Flapping Airplanes raises $1B at $6B valuation with zero product
Investors just handed $1 billion to a one-year-old AI lab with no product because they promise to train models with a fraction of the data.
Funding · Source: TechCrunch
What happened
Flapping Airplanes just raised a $1 billion Series A round. Index Ventures and Lux Capital co-led the massive deal. The one-year-old artificial intelligence research lab now holds a reported $6 billion valuation. They have absolutely no commercial product and zero revenue. The valuation quadrupled in just nine months. In January, they raised a $180 million seed round valuing them at $1.5 billion. Total funding now sits at $1.18 billion.
The founding team has serious technical credibility. Asher Spector holds a statistics doctorate from Stanford. His brother Ben Spector is a Stanford computer science doctoral student who previously founded the Prod accelerator at MIT. Ben also led the open-source GPU toolkit ThunderKittens. Aidan Smith is a Thiel Fellow who worked at Neuralink. Together, they spent their first year building new primitives for graphics processing units. They operate in person with a couple dozen people in San Francisco.
The startup is attacking the biggest bottleneck in artificial intelligence. They want to build models that learn from far less data than today's frontier systems consume. The founders built what they call the world's most complete megakernel framework. A megakernel runs an entire program on a GPU as a single kernel. The chip never loses time between steps. The framework can handle a full training run, serve inference, and play a video game simultaneously. It overlaps memory-bound inference with compute-bound training.
Key facts
- $1 billion — Series A funding raised by Flapping Airplanes
- $6 billion — Reported valuation of the one-year-old startup
- $1.18 billion — Total capital investors have put into the company
- $180 million — Seed round raised nine months prior in January
- 1 million — Times more data efficient the founders want their models to be
Why it matters
The artificial intelligence industry is running out of compute power and clean data. Training frontier models costs billions of dollars and requires scraping the entire internet. Demand for artificial intelligence will keep outrunning the supply of infrastructure. Flapping Airplanes is trying to completely reshape the cost structure of the industry. Co-founder Asher Spector says a model that is a million times more data efficient is a million times easier to put into the economy.
This funding round reveals a massive divide in venture capital strategy. Investors are financing fundamental research before a commercially established business even exists. They are betting that breakthroughs in efficiency will create enormous economic value. Markets like robotics, high-frequency trading, and life sciences desperately need more efficient artificial intelligence. If this megakernel framework works, the competitive advantage shifts away from companies hoarding massive datasets. The winners will be the teams who can do the most math with the least amount of hardware.
For builders
Efficiency beats raw data volume
Stop worrying about scraping the entire internet. The next massive opportunity is doing more with less data. Founders who build tools to make training and inference cheaper will attract serious capital. The market pays a premium for data efficiency.
Optimize hardware at the lowest level
Hardware is expensive and chips sit idle too often. Flapping Airplanes proved that writing better GPU kernels is worth billions of dollars. Engineers who can overlap memory-bound and compute-bound tasks are the most valuable people in tech right now. Nvidia chips still have untapped potential.
Hard tech requires no early product
You do not need a software dashboard to raise money. If you are solving a fundamental infrastructure bottleneck, investors will fund the pure research. Deep tech founders should focus on scientific breakthroughs instead of rushing a commercial release. Capital is available for massive technical risks.
My take
Product is not the prize right now. Efficiency is. Investors are throwing a billion dollars at a math problem because whoever figures out how to train models without eating the whole internet wins the AI war.
Original reporting: TechCrunch. This is my rewrite and opinion.