In 2025, AI hardware will return to the real world.
In 2011, venture capitalist Marc Andreessen's assertion that "software is eating the world" still holds true in the current AI wave. However, since 2025, "AI hardware" has once again become the darling pursued by capital and public sentiment.
Compared with the hype of AI-native hardware such as AI Pin and Rabbit R1 in the first round, the new round of AI hardware explosion presents three obvious characteristics:
Firstly, it is approaching the eve of the birth of a blockbuster product. Referring to the threshold standard of "one million in shipments", products such as PLAUD's AI recording hardware and Meta's Ray-Ban Meta have all announced that they have reached this threshold.
Secondly, the user base continues to expand. Compared with the attraction of AI products to geeks last year, "Silicon Research Lab" has learned that leading companies in some popular categories such as smart office tablets and AI earphones have completed the accumulation of one million users and are trying to move from the niche market to the mass market.
Thirdly, the competition routes are diverging. The new round of "buying sprees" by large companies has begun, while relatively pragmatic AI hardware startups are turning to in-depth exploration in niche markets.
Looking back at least, compared with the setbacks of early products like AI Pin, players at the forefront of the second wave of AI hardware explosion have learned a lesson: No matter how amazing the AI hardware startup story is, it is ultimately a hardware startup.
1. The "Consensus" in Blockbuster Products: Seeking Scale in Major Markets and Exploring Demands in New Categories
In 2025, in terms of both shipments and user feedback, AI hardware has finally welcomed "potential blockbuster candidates".
First, there is the AI glasses jointly developed by Meta and Ray-Ban, and then there is the AI card recorder launched by startup Plaud.AI. From these two "one-million-shipment" players, we can see a "consensus in blockbuster products": Seeking scale in major markets and exploring demands in new categories.
An investor believes that to find AI hardware products with blockbuster potential, one basically "just needs to follow the large companies", such as AI glasses and AI earphones. These two categories have entered the mainstream view with the influx of large companies and startups.
Take the ongoing "battle of a hundred glasses" in the AI glasses segment as an example. It is already crowded with dozens of players - whether they are large companies like Meta, Google, Alibaba, and ByteDance, emerging brands like Rokid and Thunderbird, or Xiaomi, which has entered the market from a cross - industry perspective. AI glasses have accelerated their entry into the mass consumer market.
Yang Longsheng, the CEO of "Inmo Technology", once mentioned in an interview that after large companies like Xiaomi entered the market, the overall sales of various manufacturers this year have tripled or quadrupled.
Glasses and earphones are not new categories. As natural entry points closer to users, they already have a certain user base. The opportunity brought by the AI explosion to such categories is to unify the fragmented experiences, systems, interactions, and supply chains in the industry in the past.
For players in categories such as AI glasses, the key to competition will be how to "expand the scale while optimizing the experience".
In contrast, new categories such as the AI card recorder represented by Plaud.AI have become blockbusters because they target a specific niche scenario with strong demand and a specific group of people, and explore demands in a market that "large companies look down on and small companies cannot handle", quickly accumulating their own advantages.
Source: Plaud.AI official website
The person in charge of the hardware business of a domestic technology company told "Silicon Research Lab" that for new categories like Plaud.AI to become blockbusters, there are three prerequisites: Firstly, the product definition must be precise enough. "One must figure out clearly in the definition how it is different from traditional mass consumer electronics products." Secondly, there must be a strong ability to educate users and iterate the product based on consumer feedback. Thirdly, there must be a complete hardware, software, and AI team.
In addition, a bit of luck is also needed.
Xu Gao, the CEO of PLAUD AI, once mentioned in an interview that in 2021, when large models had not yet exploded, they entered the voice recorder market because they saw the technological iteration space and value in the NLP field in the early stage and invested a lot of AI capabilities in long - text extraction. After GPT came out, "the large model achieved this in a way that is ten thousand times better", which confirmed the value of AI in this direction as a strong - demand factor.
2. Diverging AI Paths: Traditional Terminals "Build Central Hubs", Wearable Devices "Achieve Multi - Modal Collaboration"
Under the popularity of blockbuster products, a review of the current AI integration ideas in AI hardware also shows obvious divergence.
