Figure AI

We invested in Figure AI because Brett Adcock is building the physical foundation of a labor economy that doesn't yet exist. Most AI companies are layering intelligence onto things people already do. Figure is building the machine that can do things people won't be able to keep doing.

The team's conviction is unusual. The engineering discipline is unusual. And the customer roster, BMW, one of the most rigorous manufacturing partners any physical AI company could hope to land, is unusual. A manufacturing line is not a place for demos. The fact that Figure is deploying into one tells you something specific about what the hardware and control stack can actually do.

What sold us in diligence wasn't the demo video. It was the unit economics trajectory. The cost curve on humanoid hardware over the next five years is the single largest unknown in the labor-automation category, and Figure is one of the few teams positioned to push it down faster than competitors. Lower cost per unit unlocks deployment. Deployment unlocks data. Data unlocks the model improvements that matter.

We wrote the Series B with a long-term view of what happens when general-purpose humanoids reach production scale. If the thesis plays out, Figure isn't a robotics company. It's the category's infrastructure.

Figure AI is building the first general-purpose humanoid robot for everyday life. A machine that can walk into a factory, a warehouse, a home, and perform physical tasks that no specialized robot could do affordably before.

Figure's thesis is that the labor economy is structurally constrained by human availability in physically repetitive work, and that the general-purpose humanoid form is the right abstraction for that constraint. Specialized robots exist, but each new task requires new hardware, new engineering, new integration. A humanoid that can be re-trained on new tasks amortizes the hardware investment across everything a human can do.

The company has moved from prototype to industrial pilot to enterprise partnership faster than skeptics predicted. The BMW deployment in South Carolina is the first commercial humanoid operating on a real automotive production line, not a lab floor. That distinction, production versus demonstration, is the entire difference between a venture-backable robotics company and one that will remain a research project.

Figure is also building its own foundation model for robotics control, rather than licensing someone else's. That vertical integration is expensive in the near term and compounding in the long term. It's the same pattern we've seen in other category-defining hardware-plus-software companies, and it's one of the reasons we committed at Series A.