CES2026: Robots Edging Toward Real-World Deployment
Industry leaders said adoption, not flashy demos, were defining the next phase of physical artificial intelligence with scale and trust.
LAS VEGAS, Jan. 5, 2026 — Physical artificial intelligence is moving from spectacle to substance, as companies learn what it takes to make autonomous machines useful, safe and trusted in working environments, robotics leaders said Monday.
The assessment came during a CES Research Summit preview to the CES trade show here in a panel moderated by Ani Kelkar, a partner at global consultancy McKinsey. Why are robotics advancing now after decades of promise that failed to translate into broad adoption?
Executives pointed first to scale. Robert Playter, chief executive of Boston Dynamics, said robotics was moving beyond one-off pilots as machines entered sustained use with customers.
He said the Massachusetts-based company was expanding manufacturing with Hyundai Motor Group, including bringing more component fabrication in-house to support higher production volumes. Humanoid robots, he said, were expected to begin factory testing this year, with deployment targeted for 2028.
But Playter emphasized that deployment did not mean instant transformation. Customers, he said, typically needed two to three years to adjust operations and become comfortable with robots working alongside people. Robotics adoption, he said, was as much a cultural shift as a technical one.
That point was reinforced by Mikell Taylor, who leads robotics strategy at General Motors, the Detroit based automaker. He said reliability was the dividing line between impressive demonstrations and systems that could survive daily use.
Robots had to perform consistently in unpredictable environments such as factories, hospitals and service settings, not just under controlled conditions.
She said companies often underestimated the adoption phase, which requires a deliberate effort to educate workers. Employees needed to understand how robots perceived their surroundings and made decisions, she said, warning that erratic movement or unclear behavior could quickly erode trust and slow deployment.
Robots face strict safety constraints
From the research side, Carolina Parada, senior director and head of robotics at Google DeepMind, the Alphabet-owned AI research lab, said physical AI would not become commercially viable all at once. Unlike conversational AI systems, she said, robots faced strict safety constraints and had to operate in physical spaces shared with people.
Parada said progress depended on moving away from brittle, task specific models toward systems that learned transferable capabilities across domains. While recent advances allowed robots to perform increasingly complex behaviors, she said major challenges remained in sensing and manipulation, with a long way to go in matching the dexterity of the human hand.
Those limits shaped how the technology was being built. Nakul Duggal, executive vice president at Qualcomm, the San Diego-based chip designer, said physical AI systems would not reach their full potential on day one. Instead, he said, robots should be viewed as evolving platforms that improved through real world use.
Duggal drew a distinction between robotics and autonomous driving. While self-driving vehicles addressed a defined transportation problem, robotics offered a broader set of opportunities across manufacturing, logistics, health care and service environments. Many of those uses, he said, involved existing facilities where robots had to integrate safely with established infrastructure.
Safety and trust connected the panelists’ remarks. Duggal said human presence had to be treated as a core design requirement.
Despite viral videos of robots performing acrobatics, Playter said the real challenge was not making machines move impressively–but making them reliable enough to do useful work day after day.
Member discussion