Zinova Bets on Tool-Use Intelligence, Not Dexterous Hands, for Construction Robots
Construction robotics has long been a field where companies either achieve greatness or fail. Forbes contributor John Koetsier recently described construction as robotics' holy grail—or its white whale—but more often a graveyard.
RIC Robotics built the 20-foot-tall Zyrex robot and 3D-printed a 5,000-square-foot Walmart expansion in a week. Its spinout Zinova now offers a different answer.
Zinova cofounder Xu Ziyou said in a launch video that everything humans build—from homes to factories to spacecraft—comes from skilled tool use. The company aims to give physical AI the same ability, so robot form can change while tool-use intelligence remains.
In 2026, the humanoid industry sees dexterous hands—five fingers, tendon-driven, tactile skin—as the bottleneck and a major funding target. Zinova abandons that path.
Instead, it uses a simple rigid gripper to hold tools, plus a force-feedback sensor called TEISI that captures force, torque, vibration, resistance, and contact between tool, robot, and material.
The sensor’s logic: a carpenter knows a drill is about to bind from wrist load, sound, and experience half a second before vision. Cameras cannot see this; force feedback can. CEO Ryan Cox said the company does not care what the robot looks like, only that it brings the right tool to the right place at the right time.
In its demo, Zinova used LimX Dynamics' TRON 2 dual-arm robot to complete a scaled tilt-up concrete panel workflow—common in warehouses and data centers, where concrete walls are cast on the ground and lifted after curing.
The robot nailed two 2x4s together with a nail gun, drilled holes, installed and tied rebar, then smoothed concrete poured into a formwork. This was not one action but a continuous sequence involving different tools, heights, and hand poses.
Single-task robot TyBOT can also tie rebar at 1,200 intersections per hour, but it must run on rails and only ties rebar. TRON 2 can move, work at different heights, and switch tools.
Founded in 2025, Zinova held its first public demo. Target scenarios are housing, retail, data centers, and eventually the first buildings on the Moon and Mars.
Construction sites are highly non-standardized; each project has different environments, material specs, and processes, so uniform automation is hard. That is why many robotics companies die on sites—what works in the lab falls apart in the field.
U.S. construction data: 20% of construction workers are over 55, and over 40% are expected to retire in five years. Large projects run 20% over schedule and 80% over budget on average. Construction workers are 7.5% of the labor force but 20% of workplace deaths. This means a widening labor gap and growing demand for safer alternatives.
Zinova's choice is to bypass the traditional difficulty: rather than redesign the site, it makes robots adapt to existing tools. Tools are designed for human hands—pistol grips and triggers on drills, nail guns, saws, sprayers, and even breakers. If that is the target toolset, a robot does not need a hand that can do everything a human hand can; it only needs to grip handles and pull triggers.
If this path works, the significance is transferability. The same tool intelligence can migrate across robot forms, avoiding dedicated machines for each trade. LimX Dynamics cofounder Pang Bo said the company provides a hardware platform, development tools, and model training platform so innovators can quickly build applications on mature bodies.
Zinova has not yet been on a real construction site. The demo was scaled: a 6x6-foot wall panel, while real tilt-up panels are 20 to 40 feet and weigh tens of tons. From scaled demo to real projects, much remains. But no one had publicly demonstrated one robot using multiple tools to complete multiple processes. Zinova did that; next is real-site performance.