AWE Exhibitors Enthusiastic About "Raising Lobsters": Intelligent Appliances Still Immature, Gree Executive Says Industry Inflection Point Has Not Yet Arrived

Source: Times Weekly Report Author: Zhu Chengcheng

In recent days, the Shanghai New International Expo Center has once again been filled with exhibitors and visitors, but compared to previous years, the atmosphere is noticeably different this year.

At the China Appliance and Consumer Electronics Expo (AWE) in 2026, almost all leading manufacturers have incorporated “AI” into the core logic of their products. From TVs and refrigerators to air conditioners and cleaning robots, manufacturers are showcasing not just connectivity capabilities, but also moving beyond early “smart” features like voice wake-up and app control. Instead, they are integrating computing power, models, and sensor systems into specific devices, enabling machines to perceive environments, understand needs, and make decisions.

This shift is reshaping industry narratives. The repeated mention of “Internet of Everything” in recent years has become less prominent in this year’s exhibition halls. Instead, more specific terms like “proactive service,” “home intelligent agent,” and “embodied intelligence” are emerging.

“AI’s impact on future humanity is immeasurable. AI is no longer an option but a necessity,” said Huang Hongsheng, founder of Skyworth Group (00751.HK), at the exhibition.

In fact, the core of this change is not whether household appliances are connected to large models, but whether they truly have the ability to understand user contexts and collaborate effectively.

Zhao Siquan, senior analyst at IDC China, pointed out that home devices can break through the single-command response mode to achieve cross-scenario, multi-device intelligent coordination and demand prediction, providing more personalized automation services that align with daily life rhythms. This elevates household appliances from passive tools to smart home hubs capable of understanding, learning, and adapting.

Every evolution of a machine’s “brain” (computing method) forces its “body” (hardware device) to take on a completely different form. Tian Feng, director of the Fast and Slow Thinking Institute, told Times Weekly that the “computing paradigm” can be understood as the “soul” (computing power and interaction logic), while the “physical terminal” is the “body” (hardware device). Once the soul is upgraded, the old body cannot contain it, and a complete reshaping is necessary.

If the past decade’s keywords in the home appliance industry were “connectivity” and “intelligence,” a bigger question is emerging at this year’s AWE: after AI truly enters home scenarios, what will household appliances become?

Proactive service becomes a trend

“Unlike previous smart home systems that required users to give explicit commands, with AI capabilities, the interaction methods are changing,” said a Skyworth staff member to Times Weekly. In the past, users had to say “turn on the air conditioner” for the device to respond; now, when users say something natural like “I’m so hot,” the system can combine environmental perception and user habits to automatically adjust the temperature.

For example, the system can analyze past usage records to determine user-preferred temperatures and adjust accordingly, rather than setting a uniform temperature for everyone. If a user habitually sets the temperature to 28°C, the system will gradually learn and establish that as the default.

In this ecosystem, the TV is regarded as an important hub for smart homes. The Skyworth staff explained that the TV can connect to and control various household appliances via WiFi, acting as the “butler” of the entire home. Once data is shared across different terminals, the system can perform more complex scene linkages.

For instance, in the kitchen scene, a refrigerator with an embedded camera can recognize stored ingredients, analyze them in conjunction with user eating habits, automatically recommend recipes, and provide cooking tutorials on the TV or other devices. The system can also recommend content based on regional and user preferences.

“Integrating models into intelligent carriers is essentially the process of large models moving toward the terminal,” said Wei Jiaxing, founder of Yunfu Intelligence, to Times Weekly. The substantial improvement in products is reflected in interaction experience, such as transforming search services on TVs from mechanical responses to intelligent interactions.

However, this shift also raises higher demands on home appliance companies. Wei Jiaxing believes the main challenge lies in understanding the intelligent agent and the upstream and downstream ecosystems of products. For example, whether they can carry extensive user historical context.

Some appliance companies are cautiously advancing “proactive intelligence.”

“Chinese language and culture are complex, and the same need can be expressed in multiple ways. If the system cannot understand accurately, proactive services might become disturbances,” said Li Jie, planning manager of Midea Group’s (000333.SZ) household air conditioning division. He noted that implementing AI proactive intelligence requires a comprehensive smart home ecosystem. Currently, Midea is still overcoming key difficulties, especially in deep understanding of user language and needs.

Li Jie said that Midea’s current AI proactive intelligence mainly relies on learning from repetitive user behaviors to provide clear, essential proactive services, avoiding unnecessary disturbances. In the future, they will continue to improve in understanding user language, scene perception, and multi-dimensional applications to make proactive intelligence more aligned with user needs.

