Flapping Airplanes
Flapping Airplanes is a foundational AI research lab developing a megakernel framework that makes GPUs more efficient and expressive, enabling training, inference, and other workloads to run inside a single kernel. The lab's systems-level work aims to address the growing gap between AI demand and available data and compute resources. Its framework can already perform entire training runs, serve inference, and play video games simultaneously.
- Artificial Intelligence
- Developer Tools
- Software Only
Funding
Raised to date
$1.2BRaised to date based on public sources. This may differ from the amount the company actually raised and is based only on what is publicly available on the internet.
across 2 rounds
+5$1B · Series A
Announced Oct 2026
Disclosed cumulative: $1.1B
Founders
Founder details are not available yet.
Product
Problem
Demand for AI is increasing exponentially, outpacing the supply of data and compute. Trillions of dollars are being spent on AI, and companies are differentiated by their ability to use resources efficiently, creating a fundamental bottleneck for model development and deployment.
Solution
Flapping Airplanes is a foundational AI research lab focused on solving the data and compute efficiency problem through systems-level innovation. The lab has developed what it describes as the world's most complete megakernel framework, which makes GPUs more efficient for cheaper inference and training while also making them more expressive. This framework enables the use of unconventional architectures and optimizers, and can overlap memory-bound inference with compute-bound training inside the same kernel. The system can already perform entire training runs, serve inference, play video games, and do all three simultaneously in a single kernel, with the team believing GPUs are more flexible and valuable than most assume.
Target Audience
The primary audience is the broader AI research community and organizations that depend on advances in machine learning efficiency, though the lab's work is foundational and not tied to a specific customer segment.
Features
- World's most complete megakernel framework for GPU efficiency and expressiveness
- Ability to overlap memory-bound inference with compute-bound training within a single kernel
- Support for unconventional architectures and optimizers
- Capable of performing training runs, serving inference, and playing video games simultaneously in one kernel
- Foundational research approach exploring multiple radical technical directions rather than a single hypothesis