A couple dozen desks, a few whiteboards covered in equations, and not a single product for sale anywhere in the building.
That is the entire operation investors just priced at billions of dollars.
No app. No subscription. No customers writing checks.
Just a bet that a handful of researchers can rewrite how machines learn.
So how does a company with nothing to sell become worth more than most public firms?
Why investors are paying for ideas, not income
The mechanism is simple once you strip away the hype: venture firms are no longer pricing current products or revenue, they are pricing specific research talent and the chance that talent discovers a cheaper way to build artificial intelligence. A small founding team that previously worked at a major AI lab can now raise money purely on reputation and a technical thesis.
That thesis, in this case, is about training large AI models using less data, through a megakernel, a single GPU kernel designed to run multiple workloads, such as training, inference and even unrelated compute like video games, simultaneously rather than sequentially, so a chip keeps working instead of sitting idle between tasks.
Investors are not funding a finished tool. They are funding a wager on training efficiency that could make artificial intelligence dramatically cheaper to run everywhere else, from chatbots to the coding assistants already billing companies by the day.
Inside the deal that quadrupled a company’s price tag in nine months
Flapping Airplanes, a San Francisco AI research lab founded in late 2025, has raised a $1 billion Series A at a $6 billion valuation, Bloomberg reported.
Even the math behind that number is startling. The company has no product and no revenue, and only “a couple dozen people”, per Index. Its valuation has quadrupled in nine months, from $1.5 billion when it launched with a $180 million seed round in January.
Nine months. Four times the price. Zero paying customers.
The investor list reads like a who’s who of venture capital, led by Index Ventures and Lux Capital, with Sequoia Capital, Alphabet’s GV and Kleiner Perkins also taking part, and Nvidia counting itself among the company’s backers. A major chipmaker holds a stake too, betting that cheaper training methods could eventually sell its own hardware to a wider pool of buyers.
The hard numbers behind a brand new funding category
This is not an isolated case. Researchers now have a name for firms like this one: neolabs, a research-first startup organized around a scientific breakthrough hypothesis rather than a product roadmap, built to raise institutional capital at scale before any commercial output exists.
What makes the round remarkable is its size relative to the broader market. Dealroom data puts the round in the top 1% by amount of 3,497 comparable early-stage US deep tech rounds.
The category itself has grown fast, though trackers disagree on its exact size. A venture analysis published in June 2026 counted more than 40 such labs that had raised over $40 billion combined since 2024, most with no disclosed revenue. A separate, narrower count by the venture firm Radical Ventures put the number at a smaller cohort with a smaller total.
Other members of the group carry equally lopsided numbers. Nvidia has made investments including $8 billion in total for Safe Superintelligence, the stealth lab co-founded by OpenAI co-founder Ilya Sutskever, reportedly valuing it at $32 billion. SSI has released no public research or product since launching in 2024. Another, led by a well known former Meta scientist, raised more than a billion dollars on a similar pattern. Twenty-nine AI neolabs founded since June 2024 have raised a combined USD 23 billion from venture investors, and under 35 percent of neolabs studied have commercial offerings, meaning most of this capital sits inside research budgets rather than shipped software.
Where the bet gets complicated and who could lose
Not every neolab is purely theoretical. One competitor, Thinking Machines Lab, shows what happens when the research actually starts producing income, having scored a $2 billion seed round at a $12 billion valuation, launched Tinker and its Inkling model, and reported in September 2026 that the company’s annual revenue run rate had passed $100 million, citing a source familiar with its finances.
So the spread inside this category is enormous, running from labs with meaningful revenue to labs with none planned at all. That gap matters because investors are pricing two very different bets under the same label.
Meanwhile the surrounding infrastructure carries its own exposure. Neolabs are raising money for years of research requiring expensive compute, while the firms supplying that compute are spending billions to build capacity meant to serve them, a cycle that looks far harder to justify if the research bets stall.
Even the founders admit commercialization sits years away. “The basic thesis is demand for AI is going to outpace chip and data supply,” co-founder Asher Spector said, according to Bloomberg. “The company that’s going to be most successful, and create the most value for the world, is the one that can squeeze the most out of the inputs.” Spector said that with so much money going into infrastructure today, a 10% performance boost for every chip would be worth a great deal economically.
What it means for the next round of startups chasing capital
For ordinary founders outside this rarefied tier, the lesson is less about chasing a billion dollar round and more about what investors are now willing to fund. A credible team and a sharp technical thesis can apparently outweigh a finished product, at least for a narrow set of researchers with the right résumé.
That shift also reshapes who holds leverage in Silicon Valley. Engineers who once needed years inside a large company to prove themselves can now walk out with a founding team and raise at valuations that used to require actual customers and revenue.
Still, the math only works if someone eventually ships something people pay for. Every one of these labs is ultimately betting that a research breakthrough converts into a business before investors lose patience with billion dollar valuations built on whiteboards.
For now the wager looks remarkably confident. A couple dozen people, no revenue, and a price tag that keeps climbing anyway.
