Original article:The AI Transformation of the Chenghai Toy Cluster: How China's Toy Capital Went from Making Molds to Making AI
If China's toy industry has a heart, it beats in Chenghai, Shantou.
This district, under 400 square kilometers, accounts for close to 70% of the country's toy production capacity. From licensed Disney figures to remote-control cars on Walmart's shelves, from blind boxes to ride-on cars for children, a substantial share of the global toy supply chain passes through here. In 2025, Chenghai's toy output value passed ¥30 billion, across more than 18,000 companies employing over 200,000 people.
But in 2026, this cluster is going through the deepest change it has seen in forty years. The arrival of the AI module is not adding one more feature to a toy; it is rewriting how the whole cluster operates.
1. The Chenghai model: its peak and its ceiling
1.1 A formula that ran for forty years
The competitiveness of Chenghai's toy industry rests on three pillars:
- A complete industrial cluster: mold development, injection molding, electronics assembly, printing and packaging, logistics and distribution, all of it within a 50 km radius. A toy can go from design drawing to packed shipment in as little as 7 days
- Extreme cost control: the scale effect of the cluster puts Chenghai's BOM costs 15%-25% below the Yangtze River Delta
- Flexible production organization: large factories take the orders, smaller ones take subcontracts, and home workshops handle non-core processes, which makes capacity extremely elastic
This model ran from the 1980s into the 2020s and met almost no real challenge along the way.
1.2 Three ceilings you cannot get around
Looking back from 2026, though, the Chenghai model is running into three ceilings:
Ceiling 1: a price war on identical products, with no way out
For a talking plush toy, you can find 300 suppliers in Chenghai. The only differences are how fine the mold is, how the plush feels and how the speaker sounds. When all competition is squeezed into the single dimension of "who is cheaper", margins get pushed down to the edge of what the workers on the floor are paid.
Ceiling 2: IP licensing fees eating the margin
More and more Chenghai factories realize they need to build brands and secure IP licenses. But the licensing fees and minimum guarantees attached to top-tier IP (Disney, Marvel, Pokémon) are now high enough to put off mid-sized factories. For a toy factory with annual revenue of ¥50 million, a one-year license for a second-tier IP can cost ¥2-3 million, which takes half the profit straight off the top.
Ceiling 3: labor costs rising irreversibly
Monthly wages for workers in Chenghai have risen from ¥3,000-4,000 in 2015 to ¥6,000-8,000 in 2026, and hiring is getting harder all the time. The younger generation would rather deliver food than work in an injection molding shop. Automated lines can replace part of it, but sewing, inspection and packing for plush toys still depend heavily on human hands.
Stack these three ceilings together and they point to one conclusion:Making it cheaper no longer works. It has to be made differently.
2. AI modules arrive in Chenghai: not a feature add-on, but a category reshaped
2.1 How one module changes what a factory can charge
The pricing logic of a traditional Chenghai toy factory is:
Ex-works price = BOM cost × 1.3-1.5 (markup)A plush toy with a BOM cost of ¥20 leaves the factory at ¥26-30, giving the factory a gross margin of ¥6-10. The wholesaler adds 30%, the retailer adds another 50%, and the shelf price lands at roughly ¥60-80.
Fit that toy with an AI module and:
Ex-works price = (original BOM cost + module BOM) × 2.0-2.5 (markup)A toy with an original BOM cost of ¥20, plus a ¥30 AI module, can leave the factory at ¥100-125. Factory gross margin goes from ¥6-10 to ¥50-75. That is not a little more profit; it is an order of magnitude more.
More importantly, the AI module lifts the product out of a pricing system based on weight. When a toy can hold a conversation, tell stories, recognize objects and remember a child's name, it stops being a plush figure and becomes a companionship device. Consumers hold an entirely different price band in their heads for that category.
2.2 Three transitions under way in Chenghai
In the first half of 2026, we saw three clear transition paths in Chenghai:
Path 1: mold maker to AI solution integrator
Some of the larger mold and electronics factories have started building their own AI integration teams. They do not write the underlying algorithms; they do a second layer of integration on top of a module solution provider's platform, Nablai's for example. They line up the speaker, battery and housing suppliers they already know, assemble a finished-goods BOM "with AI in it", and sell the whole package to brand owners.
What this model earns is an integration premium: a plastic shell that used to sell for ¥3 a piece becomes an AI finished-goods BOM selling for ¥45 a set.
Path 2: contract manufacturer to brand incubator
Some factories with OEM/ODM experience have started using their supply chain capability to incubate their own brands. The path roughly runs: take a solution from an AI module provider, assemble the finished device on their own lines, run content commerce on TikTok / Douyin, gather user feedback, then iterate the SKU.
The essence of this model is using the technology premium of the AI module to skip the long climb of the traditional contract manufacturing, white-label, distribution route and go straight to DTC (direct to consumer). Several of the AI toys that broke out on TikTok in 2026 came out of exactly this kind of Chenghai factory in transition.
Path 3: traditional electronics factory to hardware supplier for the AI module ecosystem
AI modules need specific hardware components: speakers, microphone arrays, batteries, antennas, vibration motors, touch sensors. Some of Chenghai's traditional component factories have started developing products specifically adapted to AI modules, such as a micro speaker unit tuned for the inside of a plush toy, or a low-power battery management module designed around an AI module.
These factories never touch AI algorithms or the cloud, yet their components earn a 20%-30% premium simply for being adapted for AI modules.
3. Why the Pearl River Delta: the geography of the AI module supply chain
Behind Chenghai's move into AI sits a larger geographical fact:The complete AI module supply chain has closed its loop inside the Pearl River Delta.
