OpenAI has partnered with Mattel to develop AI?powered toys. Elon Musk's xAI debuted an "AI companion" feature. ByteDance's Volcano Engine launched the AI toy development kit "EchoEar", while Baidu's Intelligent Cloud rolled out the "AI Magic Star".

At the same time, former senior executives from Alibaba and Meituan have resigned to start their own AI toy ventures. AI toys are rapidly becoming the hottest target for entrepreneurs and investors alike.

From plush toys to robots, from educational tools to emotional companions — no breakout hit yet, but the race for dominance is underway.

Top Execs Leave Big Tech to Build AI Toys

Every day on social media platforms, new AI toy products appear one after another.

"Today this company; tomorrow that one…a constant stream of fresh AI toys," remarks Tom Culture's board secretary Ouyang Meizhu. "I receive daily messages asking to become agents or distributors." At the recent CBME Global Maternal & Infant Expo, Tom Cat's modest booth attracted over a thousand industry contacts—many seeking collaboration or overseas distribution.

Tom Cat

AI toys have ignited a market-driven entrepreneurial rush.

In early 2024, manufacturing centers in Shenzhen and Dongguan led the wave, proving that even plush toys with simply-integrated large-model chips could become online blockbusters. According to e-commerce data, AI early-education toy sales surged sixfold in January. Over 100 vendors are currently offering or preparing related products for release in the months ahead.

High demand, premium pricing, and strong profit margins have made AI toys a key new-growth lane for the toy industry.

This track was initially driven by small and mid-sized startups, but from the second half of 2024 onward, major players entered the fray. Talking Tom launched an AI companion robot based on its own IP. Shifeng Culture, in collaboration with Baidu Intelligent Cloud, introduced the "AI Magic Star". ByteDance's Volcano Engine together with Espressif rolled out the development kit "EchoEar". Global toy giant Mattel signed its own deal with OpenAI. Cross-sector entrants—from Tesla to Shanghai Film Industries—also joined the emerging ecosystem.

When incumbents begin to enter, it suggests the market has moved beyond its chaotic birth phase and is entering a product-definition era.

There is a growing wave of AI toy startups founded by former executives from major tech companies. Brands like BubblePal, Friend Rabbit, and AIiPal have entered the market under such founders. According to our analysis, these teams include ex?senior leaders from Alibaba, Meituan, Microsoft, and Anker, among others.

Big tech executives entering the AI toy sector

Though no mass-market AI toy has emerged yet, investors are already lining up. Ouyang Meizhu shared: "During a Tom Cat financing roadshow, many VCs attended—everyone was asking about AI toys, AI-companion products." Tom Cat's TikTok and Xiaohongshu channels now top the toy category, and platforms have actively offered promotional support including guaranteed traffic.

Notably, IP owners like Disney, Hasbro, and Mattel are exploring AI-enabled revamps of classic characters. Robosen's China GM Cai Zefeng revealed that IP holders sometimes proactively purchase AI toy products or invite Robosen to official brand exhibitions: "We're giving life back to IP through robotics."

Per data from IT Juzi, the AI toy sector has drawn investments from 96 institutional backers, including ByteDance Investment, Lenovo Ventures, JD Technology, CocaCola Ventures, SAIF Partners, and GSR Ventures China. Both deal counts and volumes surged in 2024.

Overview of AI toy company financing since 2024

From rapid prototyping to supply-chain transformation and VC engagement, AI toys have sparked a full-spectrum ecosystem frenzy.

AI Toys Still Await a Breakout Hit

Industry insiders note that AI toys began "running" once open-source models like DeepSeek emerged. These models lowered costs, increased intelligence—and made consumer-grade AI truly feasible. Compared to enterprise (To B) use cases, AI toys represent one of the easiest paths to consumer adoption in China's rapidly evolving AI landscape.

Although some tech giants with large language models have not directly entered the AI toy sector, they are actively laying out strategies through ecosystem partnerships. ByteDance is a typical example—it once drew attention with its internally developed AI-powered "Xiǎn Yǎn Bāo" (a Mid-Autumn Festival gift). A Volcano Engine representative told National Business Daily that the company does not plan to develop AI toy products in the future, but it will work with partners to build an AIoT ecosystem.

In June, Volcano Engine launched a full-stack AI hardware solution, centered on its self-developed embedded SDK to bridge edge devices and the cloud. Additionally, Espressif collaborated with Volcano Engine's Kouzi LLM team to release a smart AI development kit, "EchoEar" (喵伴), which enables full-duplex voice interaction, multimodal recognition, and agent control. The kit is designed for use in toys and other voice-interactive products that require large model capabilities.

EchoEar (喵伴)

Haivivi plans to unveil CocoMate in August, powered by an end?to?end emotional response architecture integrated with Volcano Engine. "When the model can sustain dialogue and understand emotion, the dream of 'making toys talk' becomes commercially real," says Haivivi marketing head Zhong Wenjie. The platform believes its system is among the first to enable seamless voice?to?emotion response in consumer AI toys. Competitors are now following suit.

