From mosquito-killing experts to smart pet collars: The market breakthrough of quirky AI hardware

The AI hardware sector has always been a double - edged sword.

On one hand, tech giants and star startups have poured huge amounts of money into it but often ended up failing. Humane's AI Pin, a $699 brooch with a valuation of up to $850 million at one point, faced a large number of returns just 10 months after its launch due to issues such as poor battery life, severe overheating, and a voice response delay of up to 10 seconds. Eventually, it was "sold off" to HP for $116 million. Rabbit R1 once set a record of selling 100,000 units in 4 days. However, due to its lackluster product performance and an extremely high return rate, it finally collapsed at the end of 2025 due to a broken capital chain and fell into a situation of wage arrears.

According to incomplete statistics, about 90% of AI startups failed in 2024 due to insufficient market demand and high costs, and the failure rate of AI hardware in 2025 may reach 85%.

On the other hand, some seemingly "unorthodox" products are quietly capturing the market: a laser mosquito killer that can aim at mosquitoes on its own has made middle - class families in Europe and America line up to pay; a pet collar that can translate cat meows has become a social currency among young pet owners; an AI ring for health monitoring is being snapped up in overseas markets.

On one side, capital is pouring in crazily, while on the other side, a large number of products are failing. Why has this tough sector been cracked open by these wacky designs? What should be noted when venturing into AI hardware in 2026?

Why do foreigners line up to buy a mosquito killer?

If someone had told you two years ago that the hottest hard - tech product exported by Chinese companies was a "laser cannon" specifically designed to target mosquitoes, you'd probably think it was a silly fake news story on Reddit.

But in recent days, this has not only become a reality but also driven foreigners crazy.

Photonmatrix, an AI laser mosquito killer, was launched by "Guangzhiju", a company from Changzhou.

How amazing is this thing? It can lock onto a pesky mosquito from dozens of flying insects at a distance of 6 meters, pierce its wings within 0.003 seconds, and can shoot down more than 30 mosquitoes per second.

Technically, there's no new invention: edge - side AI visual recognition + low - power pulsed laser + high - speed pan - tilt. The chip runs a lightweight recognition model locally, which is specifically trained for the wing - flapping frequency and body features of mosquitoes, so it won't harm bees. The laser power is controlled below 1 watt, making it safe for human eyes and pets. The technology combination originally used for drone target tracking has been applied to mosquito killing and achieved unexpected results.

At the beginning of the crowdfunding, the project only raised an average of $2,000 per day. The turning point came when Wang Chuan returned to his hometown in Changzhou. He recorded a real - life video of laser mosquito killing on the lawn of a riverside park and casually posted it on TikTok.

"There were only 10,000 views before going to bed. I didn't expect it to exceed 1 million views after waking up, and then soar to 20 million." The video went viral on overseas social media, and the comments from foreigners were explosive. "This is the most practical technology I've ever seen", "It should be equipped in the Amazon rainforest immediately", "There will be a day when mosquitoes are afraid of lasers."

The crowdfunding orders then skyrocketed. The original target of $20,000 was exceeded by 80 times, with a total of over $1.6 million raised. As of now, this product priced at $600 - $620 has received over 3,000 orders, with a cumulative amount of over $2 million.

The popularity of the AI mosquito killer shows that in the AI hardware circle in 2026, it's no longer just about competing on large - model parameters.

Chinese entrepreneurs now have such big ideas that they can encompass the whole universe. They are using AI as a sharp blade to unlock all kinds of niche and quirky designs.

So, what are the unorthodox ways of using AI? What kind of technical considerations are behind those mind - blowing AI hardware products?

How can niche hardware dominate the market?

If we look further, we'll find that in the past two years, the AI hardware market is not just dominated by the AI mosquito killer. More creative and innovative designs have emerged in the AI hardware sector.

Let's first look at a growing category in the smart wearables field - AI rings and pendants. Oura, a Finnish brand, is a pioneer in this field. As of 2025, its smart ring has been iterated to the fourth generation. This ring focuses on health monitoring. Through multi - modal sensors such as infrared PPG, three - axis accelerometers, and temperature sensors, it continuously tracks deep sleep, REM sleep, light sleep, and sleep efficiency, and generates comprehensive health data analysis. Data shows that Oura Ring occupied 74% of the global smart ring market in the first half of 2025, and the number of actual users has exceeded 4.4 million.

The Looki pendant, created by the former head of Meituan's smart hardware department, positions itself as a "life recorder". This 32 - gram pendant automatically takes pictures of daily life at regular intervals. AI then generates daily vlogs, serialized comics, and in - depth analysis, transforming passive recording into emotional value. This product quickly became popular after its overseas release and recently completed a Series A financing of over $20 million, led by Ant Group.

On the other hand, the AI pet collar, which has become popular in the pet market, can not only monitor physiological features such as heart rate, body temperature, and breathing rate but also accurately identify different emotional states such as happiness, nervousness, and anxiety. It can tell whether the pet is "hungry", "wants to go out", or "is in pain" and then push the information to the owner's phone, like "Your cat says you gave it too few cans today". The model of Traini, a Silicon Valley pet emotion AI company, covers nearly 120 dog breeds, and the accuracy of emotion translation reached up to 94% in internal tests.

The AI holographic family - friend box is even more touching. By uploading photos and voices of deceased relatives, it can talk to you in the relatives' tones. The underlying technology is fine - tuning of large models and voice cloning. The technology originally used in customer - service robots has been repackaged as an emotional carrier that "gives an echo to missing someone".

