Summary

The key question at this CES was not whether a product had AI, but whether AI redefined the product—and whether the hardware fundamentals held up. PLAUD uses OpenAI API to turn a roughly $100-plus card-sized recorder into a summarization tool, making it a product that would not exist without AI; but 泓君 ultimately went back to a recorder from 10 years ago because PLAUD kept dying at crucial moments. The conclusion is simple: “It is a consumer product first, and only then a smart device.”
China’s consumer-hardware advantage has expanded beyond mature supply chains into product innovation, brand premiums and iteration speeds several times to 10x faster. The shift is clearest in robot vacuums: leading players such as Roborock and Dreame are experimenting with robotic arms, while their patent portfolios and cross-licensing arrangements have weakened the constraints imposed by iRobot’s older patents. 陈哲 believes that once innovation speed reaches an order-of-magnitude gap, overseas manufacturers’ “loss of innovative momentum is almost inevitable.”
Robot-vacuum arms are currently better understood as an entry point for differentiated R&D than as an established consumer need. Current payloads are around 300–400 grams, making even a heavy slipper difficult to pick up; reliable, fast pick-and-place remains a core robotics challenge, and 陈哲 believes another 2 or 3 product generations may be needed. 泓君 argues that consumers currently care more about automatic water connection, battery life and complete cleaning. The long-term value lies in exploring the possible form factors for $100s-to-$1,000 home-service robots, not in immediately driving sales by “picking up socks.”
The AI-glasses field is crowded, but it has yet to prove suitable for standalone startups or venture capital. At least 20 companies with physical products exhibited at CES, most of them Chinese, while similar supply chains quickly produced Meta Ray-Ban copycats; AI remains mostly an assistive feature, and AR still lacks a sufficiently compelling use case. 泓君 believes the harder-to-copy advantages may instead be comfort, fit across face shapes, fashion design, SKU breadth and retail networks, while a single electronic feature “is not enough to support commercial success.”
If smart glasses need a phone for connectivity, compute and AI, their “most likely destiny is to become an accessory to the smartphone.” TWS offers a precedent: Apple, Samsung, Huawei, Xiaomi, OPPO and vivo should capture the overwhelming majority of the true-wireless-earbuds market, squeezing independent brands and traditional audio makers. Unless glasses can break free from the phone and become a new computing platform, the eventual share is more likely to accrue to phone ecosystems than to today’s large crop of standalone brands.
The winning edge in companion robots is not superior intelligence, but the emotional value created jointly by product design and “emotional intelligence.” Products such as Looi and LOVOT suggest that users may want pet-like responses, cuteness and precisely calibrated movement—not a talking smart speaker; saying “it does not need to be intelligent” does not mean AI is irrelevant, because voice, vision, identity and emotion recognition still need to support natural feedback. The problem is that demand is highly fragmented, and there is no clear winning formula yet.
The more certain industrial opportunities may lie in foundational toolchains and organizations that genuinely know how to build hardware, rather than in the hottest consumer concepts of the year. Nvidia released the Cosmos physical-AI foundation model and is using tools such as Isaac Sim to help customers train robot models, creating a feedback loop that supports GPU demand; DJI-linked companies and Bambu Lab show that product strength comes from first-principles engineering culture, talent development and continuous iteration. CES can point to the hot themes of the next 1 or 2 years, but it cannot answer “who is worth investing in,” because “many companies will disappear.”

Deep dive

1. AI Matters Only When It Redefines the Product

陈哲 began with LiberLive’s stringless guitar: it does not use particularly advanced AI, but its clear pain point, complete user experience and industrial design “single-handedly” created a new category. 泓君’s fast wine decanter made the same point—technology can trigger a product, but value is ultimately delivered in a concrete use case.
PLAUD is the AI-native example he endorses: card-sized, attachable to the back of an iPhone, capable of recording calls, priced at a little over $100, and well suited to convincing consumers directly through short-video channels such as TikTok.
The key change brought by foundation models is not transcription but summary. Traditional voice recorders may already sell 20–30M units a year, but users will not spend 1 or 2 hours replaying them; once connected to OpenAI API, summaries turn recordings from an archive into consumable information. That makes it an experience “that could not have been delivered today without AI.”
泓君’s counterexample captures the most important constraint in hardware: PLAUD’s features won her over, but its frequent battery failures made it lose to a 10-year-old recorder whose battery lasts for months. 陈哲 accepted the hit: “It is a consumer product first, and only then a smart device.”

