“Taking a step forward is a martyr, leading half a step is a pioneer.” This saying is a well-known adage in the field of technological innovation, with historical examples like Motorola’s satellite communication project, the Iridium project, and more recently, Meta’s grand vision for the metaverse. Now, it seems that Rabbit, a leading player in the AI hardware sector, is following in the footsteps of its predecessors, becoming a new “martyr.”
Recently, Rabbit’s founder, Jesse Lyu, admitted in an interview that the company is facing financial pressure due to regulatory issues that have hindered its plans to enter the Indian market, resulting in a “significant deviation” in financial forecasts. However, this does not affect the launch of the next generation of AI hardware in 2026. Yet, a recent report from overseas media indicates that Rabbit’s financial difficulties may be far worse than expected.
The report reveals that Rabbit’s employees have been in a state of unpaid wages since July of this year, leading some employees and outsourced teams to resort to strikes in protest. This seems to explain why, when their first product, the Rabbit R1, began to sell and faced negative public opinion, Rabbit appeared indifferent. It may not have been that Rabbit did not want to quell negative sentiment, but rather that internal turmoil left them with no time to address it.
Looking back at last year’s CES, the Rabbit R1, as one of the first AI-native hardware products, became the most eye-catching presence at the event. At that time, this orange cube, priced at only $199, claimed to “replace all apps,” thereby changing the way users interact with applications. On the first day of open reservations, Rabbit sold 10,000 units, and by the time it officially launched, it had sold 50,000 units within two weeks.
So, the question arises: how did Rabbit, which seemed to have a perfect start, end up in a situation of unpaid wages? In fact, the overly advanced concept of the Rabbit R1 may be key. According to Jesse Lyu, the Rabbit R1 is something “completely different from a smartphone.” Currently, developers are trying to keep users within their apps based on traffic logic, rather than efficiently helping users achieve their task intentions. The Rabbit R1 offers a healthier interaction paradigm.
Tasks such as ordering takeout, booking hotels, planning trips, translating documents, and interpreting images can all be accomplished with a single command through the Rabbit R1. To win over overseas consumers, Rabbit collaborated with the Swedish hardware design company Teenage Engineering, known for its minimalist style, to create the eye-catching orange cube design of the Rabbit R1.
However, unfortunately, Rabbit overestimated its supply chain integration capabilities and the tuning level of its AI large model, leading to an inability to realize the vision described at CES. It could even be described as a “PPT scam.” In many users’ tests, the Large Action Mode, which Rabbit touted, turned out to be a mere shell. Although the Rabbit R1 can indeed order takeout and book restaurants, it cannot be done with a single command; instead, it requires repeated interactions and confirmations with the AI.
Even setting aside the extremely low-capacity battery and weak main control performance, the Rabbit R1’s performance does not meet external expectations for AI-native hardware, which is the root cause of its high initial expectations followed by a decline. Rabbit’s initial success was due to the company’s keen insight that more and more users recognize that “smartphone dependency” is an unhealthy lifestyle, prompting the targeted release of the Rabbit R1.
However, while Rabbit recognized that many people are beginning to reflect on the alienation caused by smartphones, it failed to notice that these consumers actually want both: they want to break free from smartphone dependency while also needing a high-performance, feature-rich mobile computing device. Consumers willing to spend money to explore new things want a “phone” that does not look like a smartphone, but Rabbit clearly lacks the ability to create such a product.
The rapid decline of Rabbit reflects that the difficulty of opening up the consumer market for AI-native hardware is far greater than entrepreneurs expected. The entire industry has not yet figured out what the next general computing device to succeed smartphones will look like. Otherwise, OpenAI CEO Sam Altman would not have used vague phrases like “users will want it” and “like sitting in a beautiful cabin by the lake” to describe their first AI hardware product in a recent interview.
Sam Altman also used such descriptions when discussing the AI hardware that OpenAI is developing, stating, “Users will not feel any pressure while using it, and will be able to use it almost instinctively, existing somewhere between complexity and simplicity.”
At this point, many friends may realize that even the current leaders in the AI industry may not know what AI-native hardware should look like.
Indeed, it has been three years since the emergence of ChatGPT, yet the only products that have truly opened up the market are AI toys. This is because the companionship value provided by toys is difficult to quantify, and the target audience, children, only need a playmate. In contrast, the selling points of the Rabbit R1, such as “ordering takeout with one command,” have clear evaluative standards: it either works or it doesn’t, while the emotional value of AI toys is clearly difficult to quantify.
Moreover, the consumers and users of AI toys are separate; the users, children, lack a voice, and the opinions of parents regarding whether AI toys can accompany their children are crucial. This is why current AI toy brands emphasize the educational attributes of their products, with phrases like “AI has vast knowledge, entertaining and educational” and “learning through play” being common advertising slogans.
In contrast to Rabbit’s extravagance, AI toy companies are actually more pragmatic. They not only understand how to communicate core selling points to consumers more clearly but also choose selling points that cannot be easily disproven.In modern society, as work pace accelerates, parents often lack the time to accompany their children. In this context, an interactive AI toy that provides personalized services becomes a “lifesaver” for parents, and can, to some extent, replace them in completing some parent-child activities.
Ultimately, “usefulness” is the touchstone for evaluating the quality of an AI hardware product. After all, AI hardware requires consumers to spend real money, which is fundamentally different from free downloads of ChatGPT, DeepSeek, Doubao, and Qianwen.
[Images in this article are sourced from the internet]