Investing in Mind Robotics

Over the last decade, software ate the world. And over the last two years, AI began to reprogram it.

Now we’re in an exciting new era. Intelligence is stepping off the screen and into the physical world.

The most advanced AI systems can already reason, plan, and generate. The next frontier is action: intelligence embedded in machines that move atoms, manipulate objects, and operate in real-world environments. As generalist robotics systems come online, they won’t just optimize existing workflows: they will reshape manufacturing and the broader industrial economy on a fundamental level.

But as anyone with experience in robotics will tell you: the physical world is unforgiving.

In software, failure is often reversible. In robotics, it rarely is. Hardware doesn’t forgive hallucinations. Micron tolerances don’t care about confidence intervals. A single integration flaw can render an entire system inoperable. Embodied AI is not just a modeling challenge, but an industrial one.

Today’s industrial robots are extraordinarily good at what they were designed to do: repeatable, dimensionally-stable tasks in tightly-controlled environments. But a large share of value-additive work in a factory setting doesn’t look like that. That work requires dexterity, adaptation to variability, and physical reasoning in dynamic settings. Wiring harnesses don’t present themselves the same way twice; materials flex; parts shift. Edge cases are the norm, not the exception.

There is a structural gap in industrial automation. And closing it requires more than incremental improvement.

Building robots capable of dexterous, variable, and reasoning-intensive tasks demands a new foundation: integrated models trained on real-world data, hardware designed for fine manipulation and robustness, and deployment infrastructure that supports continuous learning and iteration. It requires designing intelligence, machines, and industrialization together.

This is why we’re thrilled to partner with RJ Scaringe and the team at Mind Robotics.

The origin of Mind Robotics is deeply practical. While scaling Rivian, RJ searched for a robotics company that could meaningfully transform factory operations: one capable of combining intelligence, precision, and industrial reliability. But he couldn’t find one. So he decided to build it himself.

There are very few founders who have built and scaled a vertically integrated hardware company in the modern era. At Rivian, RJ didn’t just ship vehicles. He built a full stack: vehicle architecture, electronics, battery systems, embedded software, manufacturing processes, and supply chains, all integrated into a cohesive system operating at industrial scale. That experience sits directly at the intersection that robotics now demands.

Mind Robotics is building an industrial robotics platform from the ground up, spanning models, hardware, and deployment systems, with a clear focus on real tasks in real plants at real scale. From day one, the company is grounded in a live manufacturing environment through its partnership with Rivian.

That partnership creates a powerful data flywheel and provides an at-scale proving ground that few robotics companies have access to. Continuous exposure to high-throughput production tasks generates rich, diverse data for training and improving models. Improvements can be deployed directly into live operations, where performance is measured against concrete economic outcomes.

Progress in robotics compounds. Deployment generates data. Data improves models and control systems. And better performance unlocks broader deployment, which in turn produces more data. We’ve seen it many times in the history of hardware innovation: the teams that tightly integrate hardware, software, and manufacturing into a unified system build advantages that are difficult to replicate from the outside.

As AI capabilities improve, the bottleneck will shift from cognition to execution. The breakthrough is no longer just teaching a model to understand the world in 3D. It’s building machines that can reliably act in it. Every major computing era introduces a new foundational layer. The cloud era required distributed systems infrastructure. The AI-native era required model and data platforms. The embodied AI era will require companies capable of delivering intelligent machines as durable industrial systems.

We believe robotics is just starting to enter its compounding phase. The teams that integrate intelligence, hardware, and deployment into a coherent industrial platform will define the next generation of enduring companies.

Mind Robotics is setting out to build that platform.

We’re excited to partner with RJ and the entire team as they bring advanced robotics into the factory floor: not as a demo, but as infrastructure for the real economy.

I would like to thank Jamie Sullivan, David George, Raghu Raghuram, Erin Price-Wright, Jacob Zietek, and Guido Appenzeller who made instrumental contributions to the deal.