- Chips (Espressif ESP32 and others): Shenzhen / Shanghai fabs plus Shenzhen distribution
- PCB design and manufacturing: Shenzhen / Dongguan / Huizhou
- Acoustic components (microphone arrays, speakers): Shenzhen / Dongguan
- Batteries: Dongguan / Huizhou
- Molds and injection molding: Chenghai
- Final assembly: Chenghai
- Cloud services: Shenzhen (Nablai AMS console)
A closed-loop supply chain like this means an AI toy can go from design to mass production in 14-21 days in the Pearl River Delta. The same cycle takes 25-35 days in the Yangtze River Delta and 60-90 days overseas.
Speed matters more than usual in the AI toy racebecause AI models and content are iterating fast, and whoever can refresh SKUs more often and respond to market feedback more quickly will hold their position in the TikTok / Amazon recommendation algorithms. The Pearl River Delta's speed advantage is turning into one of the most important factors when AI toy brands choose a supplier.
4. What AI is doing to the deep structure of Chenghai's industry
4.1 From fragmented production to platform-based integration
Production in Chenghai has traditionally been extremely fragmented: a single toy is completed by 5-8 small factories working together. The mold shop makes the shell, the electronics shop makes the board, the assembly shop makes the finished product, the printer makes the packaging. This model is very efficient for conventional toys, but it has fatal flaws for AI toys:
- Firmware and AI models have to be flashed at final assembly, and fragmented production makes quality impossible to control
- Cloud activation and account binding have to be initialized before the unit leaves the factory, and traditional contract manufacturers cannot do that
- OTA updates require the device to be online, and traditional toy factories have no concept of managing devices online
This means AI toys need a more centralized way of organizing production, or at least a platform party that handles technical integration and quality control.That is precisely the core value a module solution provider like Nablai brings to the chain: packaging AI capability, a cloud platform and firmware management into an embeddable integration layer, so Chenghai's dispersed factories can each keep doing what they are good at while the output as a whole is an AI product with a technical quality guarantee behind it.
4.2 The talent mix is being rewritten
The core talent in a traditional Chenghai toy factory means mold masters, injection molding engineers, line managers and export order staff. In 2026, new roles are showing up in more and more factories' hiring lists:
- Firmware engineers (working against the solution provider's SDK)
- Content operations (running the toy's story library and dialogue templates)
- Data analysis (reading user behavior data in the AMS console)
- Social media operations (content commerce on TikTok / Douyin)
This is not one factory adjusting its headcount; the whole cluster's demand for talent is shifting. Chenghai is turning from a manufacturing base that only needed hands into a product innovation center that also needs brains.
4.3 Who gets left behind
Not every Chenghai factory will catch this train. In 2026 we are seeing a clear split:
- Medium and large factories with annual revenue above ¥50 million and OEM/ODM experience: they transition most easily. They have the capital to invest in AI module integration, mature export channels, and an existing base of brand customers
- Mid-sized factories with annual revenue of ¥10-50 million, focused on contract manufacturing: these are in the anxious zone. They want to do AI but have no technical team, and they are afraid of being left behind if they do not. Their best answer may be to buy a semi-finished AI BOM from a solution provider rather than integrating it themselves
- Small and workshop-style factories with annual revenue below ¥10 million: the window is closing. Minimum order quantities for AI modules, the technical bar for firmware adaptation and the management complexity of a cloud platform are all hard barriers for a small factory
Industrial upgrading never spreads evenly. AI will lift some Chenghai factories into the role of smart toy solution integrator, and leave others stuck in the thin-margin zone of conventional toys, until the market no longer wants that product.
5. Some possible futures for Chenghai
Possibility 1: the birth of a "DJI of toys"
If Shenzhen's hardware plus AI ecosystem produced DJI, then the Pearl River Delta's toy cluster plus the AI module ecosystem could well produce a global AI toy brand. That brand will not necessarily come from one of today's leading toy companies; it is more likely to come from a new player with supply chain depth, AI integration capability and DTC operating skills.
Possibility 2: Chenghai becomes the infrastructure behind the world's AI toys
Even if no super-brand emerges, Chenghai is highly likely to become the underlying supply chain infrastructure for AI toys worldwide. Just as today's iPhone is assembled at Foxconn in Shenzhen while the brand is Apple's, the AI toys of 2028 may carry the brand of a European or American IP owner, a Middle Eastern royal licensing house or a local Southeast Asian startup, while the module, the final assembly and the cloud platform most likely still come out of the Pearl River Delta.
Possibility 3: once AI modules standardize, Chenghai splits a second time
As AI module solutions gradually standardize, the way Bluetooth modules have, the room to differentiate on the module itself will narrow. At that point, competition between Chenghai factories will swing back from "who can do AI" to "who has better quality, shorter lead times and lower costs". The factories that used the AI wave to upgrade their brand and build channels will win; the ones that only rode a round of technological tailwind without accumulating any brand equity will end up back where they started.
6. Closing
Chenghai's story is, in fact, a microcosm of the upgrading of Chinese manufacturing.
Forty years ago, Chenghai started out on being able to make anything. Twenty years ago, it grew on making things cheaper than anyone else. Today, it is redefining its value on making things with AI in them.
What the AI module offers Chenghai is not a technical patch that thickens the margin; it is a ticket from manufacturing to smart manufacturing. Whether a factory gets that ticket depends on the choice each owner makes today: keep waiting by the injection molding machine for the next OEM order, or go out and find a solution provider, try TikTok, and learn to read the data dashboard in the AMS console.
Those who choose the latter will probably look back from 2028 and decide it was the best call they ever made.
Learn more about Nablai LX Series AI module solution
Shenzhen Nablai Intelligent Technology Co., Ltd. — AI module specialists for smart toys
📧 contact@nablai.com.cn 🌐 www.nablai.com.cn