The physical world is powered by physical labor, and the U.S. has been running at a deficit for decades.
Today, this gap is more pronounced than ever. America is grappling with historic labor shortages across manufacturing, construction, healthcare, agriculture, and other critical sectors — amid a global push to radically transform our industrial base.
While AI software and automation platforms have gone a long way in helping humans do more with less in industrial settings, digital tools are only part of the solution. Many critical tasks still require physical labor — and always will.
We need much more than tools to augment human beings. We need an equivalent that can stand in for human labor that simply does not exist today.
The breakthrough we need: General purpose robots — intelligent machines that can work alongside us, performing dynamic tasks in the fast-changing industries that power the world economy.
This means going far beyond the limits of what industrial robots can do today. My time in the trenches of Tesla’s production hell taught me firsthand the limits of single-purpose robotics, which are only suited for stable, high-volume tasks in predictable settings. Robots with the human-level cognition, adaptability, and dexterity are required to operate in high-mix environments.
The opportunity is tremendous. Direct labor in physical sectors accounts for an estimated $30-$40 trillion of global GDP, and the majority is unautomated. As I’ve experienced firsthand scaling industrial robots in manufacturing settings, even the most automated industries of today have a low penetration of robots, with the robot-to-human ratio rarely exceeding 1:30.
Yet, with all the advances in AI and robotics over the past decade, infusing human-like intelligence into physical machines has remained a significant bottleneck. Until now.
Genesis AI is developing a universal robotics foundation model that will do for physical AI what large language models did for digital AI. While LLM developers had the advantage of more than a century worth of text- and image-based data on which to train their models, there isn’t a similar data reservoir for the physical world. Genesis is building this reservoir, pushing the limits of what was thought possible with synthetic and real-world data.
Bridging high-fidelity physics simulation, multimodal generative modeling, and large-scale data collection from real robots, Genesis’ data engine is the training platform that will make general purpose robots not only come to life, but unlock unlimited physical labor.
Eclipse is thrilled to be co-leading Genesis’ $105 million seed round. We believe physical (or embodied) AI is the opportunity of our generation, and that general-purpose robots are poised to have the most transformative impact we’ve ever seen on the critical sectors that touch every aspect of our lives.
In Genesis, we see an approach that will finally break the barrier to training robots: Their data-centric, full-stack approach takes on the core challenges of general purpose robotics head-on, and they are moving quickly. With a world class team of scientists, engineers, and builders from top-tier AI and robotics organizations including Carnegie Mellon, MIT, Stanford, Mistral AI, and NVIDIA, Genesis has the technical talent, vision, and strategy to define the next era of physical AI.
More on today's news from TechCrunch.
Follow Eclipse Ventures on LinkedIn or sign up for Eclipse’s Newsletter for the latest on the Industrial Evolution.