2. China’s Hardware Lead Now Shows Up in Product Strength and Premium Pricing

After attending CES for 2 consecutive years, 陈哲 says one of the clearest changes is that Chinese brands have moved from exhibitors to the protagonists of innovation. XBotPark’s large group of early-stage projects also reflects how Shenzhen’s mature industrial base is transferring productization capabilities to younger founders.
Robot vacuums are a case in point: China already has 4 or 5 companies that rival iRobot in product quality, brand strength and sales scale. Roborock and Dreame are launching robotic-arm prototypes, showing that competition has moved from copying mature form factors to actively exploring new structures.
The advantage is beginning to show up in pricing as well. 陈哲 observes that companies such as Roborock, Starmax and Kooma are moving into premium, high-markup segments in their respective categories and “increasingly gaining pricing power over technology products,” rather than relying on low prices to substitute for overseas brands.

3. Robot-Vacuum Arms Are Research-Led Differentiation, Not a Mature Must-Have

泓君’s consumer objection is direct: when there is obvious clutter on the floor, people will usually just pick it up; compared with slowly retrieving socks, automatic water connection, vacuum-and-mop integration and enough battery to clean an entire home are more urgent needs.
陈哲 brings the innovation back to floor cleaning. In 2024, an extendable side brush improved edge coverage, while a leg-wheel-like drive system allowed robots to cross 4–5 cm thresholds—changes that directly raise cleaning rates. Robotic arms are an attempt to let the robot arrange its environment before or after cleaning when obstacles undermine coverage.
陈哲 believes increasing battery capacity is relatively easy in new models—a linear problem. Dealing with more obstacles on the floor remains open-ended. Current arms can carry only around 300–400 grams: “Apart from socks, they can barely handle even a relatively heavy slipper.” Reliably completing pick-and-place in an unstructured home is itself a problem that industry and research are still working to solve; he estimates another 2 or 3 product generations may be needed.
Asked whether consumer products are being used to train embodied intelligence, 陈哲 does not treat valuation as the core logic. The more important task is to find low-cost, valuable home-service scenarios beyond floor cleaning. Existing robot vacuums have established a price anchor from several hundred dollars to $1,000; whether a new form factor can escape that anchor remains an open question.

4. Higher Penetration Depends on Local Demand; Patents Are No Longer the Main Constraint on Leaders

Robot-vacuum penetration in Europe and the US is around the low-teens percentage range, still far below the 80–90% penetration common for household appliances. iRobot initially solved carpet vacuuming, while the Chinese market moved from below 1–2% penetration toward nearly 10% through whole-home mapping, vacuum-and-mop integration, and innovations such as Narwal’s mopping, water connection and base stations.
陈哲 acknowledges that Chinese brands initially optimized for domestic hard-floor demand, with relatively little customization for overseas households dominated by carpet. More designs tailored to overseas use cases will emerge only as companies continue expanding abroad. This is not something that can be explained simply by a greater “determination to go overseas.”
泓君 suggests that iRobot’s early patents both blocked competition and may have slowed innovation. 陈哲’s correction is that patents hurt long-tail players, contract manufacturers and solution providers more; leading companies already hold large patent portfolios and cross-license with brands such as iRobot and Ecovacs. The more important fact today is that Chinese manufacturers iterate at speeds several times—and sometimes more than 10x—those of their overseas peers.

