Flapping Airplanes Raises $1 Billion at a Reported $6 Billion Valuation Nine Months After Its Seed

Flapping Airplanes, a year-old AI research lab, raised a $1 billion Series A led by Index and Lux, TechCrunch reported on October 21. The valuation is reportedly $6 billion, a figure TechCrunch attributes to Bloomberg.

Earlier backers Sequoia, GV and Kleiner Perkins also took part, along with Valor, Atreides, A* and Nvidia, TechCrunch said. No announcement from the company itself was found.

How the round compares with the seed

The lab raised $180 million nine months before this round at a $1.5 billion valuation, from GV, Sequoia and Index, according to TechCrunch. The Decoder dated its report of that seed January 29.

Round	Amount	Valuation	Named investors
Seed, January 2026	$180 million	$1.5 billion	GV, Sequoia, Index
Series A, October 2026	$1 billion	$6 billion, reported	Index, Lux lead

1. Seed, January1.5 USD billions
2. Talks reported by Forbes, July 315 USD billions
3. Series A, October6 USD billions

Source: TechCrunch, 2026-10-21, and runtimewire reporting of Forbes, 2026-07-31

The terms moved during the summer. Runtimewire, citing Forbes on July 31, reported talks at a $5 billion valuation led by Index and Lux, with a round in the hundreds of millions of dollars. The closed figures are larger on both counts.

Who runs it and what it says it will build

TechCrunch names three founders: Aidan Smith, a Thiel Fellow who worked at Neuralink during college, Asher Spector, a statistics PhD student at Stanford, and Ben Spector, a computer science PhD student at Stanford who founded Prod while at MIT. It gives no formal titles.

The lab's aim is models that train on far less data, described as learning "without ingesting half the internet". The Decoder reported that the company estimates humans are 100,000 to 1,000,000 times more data-efficient than current models. That estimate comes from the company's own social media post and is not independently tested.

The one engineering detail: a GPU megakernel

TechCrunch says the lab is building a GPU "megakernel" framework that bundles training, inference and other tasks into one program. The idea is to overlap memory-bound inference with compute-bound training so the chip is rarely idle.

Asher Spector argued, in TechCrunch's paraphrase, that a model a million times more data-efficient would be about a million times easier to put into the economy.

What is not known

TechCrunch gives no team size, no compute commitments and no hiring targets. The Decoder's January report named no founders and no valuation.

No benchmark, model or paper from the lab has been cited in any of the three reports. Related funding coverage on this site includes Legora's $550 million Series E and Eka's robotics round.

None of the three reports gives a product date. The next datapoint will be the first model or paper the lab releases.