A new generation of consumer electronics is beginning to take shape — not defined by screens, connectivity or form factors, but by whether the device can function at all without artificial intelligence.

That was the argument put forward by Yang Yuxin, chief marketing officer of Chinese chipmaker Black Sesame Technologies, during a panel discussion at the CES 2026 technology show.

"My definition of an AI-native device is one that loses its reason for existence without AI," Yang said. "If you remove AI and the product no longer makes sense, that is when you know the form factor is truly AI-native."

After two years in which generative AI dominated CES largely through demonstrations, concept products and speculative storytelling, this year ’ s event marked a shift toward practical deployment — a transition reinforced by Nvidia CEO Jensen Huang ’ s keynote speech promoting "physical AI," and by growing attention to robotics, wearable devices and on-device intelligence.

Yang Yuxin, Chief Marketing Officer of Black Sesame Technologies ( right )

Executives and engineers attending the show said the focus is moving away from general-purpose AI claims and toward narrowly defined use cases, hardware-software integration and real-world economic viability.

"Instead of asking what AI can do, the industry is now asking where AI actually belongs," said Liu Xiangming, co-founder and co-CEO of Chinese technology media group TMTPost, who moderated a panel featuring Yang and Dai Zhaoen, chief executive of smart-glasses maker MICROLUMIN.

Dai Zhaoen, CEO of MICROLUMIN

The distinction between traditional smart hardware and AI-native devices has become central to industry debate.

Earlier waves of smart devices were built primarily around connectivity — linking appliances, wearables or sensors to the internet so users could monitor or control them remotely. AI-native devices, by contrast, are expected to operate autonomously, interpret their environment and act without continuous human input.

"Smart hardware was about connection. AI-native hardware is about cognition," Yang said. "The device itself must perceive, decide and act."

Dai said the difference may be less visible to consumers but is fundamental for product designers.

"Users don ’ t care whether something is called AI-native or not," Dai said. "They care whether it solves a real problem better than before. But there are some products that simply would not exist without AI — those are the ones that matter."

He cited AI-powered glasses and embodied-intelligence robots as examples of such categories.

"For products like AI glasses and embodied intelligence, there ’ s no reason to buy them without AI," Dai said.

Despite the growing interest, executives cautioned that most AI-native devices remain in early stages and that many prototypes shown at previous CES events failed to reach the market.

"The problem is not technology, it is relevance," Dai said. "You cannot start from ‘ everything plus AI ’ . You have to start from one scenario where users are willing to pay."

He said sectors such as education, healthcare and tourism show stronger early demand because users perceive direct value.

"Rather than chasing a universal super-app, you break through in one vertical, solve one painful problem very well, and build from there," Dai said.

Yang agreed that scenario design — not hardware form — should lead development.

"Embodied intelligence should start from intelligence first, and then define the shape of the device based on the task," Yang said. "It doesn ’ t have to be humanoid. It could be wheeled, legged, or something else entirely."

Robotics and physical automation dominated the CES exhibition halls, as companies showed warehouse robots, service robots, industrial manipulators and consumer-oriented companions.

Many of these systems rely on AI models that interpret vision, sound and spatial data in real time — requiring powerful chips, efficient energy use and tight integration between hardware and software.

Yang said Black Sesame ’ s move beyond automotive chips into robotics and consumer electronics reflects this convergence.

"The automotive supply chain and the robotics supply chain overlap more than people realize," he said. "Safety, perception and decision-making are common foundations."

He added that as robots become more capable, safety standards will become more important.

"The more powerful the robot, the greater the potential harm if something goes wrong," Yang said. "Automotive-grade standards can help make robots safer."

One of the central technical debates is where AI computation should occur — in the cloud, on edge servers or directly on devices.

Dai said smart glasses are constrained by weight, heat and battery life, making full on-device AI unrealistic in the near term.

"The current model is cloud-edge-device collaboration," he said. "But latency and cost are problems."

He predicted that smartphones would increasingly run lightweight models locally to handle everyday tasks, reducing reliance on cloud services.

"Consumers want faster, cheaper and more convenient," Dai said. "Anything that goes against that must prove its value."

Yang said cloud computing remains essential for training and large-scale inference, while on-device processing is critical for privacy and real-time response.

"Cars cannot wait for the cloud to brake," Yang said. "Private data and real-time decisions belong on the device."

Smart glasses emerged as a focal point for AI-native experimentation.

Dai said recent advances in waveguide optics and weight reduction have made glasses viable for everyday wear.

"Glasses sit at the center of the face," he said. "They see what you see, hear what you hear, and can contextualize the world around you."

Unlike phones or smart home devices, glasses occupy a constant sensory position — giving them unique potential for personalized AI.

But Dai said current products remain immature. "Most apps are just phone apps ported to glasses, and that doesn ’ t work," he said. "We need to rethink interaction from scratch."

Yang said appearance and comfort remain critical. "People see how they look; you feel how they fit," he said. "That ’ s why Meta and Ray-Ban succeeded — they started from fashion, not from tech."

The emergence of AI-native hardware is reshaping the industry ’ s division of labor. Chip companies are no longer just component suppliers, Yang said, but platform providers.

"We provide chips, operating systems, toolchains and model support," he said. "Without software, AI chips are useless."

Dai said device makers remain closest to users. "We understand why something sells or fails," he said. "But we depend on foundational technology to move forward."

Both said collaboration — not vertical integration — would define the future. Chinese companies played a prominent role at CES, particularly in AI chips, robotics and hardware manufacturing.

Yang said China ’ s strengths lie in its supply chain and engineering talent. "The world ’ s most complete hardware ecosystem is in China," he said.

Dai said global expansion requires cultural sensitivity. "Products are not algorithms," he said. "They are tied to habits, language, aesthetics and trust."

He warned against rapid overseas expansion without local understanding. "Respect local culture first," Dai said.

Panelists said CES 2026 reflects a maturing industry. "There are fewer slogans, more engineering, more cost discussions," Liu said. The industry is moving from spectacle to infrastructure, from promise to practice.

As Yang put it: "The real AI era is not when everything claims to have AI, but when some things cannot exist without it."