Humanity's Last Machine

0:00 Over the past few years, General-purpose robotics have sold a vision of hardware that  can do possibly any task a human can. The biggest bottleneck to robotics has  been the software. As VCs and engineers at these companies, this is our day to day. We wanted to understand the components of  humanoid robotics from first principles. So we spent months breaking the  system down, piece by piece. But you can't understand the reality of this  industry by looking at a screen in California.

0:22 You have to speak to the people iterating  on the hardware. So our crew went to China… After a 15-hour flight, we landed in Shanghai. Over the next 10 days, our journey brought us  to hotspots in the supply chain like Sūzhōu, Hángzhōu, Shēnzhèn, and Dōngguǎn. To build a humanoid, you need four things:  Sensors, a Brain, a Battery, and Actuators. Very few parts need to be made  from scratch and manufacturing capabilities rests on the shoulders of giants. The development of EVs, drones,  and consumer electronics has accelerated robotics hardware manufacturing.

0:54 China produced over 12 million EVs in 2024, which commoditized the lithium-ion  batteries now powering humanoids. DJI spent a decade perfecting  high-torque motors—those same motors now drive robotic actuators. And the cameras and sensors in your iPhone are now off-the-shelf components  for any robot manufacturer. China has tremendous capabilities to drive  actual innovation. Take the dexterous hand. We met multiple teams in China building hands that  rival Tesla's—like this one from Sharpa Robotics. They’ve previously scaled Lidar  from costing $50K to $500; today they are trying to create  far more affordable robotics hands.

1:25 And the speed of it all is staggering. In Suzhou, a micro-actuator factory  turns raw metal rods into precision joints. If a part fails, they fix the  design and cut a new one by lunch. A gripper manufacturer told us every component  they need is within a 25-minute drive of his office. Founded five months ago, the company’s  already gathered 10,000+ hours of robotics data. After visiting factory after factory, we  began to see a fundamental difference: the US and China are building general-purpose  robots with completely different constraints.

1:52 First: Demand In the US, everyone’s racing to build a foundation  model that could allow a robot to enter an unseen home and clean a kitchen or make a bed, or get  deployed in any car factory and wire a harness, all from a single high-level command with minimal  fine-tuning or robot technicians involved. An $80,000-a-year wage doesn't justify a humanoid robot on economics alone until it  can perform reliably, end-to-end. In China, a $15,000-a-year wage doesn't justify  a humanoid robot on economics alone either.

2:17 But demand is actively being created—by  both the state and capital markets. Contrary to the US definition of  success, Beijing wants China to be the world’s factory for robotics. As  a result, many robotics companies here optimize less for peak intelligence,  and more by quantity of robots sold. China does this in a few ways: Provincial government owned facilities like  museums and hospitals act as anchor customers, and are encouraged to deploy humanoid robots. Robots perform at the Spring Festival Gala, at concerts, creating cultural momentum. And  quietly, there are circular supply chains.

2:45 A motor supplier might buy robots from  the OEM they supply to. It's not optimal, but it keeps factories running  and iteration loops alive. Second: Capital Robotics is often referred to as the graveyard of  venture capital, and for good historical reason. In the U.S., investors are deeply cautious about  factories optimized for first-generation robots. China’s capital system, by contrast, is  far more willing to absorb early losses. The bet is that early manufacturing  dominance matters more than early profits.

3:09 A single announcement that a company has become  a supplier to Tesla can send a stock up 50%, creating incentives that benefit both sides. Suppliers gain credibility,  volume, and capital-market upside, while Tesla sees the real time creation of  a supply chain that didn’t exist before, and one that is constantly fighting  to be faster, cheaper, and better. General-purpose robotics is a major  technological frontier where the US and China started investing at roughly the same time. There is no clear head start or clear winner.

3:34 After months of research and time on the  ground, we came away with three observations: First: Some companies are learning to  operate across both worlds—developing AI in the US while leveraging  China's manufacturing ecosystem. Second: With rising tariffs, suppliers are shifting assembly to  Thailand, Mexico, and elsewhere. Today, Chinese companies possess  the lion share of process knowledge. Companies with process knowledge that can replicate their capabilities abroad  have structural advantage over time. And third—this surprised us—Chinese  robotics research is much closer to the frontier than most people in the West realize.

4:04 The best work coming out of China is  improving fast and closing the gap. So how do we feel after ten  days of running around China, wrapping it all up on the very last day of 2025? Honestly? A mix of excitement. Accomplishment.  And this sense of calling. The more we pulled on this thread, the more we  realized there’s still so much to learn and build. As we pack our bags and head back to our normal  lives, we didn’t want this trip to just… end.

4:25 So for anyone else who dreams  of the future and feels this same pull to build intuition  on the frontier of robotics, we put together this resource  and with the help of dozens of experts who reviewed our work and kept us honest. All of that lives here, on this site. For anyone who’s interested. And for us, this is just the beginning.