Twelve AI labs have a mixed valuation bigger than Ford and GM. None of them promote something. I name them the Virgin Unicorns — valued above a billion {dollars}, however harmless of product or income.
OpenAI proved that an AI analysis lab with the fitting product might develop into one of the vital precious firms on earth. A dozen different AI labs try to repeat the trick. They’ve raised greater than $29 billion at a mixed valuation approaching $130 billion, with out delivery something a buyer should buy.
Two questions are value asking:
- Why are refined traders writing growth-stage checks to pre-companies?
- What does historical past say about how this story ends?
Prime Virgin Unicorns
| Firm | Based | Founders | Valuation | Raised | Lead Buyers | Product |
|---|---|---|---|---|---|---|
| Mission Prometheus | 2025 | Bezos, Bajaj | $38B | $16.2B | JPMorgan, BlackRock, Bezos | None |
| Secure Superintelligence | 2024 | Sutskever, Gross, Levy | $32B | $3B | Greenoaks, Sequoia, a16z, Lightspeed, DST, Alphabet, Nvidia | None |
| Pondering Machines Lab | 2025 | Murati, Schulman, Zoph, Weng | $12B | $2B | a16z, Nvidia, AMD, Cisco, Accel, Jane Road | Tinker* |
| Reflection AI | 2024 | Laskin, Antonoglou | $8B | $2.1B | Nvidia, Lightspeed, Sequoia, Schmidt, Citi, 1789 Capital | None |
| Bodily Intelligence | 2024 | Levine, Finn, Hausman, Ichter, Groom | $5.6B | $1B+ | CapitalG, Lux, Thrive, Bezos, T. Rowe Worth, Index | Demo |
| Ineffable Intelligence | 2025 | Silver, Czarnecki, Espeholt, Oh | $5.1B | $1.1B | Sequoia, Lightspeed, Nvidia, Google, UK Sovereign AI, Index | None |
| World Labs | 2024 | Li, Johnson, Mildenhall | $5B | $1.2B | a16z, NEA, Radical, Nvidia, AMD, Autodesk, Emerson Collective | Marble* |
| Recursive Superintelligence | 2025 | Socher, Rocktäschel, Tian, Clune, Tobin | $4.65B | $650M | GV, Greycroft, Nvidia, AMD | None |
| Unconventional AI | 2025 | Rao, Carbin, Achour, Lee | $4.5B | $475M | a16z, Lightspeed, Sequoia, Lux, DCVC, Bezos | None |
| People& | 2025 | Zelikman, Harik, Peng, He, Goodman, and others | $4.48B | $480M | SV Angel, Harik, Nvidia, Bezos, GV, Emerson Collective | None |
| Ricursive Intelligence | 2025 | Goldie, Mirhoseini | $4B | $335M | Lightspeed, Sequoia, DST, Nvidia, Felicis, Radical | None |
| AMI Labs | 2025 | LeCun, LeBrun | $3.5B | $1.03B | Cathay, Greycroft, Hiro, HV, Bezos Expeditions, Nvidia, Samsung, Temasek | None |
| Whole | ~$127B | ~$30B | ||||
| * Restricted analysis launch. Tinker is a fine-tuning software for researchers; Marble is a 3D-world-generation API in early associate entry. Neither is a general-availability industrial product. Sources: firm bulletins, Bloomberg, Monetary Instances, TechCrunch, Crunchbase, and PitchBook reporting from 2024-2026. Valuations replicate the newest confirmed spherical; figures for rounds in lively negotiation are usually not included. | ||||||
To reply these questions, let’s establish 4 patterns throughout this cohort of firms.
Sample 1: The pedigree premium. Each founder is a acknowledged chief of their discipline, and most come from a small set of establishments. Roughly four-fifths maintain PhDs, principally in laptop science from a handful of universities — Berkeley, Stanford, MIT, Toronto, Alberta, Cambridge, UCL — and many of the relaxation left PhDs at a kind of applications to begin their firms.
On the employer aspect, the focus is tighter nonetheless. 4 of the twelve firms are anchored by DeepMind alumni (Ineffable, Reflection, Ricursive, Recursive Superintelligence). Two are anchored by OpenAI alumni (Pondering Machines, Secure Superintelligence). AMI Labs traces again to Meta’s FAIR group, and People& attracts its founders from throughout Anthropic, xAI, and Google. Stanford and Berkeley college appointments account for many of the relaxation (World Labs, Bodily Intelligence, and Noah Goodman of People&).
4 establishments — DeepMind, OpenAI, Berkeley, and Stanford — have produced the founders of almost each Virgin Unicorn within the desk. Buyers are pricing CVs, not merchandise.
Sample 2: Nvidia as kingmaker. 9 of the twelve firms within the desk have Nvidia as an investor. The provider of the picks and shovels can also be an fairness holder within the prospectors. Nvidia will get early visibility into essentially the most bold AI bets, locks in compute commitments, and earns multiples on capital deployed at near-zero marginal price. Promoting the shovels was already a superb enterprise. Proudly owning the mines too is unprecedented.
