Most robots are clever software bolted onto dumb metal. morph wants to flip that idea on its head. The company has launched a physically intelligent soft robotics platform built around what it calls soft robotic cells — a term it coined to describe a new class of reconfigurable, deformable hardware that doesn’t just follow instructions but adapts on its own.
The pitch is deceptively simple: instead of treating intelligence as something that lives only in a controller, morph embeds sensing and adaptive control directly into the material. The result is hardware capable of changing its own morphology and stiffness in real time, reshaping itself to suit the task at hand rather than relying on rigid joints and pre-programmed motion.
This is the increasingly fashionable idea of physical AI — pushing decision-making out of the cloud and the CPU and into the body of the machine itself. A soft robotic cell that can stiffen to grip a heavy component, then go limp to slip past an obstacle, behaves less like a traditional industrial arm and more like a living organism responding to its environment.
Under the hood, morph pairs reinforcement learning with high-fidelity, physics-based simulation. That combination matters: soft materials are notoriously hard to model because they flex, fold and deform in ways rigid robotics never has to worry about. By training behaviours in a simulator that faithfully captures real-world physics, morph can teach its cells how to deform and adapt before they’re ever deployed on a factory floor.
The platform was announced on June 2, 2026, and morph is positioning itself less as a gadget maker and more as an enabler. The company is currently working with several industrial partners and plans to operate primarily as a software, design and fabrication partner for businesses looking to adopt soft robotics — handling the tricky end-to-end work of designing and manufacturing the cells rather than selling them off the shelf.
That also means there’s no consumer product to buy here, and no public pricing. This is infrastructure for the people building the next generation of machines, not a thing you’ll find in a box.
Why does it matter? Conventional robots excel in structured, predictable environments and struggle the moment reality gets messy — variable objects, delicate handling, unstructured spaces. Soft robotics has long promised a way around those limits, but it has been held back by the difficulty of controlling materials that refuse to behave like rigid links. By baking adaptive control into the material itself and learning behaviours in simulation, morph is making a credible bet that the body of a robot can be just as intelligent as its brain.
If it works at industrial scale, the line between hardware and AI gets a lot blurrier — and that’s exactly the point.
