The Token sales on Tmall during the Double 11 shopping festival have gone viral, how far is AI from becoming a public utility as common as water, electricity and coal?

On the evening of November 2, a new term called Token appeared in the product portfolio of Li Jiaqi's live streaming room for the Double 11 Shopping Festival. 

Among the 20 trillion Token welfare packages launched that day, 100,000 monthly membership cards for the basic version of Qianwen Office were sold, with each Taobao ID limited to one order, and only new users are eligible for activation. In the live stream, the familiar beauty swatch scenes for fans were replaced by office scenario demonstrations. Li Jiaqi and Chen Yusen, CEO of Qianwen Office, demonstrated on the screen one-click document generation, sorting out children's wrong questions, and turning live streaming replays into commercial analysis reports. 

Eight years ago, in the same live streaming room, he sold 15,000 lipsticks in 5 minutes. 

In the past, such order-limited welfare activities were more about drainage, and were never expected to make profits. But this time it is different. Behind the super low-price monthly card is Tmall's first move to include Token as an independent category in the Double 11 Shopping Festival, and the transaction volume of this category increased by 450% month-on-month on the first day of sale; enterprises including Zhipu AI, Alibaba Cloud, and MiniMax have queued up to open their official flagship stores on e-commerce platforms. This business looks very similar to the traffic packages that telecom operators have sold for 30 years, which turns computing power into a prepaid commodity just like phone bills. 

The problem is that computing power is not traffic. This statement determines that the Token launched in Double 11 is not just a "big news", but the real beginning of the popularization of AI consumption. Then, it is worth paying attention to why manufacturers are willing to offer discounts, why platforms are willing to set up shelves, and what real value the 100,000 monthly cards purchased by users have for the sinking of AI in China, all of which need to be counted one by one. 

01. This business was already operated by telecom operators 30 years ago 

Token is the unit of measurement for large models to process information, and also the billing unit for AI services. Every step of the model, including reading requirements, reasoning, and generating responses, consumes Tokens. There is no difference in business logic between this and consuming MB of traffic every time you watch a video on your mobile phone. What large model manufacturers moved to Tmall is exactly the shelf solution verified by telecom operators, which quantifies and packages invisible resources into prepaid products that can be purchased monthly and deducted at any time. 

On September 2, Zhipu AI's first official flagship store landed on Tmall. On September 3, Tmall launched the Token Recharge Center, connecting Alibaba Cloud, Zhipu AI, Kimi, MiniMax, and DeepSeek. On October 13, MiniMax opened its official brand flagship store. 

Up to now, Alibaba Cloud Bailian, Qianwen Office, Qoder, Moonshot AI, StepFun and other platforms have fully settled in the Tmall AI Token Station, making computing power and model services commodities that can be directly ordered. Before the Double 11 Shopping Festival, monthly cards, quarterly cards, annual cards, card passwords and direct top-up services were all ready. 

Zhipu AI's Coding Plan is priced from 118 yuan to 1078 yuan, compatible with more than 20 mainstream Agents such as ZCode, ClaudeCode, and Codex; MiniMax's M Plan is divided into three tiers: Go, Explore, and Build, with a 50% discount for the first month and a monthly subscription starting at 24.5 yuan; the packages generally retain both 5-hour rolling window and weekly quota dual traffic restrictions, which is exactly the same as the speed reduction mechanism after exceeding the limit of traffic packages. 

Si Dengyu, General Manager of 3C Digital at Tmall, put it plainly: In the past, there was always a gap between consumers and the official websites of manufacturers. "Today we put the standardized store e-commerce transaction system on the platform, allowing consumers to complete consumption just like recharging phone bills and buying memberships, which is the first step we must take." 

But this analogy only holds on the shelf, not in the cost structure. 

This is because the marginal cost of traffic approaches zero. The larger the user scale, the more sufficient the statistical multiplexing, and there is almost no cost for overselling. Tokens are different, every part of them has a cost. The cost can be reduced, but cannot approach zero. Each Token corresponds to GPU, power and depreciation, and cannot be reused at low cost. Once oversold, it will lead to queuing and traffic restriction. Earlier, Ding Ding, Deputy General Manager of Data Intelligence of China Unicom, judged that it is difficult for Tokens to simply apply the logic of buying traffic. It is precisely because of the rigid cost that the seemingly obstructive traffic restriction mechanism is essentially a product of cost constraints, not a preference for product design. 

The logic of discount is also determined accordingly. Data from the China Academy of Information and Communications Technology shows that the average daily consumption of domestic tokens in June 2026 was about 500 trillion, nearly 4 times that of December 2025. During the same period, the normalized average price rose from 6.8 yuan to 13.46 yuan, almost doubling. 

In the market where both volume and price are rising, manufacturers are willing to offer discounts to lock in certain prepaid cash flow, and exchange gross profit for certainty. 

