Legless Manipulation Outcompetes Humanoids

The core advantage is that most near term physical AI jobs are hand tasks, not walking tasks. In food prep, packing, bench assembly, and hazardous handling, the worker usually stands or sits at one station, so a platform that skips legs avoids extra joints, balance control, batteries, and safety complexity while still doing the part of the job that creates value, which is seeing, grasping, moving, and placing objects.

• Anvil is built around this assumption. Its kits focus on arms, cameras, controllers, compute, teleoperation software, and training pipelines, because many customers need to collect manipulation data and prove a workflow before worrying about full body mobility. That cuts months of integration work and keeps the hardware bill closer to the task itself.
  1 sacra
• Humanoids win when the environment cannot be reworked and the robot must move through aisles, stairs, and human sized spaces. But that benefit matters less at fixed cells and benches, where specialized systems can be simpler and cheaper. Gartner now expects very few humanoid vendors to reach production scale by 2028, while narrower polyfunctional robots win earlier in warehouses.
  2 sacra 4 gartner
• The market data also leans toward modular, task specific systems first. IFR says professional service robot sales are led by transportation and logistics, then inspection and maintenance, which reflects real budgets going to robots that solve one concrete workflow well, rather than one machine trying to replicate the whole human body.
  3 ifr

The next few years should reward companies that turn manipulation into packaged workcells and vertical solutions. As models improve, more deployments will add mobility only when the workflow truly needs it. That keeps the center of gravity of commercial robotics on simpler bodies with better hands, better sensing, and tighter integration into a specific station or process.