We can divide AI hardware terminals into two categories: one is traditional hardware terminals, such as mobile phones, PCs, and tablets; the other is wearable devices, with typical forms such as glasses, TWS earphones, smart watches, etc., and emerging categories such as smart rings.
In the implementation of AI on traditional terminals, mobile phone and PC manufacturers are regarding the intelligent agent as the system hub. Taking mobile phones as an example, they let the intelligent agent take over the permission interfaces of the camera, positioning, communication, etc., to achieve cross - application and cross - device experiences. To some extent, the intelligent agent has become the "new OS" for mobile phones, PCs, etc.
Evolving from a simple dialogue assistant to a collaborative system for complex tasks and upgrading the AI voice assistant to an intelligent agent that can recognize, understand, and execute user tasks has become a common move among mobile phone manufacturers.
For example, Samsung has connected 30 local applications and more than a dozen third - party applications in its S25 series, enabling its AI assistant Bixby to have stronger cross - application execution capabilities. Domestic mobile phone manufacturers such as Huawei, Xiaomi, OPPO, and VIVO are also optimizing the layout of intelligent agents in their customized system frameworks and demonstrating AI capabilities in high - frequency scenarios such as travel and social messaging.
Compared with the systematic projects of traditional terminal manufacturers, wearable device manufacturers, who are still building their own entry points, are more concerned with multi - modal collaboration. At present, their AI paths still focus on leveraging the advantages of their own hardware forms and encapsulating AI capabilities into specific functional experiences and scenarios.
Take earphones as an example. Their original basic functions are sound quality and noise reduction. After the access of AI capabilities, they can not only optimize the basic capabilities but also bring new extended values, such as real - time transcription and translation for meeting scenarios and new functions such as heart rate recognition in the health scenario.
3. 2025: Demystifying the AI Story and Returning to the Real Hardware World
Although the AI hardware market is bustling, 2025 is destined to be a watershed for AI to return to the real world.
On the one hand, the fierce competition in popular markets has raised users' expectations for AI hardware products. Xu Chi, the founder of XREAL, recently poured cold water on the AI glasses market: "The current AI glasses are more like a 5 - year - old child in terms of intelligence and do not have the stickiness to make consumers wear them continuously every day."
On the other hand, AI hardware cannot escape the "survival rules of hardware" - Hardware, software, algorithms, and scenarios, the four elements are indispensable.
In other words, AI hardware is ultimately a cruel hardware adventure. Take the AI glasses popularized by Meta and Ray - Ban and reaching one million in shipments as an example. In addition to the AI story, it is largely because they have injected a fashionable attribute into AI glasses. For hardware products, fashion can be an influencing factor, but it is difficult to become the underlying moat for hardware products.
The person in charge of the hardware business of the above - mentioned domestic technology company told "Silicon Research Lab": "One should have a sense of awe when doing hardware, especially when doing consumer - grade products, one should be restrained about growth." On the one hand, they should maintain a sense of awe for technology, and on the other hand, they should maintain a sense of awe for users. They have divided their R & D and product manager teams into logical function groups and AI effect groups internally, and at the same time require all employees to maintain a connection with users.
Returning to the first - principle thinking, AI is an adjective, and hardware is the essence. Different from other sectors, AI hardware is a typical sector with an "end - game mindset". Whether it is the ambition to become the next "computing center" or to become the leader in a category, user experience is the consensus on the way to the end - game.
Above the bustling and prosperous surface of the AI hardware market, Steve Jobs' experience of "leaving complexity to oneself and simplicity to users" is still worth learning by more people.
References:
1. NetEase Technology: Stop Daydreaming! Xu Chi of XREAL: The Current AI Glasses Are as Smart as a 5 - Year - Old Child and Lack the Stickiness for Daily Wear
2. JQ Capital: Interview with the Founder of PLAUD: AI Recording Hardware, with Over 200,000 Global Users, How Did It Achieve That?
3. Soochow Securities: Edge - Side AI Reconstructs the Terminal Ecosystem: Hardware Upgrades Drive the Replacement Wave, and Multi - Device Collaboration Spurs the Leap of Business Models
This article is from the WeChat official account "Silicon Research Lab", and the author is kiki. It is published by 36Kr with authorization.