Open hardware is reaching humanoids
A full hardware design, BOM, and simulation environment are now public, with a kit targeted at $15,000.
Born too late to explore the earth, born too early to explore the stars, born just in time for humanoid robot lego sets.
Asimov is open-sourcing a humanoid platform you can build yourself from a full bill of materials and simulation stack.
Source: Asimov / Menlo research
Menlo Research, a Singapore-based applied R&D lab, today published the full mechanical design and simulation files for Asimov v1 and recently opened pre-orders for a $15,000 DIY kit. The full bill of materials is public. Parts can be sourced independently or bundled into the kit with shipments targeted for end of summer 2026. This pushes the mechanical layer toward a repeatable baseline.
Source: Asimov / Menlo research
Menlo's stated motivation was frustration with closed platforms. The team was using a Unitree G1 for research when a knee joint failed; the replacement took two months to arrive, halting work entirely. Building on a closed system means iteration speed is bounded by the vendor's supply chain. Asimov's design lets builders see every component, source off-the-shelf replacements, and 3D-print structural parts when something breaks, which changes the daily reality of running a humanoid research programme.
Unitree is one of the highest-volume humanoid robot manufacturers and Unitree vice president Jiang Liming said the company develops more than 90% of its hardware itself, including key robot components such as motors, reducers, controllers, LiDAR, perception, and motion-control algorithms. That vertical integration gives tight control over performance, cost, and supply, at the expense of openness.
The $15,000 price is close to bill-of-materials cost by design. Menlo's longer-term revenue model targets proprietary software components (robot skills, full-body teleoperation), cloud services (simulation, data collection), and OEM licensing rather than hardware margins. The $30,000 annualised total cost of ownership is framed as a design constraint rather than a forecast; the argument is that a humanoid needs to be economically comparable to a minimum-wage worker in developed markets before it becomes viable as infrastructure rather than a research object. If enough systems are built, the hardware layer becomes a shared baseline and differentiation shifts to what runs on top of it.
Because the design is openly licensed, any manufacturer can pick it up and produce units without a licensing agreement with Menlo. They frame Asimov as a reference design for an open manufacturing partner ecosystem, where multiple producers can build compatible systems in parallel, with parts sourced locally and supply chains distributed rather than centralised. Cost pressure moves into production and sourcing rather than the design itself, with competition at the manufacturing level rather than control of the hardware.
Asimov is not arriving alone. RoboParty open-sourced ROBOTO ORIGIN earlier this year, a 1.25-metre, 34-kilogram bipedal platform with hardware, software, and engineering documentation released for developers. ROBOTIS has now introduced AI Sapiens K0, a 1.3-metre, 34-kilogram research humanoid with 23 degrees of freedom, DYNAMIXEL-Q actuators, CAD, BOM, and simulation support. Berkeley Humanoid Lite takes the same accessibility problem further down-market, using 3D-printed gearboxes and off-the-shelf components to keep hardware cost below $5,000.
K-Scale Labs is the cautionary comparison. It offered the K-Bot Founder's Edition at $8,999 before shutting down after failing to raise the capital needed to scale manufacturing. Menlo is pricing closer to BOM and betting that a community-built hardware layer can support a proprietary software business above it.
Asimov’s model commoditises the mechanical layer and shifts the constraint upward. What it does immediately is give researchers and small teams a reproducible bipedal platform they can modify, repair, and train on without owning the full stack or waiting on outside teams.