Sample 3: The cap tables are unusually huge. Every spherical within the desk features a syndicate of ten to twenty traders — enterprise companies, company strategics, sovereign wealth funds, and people. Sequoia and a16z nonetheless lead. However the rounds are giant sufficient that they require balance-sheet capital — from JPMorgan, BlackRock, Alphabet, the UK Sovereign AI Fund, Samsung, Temasek, ADIA, and Bezos personally — to fill out. That makes these rounds structurally totally different from classical enterprise financings.
Sample 4: A post-LLM thesis. Each firm is arguing, in some type, that the present paradigm isn’t sufficient — that scaling LLMs received’t attain AGI, and that one thing else (world fashions, reinforcement studying, agentic programs, AI scientists, novel chips, formal mathematical reasoning) is required. The thesis is the product. The product is a promise.
Others have dissected these unicorns:
- Howard Marks, in his December 2025 Oaktree memo Is It a Bubble?, described investor habits as “lottery-ticket considering” — traders backing startups with no product on the dream of an unlimited payoff regardless of an amazing chance of failing.
- Derek Thompson, writing in October, framed the identical dynamic by reporting {that a} Pondering Machines pitch assembly was described by one investor as “essentially the most absurd pitch assembly” as a result of Mira Murati “couldn’t reply any questions” about what she was constructing.
- GeekWire’s personal year-end survey of regional enterprise traders discovered the identical skepticism nearer to residence: the bubble, they stated, is most pronounced on the early phases, the place AI storytelling can substitute for actual traction.
The lottery-ticket framing is now standard knowledge. However will this lottery pay out? One technique to handicap the chances is to look to the previous.
What historical past teaches us
The closest historic parallel just isn’t the dot-com period. Webvan, Pets.com, and Boo.com failed not as a result of they had been pre-product, however as a result of they’d merchandise and dangerous enterprise fashions. These firms burned capital on infrastructure and advertising and marketing, not on analysis.
The nearer cautionary tales are the celebrity-founder pre-product flops of the final fifteen years.
- Magic Leap raised $3.5 billion over 9 years on the energy of Rony Abovitz’s prior exit and shipped a flop.
- Quibi raised $1.75 billion on Katzenberg and Whitman’s pedigree and lasted six months.
- Inflection AI raised $1.5 billion on Mustafa Suleyman and Reid Hoffman and was successfully absorbed into Microsoft in 2024 — its staff employed, its expertise licensed, its firm hollowed right into a shell.
In every case, founder credentials raised the cash. The product by no means justified the valuation.
The structurally closest analogy, although, is biotech. Roughly 80% of 2021 biotech IPOs had been pre-revenue. The chance {that a} pre-clinical drug reaches commercialization is below 10%. Improvement takes a decade and prices $1 billion. But a Bentley College examine of 319 biotech IPOs from 1997 to 2016 discovered that the cohort produced over $100 billion in internet shareholder worth regardless of a failure fee above 50%. The winners had been giant sufficient to hold the portfolio. And most of the most profitable biotechs had been acquired earlier than reaching profitability.
The Virgin Unicorns are biotech-shaped companies. Pre-revenue, science-driven, decade-long timelines, binary outcomes, acquisition as the standard exit. However they aren’t financed like biotechs. Biotech traders launch capital in milestone tranches tied to particular scientific outcomes, and so they count on most candidates to fail. Virgin Unicorn traders launch capital in a single giant spherical on the energy of a CV, and worth for fulfillment. Identical form of enterprise, reverse financing logic. That mismatch is the place the frustration will come from.
Why Sequoia invests anyway
The OpenAI story counters the biotech analogy. From its 2015 founding to the ChatGPT launch in late 2022, OpenAI appeared precisely like a Virgin Unicorn — pre-consumer-product for seven years, billions in capital, and solely analysis to point out for it. Then ChatGPT shipped and income went from zero to over $10 billion in three years. No biotech has ever scaled like that.
Sequoia and different traders writing checks to as we speak’s Virgin Unicorns aren’t pricing for biotech outcomes. They’re pricing for the second coming of OpenAI.
The desk above makes the scale of that wager legible. Early-stage enterprise traders goal for a 10x return. Most of those twelve will return zero, so the one winner has to hold the opposite eleven by itself. At a $127 billion mixture marked-up worth, which means the winner alone has to supply one thing like $1.3 trillion in worth.
That’s not a forecast — it’s the wager the VCs have already positioned. Sequoia and a16z made precisely this type of wager on OpenAI and Anthropic, and the on-paper returns have already vindicated it many instances over. Anthropic itself appeared like a Virgin Unicorn in 2022 — after which it shipped Claude and constructed income.
The historic document suggests some skepticism. However bubbles have a approach of manufacturing the occasional Amazon or Google amid the wreckage. Figuring out which Virgin Unicorn will develop into a trillion-dollar firm — a “kilocorn,” a thousand unicorns in a single — is hard. Which one would you wager on?
