Flapping Airplanes Raises $1 Billion at $6 Billion Valuation
Flapping Airplanes raises $1 billion at a $6 billion valuation to build a GPU megakernel designed to train and run AI models using far less data and compute.
Summary
On Wednesday, October 21, 2026, one-year-old Flapping Airplanes announced a $1 billion Series A, co-led by Index and Lux, reportedly at a $6 billion valuation. The financing follows a $180 million seed round nine months earlier from GV, Sequoia and Index at a $1.5 billion valuation, meaning the valuation quadrupled. Sequoia, GV and Kleiner Perkins joined the new round. The investor roster also includes Valor, Atreides, A* and Nvidia.
Aidan Smith founded the company with brothers Asher and Ben Spector. Smith is a former founder and Thiel fellow who worked at Neuralink during college. Asher was a Stanford statistics PhD, while Ben was a Stanford computer science PhD student and founded Prod at MIT. Sequoia partner and investor David Cahn described the five-year-old, student-run Prod as a major Silicon Valley talent incubator and called the founders a young person’s AGI lab.
Flapping Airplanes says its megakernel, a single GPU program bundling many tasks, can complete training runs, serve inference and play video games simultaneously. It claims the framework overlaps memory-bound inference with compute-bound training, reducing chip idle time and potentially creating intelligent systems without ingesting half the internet. The goal is cheaper training and operation. Asher Spector said millionfold data efficiency could make a model millionfold easier to introduce into the economy. The company will use its new capital to hire and try to validate that premise.
Positives
- The $1 billion Series A gives Flapping Airplanes substantial capital to hire and develop its GPU megakernel framework.
- The $6 billion valuation is four times the $1.5 billion valuation attached to its seed round nine months earlier.
- Index, Lux, Sequoia, GV, Kleiner Perkins, Valor, Atreides, A* and Nvidia provide extensive venture and industry backing.
- The megakernel claims to train models, serve inference and play games simultaneously within one GPU program.
- Overlapping memory-bound inference with compute-bound training could reduce idle GPU capacity and lower AI costs.
Risks & concerns
- The $6 billion valuation arrives after only one year and before the company has demonstrated that its efficiency goal works at scale.
- The megakernel’s performance and data-efficiency assertions remain founder claims, with independent results or benchmarks still limited.
- The company must convert $1 billion in financing into hires and a working system capable of delivering cheaper training and inference.
- Millionfold data efficiency is an ambition described by Asher Spector, not a result established in the available details.
