The lost gears and the blank moments: Why does the trillion-dollar AI toy market only hear the thunder but not the rain?

Recently, at industry exhibitions and private discussions, an awkward silence has been spreading. Although the concept of “AI + toys” is extremely popular in the capital market and the PPTs at the press conferences are extremely cool, the data from the front-line sales shows a dismal picture. A question that has been repeatedly raised is: “Why are AI toys so popular, but the ones that can actually sell are very few?”

After having in-depth conversations with dozens of practitioners, the answers were surprisingly consistent: They didn’t fail to do AI, but they stopped doing it soon after. Some manufacturers took over the voice module, some tried simple conversation functions, and even some leading enterprises had already spent a fortune to access the so-called large model capabilities. But the results were surprisingly similar: The BOM cost (material list cost) increased, the supply chain became more complex, and the product manual became thicker. However, the users’ purchasing enthusiasm did not ignite accordingly, and the repurchase rate was even pitifully low.

Initially, the industry habitually attributed this frustration to “the technology is not mature enough”, believing that as long as we wait for the emergence of GPT-6, 7, and 8, everything would be resolved. But at the exhibition site, looking at the piles of “intelligent junk” on the exhibition stand, my judgment fundamentally reversed: The problem with AI toys is not that the technology cannot be done, but that no one can define – when should it be used.

I. Missing anchor point: All failed products cannot find the “use moment”

We can conduct a simple thought experiment. Place an AI toy in front of the user and ask a simple question: “At which specific moment in a day would you reach out to take it?”

Is it in the early morning when waking up? On the way to work? During breaks at work? Or at the moment when turning off the lights before going to bed? Or when the child comes home from school, throws down their schoolbag? Or when adults face dinner alone and feel no one to talk to?

If even the product manager themselves cannot answer this question clearly, then this product is doomed from the start. Because a product without a “use moment” is like a key without a lock hole, no matter how precise the teeth are, there is no reason to be inserted and turned.

Nowadays, the common problem with many AI toys is that they “can do many things” – recite Tang poems, do arithmetic, tell jokes, control smart home devices. But in the users’ daily lives, there is no single thing that “must and can only be completed by this toy”. This all-purpose pseudo-demand ultimately leads to the result that: Users buy it to play for two days, find it fresh, and then let it lie in the corner to rot.

II. Misaligned logic: Technology accumulation vs. user excuses

There is an extremely fatal gap in the current industry – the mismatch between technology supply and user demand.

The logic of technology companies is linear and arrogant: “I can have a conversation”, “I can remember”, “I can recognize emotions”, “I can even chat like a human”. They assume a premise: the stronger the ability, the more the product will succeed. It seems as if by filling the parameter table with functions, users will naturally flock to it.

But the logic of users is completely different. When facing a new thing, the first thought that flashes through their minds is always: “Why should I use you at this time?” Or more bluntly, “What use do you have for me?”

This dislocation leads to an absurd situation: the product’s feature list gets longer and longer, but the opening frequency gets lower and lower. Because users never find a clear, un-rejectable reason to use it. Just like a nanny with a PhD, if she can’t solve your specific problem (such as soothing a child to sleep) at a specific moment, then her ability is redundant noise for you.

III. The confusion at the manufacturing end: “AI anxiety” in Chenghai Factory

This confusion does not only exist in internet companies, but also spreads down to the very end of the industrial chain – the toy manufacturers in Guangdong Chenghai.
As the heart of China’s toy manufacturing, the bosses in Chenghai are extremely sensitive, but they have also fallen into deep confusion. Their path is usually a follow-the-leader approach: “When others add voice, I add it too”; “When others can have a conversation, I add a module”; “When ‘AI’ is very popular at the investment promotion meeting, my packaging must also have it”.

This is a defensive “arms race”. But when you pull them aside and ask a painful question: “Mr. Wang, after adding AI to this toy, when will users use it?” Most people will be stunned and then stammeringly answer: “Um… it’s for fun, it’s high-tech.”

