Feather exits stealth: its $29,990 dual-arm Physical AI robot is shipping now
NEWS + ANALYSIS | 24 September 2026
Feather Robotics is coming out of stealth with an unusually concrete proposition: instead of promising a humanoid "later," it is offering a wheeled dual-arm Physical AI platform that can already be configured and ordered. The promotional Standard price is $29,990. The company emphasises open software, ROS 2, teleoperation and support for external models rather than locking customers into a single robot "brain."
- Feather says "Shipping today", with a public configurator and purchase path.
- The Standard configuration is listed at a promotional $29,990; Performance is $42,540 and bare/custom chassis options start at $15,995.
- The robot uses a holonomic wheeled base, two 7-DoF arms and 23 total degrees of freedom.
- The company lists 600 mm of vertical travel, about 1 m arm reach and a maximum base speed of 1.3 m/s.
- Feather supports Python 3.9+, ROS 2, teleoperation, inverse kinematics, collision detection and several NVIDIA Jetson compute options.
- The company claims up to 10 hours of operation from two hot-swappable 48 V / 50 Ah packs; independent endurance testing is not yet available.
| Item | Confirmed | Status / caveat |
|---|---|---|
| Standard | $29,990 promo | Public configurator; regular price shown as $32,990 |
| Performance | $42,540 | Higher configuration |
| Bare/custom chassis | from $15,995 | For integrators |
| Mass | 105 kg | Manufacturer specification |
| One-arm payload | 1.8 kg rated / 6.8 kg peak | Manufacturer figure; configuration-dependent |
It is not another humanoid - and that is the point
Feather does not try to reproduce the human body at any cost. It replaces legs with a mobile base and concentrates complexity on bimanual manipulation and a repeatable hardware platform. In a laboratory, kitchen, light-manufacturing line or warehouse, that compromise can be practical: fewer balance problems, simpler power management and more of the energy budget available for useful work.
It also tests the idea that Physical AI may develop a hardware layer comparable to a reference platform: body, sensors, actuators and SDK, while the integrator selects the control model. Feather explicitly promotes compatibility with outside software rather than presenting a proprietary foundation model as the only intelligence available to the robot.
A $30,000 price resets the benchmark
$29,990 is still meaningful capital for a small company, but in the market for advanced dual-arm mobile robots it is low enough to attract laboratories, integrators and application teams that cannot justify building bespoke hardware. Just as importantly, the price is public and the configurator goes beyond a generic "contact sales" form.
The base price should not be confused with the cost of a finished autonomous worker. Grippers, workstation safety, task software, training data, support and maintenance can materially increase total deployment cost. The public figure is therefore a platform price, not a promise of a complete $30,000 employee replacement.
What we know about real deployments
TechCrunch reports that Feather has passed $1 million in revenue and previously sold small batches of robots that the company says are operating in restaurants and laboratories. Customers and the exact number of deployed units have not been disclosed. There are also no public uptime, MTBF or full-shift productivity datasets.
For now, the product, configuration, price and sales channel are verifiable, while industrial fleet maturity is not. "Deployment-grade" should be read as Feather's product positioning rather than an independently validated quality certification.
Europe: promising hardware, unclear local support
For European integrators, open software and the ability to connect their own models are the most compelling aspects. Local inference and stack control can help projects where data privacy matters. Feather, however, is currently a clearly US-centric company, and EU service conditions and spare-parts availability are not as transparent as the online configurator.
RoboMorrow will watch for three concrete triggers: named customer deployments, independent endurance testing and an announced European service channel. Those will determine whether Feather becomes a practical integrator platform outside the US or remains primarily a compelling tool for research teams and startups.
An "open robot" is not a finished application
Feather lowers the hardware entry barrier, but a buyer still has to supply or integrate perception, task policy, safety, teleoperation/recovery and process logic. That distinction is essential: ROS 2 and an SDK can shorten engineering work without turning the platform into an autonomous worker by themselves. The more useful benchmark will be time from unpacking to a task that runs reliably for many hours.
RoboMorrow is tracking the same issue on the software side. Isaac ROS 5.0 and Intrinsic Core illustrate the expanding open robotics stack, while Universal Robots Gen 7 exposes open APIs and ROS 2 in an industrial platform. Feather is interesting as the body for those layers; customer integrations, uptime and support cost will determine its real advantage.
What would justify an update
RoboMorrow will not create another story simply because a second publication covers the same event. A material update requires a real status change: commercial shipping or production, a new price, a signed customer, a major deployment, entry into a new market, independently verified performance data or a regulatory decision. Smaller clarifications should update this canonical owner instead.
Sources and methodology
RoboMorrow verification: 25 Sep 2026. Manufacturer specifications are treated as claims unless stated otherwise.