Flapping Airplanes raises $1 billion to pursue more efficient AI training

Flapping Airplanes has secured $1 billion for a research effort aimed at making AI less costly to train and run. The startup announced its Series A on October 21, co-led by Index Ventures and Lux Capital. Bloomberg reported a $6 billion valuation, including the new money, for the company founded in late 2025.

The financing backs a lab still focused on research rather than a commercial product. Bloomberg places Flapping Airplanes among a newer group of AI research startups that have yet to introduce commercial offerings. Its technical pitch starts with software that changes how work is organized on a graphics processing unit, or GPU.

One GPU program, several kinds of work

The founders describe their approach as a megakernel framework. A kernel is a program that runs on a GPU; a megakernel combines many tasks within one program. According to TechCrunch, the founders say their framework can perform entire AI training runs, serve inference—the work of running a trained model—and play video games, even doing all three simultaneously in a single kernel.

The proposed efficiency gain comes from pairing tasks that use the chip differently. The founders say the framework can overlap inference constrained by memory access with training constrained by computation. Put simply, one task spends much of its time waiting for data, while the other spends much of its time doing math.

Combining those workloads is intended to leave the GPU idle less often. The company’s premise is that AI systems have barely tapped what GPUs can do. This is a software-centered approach to extracting more useful work from the hardware, with the stated goal of reducing the cost of training and running models.

Investors reprice the lab

Sequoia Capital, Alphabet’s GV and Kleiner Perkins also participated in the new financing, Bloomberg reported. The round follows a $180 million seed investment from GV, Sequoia and Index nine months earlier, according to TechCrunch. That earlier financing accompanied the startup’s launch; the Series A now gives the research effort a much larger pool of capital.

The financing is not a commercial-product debut. It supplies money to continue the research and hire, as TechCrunch describes the company’s next task. The scale of the investment is therefore a commitment to developing the approach, while the founders’ descriptions of what the framework can do remain company claims.

The people and the less-data ambition

Flapping Airplanes was founded by Aidan Smith and brothers Asher and Ben Spector. Their backgrounds span startup building, statistics and computer science:

• Aidan Smith is a former founder and Thiel fellow who worked at brain-implant company Neuralink while in college.
• Asher Spector was pursuing a statistics doctorate at Stanford before co-founding the company.
• Ben Spector was a computer science doctoral student at Stanford and founded the student-run startup accelerator Prod while at MIT.

Beyond better GPU use, the founders want intelligent systems that learn from less data. Asher Spector described that ambition to TechCrunch shortly after launch: “A model that’s a million times more data efficient is probably a million times easier to put into the economy.” That was an argument for pursuing data efficiency, not a statement that the company had achieved such an improvement.