CES 2026, my favourite January event, offered plenty of spectacle—rollable screens, AI-powered mirrors, a parade of concept vehicles. I think this year the real signal was in the side halls and partner demos, where humanoid robots appeared in numbers that would have seemed implausible two years ago. Unitree, Figure, Agility, Apptronik, and a dozen Chinese entrants all showed working prototypes. The implicit message: general-purpose robotics is no longer a research curiosity. It’s a product category in formation.
Yet walk into any factory, warehouse, or retail environment today and you’ll find precisely zero humanoids doing useful work at scale. The gap between demo and deployment remains vast—and understanding why matters for any executive weighing robotics investments.
Focus on: the four bridges to commercial reality
A humanoid must cross four thresholds simultaneously before it becomes economically viable outside controlled pilots.
Safety without fences
Industrial robots have operated behind cages for decades. Humanoids, by design, work alongside people. Existing standards were written for fixed arms and collaborative cobots; humanoid-specific certification frameworks remain under development. In Europe, the Machinery Regulation (EU) 2023/1230 applies from January 2027, which means any manufacturer serious about European deployment must build compliance into the architecture now. Regulators are yet again defining the rules of the game, and those rules favour documented safety engineering over move-fast iteration.
Uptime economics
Most current prototypes operate two to four hours per charge. An eight-hour shift requires either hot-swap battery logistics or redundant units—neither of which improves the business case. Labour substitution only makes sense when the machine reliably covers the shift it’s meant to replace. Until uptime approaches human norms, buyers will prefer robotics-as-a-service models with performance-based fees and uptime SLAs. The risk of idle capital is simply too high.
Dexterity and cost
The human hand has roughly twenty-seven degrees of freedom. Replicating that mechanically, with the sensorimotor integration to use it, remains expensive. Actuation accounts for 40–60% of a humanoid’s bill of materials. Prototypes sit at $150,000–$500,000 per unit; broad B2B commercial viability likely requires $20,000–$50,000, for scaled B2C adoption the cost needs to drop below $5,000. That’s an order-of-magnitude reduction, achievable only through architectural simplification, modular integration, and supply-chain learning curves that haven’t yet materialised.
Where pilots are actually working
The deployments that exist cluster predictably: structured environments, limited task scope, mobility over fine manipulation. BMW is testing Figure humanoids at its Spartanburg plant. Amazon has trialled Agility’s Digit for repetitive tote handling. Mercedes-Benz and Apptronik announced a pilot agreement. These are real, but they’re also bounded—proofs of concept, not scaled operations.
I believe that near-term traction will concentrate in automotive and warehousing, where environments are controlled and tasks are well-defined. The leap to healthcare, hospitality, or domestic settings requires advances in perception, safety certification, and cost structure that remain several product generations away.
What this means for 2026
I expect the coming year to deliver two things: continued rapid improvement in demonstration capability, and a sharper bifurcation between companies building certifiable, deployable systems and those optimising for investor attention. The former will accumulate the data, the incident logs, and the assurance cases that unlock procurement gates. The latter will struggle to convert pilots into contracts.
For marketing leaders, the implication is clear: humanoids are becoming trust products. Brand exposure rises when machines operate near humans, and the reputational penalty for a safety incident will exceed any first-mover advantage. As a result, messaging should anchor in certifiability and operational controls, not novelty.
The humanoid moment is real. But crossing from CES floor to factory floor requires engineering discipline, regulatory patience, and unit economics that most entrants haven’t yet demonstrated. Those who solve all four problems together—not sequentially—will define the next decade of robotics. I remain optimistic we’ll see meaningful progress by the end of 2026.
Follow me
To keep up with the latest in generative AI and its relevance to your digital transformation programs, follow me on LinkedIn or subscribe to this newsletter.
Disclaimer: The views and opinions expressed in Chronicles of Change and on my social media accounts are my own and do not necessarily reflect the official policy or position of S&P Global.