Morph Launches Shape Shifting Soft Robots for Physical AI

An octopus has roughly 500 million neurons, and about two-thirds of them sit in its arms rather than its brain. A London startup called morph just raised money on the idea that this is the blueprint everyone building robots has been ignoring. Instead of bolting smarter software onto rigid metal limbs, morph wants the body itself to do part of the thinking.

What Actually Happened

On June 2, 2026, morph came out of stealth and launched what it calls the world's first shape-shifting soft robotics cells platform, a system for designing and manufacturing programmable soft robotic components for physical AI. Rather than producing a single finished robot, morph sells a modular platform of deformable cells that teams can configure, simulate, and iterate into products. The pitch is that sensing and adaptive control are embedded directly into reconfigurable, deformable materials, so the hardware bends, grips, and adapts in ways a jointed mechanical arm physically cannot.

The company was founded by Dr. Jean Nehme, a former reconstructive surgeon who previously built and sold the surgical AI company Digital Surgery, which Medtronic acquired in 2020. That background matters, because it pairs a clinician's understanding of soft tissue and safety with a track record of shipping regulated medical AI. morph's platform fuses reinforcement learning with high-fidelity, physics-based simulation, letting engineers program the behavior of a soft cell in software, test it against real-world conditions virtually, and compress the path from concept to physical product.

The launch is backed by a roster heavy on operators rather than only venture funds: deep-tech investor 8VC, the investment vehicle of musician and entrepreneur Pharrell Williams, Harvey Spevak of the Equinox Group, Qubit Ventures, Valia Ventures, Copper, and Blue Lion. morph says its first applications will target athletic performance, injury prevention, and mobility support, with a stated roadmap that later expands into healthcare, automotive, and industrial safety. The company has not disclosed the size of the round, a detail that itself says something about where it sits on the risk curve.

Why This Matters More Than People Think

The robotics money of the past two years has flowed almost entirely into rigid, human-shaped machines. Figure reached a reported $39 billion valuation, Skild AI vaulted to roughly $14 billion, and Tesla, Boston Dynamics, and Unitree have all poured resources into bipedal humanoids. morph is a bet that this entire wave optimized for the wrong variable. Humanoids spend enormous compute and capital making rigid bodies behave gently. A soft body starts gentle and adaptive by default, which inverts the problem.

This is the concept roboticists call morphological computation: the idea that a cleverly designed body can perform tasks that would otherwise require heavy sensing and control. A soft gripper conforms to an irregular object without calculating a single grasp point. If morph can productize that principle, it reduces the very thing that makes humanoids so expensive, the need for massive models and sensor stacks to manage contact with the messy physical world. The cost and data burden of physical AI could shift from the brain to the body.

Zoom out to the labor implications and the stakes grow. The humanoid pitch is explicitly about replacing human physical work in warehouses and factories, which is why it draws both capital and anxiety. morph aims its first markets in the opposite direction: a runner who avoids a hamstring tear, a patient who regains mobility, a worker whose exosuit prevents a back injury. That framing sidesteps the political backlash gathering around job-displacing automation and targets buyers who want to enhance people rather than retire them. It is a quieter thesis than humanoid labor replacement, and possibly a more durable one, because it sells into health and performance budgets instead of fighting over headcount.

There is also a market-timing logic. By starting in athletic performance, injury prevention, and mobility, morph aims at consumer and clinical wearables, categories with shorter sales cycles and clearer willingness to pay than industrial automation. A soft, body-worn device that adapts to a knee or a spine is a far easier first product than a general-purpose factory robot. Nehme's surgical past suggests the deeper prize is healthcare, where soft, tissue-safe robotics for rehabilitation or assistance could command medical pricing, but the early revenue is meant to come from sport and movement.

The Competitive Landscape

morph is not entering an empty field, but it is entering an oddly under-commercialized one. Soft robotics has thrived in academia for two decades, from Harvard's Wyss Institute to countless university labs, yet almost none of it has become a scaled business. The closest commercial cousins are pneumatic soft grippers from companies like Soft Robotics Inc., which found a niche in food handling, and a scatter of exosuit and wearable-assist startups. None has built a horizontal platform for designing soft robotic behavior, which is the lane morph is trying to own.

There is also a defensibility question that favors a platform. Single-product soft robotics companies have historically been bought for one niche use case and absorbed, their technology frozen at a single application. By selling the design and simulation layer rather than one device, morph is attempting the move that turned industrial tooling companies into durable franchises: own the workflow every soft-robot builder needs, and capture value across applications it never has to build itself. Whether enough independent teams want to build soft robots to sustain a platform is the open question, but the structural logic mirrors how Autodesk and Cadence came to sit underneath entire industries.

The louder competition is philosophical, against the physical-AI establishment. Nvidia is pushing Cosmos world models and GR00T for humanoids, Skild and Generalist are building general robot policies, and Physical Intelligence is chasing a universal control model. Every one of those bets assumes a capable rigid body controlled by an ever-larger neural network. morph argues the body should absorb complexity the network currently has to learn. If it is right, some of the compute arms race in robotics is solving a problem better hardware would make smaller.

