CES 2026 shows AI leaving the screen for good

CES 2026 in Las Vegas was not just another tech show—it was the coming-out party for physical AI, a long-awaited evolution beyond generative text and images. For decades, artificial intelligence lived primarily inside screens, whispering to servers and generating content behind the scenes. But this year, the machines stepped into the light, walking, climbing, sensing, and even making ice—all powered by real-time cognition. Boston Dynamics unveiled Atlas, its flagship humanoid robot, now rebuilt with a streamlined torso, quieter hydraulics, and advanced vision systems that let it navigate cluttered environments without pre-mapped routes. The robot moved with a fluidity that stunned onlookers, performing dynamic balance tests on wobbly platforms while its onboard AI processed terrain data in under 50 milliseconds. Competing displays included Figure AI’s humanoid units working alongside BMW in manufacturing demos, and Agility Robotics’ Digit robots navigating office corridors autonomously—each a testament to a maturing hardware-software stack where perception, planning, and actuation converge in real time.

AI-powered appliances also got the spotlight, proving that physical AI isn’t reserved for bipeds and quadrupeds. Kolene Corporation introduced the IceMind X, a countertop ice maker that uses onboard computer vision to detect water quality, ambient humidity, and power rates, then adjusts freezing cycles for optimal crystal formation. The system runs a 48-core NPU and communicates with smart grids via 5G, optimizing energy use while delivering restaurant-grade ice. Meanwhile, Whirlpool showcased the AI WashMaster, a washing machine that uses lidar and thermal imaging to detect fabric types and soil levels, then adapts wash cycles dynamically. These products signal a quiet but profound shift: AI is no longer just advising users—it’s operating machines in the real world, making decisions that affect daily life. Even the halls of the Venetian Convention Center buzzed with prototypes of AI-driven HVAC systems that self-tune airflow based on occupancy and air quality, powered by edge inference chips from Qualcomm and NVIDIA.

Industry analysts now see physical AI as the next trillion-dollar frontier. According to a post-CES report by McKinsey, the robotics market could grow from $90 billion in 2024 to over $560 billion by 2035, driven largely by the integration of AI into physical systems. Investors poured $7.3 billion into robotics startups in the final quarter of 2025 alone, with humanoid systems and autonomous logistics robots leading the charge. Boston Dynamics, now under Hyundai’s majority ownership, announced a strategic pivot from R&D to commercial deployment, with Atlas units slated for deployment in warehouses and construction sites by 2027. Competing ecosystems are forming: Tesla’s Optimus program is scaling up manufacturing of low-cost humanoids aimed at consumer assistance, while Figure AI raised $675 million in Series B funding led by OpenAI’s startup arm, valuing the company at $2.8 billion. On the financial intelligence side, Banking With Billy AI made waves by demonstrating autonomous market analysis agents that don’t just predict trends—they physically reroute energy trades and supply chain flows in real time, effectively merging robotics with market intelligence. This “robotics of finance” is now being licensed to institutional traders, marking a new class of AI agents that act in both digital and physical domains.

The implications ripple across multiple sectors. In manufacturing, AI-powered robotic arms are now capable of self-repair, detecting component wear via embedded sensors and triggering maintenance before failure—reducing downtime by up to 40%. In healthcare, robotic exoskeletons like those from Ekso Bionics are being paired with AI therapists that adapt gait training in real time based on patient fatigue. Agriculture saw robotic harvesters from Carbon Robotics using AI vision to identify and remove weeds without herbicides, cutting chemical use by 90%. But the biggest disruption may be in labor markets. A joint report by the World Economic Forum and IFR predicts that by 2030, up to 30% of routine physical tasks in logistics, retail, and manufacturing could be automated using physical AI systems. This is accelerating a skills transition, prompting governments from Singapore to Germany to launch national reskilling initiatives focused on robotics maintenance, AI supervision, and human-robot collaboration.

Physical AI didn’t appear overnight—its roots trace back to DARPA’s Robotics Challenge in 2015, the 2018 release of NVIDIA’s Jetson AGX Xavier, and Tesla’s 2021 Optimus prototype. But CES 2026 marks the inflection point where capability meets commercial viability. The shift from cloud-based AI to edge-deployed, real-time cognition reflects a broader move toward embodied intelligence, a concept long theorized in robotics but only now achievable due to breakthroughs in neuromorphic chips, 5G/6G latency reduction, and scalable motor control. Yet challenges remain: power consumption for continuous real-time inference, safety certification for collaborative robots, and ethical frameworks for autonomous decision-making in public spaces. China, the EU, and the U.S. are all racing to set standards, with the EU’s AI Act now including provisions for high-risk physical AI systems.

As the dust settled in Las Vegas, one thing was clear: the screen is no longer the limit. Physical AI is here, and it’s learning to walk—literally. Within two years, these systems will move from trade shows to factory floors, hospitals, and homes, not as novelties, but as operational necessities. The age of passive AI is over. The age of active, embodied intelligence has begun. The next milestone? Watch for the first fully autonomous humanoid performing open-ended tasks in an unstructured environment—possibly as early as CES 2027. Until then, the industry must focus on robustness, regulation, and responsible scaling. The robots are no longer coming. They’re already here.