5. AI Glasses Have Split into Camera and AR Products, but AI Remains an Assistive Feature

陈哲 divides the products on display into 2 tracks. One is close to Meta Ray-Ban: glasses or sunglasses that provide cameras, voice and some onboard processing. The other includes products such as Rokid and XREAL with AR displays, targeting spatial computing, video viewing or gaming.
Gyges Labs’ Halliday AR glasses use a projection method different from conventional AR, offering a new approach to display architecture. Overall, brightness, field of view and weight have improved materially, but these metrics remain highly dependent on the supply chain.
At least 20 smart-glasses companies with physical products exhibited at CES, and the overwhelming majority were Chinese brands. Mature solutions, contract manufacturing and production capabilities allow a team with a design idea to ship quickly, but they also drive convergence in display architecture and feature definitions.
陈哲’s core view is that “AI in today’s smart glasses may all be an assistive function.” Long-term wearability still rests on core features such as photography or video viewing; no brand has yet delivered a decisive answer for how multimodality and foundation models can create an irreplaceable glasses experience.

6. The Real Moat in Glasses May Be Traditional Product Capability, Not Electronics

Meta’s Orion represents a meaningful breakthrough in display and optics, while Apple and Meta have the budgets to push supply chains toward performance limits. But 陈哲 warns that these structures are often not exclusive; once the supply chain matures, it empowers third parties at the same time, allowing Meta Ray-Ban copycats to appear quickly.
VR has at least found product-market fit in gaming and solo entertainment. AR glasses have yet to reach the point where consumers feel they “must buy” them. Meta Ray-Ban shows that $200–$300 camera glasses with limited AI and voice capabilities can already attract sustained use from some consumers, but it does not prove that the full AR form factor works.
泓君’s inference is more consumer-focused: European and Asian faces differ in shape, interpupillary distance and nose-bridge structure, making comfort a barrier in its own right. Appearance, SKU breadth and Ray-Ban’s retail network built over decades also matter. “Smart glasses are glasses first.” Building a solution and getting consumers to buy and recommend it are 2 different logics.
陈哲 therefore has not invested directly in the category; even upstream, he has yet to see a clear moat. If 1 or 2 features from a startup are insufficient to establish user value and a commercial moat, it is “most likely not a good venture-capital opportunity.”

7. Phone-Dependent Glasses Are Likely to Repeat the TWS Industry Outcome

陈哲 has consistently argued that wearables too close to phones in function and value will become phone accessories. In TWS, Apple, Samsung, Huawei, Xiaomi, OPPO and vivo should take the overwhelming majority of the market, while traditional audio brands such as Sony, Bose and Sennheiser are also being squeezed by phone makers.
When he worked on Google Glass from 2014 to 2016, the team already understood that phone dependence was a constraint, so it added Wi-Fi to let the glasses connect directly to the internet. But wireless and battery technology at the time could not accommodate cellular connectivity. Today’s glasses remain constrained by compute, battery and connectivity, and the prospect of an independent computing platform is still very limited.
Even with Meta Ray-Ban, the workflow of syncing captured content to a phone or the cloud before sharing and operating on it is “not particularly smooth.” 泓君 adds that Meta’s product team is continuing to explore transmission and interaction between glasses and phones, while the ideal state would make the entire process transparent, natural and seamless.
After Google and XREAL announced their partnership in December 2024, the more likely strategy is to replicate Android and Pixel: first build reference projects that demonstrate how models, the cloud and an operating system can empower glasses, then attract 10 or 100 hardware makers. 陈哲 is “very pessimistic” that Google or Meta can turn the relevant division into a profit center, and estimates that each pair of Meta Ray-Ban glasses likely brings a meaningful loss to the group.

8. Companion Robots Sell Emotional Connection, Not Task Efficiency

Another area of activity 陈哲 saw was companion robots: products emerging simultaneously for humanoid, biomimetic, specific-user and specific-scenario applications. Looi attaches an iPhone to a mobile stand and turns it into a desktop robot; the team observed that in private one-on-one spaces, users summon and actively interact with the robot at markedly different frequencies than in shared public settings.
Japan’s LOVOT offers another proof point: around 40 cm tall, egg-shaped, mobile and huggable, with clothing, sensors, expressions and movement. A large share of its most devoted and dependent users are women over 40. These products may not need to speak at all, because their benchmark is not the smart speaker but a cat or dog.
“It does not need to be intelligent” means it does not need to display intelligence, not that it needs no technology. Cuteness, plush texture and dynamic posture trigger emotional responses; these products are closer to smart toys. 陈哲 points to WALL-E, which barely speaks yet is highly affecting, as evidence that character, movement and industrial design can themselves be the product.
AI’s job is to make the response precisely calibrated: recognize sound, video, environment, space, identity, speech and emotion, then produce a brief response. 泓君 summarizes it as “not superior intelligence, but superior emotional intelligence.” Because companion users and scenarios vary so widely, the industry still has no clear winning formula.