Smart home appliances with AI agents are still in early stages

Recently, a product called OpenClaw, an AI agent, has become popular among developers worldwide and is one of the frequently mentioned technological concepts at this year’s AWE. Exhibitors set up dedicated “shrimp farming” experience zones and released related products or demos to showcase the potential integration of household appliances and AI agents.

However, from actual demonstrations, appliance companies are still exploring applications of this technology.

For example, Ecovacs (603486.SH) showcased the “Eight Realm Robot,” claiming it is equipped with OpenClaw. But from the live demo, its system mainly functions as a device feature, without demonstrating a complete “lobster-style” intelligent agent structure.

Technically, OpenClaw is an open-source, “local-first” AI agent emphasizing local execution. Unlike traditional chatbots, it functions more like an automation proxy framework. It can interact with users via messaging apps, invoke tools on local devices, and perform real-world operations.

A complete OpenClaw system typically requires several key capabilities: local-first execution, multi-channel interaction interfaces, multi-agent collaboration mechanisms, and an expandable Skills system to continuously add new tasks.

However, in this year’s AWE, these capabilities have not yet formed a complete system. Products like “Eight Realm Robot” still mainly focus on single-device feature upgrades, and there is a clear gap from true OpenClaw household appliances.

“OpenClaw has been successful on computers because Windows and macOS provide highly standardized APIs and DOM trees for it to call,” said Tian Feng. The Chinese home appliance market is highly fragmented, with brands like Midea, Haier, Gree, Huawei, and Xiaomi each having their own closed communication protocols. In such a fragmented environment without a unified language, an agent is like a high-level commander blindfolded and restrained, unable to exert its cross-domain coordination capabilities.

Tian Feng believes that Huawei (HarmonyOS Smart Home) and Xiaomi (Full Ecosystem for Vehicles and Homes) have foundational OS-level coordination capabilities. But traditional white and black appliance giants tend to operate independently, lacking the software engineering DNA for cross-brand, cross-protocol API integration. Without underlying connectivity, large models’ “proactive services” become ineffective.

" Lobster AI agents are not yet mature products. The more important goal now is to enable household appliances to gradually understand user intentions," said Wei Jiaxing. The key challenge remains in privacy and security.

Tian Feng further pointed out that OpenClaw is currently popular on PCs because the digital world has higher fault tolerance—mistakes like sending an email to the wrong person can be retracted, and web searches can be redone. Household appliances involve the physical world, and if an open-source agent embedded in an oven or gas stove experiences “AI hallucination” and autonomously decides to start high-temperature baking, the consequences could be disastrous, like a house fire. Safety and compliance must be 100% certain in home appliances—this is the red line for autonomous exploration of any intelligent agent.

“Don’t just create hype”

As AI technology continues to penetrate the home appliance industry, how to truly embed algorithm capabilities into manufacturing and product details remains a practical challenge for companies.

“AI performs well in academic and research problems, but implementing it in manufacturing involves countless details,” said Zhu Lei, CMO of Gree Electric Appliances (000651.SZ), in a group interview. He noted that they are moving in this direction, but a major turning point has not yet arrived, mainly due to limitations in production technology.

In contrast, Haier Smart Home (600690.SH) is actively promoting the concept of “unmanned household chores.” After launching the “AI Eye” technology last year to transition appliances from “helping with chores” to “unmanned chores,” this year Haier upgraded the technology to version 2.0 and expanded application scenarios.

According to Haier, the new “AI Eye” has expanded recognition capabilities. For example, the “Refrigerator AI Eye” now recognizes nearly all common ingredients, up from 230 types before; the recognition area has expanded from just the refrigeration zone to the entire “refrigeration + freezing” area. The system’s recognition accuracy has also improved; for instance, the “Washing Machine AI Eye” combines visual and voice recognition to better understand user needs.

However, Zhu Lei believes that truly applying AI capabilities to specific home appliance scenarios is not easy. “Some brands claim their AI wisdom eyes can identify stains on laundry, but even with the naked eye, it’s hard to distinguish whether a piece of clothing is wool or cashmere. Achieving this with AI is very difficult.” He emphasized that AI is a direction everyone is pursuing to make products smarter, but it should not be about creating hype.

He sees the ultimate goal of a smart home not as more interactions, but as less presence.

“The core of future smart homes is health,” Zhu Lei told Times Weekly. From a user experience perspective, the ideal state should be “imperceptible.” Current products still require wake-up commands, voice, or other interactions, but this is just a transitional phase. “The final goal will definitely be seamless and invisible.”