CES 2026 defines the moment that AI stepped out of the screen to enable motion, decisions, and physical actions in the real world. The show was dominated by announcements from NVIDIA, AMD, and a massive wave of consumer robotic products; a sharp glimpse into the future of infrastructure that enables autonomous systems in the real world. While the previous years were about chatbots and generative AI, the shift toward Physical AI at CES 2026 represents the new “hard reality” of the industry.
Here are my top 10 picks from CES 2026, framed around the three “bedrock” foundations of the Physical AI era.
The Bedrock of Physical AI
1. NVIDIA’s “Reasoning” Ecosystem (Rubin, Alpamayo & Drive)
NVIDIA used CES 2026 to present a single narrative across training, inference, autonomy, and robotics: a unified stack designed for systems that perceive, reason, and act. Jensen Huang’s keynote centered on the Vera Rubin Platform, the successor to Blackwell, promising a 5x leap in inference performance.
NVIDIA introduced Alpamayo 1, a 10-billion parameter open-source VLA (Vision-Language-Action) model. Unlike previous AV stacks that relied on simple object detection, Alpamayo excels at chain-of-thought reasoning. It understands that a fire hose on the road signifies an “emergency scene” requiring caution, rather than just a generic “obstacle.” Coupled with Level 4 Robotaxi trials and Mercedes-Benz’s integration of this reasoning AI into the production CLA line, NVIDIA has effectively showcased its top-of-the-line AI products for the physical world.
2. AMD’s Ryzen AI & The Local Compute Pus
AMD’s announcement of the Ryzen AI next-gen NPUs (Neural Processing Units) focuses heavily on privacy and local inference. They expanded the Ryzen AI lineup (Ryzen AI 400 and Ryzen AI PRO 400) and emphasized a key point that the PC is being redefined around on-device AI.
AMD highlighted configurations delivering up to 60 TOPS of NPU compute for Copilot+ PCs, plus broader platform expansion across OEMs through 2026. The pitch centers on privacy and latency: with these chips, you can run powerful AI assistants and enterprise-grade analytics directly on a laptop without an internet connection. This eliminates the security risks of the cloud and allows for real-time data processing in the field.
3. Boston Dynamics Electric Atlas
The retirement of the hydraulic Atlas marks a strategic pivot for Boston Dynamics toward commercial scalability. The fully electric Atlas made its production-ready debut, showcasing a new swiveling joint design with 56 degrees of freedom. This allows for fluid, human-like motion and in some cases, omnidirectional movements that exceed human capabilities.
This electric architecture solves the noise and maintenance issues of previous generations, making the unit viable for indoor factory floors. Boston Dynamics also introduced Orbit fleet management software, designed to integrate Atlas into existing digital transformation workflows. Combined with AI-based reinforcement learning, Atlas is positioned as a versatile “super worker” for industrial settings, with initial deployments expected in Hyundai manufacturing facilities.
The Consumer AI Showcase
4. LG CLOiD: Autonomous Mobile Robots Enter the Home
LG’s CLOiD represents the transition from digital assistants to functional household robotics under their “Zero Labor Home” vision. The robot features a head with a display/camera array, a torso with two articulated arms, and a wheeled base.
Each arm possesses 7 degrees of freedom and a five-fingered hand with independent actuation for fine-motor tasks like folding laundry or loading a dishwasher.
Built on the “Affectionate Intelligence” platform, it uses Vision-Language-Action (VLA) models to translate verbal requests into physical actions.
Integrated with the ThinQ ON hub, CLOiD learns user patterns to proactively manage connected-home appliances.
5. Roborock Saros Rover
Roborock introduced the Saros Rover, featuring a world-first wheel-leg architecture designed to solve the “staircase barrier” in domestic robotics. The chassis can independently raise and lower each wheel-leg, allowing the unit to navigate and clean stairs, slopes, and complex room thresholds while keeping the main body level.
Combines LiDAR with RGB cameras to recognize 201 distinct objects and generates multi-floor 3D maps.
Features a record-breaking 35,000 Pa HyperForce® motor.
Supported by the RockDock®, which utilizes 100°C (212°F) hot-water washing for a fully hands-off cleaning cycle.
6. Automotive as a Data Platform: Sony-Honda Afeela & Mercedes-Benz
Automotive tech at CES 2026 highlighted the car as a high-compute “Experience Platform.” The Sony-Honda Afeela 1 uses a staggering 800 TOPS of on-board compute to run a panoramic interior screen powered by Unreal Engine 5, rendering real-time 3D visualizations of the car’s surroundings.
Simultaneously, Mercedes-Benz demonstrated its Drive Assist Pro integration. In collaboration with NVIDIA, Mercedes is utilizing reasoning-based AI to handle “novel” driving scenarios like construction zones, interpreting a traffic officer’s hand signals by reasoning through visual data rather than following pre-programmed rules.
7. Razer Project Motoko: Multimodal AI at the Edge
Razer took a familiar form factor and gave it a new job. Project Motoko is a Snapdragon-powered wearable AI headset featuring dual eye-level cameras for augmented computer vision. It uses multimodal AI to process visual data in real-time and provide auditory feedback. While designed for gaming coaching, the “heads-up” utility for professional tasks like hardware repair or assistance in household context, where users need information without a screen obstructing their view which makes it a compelling use case for “heads-up” productivity.
8. The Smart Ring Explosion
The smart ring segment has matured into a clinical-grade hardware category. Oura, RingConn, and the Bond Ring all showcased devices that miniaturize PPG (Photoplethysmography) and skin temperature sensors into jewelry. The trend is the shift toward continuous longitudinal data – tracking sleep, HRV, and SpO2 for 7+ days straight which is critical for the next generation of proactive health analytics and early anomaly detection.
9. Next-Gen Display Tech
Two display announcements captured the “screens as objects” direction CES has been building toward.
Samsung Micro RGB: Spanning 55 to 130 inches, this tech uses sub-100 μm LEDs that emit light independently. The Micro RGB AI Engine Pro manages frame-by-frame upscaling to maintain 100% of the BT.2020 color gamut.
LG Signature OLED T: A 77-inch transparent 4K OLED screen that “disappears” when idle. The Alpha 11 AI Processor allows users to toggle between a transparent “digital canvas” and an opaque, high-contrast cinema mode.
10. Amazon Alexa+
Amazon officially unveiled Alexa+, a fundamental LLM-based re-architecture powered by Amazon Nova and Anthropic’s Claude. It transforms Alexa from a “skill-based” assistant into a conversational partner capable of agentic task completion, such as navigating websites to fill out forms or booking service appointments via Thumbtack. Amazon confirmed that 97% of its 600M+ existing devices are compatible with the new stack via hybrid edge-cloud coordination.
Join the Conversation
Hope you enjoyed reading the post. Share your take in the comments, or continue the discussion with me on LinkedIn, X or on Threads. If this roundup helped you cut through the CES noise, feel free to pass it along to someone who may enjoy reading it.
Leave a Reply