From 2025 to 2026, these niche AI hardware products have carved out a path in the hardware sector. Through the above analysis of these quirky AI products, we can also see the common logic behind the successful breakthrough of AI hardware under its quirky appearance.

First, they all target a scenario with clear pain points. Successful quirky AI hardware products can often accurately capture a certain demand gap of users in the physical world or at the spiritual level. The troubles caused by mosquitoes, the unique anxiety of modern people, and the inefficiency of traditional service processes are all pain points.

It's difficult to swat mosquitoes, people can't sleep well, and it's too troublesome and expensive to take pets to the hospital for health check - ups... The value of AI lies in turning things that people don't want to do, can't do well, or are too lazy to do into something interesting and simple with the help of technology.

Whether it's Photonmatrix's breakthrough in identifying tiny targets or the AI pet collar's accurate emotion recognition, these niche AI hardware products all rely on underlying "hard - core technologies".

These technological capabilities are skillfully applied to solve a specific and long - standing pain point, thus creating an obvious technological premium. Users are willing to pay a higher price for such efficient, intelligent, and sometimes even futuristic solutions.

Second, they all have a high degree of topic freshness and social - media spreadability.

These products often attract the public's attention quickly with their novel concepts and unique implementation methods. The health screenshots of the AI ring are naturally social currency. The video of the laser mosquito killer "killing 30 mosquitoes per second" is a hit material on TikTok. These products have all gone viral on social media in a short time. Market freshness is an important driving force for products to gain attention and accumulate users in the early stage, especially on crowdfunding platforms and when starting up quickly through video marketing.

On the contrary, what did those failed AI hardware products do wrong?

An AI universal remote control from a tech giant can control TVs, air conditioners, lights, and curtains, but it does a poor job in each aspect. In the end, users still picked up their original remote controls. AI Pin and Rabbit R1 made the same mistake: they tried to be all - encompassing. They wanted to solve all problems for all people but ended up solving none of them well.

Industry observers point out that many projects don't fail because of lack of innovation but because of "over - innovation". They innovate for the sake of innovation, pursue technological or conceptual disruptions, but ignore real market demand.

Where is the future of AI hardware?

At this point, a more fundamental question emerges: What is the dividing line between failed and successful products in the AI hardware sector?

The answer is actually not complicated: Between a gimmick and a real need lies the question of "how much pain the user is in".

The failures of AI Pin and Rabbit R1 are essentially typical cases of using AI for the sake of AI. Their product logic is like this: We have a cool AI technology (laser projection, large - model dialogue), so let's turn it into a hardware product and tell users "you need it". They skip questions such as what pain points users originally have, how this hardware is better than existing solutions, and whether users are willing to change their habits to use it.

In contrast, the logic of those successful products is the opposite: First, find a real and long - standing problem that troubles users, such as not being able to swat mosquitoes or understand what cats are meowing... Then use AI technology to provide an "outrageous but effective" solution.

In simple terms, for products driven by gimmicks, technology is the protagonist and users are the supporting roles. Entrepreneurs spend a lot of time answering "what we can do" but rarely ask "what users need". For products driven by real needs, users are the protagonist and technology is the tool. Based on this observation, the team can extract several more practical insights for the AI hardware sector.

First, abandon the "big and comprehensive" fantasy and focus on a very small point.

Those failed AI hardware products often try to solve too many problems. An AI universal remote control from a tech giant can control TVs, air conditioners, lights, and curtains, but it doesn't control any of them well. In the end, users still picked up their original remote controls. AI Pin and Rabbit R1 also made the mistake of "trying to do too much". They tried to replace mobile phones but couldn't even solve basic problems such as battery life and overheating. In contrast, PhotonMatrix only targets mosquitoes, not flies or moths; Traini's pet collar focuses on behavior and emotion translation. This shows that perfecting a single experience may be more effective in building user stickiness than piling up functions.

Second, prioritize user experience over technological show - off.

The team shouldn't always think about using the most expensive chips and the largest screens. An obvious trend is that the ideas for a popular AI hardware product often come from small startups with natural technological advantages rather than tech giants. Oura, a Finnish brand, gave up the screen and notification push on its smart ring to make room for sensors and batteries, resulting in longer battery life and more continuous data collection. Currently, the number of actual users has exceeded 4.4 million. Users are more willing to pay for actual experience than for technological gimmicks.

Third, verify the demand before mass - producing.

Projects such as PhotonMatrix, Traini smart collar, and Looki pendant all completed market verification through crowdfunding or pre - sales. Before large - scale mass production, crowdfunding can not only raise funds but also verify whether the target users are willing to pay for the product concept. If the data is poor after the product page has been up for a week, it means the product direction needs to be adjusted. If there is a warm response, approaching the supply chain with orders will significantly reduce the risk.

The future growth of AI hardware will highly depend on whether the cutting - edge AI capabilities can be effectively applied to vertical scenarios in consumers' daily lives that are not fully met.

This means that developers need to abandon the generalization mistake of trying to solve all problems with one device and instead focus on a single point, a specific group of people, and a particular scenario, dig deep into the pain points, and provide an excellent solution.

From the AI mosquito killer to the smart pet collar, these seemingly out - of - the - ordinary AI hardware products have carved out a possible hardware demand path. In the next five years, as the cost of edge AI chips continues to decline, the inference ability of large models continues to migrate to the terminal, and the cost of 3D vision modules gradually decreases, such "special