Flapping Airplanes has raised a $1 billion Series A for its effort to wring more useful work from the GPUs that train and run AI models. TechCrunch reported that the year-old company announced the financing Wednesday, with Index Ventures and Lux Capital co-leading. The publication put the valuation at a reported $6 billion.
At the center of the company is what its founders call a megakernel framework. A GPU kernel is a program that runs on the chip; Flapping Airplanes bundles multiple kinds of work into one program that can train models, serve responses and even run games simultaneously. For AI teams, the framework is meant to coordinate those jobs instead of treating them as isolated workloads.
That scheduling matters because different jobs stress different parts of a chip. The company says its framework can pair inference work that waits on memory with arithmetic-heavy training, keeping more of the GPU occupied. Its larger bet is that better utilization can lower the cost of training and running models while reducing the amount of data they require.
Flapping Airplanes was founded by Aidan Smith and brothers Asher and Ben Spector. Smith previously worked at Neuralink while in college. Asher studied statistics at the doctoral level at Stanford, while Ben was a Stanford computer-science doctoral student and previously founded the student-run startup accelerator Prod.
The Series A comes only nine months after a $180 million seed round that valued Flapping Airplanes at $1.5 billion. On the reported new figure, its valuation has quadrupled. Earlier backers Sequoia, GV and Kleiner Perkins joined the latest financing, which TechCrunch said will support hiring as the company tries to turn its GPU-efficiency thesis into a working AI system.