Flapping Airplanes, an AI research lab founded about a year ago, announced a $1 billion Series A round on Wednesday, October 21, co-led by Index and Lux, TechCrunch reported. The valuation is reportedly $6 billion, four times the $1.5 billion figure from its seed round just nine months earlier. The company’s stated mission is to find ways to train large models with much less data than today’s systems consume, which makes its funding a bet against the idea that progress requires ever larger datasets.

The round

TechCrunch reports that the $1 billion Series A follows a $180 million seed round, led by GV, Sequoia and Index, that valued the company at $1.5 billion. Beyond the co-leads, participants named in the report include earlier backers Sequoia, GV and Kleiner Perkins, along with Valor, Atreides, A* and Nvidia. The $6 billion valuation is described as reported, not confirmed by the company in the article.

The team and the pitch

The company was founded by Aidan Smith and brothers Asher and Ben Spector. TechCrunch describes Smith as a former founder and Thiel fellow who worked at Neuralink during college; Asher Spector as a former Stanford statistics PhD student; and Ben Spector as a Stanford computer science PhD student who founded Prod, a student-run startup accelerator at MIT. Sequoia partner David Cahn, an investor, has called the group “the young person’s AGI lab,” according to TechCrunch.

The founders’ stated goal is intelligent systems built “without ingesting half the internet.” Asher Spector told TechCrunch shortly after launch: “A model that’s a million times more data efficient is probably a million times easier to put into the economy.” That claim is a founder’s aspiration, not a result; the company has not published a model or benchmark scores in the coverage we reviewed.

What a megakernel is

The founders describe their technical work as “the world’s most complete megakernel framework,” which they say can run entire training runs, serve inference and even play video games, all in a single GPU kernel. A kernel is a program that runs on a GPU; a megakernel bundles many tasks into one program instead of launching many small ones.

The company claims it can overlap memory-bound inference, which spends much of its time waiting for data to arrive, with compute-bound training, which spends its time on arithmetic, so the chip is rarely idle. The premise is that today’s software leaves much of a GPU’s capability unused. These are the company’s claims as relayed by TechCrunch; we have not seen independent measurements.

Why it matters

The round shows that investors are still willing to fund very large checks for early-stage labs with no public product, a pattern also visible in the reported OpenAI raise at a $1.25 trillion valuation. It also signals interest in a different axis of competition: efficiency rather than scale. Labs that need less data and less compute per capability could change who can afford to build frontier systems. For a primer on how rounds like this work, see how AI startup funding works.

There is a counterpoint: data efficiency has been a research goal for years, and no lab has yet shown a method that replaces large-scale pretraining at the frontier. The company’s progress will have to be judged on published results.

What to watch next

  • Whether Flapping Airplanes releases a model, paper or benchmark results to back its efficiency claims.
  • Hiring: how fast the company grows its research and kernel-engineering teams with the new capital.
  • Whether the company discloses its compute arrangements, including any link to Nvidia as an investor.
  • Competing efforts on data-efficient training from larger labs and other startups.
  • Confirmation of the valuation from the company or its investors.

Key takeaways

  • Flapping Airplanes raised a $1 billion Series A, co-led by Index and Lux, nine months after a $180 million seed.
  • The valuation is reportedly $6 billion, up from $1.5 billion at the seed.
  • The company says it aims for models that need far less training data and far less cost.
  • Its megakernel framework is meant to keep GPUs busy by combining training and inference; no independent results have been published.

Frequently asked questions

What does Flapping Airplanes do?

It is an AI research lab that says it wants to train large models with less data and lower cost, using a GPU software framework built around megakernels.

Who invested?

Index and Lux co-led the round, according to TechCrunch. Other investors named include Sequoia, GV, Kleiner Perkins, Valor, Atreides, A* and Nvidia.

Has the company released a model?

Not in the coverage we reviewed. Its claims about data efficiency remain unverified.