The FF Nova robot is small enough to carry in one hand and, according to its manufacturer, priced at $2,490. The 14 October 2026 product launch by Faraday Future and its robotics partner puts a walking humanoid within the budget range of serious hobbyists and teaching laboratories, though not yet on ordinary shop shelves, as detailed in Faraday Future's product launch. Its first phase is invitation-only and intended for developers and educational institutions.
That distinction is the difference between a striking technology demonstration and a promise that everyone can buy a useful robotic servant. Nova is a platform for teaching physical AI to move, sense and respond in the real world. It is not advertised as a proven replacement for a cleaner, carer or warehouse worker.
The FF Nova robot has a small body and an ambitious brief
Nova stands about half a metre tall and weighs about 2.8 kilograms, according to the product specifications. The manufacturer lists a 20-joint body, sensors for depth and distance, a camera and a microphone array intended to help it detect where a sound comes from. It can operate on legs or use a detachable wheeled base, which may be useful for work that needs stability more than a walking demonstration.
The product description describes support for voice AI, on-device computing and a software development kit. Those features should not be confused with evidence that the robot can carry out every task imagined in a promotional picture. Developers will have to create, test and refine specific behaviours. The robot’s real-world performance will depend on its software, surroundings and physical limits.
LiveAIWire recently covered a much larger-scale question in robots assembling furniture: what happens when a robot is given an instruction to build a physical object? A platform such as Nova is designed to let more people explore smaller versions of the same challenge.
The opening price does not tell the whole story
The launch material lists the $2,490 device and an optional $390 developer skills package. Access during the first phase is restricted to a limited group, rather than open sales to anyone who wants a novelty for their living room. The manufacturer says its initial allocation consists of 100 units and that subsequent phases will broaden access.
There is also a smaller Nova Mini, presented separately as an educational product. Mixing prices or capabilities between those two models would make the offer sound more generous than it is. Nor is a low headline price a guarantee that replacement parts, training, services or long-term support will be inexpensive.
A prospective buyer should establish exactly which hardware is included, whether documentation and updates will remain available, what return or warranty rights apply and what access is required for cloud-based features. For now, the product’s realistic audience is someone who wants to experiment with robotics, not someone trying to avoid doing the washing up.
Why the wheels may be more interesting than the walking
Humanoid robots attract attention because they resemble people, but walking uses energy and demands continuous balance. Wheels can be simpler on suitable floors. Giving a small robot both modes acknowledges that the most photogenic way to move is not always the best way to get a job done.
The manufacturer says its robot supports a training process in which developers learn movements in a virtual simulation, export the results and test them on a real machine. That is often called simulation-to-reality transfer. It reduces the need to attempt every experimental movement directly on vulnerable hardware, although a behaviour that works on screen can still fail on carpet, uneven ground or in poor light.
The difficulty of translating vague human requests into physical action was explored in LiveAIWire’s robots following vague instructions. A robot may appear to understand a sentence yet still need precise definitions of what should happen next.
A classroom robot could matter more than a domestic one
A relatively affordable walking machine gives students something tangible to programme. They can see the consequences of a weak decision, a missed obstacle or an unstable step without needing an industrial laboratory. That is potentially valuable precisely because real objects do not behave as neatly as software simulations.
Faraday Future’s launch also includes teaching material and a developer community, but these are company-provided plans, not independently reviewed education outcomes. Schools would still need to consider supervision, safety, budgets, accessibility and the amount of specialist support required to use the platform well.
The appeal overlaps with the challenges LiveAIWire has described in robot training in virtual homes, where machines are being taught to practise tasks inside virtual environments. The important educational lesson is not that every student must own a humanoid. It is that artificial intelligence eventually has to meet the laws of physics.
Why comparisons with full-size robots need care
A compact educational robot may be easy to photograph beside a person, but it cannot be compared with full-sized industrial machines on price alone. The larger machine might carry heavy loads, work long shifts or operate around specialised safety barriers. Nova has a different purpose: inviting students and developers to try embodied software on a physical platform at a smaller scale.
The manufacturer’s detailed specifications are useful for understanding what comes in the device, but specifications do not measure reliability after months of use. Buyers and schools should distinguish a claimed sensor capability from an independently demonstrated task. The exciting achievement is that more developers may be able to experiment with real movement, not that a small machine is ready for all the jobs performed by a human.
The long road from clever machine to useful assistant
A robot that turns towards a voice or walks across a desk-side floor is memorable. Repeatedly doing a useful chore in a busy home is a harder standard. Real households contain pets, furniture, dust, stairs, delicate objects and people who do not behave like carefully arranged demonstration scenes.
The small Nova suggests that entry-level hardware may make physical AI easier to study. Whether that leads to dependable mass-market assistants remains a separate commercial and technical question. For anyone tempted by the price, the wisest starting point is to treat it as a development kit with a body, not a new member of the domestic workforce.
About the Author
Stuart Kerr is Technology Correspondent at LiveAIWire, covering artificial intelligence, cybersecurity and the social impact of emerging technology. LiveAIWire is an independent, human-led technology publication using AI-assisted research, editorial production and original AI-assisted editorial illustrations under his direction.