NVIDIA Pavilion Marks Humanoid Robot Shift to Production at Automate 2026

What is happening at Automate 2026?

Automate 2026, North America's largest robotics trade show, opens Monday at McCormick Place in Chicago and runs through June 25. The 50th edition of the show draws more than 50,000 attendees and over 1,000 exhibitors across 450,000 square feet, making it the biggest in the event's history. For the first time, a dedicated Humanoid Robot Pavilion anchors the show floor, sponsored by NVIDIA, with more than 20 humanoid robot organizations from around the world demonstrating their platforms there.

The pavilion is paired with a separate paid Humanoid Robot Forum running Tuesday and Wednesday afternoons, June 23 and 24, featuring leaders from Boston Dynamics, NEURA Robotics, NVIDIA, and Toyota Research Institute. Sessions cover real-world commercialization, safety standards, and the engineering gaps that still separate a compelling demo from a reliable production deployment. The show's central theme is physical AI, a training approach in which robots learn by observing human demonstrations rather than following hand-coded movement scripts.

The question at Automate 2026 is no longer whether humanoid robots are ready, but how fast industry can absorb them now that production units are shipping.

Why does NVIDIA's pavilion sponsorship matter for the industry?

NVIDIA's decision to sponsor the humanoid pavilion is a strategic signal, not just a marketing spend. The company's Isaac GR00T training stack and Jetson Thor inference chip have become the dominant computational backbone for robot AI development. By anchoring the show's most prominent new section, NVIDIA is telling the industry that humanoids represent a meaningful share of its addressable inference market going forward.

The architecture driving most current deployments is the vision-language-action model, a single end-to-end learned system that takes a camera feed and a natural-language instruction and outputs continuous motor commands, with no separate perception pipeline or planning module. Figure AI's Helix, NVIDIA's Isaac GR00T N1, and Google DeepMind's Gemini Robotics are all implementations of this approach already in live industrial use. NVIDIA's pipeline combines Isaac Sim for physics rendering, Isaac Lab for GPU-parallel policy training, Cosmos world foundation models for synthetic training data, and the Newton physics engine for contact dynamics. The on-device inference module running inside the robot is the Jetson Thor chip, separate from the Vera Rubin data-center platform used for upstream training.

The central unresolved problem the show will spend four days debating is the sim-to-real gap: the disconnect between how an AI policy performs in simulation and how it performs on an actual factory floor, where friction, material variation, and contact dynamics behave differently than any simulator perfectly replicates. Whether NVIDIA's pipeline narrows that gap enough for the reliability levels production manufacturing requires remains the defining open question in industrial robotics.

Which commercial deployments are already running?

Three deployments crossed from pilot to commercial production in the months before the show opened, and each gives the industry a concrete data point to work from.

• Figure AI grew its BotQ facility in California from one Figure 03 unit per day to one per hour in under 120 days, a 24x throughput increase, with more than 350 units delivered and end-of-line first-pass yields above 80 percent. The ramp also functions as a data-generation engine: each deployed robot feeds real-world operational data back into the training pipeline.
• Boston Dynamics began commercial shipments of its electric Atlas humanoid, which features 56 degrees of freedom, a 110-pound lift capacity, 360-degree torso rotation, and autonomous battery swapping. The entire 2026 production run is committed to Hyundai's Robotics Metaplant Application Center and Google DeepMind, with additional customers planned for 2027.
• Agility Robotics signed a commercial Robots-as-a-Service contract with Toyota Motor Manufacturing Canada for seven Digit humanoids at the Woodstock, Ontario plant producing the RAV4 and RAV4 Hybrid, following a year-long pilot. The RaaS model converts what would otherwise be a capital expenditure into an operating expense, lowering the financial barrier for manufacturers uncertain about long-term utilization.

A constraint runs across all three deployments. Humanoid robots in 2026 remain well-suited to structured, repetitive industrial tasks but cannot yet reliably generalize to complex, variable, or safety-critical environments. That boundary, and how fast it moves, is what the industry is gathering in Chicago to figure out.

What does this mean for manufacturers considering humanoid adoption?

Standard Bots CEO Evan Beard, speaking in a Wednesday keynote, will argue that roughly 99 percent of real-world manufacturing tasks have resisted automation because the programming burden for each variation is prohibitive under traditional approaches. Physical AI, which moves the model from explicit programming to demonstration-based learning, is the proposed answer. A human performs a task via teleoperation, physical guidance, or a head-mounted camera, and the robot's AI model learns the underlying policy from that observation data.

For manufacturers, the practical implication is that the cost and time required to deploy a robot on a new task could fall significantly if demonstration-based training proves reliable at scale. The RaaS pricing model that Agility and Toyota Canada used adds another lever: manufacturers can trial humanoids without committing to a large capital purchase. Both factors lower the entry point, but neither eliminates the reliability question. Until humanoids can handle the variability of real production environments consistently, adoption will stay concentrated in the structured, repetitive tasks where the technology already works.

The Automate 2026 show floor and the Association for Advancing Automation conference sessions this week will provide the clearest public picture yet of where that reliability threshold actually sits.