Flapping Airplanes secures $1B Series A at reported $6B valuation
Flapping Airplanes, a year-old AI startup, has announced a $1 billion Series A round at a reported $6 billion valuation. Index and Lux co-led the financing, which arrives nine months after the company launched with a $180 million seed round that valued it at $1.5 billion.
Sequoia, GV and Kleiner Perkins returned as investors. Valor, Atreides, A* and Nvidia are also investors in the company. The latest valuation is four times the level reported for the seed financing.
A bet on more efficient model training
Founded by Aidan Smith and brothers Asher and Ben Spector, Flapping Airplanes is pursuing methods to train large models with less data. The company’s premise is that AI workloads have not yet exhausted the capabilities of GPUs, and that more efficient use of those chips could reduce the cost of training and operating intelligent systems.
The founders describe their technology as a “megakernel” framework: a kernel is a program that runs on a GPU, while a megakernel combines multiple tasks within one program. Flapping Airplanes says its framework can perform complete training runs, serve inference, play video games, and do all three at once in a single kernel.
Combining GPU workloads
A central claim is that the framework can overlap memory-bound inference with compute-bound training. In practical terms, the company says a workload that spends much of its time waiting for data to be fetched can run alongside one focused primarily on mathematical computation, reducing periods when the chip would otherwise be idle.
Asher Spector told TechCrunch shortly after the company’s launch that a model that is a million times more data efficient would be a million times easier to put into the economy. The new capital gives the company resources to hire and pursue that objective.
What the round signals
The financing places substantial investor backing behind an infrastructure-led approach to AI development rather than a disclosed model release. Flapping Airplanes will now have to demonstrate whether its framework can translate its stated GPU-utilisation and data-efficiency goals into systems that cost less to train and run.
For businesses assessing AI investments, the round is a reminder to evaluate the infrastructure economics behind model deployment: training data requirements, GPU utilisation and inference costs can be as consequential as headline model capabilities.