9. Nvidia Is Locking In the Robotics Ecosystem Through Its Toolchain

Among the major-company launches at CES, 陈哲 paid closest attention to Nvidia’s Cosmos physical-AI foundation model. He believes it could help develop and train autonomous-driving and mobile-robot systems, supporting the rollout of more intelligent robots over the next 1 or 2 years.
Nvidia’s ecosystem position is not to train one standardized foundation model for customers, but to provide development, simulation and training tools that enable customers to produce more models—and then purchase more GPUs and processors. “When its toolchain becomes something every company must use to develop a new type of robot product,” the chip business gains a much stronger foundation.
Many researchers complain about Isaac Sim’s maturity and usability, but 陈哲 sees it as the most fundamental reason for the progress of legged robots over the past 2 years. Simulation environments allow teams to use reinforcement learning to train control algorithms and policies quickly. The long-term value of the toolchain may therefore exceed that of simply selling a single model.

10. A Good Hardware Founder Must Understand Users, Engineering and Organizations

陈哲’s first criterion for evaluating a founder is attention to “the user’s real needs, not their own fantasies”: empathy matters, and ego cannot be too large. Software can continue to be upgraded, but when hardware gets the need definition wrong, the entire supply chain must be rebuilt, with longer validation cycles and higher failure costs.
MVPs are especially important in hardware. Mock-ups, 3D printing and simple demos can validate the core selling point, while crowdfunding and user testing can shorten the feedback loop. When the overall direction is wrong, the probability of hardware success is very low.
Founders also need first-principles thinking, technical decomposition, spatial understanding, engineering execution and aesthetic judgment, along with the ability to organize a full-stack team spanning design, production, mass manufacturing and sales. 陈哲 warns that a brilliant product manager may generate RMB1B in revenue from a single product, but without management and sustained R&D capabilities, still fail to build a durable company.
陶冶 of Bambu Lab is a rare example in his view: he worked at DJI on Mavic, early propulsion systems and product definition before bringing an established team into 3D printing. Bambu Lab achieved more than 50% market share in 1 year and gained pricing power—an outcome 陈哲 attributes to organizational capability, not a one-off product hit.

11. CES Is a Cross-Section of the Annual淘汰赛, Not an Investment Answer

泓君’s fatigue with CES comes from the trade show’s incentives: companies amplify new features for exposure, fundraising and media attention, sometimes ignoring consumer needs, which makes the products on display look rough. She focuses on disruptive technologies, but products truly sold to consumers usually need to rely on “mature technology and mature solutions.”
陈哲 agrees that CES is a marketing event, but believes repeated observation reveals the淘汰 process. A large number of early-stage humanoid-robot companies appeared the previous year, but fewer booths remained the next year, while survivors such as Unitree and DEEP Robotics showed greater maturity. Lawn-mowing and pool-cleaning robots likewise went from crowded fields of more than 10 exhibitors to a handful of leaders with stronger sales, brands and booth footprints.
AI glasses are likely to follow the same path: “By this time next year, many companies will have disappeared.” The survivors may be the ones able to produce better products. CES can therefore indicate trends and market reactions over the next 1 or 2 years, but one exhibition cannot determine who is worth investing in.
The scale of frontier research is not the same as the scale of consumer innovation. Adding an arm to a robot vacuum may be a small experiment in general-purpose robotics research, but a major breakthrough for a mass-produced robot vacuum. The entire conversation ultimately returns to the same product discipline: AI can create variables, but whether they become a business still depends on demand, reliability, cost and sustained iteration.

Eran Zinman, Co-Founder & Co-CEO @Monday.com: Going Upmarket, International and Multi-Product |E1247 AI’s Uneven Arrival, TikTok’s Potential Departure, Xiaohongshu and the Delights of Cultural Exchange