Ouyang remarks that existing large models sufficiently support toy interactions, and at times, output must be restricted to ensure concise, child?appropriate responses. "The industry's missing ingredient is a breakout product. As soon as one AI toy sells in the millions, everything will shift."

"Right now, all runners are watching the televised race—not yet storming the finish line," Ouyang said, using a track analogy to describe the sector's early stage.

Consumer perception still lags—many parents only worry if the toy can "tell stories". Few recognize deeper value: shared creativity, memory, emotional companionship. Most toy products represent only superficial AI "shells"—without tailored scenario logic or nuanced model tuning.

Interviews reveal the industry is now customizing vertical models focused on long-term memory, multimodal responsiveness, ethical safety, and emotional stability. According to Cai Zefeng, generic large models may too often hallucinate or mishandle sensitive child interactions; vertical specialization is vital for safe, regulated companion use.

AI Toys Face a Defining Moment

In the early days of any breakthrough, uncertainty prevails. Just as few could foresee the rise of the automobile in the age of horse-drawn carriages, the definition of an AI toy remains elusive even as the market accelerates。

Cai further clarifies: "Some companies have tech but no IP; others have IP but no tech." Product leadership now hinges on combining tech, creative content, and emotional insight.

Contemporary product categories range from plush, story machines, and robot toys, to AI pets. Markets vary—education, entertainment, emotional companionship. Cai identifies three clear sub-paths: AI children's toys, AI pets, and small-bodied intelligent robots.

Ouyang emphasizes that winners won't just attach a chip to a plush toy—they'll emotionally respond to children's cues. "In toys, users are kids but buyers are parents. To succeed, a company must understand the emotional needs of families," she says. That requires deep data accumulation and long-term service experience—an advantage most tech-only startups lack.

She adds: "AI toys record past interactions and adapt to personal preferences. They evolve. It's not powerful models that win—it’s emotionally intelligent products."

Ouyang sees today's market as a data-acquisition window: "Whoever captures users first gets the most data. Better models, richer content, more precise iteration." As a result, many companies invest heavily in marketing to get early traction, then use feedback to refine.

Robosen, with 20 years in robotics, touts its hardware advantage—its latest Buzz Lightyear product alone integrates 23 joints, 75 chips, and micro-servos to deliver realistic micro-expressions. Cai asserts: "These systems are difficult for conventional toy companies to replicate quickly."

Hardware automation enables autonomous transformation. AI brings voice control and responsive motion. "If I say 'come closer', the toy activates every joint and walks toward me," Cai notes. AI toys could rewrite industry logic altogether.

While Robosen envisions AI toys as a fusion of entertainment, companionship, and education, some experts caution that assigning too many functions—like teaching or caregiving—may dilute the essence of what a toy is meant to be.

A Trillion Yuan Market Blueprint, China—The Innovation Engine

"AI companionship is a trillion-yuan opportunity," Ouyang says. According to HUAAN Securities research, the global toy market reached ￥773.1 billion in 2023 and is expected to hit ￥993.7 billion by 2028. China contributed ￥104.9 billion in 2023 (14% global share); projected to grow to ￥165.5 billion by 2028, with a 9.5% CAGR.

Compared to traditional toys, AI toys command higher prices and margins. According to Douyin e-commerce data, pricing tiers below ￥30 dominate volume sales, and ￥10–￥300 contribute most GMV. AI toys, in contrast, often cost multiple times more—one Talking Tom plush costs about ￥69, versus an AI robot at ￥1,799; margins typically range 70–80%, with top-tier models exceeding 90%.

Pricing power corresponds to more advanced functions—capacitive sensing, head-tracking, sophisticated displays. "Speech is simple—but responsiveness is expensive," quips Ouyang. Japan's premium AI? companion LOVOT, featuring full-body capacitive feedback, retails for over ￥30,000—proof that consumers will pay for emotion-enabled design.

Guangdong's Department of Industry & Information Technology projects AI + trend toys could add over ￥100 billion in industrial output. Shenzhen—home to 14.4% of China's collectible toy manufacturers—is emerging as a key innovation zone.

AI toys are expanding beyond children into adult demographics, branching into two main trajectories: collectible "toy-style Pop Mart" items and family companion robots. Ouyang predicts: "Future frontline toy brands will all embrace AI."

According to a research report by HUAAN Securities, China is one of the world's largest toy producers and consumer markets. In terms of production, China manufactures around 70% of the world's toys, with plush toys accounting for approximately 75% of the global market. In terms of consumption, China represents nearly one-sixth of the global toy market. The country also boasts a mature traditional toy supply chain. Currently, many AI toys are built using a simple combination of plush exteriors and embedded AI modules, which require relatively low manufacturing complexity and can be produced quickly at scale.

A representative from a global toy brand told National Business Daily that the Chinese market is of strategic importance.

Ouyang Meizhu believes China is best positioned to produce the first breakout AI toy. "AI toys are an application-layer product, and China is the world’s largest application market," she said. The country's massive user base offers fertile ground for consumer success. At the same time, high user activity and diverse use cases provide the data needed for continuous model optimization.