Unitree Unveils a 28 MPH Humanoid Two Days Before Its $905 Million Market Debut.

On August 17, Reuters reported that Unitree Robotics unveiled a high-speed humanoid it calls “Superman,” claiming the machine can run at 12.66 meters per second—about 28 miles per hour—and jump two meters from a standstill. The timing is as important as the spectacle: Unitree is scheduled to begin trading on Shanghai’s STAR Market on August 19 after raising roughly $905 million.

For business leaders, the event is a signal that humanoid robotics is moving from laboratory demonstrations into a more demanding phase shaped by manufacturing volume, public-market scrutiny, cost, safety, and measurable work.

Speed Is the Headline; Control Is the Capability

A fast robot makes an irresistible video. It does not automatically make a useful employee.

The reported speed and jump are company claims carried by Reuters, not yet an independently standardized benchmark.

What matters underneath them is the control stack required to keep a full-size biped balanced while its feet repeatedly leave the ground.

That means sensing the environment, estimating body position, coordinating motors, correcting instability, and making those adjustments fast enough to prevent a fall.

Unitree’s existing H1 documentation helps explain the engineering foundation. The company describes 3D LiDAR and depth-camera sensing, a roughly 47-kilogram body, high-torque joints, and a replaceable battery. Its public specifications list the earlier H1 at 3.3 meters per second, with potential mobility above 5 meters per second.

A claimed leap to 12.66 meters per second would therefore represent a major change in performance, but the company has not yet published enough standardized test detail to compare endurance, payload, surface conditions, or repeatability at that speed.

For a warehouse, factory, inspection team, or emergency operation, those missing details matter more than the record itself. A machine that can sprint once is different from one that can work safely for hours, recover from contact, carry useful weight, and repeat the same task without damaging products or people.

Eighteen Thousand Robots Change the Question

Reuters reported that Unitree says it had produced and delivered approximately 18,000 bipedal humanoid robots across its models by July. That number is consequential because robotics improves through more than brilliant prototypes. It also improves through production engineering: sourcing motors and sensors, reducing assembly variation, collecting field failures, simplifying maintenance, and learning which parts break outside controlled demonstrations.

Manufacturing volume can bend the cost curve. It can also expose problems that a small pilot never reveals.

A joint that performs perfectly in fifty units may behave differently across thousands.

A calibration process that takes an engineer two hours becomes an operational bottleneck at scale.

Software updates, replacement parts, technician training, warranties, and remote diagnostics become part of the product.

This is where the robotics race begins to resemble the evolution of computers, smartphones, and electric vehicles. The visible machine attracts attention, but the durable advantage often comes from the system around it: suppliers, tooling, data, service, software distribution, and the ability to improve the installed base without rebuilding everything.

The IPO Turns Engineering Claims Into Business Commitments

Unitree’s market debut adds a second kind of pressure.

Public investors will not evaluate the company only by its most dramatic motion. They will eventually ask how much revenue each product line produces, what margins survive competition, how quickly research spending converts into dependable products, and whether buyers return after the first demonstration.

A Shanghai Stock Exchange summary of Unitree’s filing reported 2025 revenue of 1.699 billion yuan, adjusted net profit of 590 million yuan, and a 60.13% gross margin for its core businesses.

The filing also described investment plans spanning robot technology, large AI models, new products, and manufacturing capacity.

Those figures make Unitree unusual in a field where many humanoid programs remain expensive research projects.

They do not remove the central uncertainty.

Investor enthusiasm can fund better engineering and larger factories, but it cannot create a commercially necessary task.

Humanoid robots still need to outperform less glamorous alternatives: conveyor systems, fixed robotic arms, wheeled platforms, specialized inspection devices, or redesigned human workflows.

What Businesses Should Watch Before Buying

Most companies should not begin with the question, “Can we buy a humanoid?” They should begin with a bounded job.

The strongest early candidates are repetitive, physically demanding, dangerous, or difficult to staff—and take place in environments already designed for people. A useful pilot might involve moving containers between two fixed stations, inspecting a hazardous area, tending a machine during an overnight shift, or collecting structured training data under supervision.

Four measurements matter immediately: successful task completions, human interventions, total cycle time, and the full cost per completed task.

Safety incidents, near misses, downtime, charging, maintenance, and integration labor belong in the calculation.

A robot that completes 90% of a job but requires constant rescue may be less economical than a simpler machine with narrower capability.

The OrionPilot point of view is to treat this as an operating question rather than a spectacle: which task, data loop, safety boundary, and business metric would justify a pilot?

The Real Race Is From Spectacle to Useful Work

Unitree itself warns on its H1 product page that the global humanoid industry remains in an early stage and urges buyers to understand current limitations. That caution is more informative than the usual science-fiction framing.

The next competitive milestones will not be another dance, backflip, or sprint alone. They will be longer autonomous operation, reliable manipulation, safer interaction around people, easier programming, lower maintenance, and proof that a deployment produces value after the camera leaves.

“Superman” makes the robotics race visible. Unitree’s production volume and public listing make it commercially serious.

The companies most affected today are not only robot manufacturers. Logistics operators, factories, insurers, software vendors, training providers, safety specialists, and service businesses all have a reason to watch.

The opportunity is real, but so is the discipline required: separate demonstrated motion from dependable work, and separate investor excitement from an operating result.