Machina Labs Wants to Make Factories Programmable

A damaged C-17 recently returned to the air with a RoboFormed structural repair part. Machina Labs made it directly from a digital design, using robots instead of dedicated tooling.“Qualifying a part from a new manufacturing process for flight can take years, time these critical programs don’t have time to spare,” Mehr said in the company’s August 18 announcement.A conventional factory usually begins with a decision: What will this factory make?That choice determines nearly everything that follows, from the machinery and tooling to the layout of the building. Once the investment is made, the factory becomes very good at producing the same thing, over and over.Mehr wants to loosen that connection between the factory and the product.“Our goal is to create a platform that can make car parts in the morning,” he told me. “But then in the afternoon, without any major physical change in the platform, it can start doing aircraft. And at night, autonomously, it can do missile components.”Machina calls its system RoboCraftsman. It is a general-purpose manufacturing platform designed to perform different jobs through software rather than through a physical rebuild of the factory.The company became known for RoboForming, its process for turning a flat sheet of metal into a finished shape with two industrial robots working from opposite sides. Watching the process brings to mind a potter molding clay, except the material is metal and the hands belong to robots.“People don’t realize that metal is very, very plastic,” Mehr said. “It’s very impressionable.”The robots supply the force and precision needed to shape the material incrementally. By applying compression from both sides, they can manage how the metal stretches and hardens as it takes form.The underlying craft is not new. Metalworkers have shaped material this way for generations. What has changed is who, or what, is holding the tools.“It’s almost the way craftsmen did it a while back,” Mehr said. “But now robots are doing it.”A different kind of scaleTraditional manufacturing excels when enormous numbers of identical products are required. Mehr points to Toyota, which can produce a vehicle roughly every 60 seconds in the United States.“What traditional manufacturing cannot do is make a different car every 60 seconds,” he said.That is the opening Machina is pursuing.In its work with Toyota, the company can customize parts that have already moved through the automaker’s conventional production system. A Tacoma hood or tailgate, for example, can be given a design chosen by its owner.Toyota keeps the efficiency of mass production while adding a level of individual expression that conventional tooling makes difficult or expensive.Manufacturing has traditionally relied on economies of scale: Invest in specialized equipment, then make enough copies of the same product to recover the cost. Mehr believes programmable factories could make variety economical, too.The result could be a world in which cars, buildings and other physical products no longer look so much alike.“That’s actually a limitation of manufacturing,” he said. “We can only scale and make cheap stuff today if we’re making the same thing over and over again.”Machina describes itself as a physical AI company because its larger goal goes beyond automating an existing production line. Mehr compares the idea to the way a person can express an intent to a large language model and receive text or an image in return.Here, the output is a physical object.Lessons from SpaceXMehr’s interest in flexible manufacturing grew out of his time at SpaceX.He remembers arriving at its Hawthorne factory and feeling like “a kid in a candy shop.” Rockets, which he had associated with governments and national space programs, were being built by a company in Southern California.He also saw how the physical constraints of manufacturing slowed even the fastest hardware companies. Every new product could require a different facility, a new set of tools and another large capital investment.“My focus has been on bringing this agility back,” he said, “bringing software-like agility to manufacturing.”Defense has become an early adopter because it presents precisely that problem. The sector needs many different systems, often in relatively small quantities, and must be able to respond quickly when demand changes.Lockheed Martin’s relationship with Machina spans investment and production. Lockheed Martin Ventures invested i