Walk the floor of any warehouse automation conference and the crowd forms around the same thing: a walking humanoid robot. Autonomous mobile robots (AMRs), the wheeled carts and forklifts already shuttling pallets in the background, draw far less attention, even though they are the machines actually moving goods in large numbers today, according to Robotics & Automation News. That gap between spectacle and deployment is what operators planning 2026 budgets need to resolve.
The practical case for AMRs right now
Robotics & Automation News frames the humanoid-versus-AMR choice for warehouse automation as an open question heading into 2026, and the facts on the ground favor AMRs for most operations today. They are already deployed in large numbers, handling the transport layer warehouses depend on. Buyers can check current specs and pricing against our humanoid spec sheet and price tracker to see how far humanoid hardware still has to go on cost and maturity.
The AMR side of the market is moving too. Aformic, which builds self-driving forklifts and carts for factories, raised a $6.95M seed round and reports more than $34M in existing orders, according to The Next Web. The company is now developing new robots aimed at AI data centres, a sign that AMR vendors are expanding into adjacent facility types rather than waiting for humanoids to catch up on the warehouse floor.
The Robot Report's Robotics Startup Radar 2026 points to a wave of young companies building hardware, software and business models for commercial robotics. For operators, that means the supplier base for logistics robotics keeps getting more competitive, worth weighing against the total cost of ownership on any humanoid program.
What humanoids are still missing
The reason AMRs win the transport job while humanoids remain mostly a demo is manipulation. Walking is the easy part to show off. Picking, placing and handling varied items reliably is the hard part that determines whether a humanoid can do useful work in a warehouse. That is the gap LiberAI is targeting with Liber-0, a new world action model.
According to The Robot Report, LiberAI unveiled Liber-0 as a world action model pretrained on more than 2.2 million hours of human experience, built to support zero-shot robot manipulation: handling tasks without task-specific retraining. Robotics & Automation News describes LiberAI's framing of Liber-0 as helping robots learn to act from human experience, positioning it as a foundation system for general-purpose robots.
That distinction matters for buyers. A model trained once on a large corpus of human activity and applied across many manipulation tasks is a different proposition than a robot programmed for one repetitive pick-and-place job. If Liber-0 and systems like it deliver on zero-shot manipulation in real deployments, that is the development that would start shifting the economics toward humanoids. A walking demo on a conference stage is not.
Infrastructure is part of the buying decision
Manipulation software is only one piece of what it takes to run fleets of robots, humanoid or AMR, at scale. The Robot Report also discusses wireless charging strategies, including opportunity charging and battery life, as infrastructure needed to keep growing robot fleets running. Operators evaluating either category should factor charging logistics into total cost of ownership, not just the sticker price of the hardware.
What to actually watch in 2026
For warehouse and logistics operators deciding where to put 2026 automation budget, the signal to watch is not stage demos. It is whether foundation models like Liber-0 move from pretraining claims to documented, repeatable manipulation performance in real warehouse conditions, and whether that performance holds up outside a single showcase task. Until that evidence accumulates, AMRs remain the lower-risk, already-proven buy for warehouse movement, while humanoid programs are still best treated as a research preview rather than a deployment-ready product. Our best humanoid robots ranking and ongoing robotics coverage will track how that evidence develops.