Jensen Huang Invests in Another AI Startup: Flapping Airplanes Valued at $6 Billion One Year After Founding, Aiming to Boost GPU Efficiency

Zhihui Dongxi October 23 news, yesterday Bloomberg reported that an AI startup founded about a year ago, which has not yet launched a commercial product, has raised $1 billion in funding, with a company valuation reaching $6 billion, about 42.6 billion yuan. This company is called Flapping Airplanes, and what it wants to do is train AI models with less data and compute.

This Series A round was co-led by Index Ventures and Lux Capital, with participation from Sequoia Capital, Alphabet's GV, Kleiner Perkins, Valor, Atreides, and Nvidia, among others. In the previous seed round, the company's valuation was still $1.5 billion, and now it has become four times the original.

Flapping Airplanes is headquartered in San Francisco and belongs to the "AI neolabs" that do research first and talk about products later. The company independently develops software used to drive Nvidia GPUs, claiming it can improve chip efficiency and reduce AI training and operating costs. However, its technical capabilities currently mainly come from the founders' descriptions, and independent tests, customer cases, and commercialization progress have not been made public, so follow-up remains to be seen.

1. Raised $1 billion in one year, valuation rose to about 42.6 billion yuan

Flapping Airplanes is headquartered in San Francisco, USA, and was founded at the end of 2025. The company's name comes from an ancient "ornithopter" concept, that is, an aircraft that flies by flapping its wings. In co-founder Spector's view, this is also a metaphor, namely building technology that combines the advantages of natural intelligence and artificial intelligence.

Flapping Airplanes is a so-called AI "neolab." Such companies are more willing at the beginning to spend energy on research rather than rushing to launch AI products and quickly generate revenue. Although the company has not yet produced a product, top investment institutions are still optimistic about betting on it.

According to Bloomberg, this $1 billion funding round was co-led by Index Ventures and Lux Capital, with participation from Sequoia Capital, Alphabet's GV, Kleiner Perkins, Valor, Atreides, and Nvidia.

TechCrunch added earlier funding information. The company started nine months ago with a $180 million seed round, about 1.3 billion yuan, with investors including GV, Sequoia Capital, and Index. At that time, the valuation was $1.5 billion, about 10.7 billion yuan. That is to say, from the seed round to this Series A round, Flapping Airplanes' valuation rose from $1.5 billion to $6 billion, quadrupling.

2. The founding team is very young, and the technical route focuses on "less data, less compute"

The company was founded by three people, namely Aidan Smith, Asher Spector, and Ben Spector.

▲The three founders (source: Bloomberg)

Smith is a recipient of the Thiel Fellowship and worked at the brain-computer interface company Neuralink during college. Asher Spector is a PhD in statistics at Stanford University. Ben Spector is a computer science PhD student at Stanford University and founded Prod while at MIT.

Prod is a five-year-old student-run startup accelerator, and Sequoia Capital partner David Cahn called it "one of the strongest talent incubators in Silicon Valley."

Cahn also described these founders as building "young people's AGI lab."

He believes that "AI demand will exceed chip and data supply. The company that can succeed most and create the greatest value for the world will be the one that can squeeze the most output from inputs." He also said that the current AI boom is spending too much money on infrastructure, and if each chip can deliver 10% more performance, the future economic value will be very large.

Smith said the company has been developing a set of software used to drive Nvidia GPUs. GPU stands for Graphics Processing Unit, the core hardware for training and running AI models, and it is good at parallel computing. With this software, Flapping Airplanes can make chips more efficient and also invoke these chips in more flexible new ways.

As the team said in a post on X, "Over the past year, we have been developing new primitives for GPUs. We believe this will enable new algorithms to change the way machines learn, open up new markets, and substantially reshape the cost structure of AI."

▲Flapping Airplanes' post on X

TechCrunch disclosed that the founders described this work as "the world's most complete megakernel framework." A kernel is a program running on a GPU, and a megakernel packs many tasks into one program.

The founders claim that this framework can execute the complete training process, provide inference services, play video games, and even complete all three at the same time in a single kernel. In addition, it can overlap "memory-bound inference" and "compute-bound training"—the former spends most of its time waiting for data reads, and the latter spends most of its time doing mathematical calculations. By stacking the two, it can keep chips from idling as much as possible.

3. Nvidia and Sequoia both bet, but commercialization remains to be verified

As a major supplier of AI chips, Nvidia also appears in the investor list of this funding round. Lux partner Brandon Reeves said that as AI development moves toward the "next frontier," "getting more output from existing resources" will be very critical. He believes that Flapping Airplanes "may be the company best positioned to do this."

Sequoia Capital partner David Cahn has known Spector and his brother Ben Spector for many years. He said this latest funding round shows that investors highly evaluate the company's founding team. "There are too many AI labs like this, and it is easy to drown in the noise," Cahn said. "But this company is not 'one among many,' but 'one of a kind.'"

However, capital bets do not mean the technology has been delivered. Spector hopes the company's work can make it easier for other enterprises that want to use AI to obtain AI, and the company also plans to eventually commercialize the technology. TechCrunch said Flapping Airplanes now has $1 billion in new funds to hire and try to turn its goals into reality.

But judging from the two reports, the company's "megakernel" framework, overlapping execution of training and inference, and other capabilities currently mainly come from the founders' descriptions. Neither report disclosed relevant information such as independent benchmark tests, customer cases, revenue, or product timelines, and its technical capabilities and commercialization progress remain to be seen.

Conclusion: The story of saving compute must be proven with products

When the marginal returns of "stacking compute" begin to be questioned, using GPUs more fully, as Flapping Airplanes does, and training usable models with less data are becoming the focus of many AI startups.

Nvidia invested, which to some extent also shows that Flapping Airplanes' research direction in GPU efficiency has been recognized by the chip manufacturer.

But whether Flapping Airplanes' megakernel framework can remain stable at larger scale, and whether improvements in data efficiency can be transformed into products customers are willing to pay for, currently still lack public verification.

Source: Bloomberg, TechCrunch