IoT and cybersecurity expert Phillipe Le Berre reflects on the visit to the seminal technology show in Las Vegas. The most striking presence at CES was NVIDIA. Its logo was everywhere - on screens, booths, and partner demonstrations - yet the company itself was not overtly promoting on the exhibition floor. Instead, NVIDIA’s influence felt structural rather than promotional. The infrastructure is now in place, and an entirely new superstructure is emerging.
That superstructure is what many are now calling Physical AI, a new generation of IoT where intelligence is physically embodied in machines that act in the real world. NVIDIA appears to have completed its transition from component supplier to ecosystem enabler, delivering proprietary hardware, software platforms, and an end-to-end stack capable of supporting this evolution.
This role was clearly demonstrated through NVIDIA’s deep vertical integration strategy, including its partnership with Mercedes-Benz to deliver a full stack for the software-defined vehicle (SDV). The Mercedes-Benz CLA, the brand’s first vehicle featuring proprietary infotainment, the MB.OS platform, introduces advanced driver-assistance features powered by NVIDIA’s full-stack Drive AV software, AI infrastructure and accelerated computing.
Just as importantly, NVIDIA and other players highlighted the hardware catalysts necessary for Physical AI to succeed: low-power AI processors and solid-state, long-life batteries designed specifically for robotic and humanoid systems. Donut Labs launched a production-ready all-solid-state battery with what promises to be a dramatically faster charging in around five minutes and significantly higher energy density up to 400 Wh/kg, vastly longer lifespan with tens of thousands of charge cycles, and improved safety thanks to the elimination of flammable liquid electrolytes. Together, these gains translate into longer range, lower total cost of ownership, faster turnaround times, and safer, more durable electric vehicles and robotic platforms.
Edge AI comes of age
Another clear signal from CES 2026 was that many of the historical limitations of edge AI are rapidly disappearing. Performance, power efficiency, and cost are converging in ways that make sophisticated, autonomous systems viable outside of the cloud.
Axcelera, for example, demonstrated impressive AI-accelerating M.2 chips capable of delivering significant inference performance at the edge. These advances are unlocking new use cases across robotics, industrial automation, and mobility without the latency or connectivity constraints that previously held edge deployments back.
Robotics becomes modular and accessible
Robotics was one of the most dynamic areas of the show. Chinese OEM Ubtech stood out with a full lineup of robotic arms, quadrupeds, and humanoids. What was particularly notable was Unitech’s business model. Rather than offering a closed, vertically controlled system, the company provides the mechanical hardware: the actuators, gyroscopes, and physical units, while leaving control of firmware, AI chips, system logic, device management, and security entirely in the hands of the customer.
End users increasingly want full governance over robotics systems, including software architecture and cybersecurity. By enabling this flexibility, Ubtech is positioning itself as a foundational hardware partner rather than a controlling platform provider.
Just as important is economics. The cost of robots is falling rapidly, and when combined with mature edge AI inference capabilities, this creates a step change in accessibility. Startups, scaleups, and large enterprises can now acquire advanced robotic platforms and tailor them to their own objectives.
However, this is not without risk. A compromised robot represents a fundamentally different threat than traditional malware, introducing real, physical danger. As Physical AI scales, cyber risk will increase accordingly, driving substantial investment into industrial and robotic cybersecurity. Kudelski Labs illustrated this risk with KLARQ, a secured robot-dog showing how off-the-shelf robotics can be made trustworthy through embedded device identity and lifecycle security rather than closed ecosystems.
From bionic wearables to commercial pragmatism
CES also featured more unconventional and revealing innovations. One OEM, ULS Robotics’ VIATRIX, demonstrated a wearable exoskeleton capable of enabling bionic performance, allowing users to run at speeds of up to 20 km/h. While dramatic, it underscored how rapidly human - machine augmentation is advancing.
At the other end of the spectrum was a far more mundane, yet arguably more telling innovation: a toilet-cleaning robot. Its significance lay not in novelty, but in commercial logic. The problem is clear, the value proposition is obvious, and the scaling potential is real. It was a reminder that successful innovation is increasingly about practical deployment, not spectacle.
Legacy consumer giants lose the spotlight
In contrast, traditional consumer electronics giants such as LG and Samsung appeared comparatively subdued. Their interpretation of AI often felt incremental, smart appliances responding to voice commands, or refrigerators opening and closing automatically. Against the backdrop of humanoids and Physical AI, these use cases felt disconnected from the transformative momentum visible elsewhere on the show floor.
Rethinking automotive, healthcare, and intelligence itself
Perhaps the most profound implications of CES 2026 lie in how Physical AI reshapes entire industries. In automotive, the concept of the software-defined vehicle now appears mature - but also potentially challenged. When humanoid robots can physically drive cars, the rationale for embedding ever more intelligence into the vehicle itself becomes less clear. It may be easier to upgrade a robot than to retrofit or redesign a car.
The same logic applies to healthcare. As robotic precision advances, the industry may need to rethink its focus on “smart tools.” Rather than intelligent scalpels designed for human surgeons, the future may belong to intelligent robots that wield standard instruments with superhuman accuracy. Several demonstrations hinted at the transformative potential of robotics in surgery and medical procedures.
The insurance question and systemic change
One of the most significant unresolved issues is insurance and liability. As robots become active participants in physical operations, responsibility and risk allocation become far more complex. Some insurers are already limiting coverage related to AI-driven systems due to systemic uncertainty.
How insurance models adapt may become a key constraint or enabler for corporate adoption of robotics and Physical AI. Premiums, liability frameworks, and regulatory responses will shape how quickly organizations integrate these technologies into their operations.
A systemic shift Is underway
CES 2026 was overall a signal of systemic change. Physical AI is forcing innovation across supply chains, value chains, and regulatory frameworks. From infrastructure providers like NVIDIA, to modular robotics OEMs, to cybersecurity firms and insurers, every part of the ecosystem will be affected.
As Philippe Le Berre observes, this is no longer about imagining the future. The pieces are now in place and the consequences will be felt across every industry that touches the physical world.
Contact Philippe at pleberre@consulare.ch