4.8 percent growth against 14.2 percent a year earlier is the whole Singles Day story told in two numbers, and through November 13, 2026 that gap is the only figure category teams can compare across platforms. Sales reached 1.78 trillion yuan in the October 7 to November 11 window, neither Alibaba nor JD.com published a headline gross merchandise value, and both pointed instead at categories and engagement. That substitution is the story: the festival is being reported in a different unit than the one it grew up with.
The growth number lost its denominator
A growth rate needs a stable denominator to be read as demand. When the shopping window stretches from one night to five weeks, the same 4.8 percent can describe a market that is holding steady or one that has simply moved spending earlier. The number is not wrong, it is answering a narrower question than the one investors and category teams are asking.
The practical consequence is that the unit of comparison has shifted from the event to the category. Platforms now publish category leaders, brand counts and engagement metrics because those survive the change in window length. For a brand, that means the annual review has to be rebuilt around category growth and repeat rate inside the window, not around a position on a leaderboard that no longer exists. A single consumer demand total cannot carry that review, because it blends hardware, services and metered usage into one line that no team can act on.
AI products are a category, tokens are a basket
Hardware doubled, services multiplied
JD reported that transaction value for AI PCs, workstations, cameras, glasses and recording devices more than doubled year over year, and that the platform generated more than 300 million service orders. Those are two different kinds of growth: hardware is a replacement cycle that a category team can forecast, while services are an attach rate that depends on how the purchase is packaged. Treating them as one AI number hides which of the two is carrying the quarter.
Tokens turned computing into a stocked item
The more interesting shift is that computing power became something a shopper can put in a cart. China Daily reported that a single livestream session sold about 100,000 monthly subscriptions to an AI office service in thirty minutes, equivalent to roughly 2 trillion tokens, and that token sales on Tmall rose more than 700 percent between mid October and early November. Large model developers opened stores on the platform, which means the category now has listings, discounts and stock keeping units like any other product.
Once a subscription behaves like a stocked item, it inherits the metrics of one: replenishment cycles, discount depth and churn. The difference is that consumption is metered, so a promotional token package can be used over months while the revenue is booked in the window. That mismatch is the reason token sales are a poor proxy for demand and a good proxy for how aggressively platforms are buying habit.
What category teams should read
| Signal | Reported | What it supports | What it cannot show |
|---|---|---|---|
| Total sales | 1.78 trillion yuan, up 4.8 percent | Market size inside the window | Whether demand moved earlier |
| AI hardware | Transaction value more than doubled | Replacement cycle is live | Repeat purchase rate |
| Service orders | More than 300 million at JD | Attach rate is rising | Margin on each order |
| Token volume | Searches up thirteen fold | New stocked category | Revenue timing |
| Brand cohort | 333 young brands led categories at Alibaba | Entry costs are falling | Durability after the window |
The table is deliberately split between what a number supports and what it cannot show. Most post event reporting collapses the two columns, which is why the same figures get used to argue both that consumption is weak and that a category is booming. Both readings can be true at once when the window and the category move in opposite directions.
There is also a sourcing point that category teams tend to skip. AI hardware demand is concentrated in a small number of brands with supply contracts, so a doubling in transaction value mostly reshuffles share inside that group. A brand outside it learns more from attach rates on services than from the hardware total, because services are where a mid sized player can still enter without a factory.
Three reads before budget season
The three reads below are meant to be done in the order listed, because each one changes the base for the next. The first fixes the window, the second fixes the unit, and the third fixes the timing of revenue. Skipping the first makes the other two look precise while resting on a comparison that the calendar already moved.
FAQ
Is 4.8 percent growth weak for Singles Day?
It is slower than the 14.2 percent of the previous cycle, but the window also lengthened, which pulls spending forward from adjacent weeks. Comparing the same number of days is the only way to separate a slower market from a longer calendar, and few published figures do that.
Why do token sales matter for a product category?
Because they turn computing into a stocked item with listings, discounts and replenishment. Once that happens, the category behaves like any other subscription product, with churn and discount depth deciding profitability rather than the headline volume sold during the event.
Should a mid sized brand chase AI hardware?
Usually not directly. The hardware cycle rewards supply contracts and brand concentration, while services reward packaging and attach rate. A mid sized brand is more likely to find category growth in bundling a service with an existing product than in entering a hardware category late.
