If your fourth-quarter category plan still lists only hardware and home appliances, the shelf has already moved past that list. As of 13 November 2026, JD.com said it sold more than 50 million AI-related products between 12 October and 11 November, generated over 300 million service orders, and saw searches for the word “token” rise more than 13-fold against the preceding period, with token package sales up more than tenfold. Syntun put total festival sales at 1.78 trillion yuan, up 4.8 percent. The interesting number is not the growth rate; it is that a computing credit became a shelf item.
Tokens on a product page change the listing
Prepaid token packages behave like neither hardware nor content. They have no shipping weight, no return logistics and no inventory risk, yet they need a product page with a price, a specification and a delivery promise. That forces marketplace teams to write listings for something that is consumed rather than shipped, and it forces the search layer to rank a credit against a physical good in the same results page.
The service side shows the same pressure. First-time buyers in the promotion ordered an average of six items each, and service orders ran past 300 million. Services and credits both reduce the inventory a platform has to fund, which is why they appear in the same promotional narrative as AI hardware — the growth story and the margin story point in the same direction. Consumer demand for credits is real, but it is also easier to satisfy than demand for goods, which is why the mix moved so fast in a single season.
What a brand should check before listing a credit
The specification problem
A token package has no dimensions, so the fields that normally carry a listing — size, material, compatibility — do not apply. What replaces them is validity, scope of use and refund rules. Buyers compare on those three, and a listing that omits them gets questions that turn into disputes rather than sales. The market trend here is not that credits are popular; it is that they are being sold with goods-grade page furniture.
The rank-and-compare problem
If credits sit in the same results page as hardware, comparison logic breaks: a shopper searching for a device may be shown a subscription, and an AI assistant answering the same query may treat both as candidates. Brands need to know which query types surface their credit at all, which is a measurement question before it is a marketing one.
A listing check for consumption goods
The table below is the minimum set of fields to confirm before a credit goes live. Sample values come from the reporting cited here; replace them with your own catalogue data.
| Check | Field example | What it prevents |
|---|---|---|
| Validity window | Months from activation | Refund disputes after expiry |
| Scope of use | Models, regions, devices | “Works everywhere” claims |
| Consumption metering | Units per request type | Silent overrun complaints |
| Refund rule | Unused balance policy | Chargebacks and reviews |
The third row is where most disputes start. Metering feels like an engineering detail until a buyer discovers that one long conversation consumed a month of allowance. Publishing a unit table does not make the product cheaper, but it moves the argument from “you hid the cost” to “we priced it”.
FAQ
Is 50 million AI products a reliable category number?
It is a company-reported figure covering a defined window and a wide definition of AI-related hardware. It is useful as a directional signal, not as a market size, because the category boundary is set by the retailer rather than by a standard.
Why do services and credits matter more than the growth rate?
Because they change the cost structure. Inventory, warehousing and last-mile delivery are all avoided, so each extra yuan of demand converts to cash more efficiently than first-party goods. That is why platforms now lead with them.
Does listing a credit hurt the main catalogue?
It can, if the credit competes for the same query. The practical fix is to separate query types rather than to discount: a shopper looking for a device and a shopper looking for a subscription should not land on the same listing.
What should a smaller merchant do first?
Write the four fields before spending on promotion. Validity, scope, metering and refund rule are the four questions buyers ask in the first week; answering them on the page removes most of the pre-sale support load.
