Flapping Airplanes, a startup that has captivated Silicon Valley, boasts an impressive founding team and a mission focused on developing methods to train large models using less data. On Wednesday, the company revealed it has successfully secured a $1 billion Series A funding round, co-led by Index and Lux, with a reported valuation of $6 billion. This funding comes just nine months after the company’s inception, during which it raised $180 million in a seed round from GV, Sequoia, and Index. The initial seed round valued the company at $1.5 billion, marking a significant increase in valuation.
Other contributors to the latest funding round include previous investors such as Sequoia, GV, and Kleiner Perkins. New investors also include Valor, Atreides, Kleiner Perkins, A*, and Nvidia. Flapping Airplanes was co-founded by Aidan Smith and brothers Asher and Ben Spector. Smith, a former founder and Thiel fellow, gained experience at Neuralink, a brain implant company, during his college years. Asher Spector holds a PhD in statistics from Stanford, while Ben Spector was a PhD student in computer science at Stanford and established Prod during his time at MIT. Prod is a relatively obscure, five-year-old student-run startup accelerator that has been described as “one of the most powerful talent incubators in Silicon Valley,” according to Sequoia partner David Cahn, who is also an investor in Flapping Airplanes.
Together, the founders are creating what Cahn refers to as the “young person’s AGI lab.” They characterize their work as “the world’s most complete megakernel framework,” capable of executing entire training runs, serving inference, playing video games, and even performing all three tasks simultaneously within a single kernel. (A kernel is a program that operates on a GPU, and a megakernel consolidates multiple tasks into one program.) The founders assert that their framework “can overlap memory-bound inference with compute-bound training,” allowing it to execute tasks that primarily wait for data retrieval alongside those that focus on computations, ensuring that the chip remains active.
The underlying belief is that artificial intelligence has only begun to explore the full potential of GPUs, and they are confident that it can develop intelligent systems “without ingesting half the internet.” Their objective is to create systems that are more cost-effective to train and operate. “A model that’s a million times more data efficient is probably a million times easier to put into the economy,” Asher Spector stated shortly after the startup’s launch. With a fresh $1 billion in funding, they are now poised to hire and work towards realizing that vision.
