Analysis
Flapping Airplanes, an AI lab founded about a year ago, has raised a $1 billion Series A, according to TechCrunch.
The round values the company at roughly $6 billion, Bloomberg has reported, up from the $1.5 billion valuation it carried after a $180 million seed round just nine months earlier.
Index and Lux led the new round, with Sequoia, GV and Kleiner Perkins — all seed investors — returning, alongside new backers Valor, Atreides Management, A* and Nvidia.
“Index and Lux led the new round, with Sequoia, GV and Kleiner Perkins — all seed investors — returning, alongside new backers Valor, Atreides Management, A* and Nvidia.”
Who's behind the 'young person's AGI lab'?
The company was founded by Aidan Smith, a Thiel Fellow and former Neuralink employee, along with brothers Asher Spector, a Stanford statistics PhD, and Ben Spector, a Stanford computer science PhD student who also founded Prod, a student-run startup accelerator, while at MIT.
Prod has built a reputation inside Silicon Valley as a talent pipeline rather than a typical accelerator. Sequoia partner David Cahn, who has backed the fund, said Prod "has become one of the most powerful talent incubators in Silicon Valley" and described the founders as building "the young person's AGI lab."
The company's pitch centers on data efficiency: training large models without needing to ingest, in Spector's words, "half the internet." The team has built what it calls a megakernel framework — a program that bundles many GPU tasks into a single kernel so training, inference and other workloads can run together, overlapping memory-bound inference with compute-bound training rather than running them as separate, sequential steps.
"A model that's a million times more data efficient is probably a million times easier to put into the economy," Asher Spector told TechCrunch.
The efficiency framing echoes arguments other labs, including DeepSeek, have made about training capable models more cheaply — though Flapping Airplanes is the first to raise at this scale on an unproven version of the thesis.
The approach runs counter to the scale-above-all doctrine that has driven OpenAI, Anthropic and Google DeepMind to spend hundreds of billions on compute rather than on data efficiency. That makes Flapping Airplanes a direct bet against the current industry consensus — a bet that remains unproven at frontier-model scale, since the company has not yet published benchmarks showing its efficiency claims hold up against GPT-class or Claude-class models trained the conventional way.
What the headline $6 billion figure doesn't capture is that the company is pre-product in any commercial sense; the round is pricing a research thesis and a founding team's pedigree, not revenue or deployed models.
