Hardware companies are difficult in a way software companies are not. Factories, supply chains and physical delivery add a category of problem that cannot be patched after release, and RJ Scaringe has spent the better part of twenty years demonstrating that a business sitting across AI, software, robotics, manufacturing and transport can nonetheless survive. He will be talking about that at TechCrunch Disrupt in October, and his argument is that the intersection he occupies is about to get considerably busier.

The vehicle business is only part of what he now runs. Rivian has published a route to full Level 4 autonomy by 2028, and is building a charging network with the stated ambition of ranking among the largest in the United States.

Alongside that, Scaringe founded Mind Robotics, a humanoid robotics company that has raised nine hundred million dollars this year. He serves as its executive chair and is currently acting as chief executive too. Rivian holds a substantial stake and is also its first customer.

The two are meant to feed each other. The plan is to use Rivian's plant at Normal, Illinois as a testbed for people and machines working the same assembly line, which Scaringe frames as a response to a shortage of human labour he expects to arrive.

The nearer-term question is the R2. The roughly fifty-eight thousand dollar SUV has started reaching customers, and it carries the job the R1T and R1S could not do — taking Rivian out of a narrow, expensive corner of the market. Scaringe has described it as possibly the most important launch the company has made.

That matters more than it would have two years ago. Demand for the pricier end of the electric market has cooled, and pressure from Chinese manufacturers selling considerably cheaper vehicles has grown.

The through-line he is expected to draw is about sequence. Plenty of companies begin with artificial intelligence and then attempt to build something physical around it. Rivian has spent two decades doing the reverse — learning to make things at scale first, and bringing the software layer in afterwards. Whether that ordering turns out to be the advantage he thinks it is remains the open question.