's AI gap is one of timing, not talent.

Europe has yet to produce a Google, a Microsoft, a Meta, an Amazon. Why not? It is certainly not talent. Talent does not discriminate; it is everywhere. And the strengths run both ways: Europe holds ground America does not, automotive and luxury off this chart, biotech and fintech on it. What separates the two maps are the external forces: the political, economic, and legal structures that decide whether someone is crazy enough to start a company, and whether that company can reach American scale. Mapping those forces against this data is an article of its own.

The scale difference is just as striking. Europe's total capital spawned is $81.8 billion. The global figure is $616 billion. Remove DeepMind's $14.7 billion contribution and Europe drops to $67.1 billion, roughly 11% of the global total. The sectors Europe is strong in (fintech, health tech, professional services) tend to have lower barriers to entry. The sectors where the biggest companies are being built (AI, aerospace, hardware, energy at scale) are the ones where Europe is underrepresented.

If you look at the three-year window, the trend is not improving. DeepMind still dominates, with a three-year spawner rate nearly twice the next European company, Enpal, a solar and battery installer. Without DeepMind, the three-year picture is energy, fintech, and health tech. The data raises a question about whether European engineering talent is being absorbed by American-headquartered companies rather than spinning out into European-founded startups.

And then there is the flight of tech. Even when Europe does produce something extraordinary, it struggles to keep it. ARM, whose chip architecture runs in most of the world's smartphones, was designed in Cambridge. It is still headquartered there, but SoftBank-controlled and Nasdaq-listed since 2016: the capital and control left even if the company didn't. Stripe was founded by two brothers from Limerick. It has always been an American company. DeepMind was acquired by Alphabet.

The more optimistic reading is on the same leaderboard. The winners of Europe's last wave are already becoming factories. Revolut has produced 38 funded founders, a spawner rate of 0.71%, nearly double Google's; Qonto and Enpal run at close to three times it. Klarna has produced 28, led by MODIFI ($354M). The Founder DNA rule holds in Europe too: King, the studio behind Candy Crush, spawned Tripledot Studios ($252M) and Resolution Games, and Spotify spawned Podimo ($222M) and Slang AI. Europe's champions are producing founders at rates traditional big tech cannot match, and the grandparent chains are starting to form here as well.

For European investors, this is the opportunity. The founders come from here, the formative years happen here, and the early rounds are raised here. The challenge was never ambition or ability; brilliant companies start in Europe every year. It is whether they can compound here: whether capital depth, regulatory speed, and patient owners give them enough reason not to sell early and not to drift to San Francisco. Getting that right is the work.

Schools Get the Spotlight. Companies Do the Work.

There is no shortage of research into which universities produce founders. Stanford is always near the top. Cambridge and MIT trade places depending on who is counting. The media covers it extensively because the data is neat and the narrative is obvious.

None of that research is wrong. It stops one step short.

What we see in our data is that funded founders come from a wide range of schools. It is not exclusively the Oxbridge elites or the Ivy Leagues that produce great founders. A person might attend any number of universities, but if they then spend five years at DeepMind, or Palantir, or Stripe, something happens. They learn how to build product, how to hire, how to operate at pace and ship under pressure. They watch their employer raise capital and learn how that works too. These are things you do not learn in a lecture hall.

The university gets the credit because the data is easy to collect. But there is an entire industry built around ranking universities, and comparatively very little effort studying which companies produce the best founders. Our data suggests the company matters at least as much.

Why This Matters

We built this because we are trying to spot the next unicorn first. The spawner data gives us an outcome-based signal for doing that. It tells us: this company, with this many alumni, produces funded founders at this rate, in these sectors. When someone leaves a high-rate spawner and starts building, we want to know about it early.

There are companies on our leaderboard that most people outside the tech industry have never heard of. Brex (rank #6), Flatiron Health (#9), Plaid (#10), Scale AI (#11). These are not household names, but they produce founders at rates that dwarf the companies that are.

The data doesn't tell us who will be the next great founder. But it narrows down where to look, and in venture, getting there first is what matters.

Explore the Data

The full spawner dataset is available on our website. You can filter by region, time window, and sector, explore individual company profiles with founder timelines and lineage graphs, and see which companies are spawning the next generation of founders right now. The data is updated weekly.

Explore the Spawners leaderboard →