The AI Hardware Bubble: The Harsh Truth Behind the Exhibition’s Excitement, 99% of Products Are Just Transitional

“Today at the venue, I believe that less than 1% of the products are truly doing it right; most are still transitional products.”

—— Qiu Ze, Founder of Craft Society

Chapter One: The AI Carnival and Illusions at the Exhibition

On October 27, the Shanghai West Bank International Technology Consumer Carnival resembled a meticulously orchestrated technology feast. The exhibition hall was bustling with people, showcasing a variety of cute AI robots, smart wearable devices, and entertainment gadgets, all infused with the air of future technology.

In front of Ropet’s booth, a group of young women surrounded a large-eyed plush AI pet robot, exclaiming in amazement. This product, priced at 1472 yuan, became a KOL favorite during the exhibition due to its adorable appearance and reasonable price. The staff enthusiastically introduced, “We sell about 20 units per day on average during the exhibition.”

Ropet Mengyou Intelligent, established in 2022, focuses on AI emotional companionship and was founded by a former ByteDance product designer. The product features a self-developed offline small model that can collect data through cameras and gravity sensors to recognize faces and emotions.

Not far away, the Japanese Lovot robot attracted another group of spectators with its high price of 33,000 yuan and exquisite design. The staff patiently explained, “Lovot’s main feature is to provide companionship and emotional interaction; other functions, such as using generative AI for conversation, are currently not achievable.”

However, behind the excitement of the exhibition lies a harsh reality. As Qiu Ze candidly stated during a roundtable discussion, “The true era of AI hardware has not yet arrived.”

Chapter Two: The Dilemma of ‘New Wine in Old Bottles’ in AI Hardware

The business director of an AI startup, Mi Siqi, revealed the truth: “Most AI smart hardware products on site are ‘new wine in old bottles’, merely updating the UI with current AI logic without achieving true intelligence.”

This phenomenon of homogenization was evident throughout the exhibition. From AI pet robots to smart wearable devices, from learning machines to entertainment devices, the product forms varied widely, yet their core functions were astonishingly similar. Many products exhibited a low level of AI, remaining at the level of voice interaction, differing only in response speed.

The Truth Behind Homogenization

• Low entry barriers; the market appears to be a red ocean but is actually fiercely competitive.
• Rapid product iteration, typically once a year.
• Limited AI functionality, relying more on design to attract consumers.
• User understanding of AI remains at the level of “can it converse?”

Yang Dongyun, co-founder of Wujie Ark, pointed out during the roundtable discussion, “AI hardware seems like a red ocean because of low entry barriers, leading many to believe the market is large.” She believes that the ability to discern core user needs is key to whether a product can break through.

However, identifying user needs is not easy. Ropet staff admitted, “Current market understanding of AI still lingers at the level of ‘can it converse?’. When they see AI features, they ask if we can converse. But features like emotion recording are also part of AI functionality.”

Chapter Three: The Gap Between Technical Bottlenecks and Market Expectations

Hugging Face showcased the desktop humanoid robot Reachy Mini at the exhibition. Although small in appearance, this robot is “small but complete” with a comprehensive system framework for structural design and AI integration. Backed by the Hugging Face open-source AI platform, users can directly invoke over 15 preset actions built into the Hugging Face center for secondary development of this small desktop robot.

However, Hugging Face staff also acknowledged, “The first-generation product has not fully integrated visual recognition technology.” They believe that the AI hardware field currently presents a market phenomenon of “supply exceeding demand”. Despite many companies investing in R&D, few can create software systems that deeply match the hardware.

Analysis of Technical Bottlenecks: Existing AI voice assistants face issues of limited functionality development and insufficient practicality. When users have actual needs, current AI hardware cannot provide a feasible solution. The main reason lies in the AI model layer, which still lacks the ability for self-recognition, self-judgment, and self-decision-making.

Another startup focusing on “emotion-sensing intelligent pendants”, NUNA, stated that their product can collect physiological data through sensors and microphones, analyze emotional states, and push reports. However, the entire sector is still in the market exploration stage and has not truly captured core user needs.

Chapter Four: The Path to Breakthrough for Next-Generation AI Hardware

Faced with an increasingly competitive AI hardware landscape, the industry has developed new thoughts and insights on next-generation AI smart hardware products.

Duan Yu, head of consumer electronics e-commerce at Guangyi Technology, suggested that AI product development should achieve decoupling of software and hardware. Using voice recorders as an example, he pointed out, “The hardware itself can remain stable, and by continuously upgrading models or software through Bluetooth and other interfaces, performance can be enhanced and the lifecycle extended.”

Yang Dongyun’s Three Curve Theory

• Market Scale Curve: Understanding the capacity and growth potential of the target market.
• Technical Barrier Curve: Establishing sustainable technical advantages.
• Industry Trend Curve: Grasping the development direction for the next 2-3 years.

Qiu Ze’s Three Elements for Survival

• Strong AI Application Layer: Robust technical strength.
• Precise PMF: Perfect product-market fit.
• Reasonable Carrier Design and Pricing: Creating brand momentum.

Hugging Face staff told Pengpai Technology that the core demand for AI hardware is flexibility, intelligence, and safety, while the product design, shape, and color should facilitate user use.

However, Xu Xiaoyin, founder and CEO of the Silicon Valley startup AI agent company Hey Boss, provided a more sobering judgment: “AI products do not have a real moat. It is too easy for peers to replicate, and homogenization is severe. The so-called ‘relying on innovation to create a moat’ is actually an illusion; you can never guarantee that every step is more innovative, and your innovation can never be preserved.”

Chapter Five: The Future Path of AI Hardware

The noise at the exhibition starkly contrasts with the calmness of industry insiders. While attendees are captivated by cute AI pet robots and dazzled by cool smart devices, practitioners are pondering deeper issues.

Staff from Ropet revealed that the team iterates and updates based on user feedback every month, stating, “We have updated our current version over 40 times.” This rapid iteration reflects the industry’s ongoing exploration of product forms and market demands.

Industry Consensus: The True Era of AI Hardware Has Not Yet Arrived

It requires entrepreneurs to rediscover their PMF (product-market fit) with each technological wave.

From the excitement at the exhibition to the calm analysis by industry insiders, we see the true face of the AI hardware industry: behind the surface prosperity lies a multitude of challenges, including technical bottlenecks, homogenization competition, and market perception biases.

The true era of AI hardware requires not only technological breakthroughs but also a profound understanding of user needs, the establishment of sustainable business models, and finding one’s position within the entire ecosystem. As Qiu Ze stated, “The true era of AI hardware has not yet arrived; it requires entrepreneurs to rediscover their PMF with each technological wave.”

The excitement of the exhibition will eventually fade, but the path of AI hardware development has just begun. In this competitive landscape filled with opportunities and challenges, those who can truly understand user needs, break through technical bottlenecks, and establish sustainable business models will stand out in the next round of competition.

The future of AI hardware lies not in the noise of exhibitions but in meeting the real needs of every user.