So, the market is filled with such “patchwork” products: they can talk, interact, and even “look very intelligent”. But after users buy them, due to the lack of a clear scene guidance, they are quickly classified as “electronic garbage”. Not because the product is poorly made, but because it simply doesn’t fit into the already crowded life rhythm of users.

IV. The broken structure: vacuum of ability, hardware and scenarios

If we break down the entire AI toy industry, we will see a clear “sandwich structure”:
The upper layer is the hot AI capabilities (large models, speech recognition, emotional computing);
The lower layer is the extremely mature hardware and supply chain (injection molding, assembly, logistics);
The middle layer is the usage scenarios – this layer is almost blank.

The current situation is that the upper layer is very hot, and capital is burning; the lower layer is very mature, and production capacity is waiting. But the connection layer in the middle is missing. This leads to a cruel result: no matter how powerful the AI capabilities are, they cannot be transformed into real user value. Just like how strong the current is, if there is no appliance (scenario) to receive it, it is only a dangerous lightning instead of a light that illuminates life.

V. Historical Lessons: All great toys occupy a certain moment

If we look back and examine those toys that have survived cycles and truly succeeded, we will find that they all precisely occupied a “usage moment”:
Cuddly toys: occupied the “bedtime” and “safety” moments. It was the last hug for children before they went to sleep.
Game consoles: occupied the “boredom” and “entertainment time”. It was a refuge for commuting or weekend relaxation.
Card games (such as Ultraman cards): occupied the “social” moments. It was the hard currency and social currency for boys on the playground.
Blind boxes: occupied the “instant reward” moments. It used dopamine stimulation to fill the small gaps in the mundane life.
They succeeded not because of the strongest functions, but because in a specific moment, you would naturally think of it. But now, AI toys, like a homeless ghost, float in the user’s life and cannot find a “must-have” moment where they can settle down.

VI. The Solution: AI should not become stronger, but should “fit into” an instant

Following this logic, we will reach an counter-intuitive conclusion: The breakthrough point of AI toys is not to become smarter than humans, more like humans, or more complex.

On the contrary, it lies in “de-intellectualization” – making oneself extremely focused, concentrating on a tiny moment in the user’s life.

The future winners will definitely be those who find the scenarios where “AI is not possible without” them. For example:
“Goodnight Ritual” scene: Before a child goes to sleep, they need a companion who won’t get impatient, can tell stories infinitely in a loop, and can adjust the speaking speed according to the child’s breathing.
“Emotional First Aid” scene: Adults need a safe, non-judgmental, and immediately responsive “safe haven” in situations such as on the subway, after overtime work, or after being scolded by the boss.
“Low-Stress Interaction” scene: Users don’t want intense real-person social interaction, but they can’t stand absolute loneliness. At this time, an AI toy that can crack jokes and make light-hearted comments together is the perfect buffer zone.
The common point of these scenarios is: What users need is not an “all-powerful god”, but a “companion who understands me”. What they need is not a solution to problems, but the feeling of being seen and responded to.

VII. Conclusion: The War for Time

The future competition will not occur between model parameters, nor will it be about whose voice is more natural. The real war occurs in “who defines that moment first”.
Once a brand defines the “bedtime moment” first and makes users form a conditioned reflex – touching this toy before turning off the lights, then no matter how much other competitors invest, it will be of no avail. Because once this “moment” is occupied, the product has usage frequency, and users will develop a muscle-memory-like dependence, and commercial repeat purchases and scale will truly arrive.

In the noisy crowd at the exhibition, I suddenly realized: We spent too much time discussing what AI toys “can do”, which is like studying how many nuts a hammer can crush. But the real question to ask is: In the user’s long and trivial life, when should this hammer be picked up conveniently?

What determines the life of a product is not how sharp it is, but whether it has a coordinate on the user’s life map and a position that belongs to it.