The historical parallel is aviation's long detour. Early flight pioneers tried to imitate birds with flapping wings, and progress only came when engineers abandoned strict biomimicry for fixed wings and propellers. Robotics has spent decades on the opposite mistake, insisting on rigid, anthropomorphic machines while nature solved manipulation with soft, distributed, boneless systems like the octopus and the elephant's trunk. morph is wagering that the field is finally ready to copy biology's actual solution rather than its silhouette, and that the timing, with cheap simulation and reinforcement learning now mature, is what makes it buildable today.

Hidden Insight: The Body Is the Model Nobody Is Scaling

The non-obvious story here is not about robots at all. It is about where intelligence lives. The dominant AI narrative says capability comes from scaling the model: more parameters, more compute, more data. morph implicitly challenges that by proposing that a large share of useful physical intelligence can be baked into matter, not learned by a network. A soft cell that passively conforms to a surface has, in a sense, precomputed a solution that a rigid robot would need a large model and live sensing to approximate.

This reframes the cost structure of embodied AI. The humanoid thesis is brutally capital-intensive because it loads everything onto the brain and the sensor suite, then needs oceans of teleoperation and simulation data to train contact-rich skills. If morph can offload part of that to material design, it attacks the single hardest bottleneck in physical AI, which is data efficiency for manipulation. Fewer demonstrations, less compute per task, and safer failure modes are the prize. That is a different and potentially cheaper road to useful robots than the one Silicon Valley is currently funding.

The founder profile sharpens the bet. A surgeon who exited a regulated medical-AI company is unusually well-suited to thread soft robotics through healthcare's approval gauntlet, where tissue safety and adaptive compliance are features rather than nice-to-haves. The same softness that makes these systems hard to control precisely is exactly what makes them safe to wrap around a human body. morph is positioning a weakness of the technology as its wedge, which is often how genuinely new categories find their first defensible market.

There is a subtler advantage in data. Rigid-robot companies are locked in a costly hunt for manipulation data, paying for teleoperation farms and synthetic video because contact-rich skills are so hard to learn. A soft system that succeeds through compliance needs fewer perfect demonstrations, because the material forgives error that a stiff gripper would punish. If morph cells reduce the demonstrations required to reach a working behavior, the company is not only selling hardware, it is selling a shortcut around the most expensive input in all of physical AI. That kind of structural edge compounds, because every task learned cheaply widens the gap against rivals still paying full price for data.

The bear case, however, is that soft robotics has been five years away for twenty years. Skeptics point out that deformable systems are notoriously hard to manufacture at consistent quality, difficult to model precisely, and limited in the force and repeatability that industrial buyers demand. The risk is that morph builds a beautiful platform that remains a lab curiosity, impressive in demos but unable to match the precision of rigid actuators or the cost curve of mass-produced wearables. An undisclosed round and a roadmap that starts in the relatively soft market of athletic performance can be read two ways: disciplined focus, or an admission that the harder markets are not yet reachable.

Step back and morph is a small wager against a very large consensus. Tens of billions of dollars now assume the path to useful robots runs through human-shaped machines and ever-bigger control models. If even part of that intelligence can be moved into engineered matter, the capital efficiency of the whole field changes, and the companies that learn to design with softness rather than fight it gain an edge no amount of GPU spending replicates. That is a bold claim from a startup that has not disclosed its round, yet it is exactly the kind of orthogonal bet that occasionally resets a field while the incumbents are busy scaling the obvious answer.

What to Watch Next

Over the next 30 days, watch whether morph discloses the round size and names its first design partners. A platform business lives or dies on whether other teams actually build on it, so early customer logos in sport, footwear, or medical devices would be the strongest validation. Silence on funding and partners would suggest the launch is more narrative than traction, while a marquee athletic or clinical partner would signal real demand for programmable soft components.

Over the next 90 days, the marker to track is a shippable reference product, not a concept video. Soft robotics is famous for mesmerizing demonstrations that never reach manufacturing, so the credible signal is a device with stated specifications: cycle life, force output, and a repeatable production process. Watch too for hires from manufacturing and materials science, because that is where this category historically breaks. A team that adds polymer and process engineers is preparing to ship, while one that only adds AI researchers is still doing science.

By the 180-day mark, the decisive questions are regulatory and strategic. Any move toward a healthcare pathway, a clinical pilot, or an FDA conversation would confirm that the medical endgame is real and not just investor framing. Equally telling will be the response of the rigid-robot incumbents. If Nvidia, Figure, or a humanoid leader partners with or tries to acquire morph, it validates morphological computation as a missing piece of physical AI. If they ignore it, the market is betting that soft robotics stays a fascinating footnote to the humanoid era.

Everyone is racing to give robots a bigger brain. morph is betting the winners will be the ones that gave them a better body.

---