Electricity is sold by kilowatt-hour, traffic is sold by GB, and now computing power is sold by Token. Once the unit of measurement can be selected by package, the technology will change from professional supply to mass consumption. However, before the market is established and the measurement standards are unified, there is friction in the sinking process. The Token consumption of the same text in different models may differ by 20% to 50%, and the billing rules of each manufacturer are varied; the price gap is also huge: calculated per million Tokens, the input of DeepSeek-flash in idle hours is as low as 1 yuan and the output is 4.5 yuan, the input of Zhipu AI GLM-5.3 is 8 yuan and the output is 28 yuan, the input of Moonshot AI kimi-k3.1-pro-highspeed is 26 yuan and the output is 136 yuan, and the difference between the most expensive and the cheapest is dozens of times. 

The dual non-unification of measurement and pricing is an unavoidable hard problem in this business, and settling in e-commerce may be the optimal solution. 

02. Official websites sell capabilities, while e-commerce sells trust 

Before Tokens were launched on Tmall, this business has always existed, but with thresholds. 

Users need to register on the open platform of the official website, complete real-name authentication, create an API Key, and fill the key into the code to get the basic conditions for use. This logic splits "buying Tokens" and "using Tokens" into two separate things, and each step screens out non-technical users. 

In this scenario, e-commerce compresses this process into three steps: selecting a package, making payment, and one-click binding, which shortens the process and makes it more convenient. The API Key is replaced by monthly cards and annual cards. Low threshold is only the admission ticket, and the deeper value lies in the other two layers. 

One layer is the transfer of pricing power.

In the past, the price of Tokens was posted in the API documents of each manufacturer, and only developers would read them, while developers were generally not very sensitive to prices; now the packages of Zhipu AI, MiniMax and Qianwen are displayed on Tmall pages, and the price comparison pressure returns from the demand side to the supply side. 

As of the first half of 2026, the number of generative AI users in China has exceeded 700 million, with a penetration rate of over 50%, but the habit of C-end users paying for AI tools is far from established. For these users, comparable price itself is the cultivation of consumption habits. Tmall has about 64 million 88VIP members, whose per capita annual consumption is 9 times that of non-members, and they purchase an average of 19 categories and services, making them the first batch of retail customers targeted by this business. 

Every search and every page stay of consumers is a mental touch that manufacturers cannot buy even with large advertising investment. 

The other layer is the information revealed by the channel structure.

For the same move of launching products on e-commerce platforms, Zhipu AI, MiniMax and Alibaba Cloud choose self-operated flagship stores, which means they choose to independently undertake product listing, customer service reception, pricing strategies and marketing activities, and also need to pass the platform qualification review. It is understood that Zhipu AI has even recently started recruiting "brand narrative managers"; Kimi and DeepSeek choose to only authorize agency distribution, and keep pricing, delivery and after-sales services within their own systems. This shows that the first three manufacturers are trying to do long-term brand operation, while the latter are testing the water through channels. The division of the two models corresponds to the differentiation of the organizational capabilities of manufacturers. 

In addition, the qualification review, order retention, after-sales consultation and user evaluation of the e-commerce platform provide delivery certainty endorsement for the "invisible and intangible" virtual consumption of Token purchase. For a consumption where payment is made first and the quota is stored on the manufacturer's server later, what consumers are willing to trust is the payment habits and platform after-sales service they have used for 20 years, rather than a promise on the official website page. 

The farther the virtual commodity is from being "visible", the higher the premium of platform credit. 

For consumers, the value is very specific. Buying Tokens on Tmall is backed by the platform for after-sales services, there are exclusive packages for e-commerce channels to choose from, and you can also enjoy great discounts during the Double 11 promotion. The official website price of Qianwen AI Lite monthly card is 60 yuan, and the promotional price is 39 yuan; the official website price of Qoder professional version is 88.99 yuan, and the promotional price is 58.98 yuan; there is no need to change the consumption habits that have been used for 20 years, and you don't have to take a technical course first before deciding whether to pay for AI. 

There is a detail that in the Taobao Recharge Center, "Recharge Token" has been listed next to recharge phone bills, recharge traffic, and recharge utilities. Putting Tokens in the queue of daily consumer goods is Tmall's strategic positioning for this business. 

03. The value of 450% growth needs to be calculated separately 

Data shows that on the first day of Double 11 sale, the transaction volume of Tmall's Token category increased by 450% month-on-month; 48 hours after Zhipu AI settled in Tmall, brand searches skyrocketed 50 times month-on-month, driving the transaction volume of AI subscription products on the platform to increase by more than 160%. 

In Li Jiaqi's live streaming room on November 2, 100,000 monthly cards for the basic version of Qianwen Office were sold in half an hour. However, the high growth rate includes the factor of low base. It should be noted that this category was only launched in September, and the high month-on-month growth compared with last month is the natural curve of new categories, which is more symbolic than actual transaction volume. The account of Li Jiaqi is more worthy of detailed calculation: each Taobao ID is limited to one order, and only new users can activate it. The so-called "20 trillion Tokens" does not correspond to revenue, but 100,000 potential paying users; the return depends on the renewal rate of monthly cards at the original price from the second month onwards. 

What Li Jiaqi sold is not just 20 trillion Tokens. The retention rate of AI office users is the real denominator of this user acquisition account. 

The signals in the financial statements and industrial chain structure are more solid. 

In the first half of 2026, Zhipu AI's Token call revenue was 825 million yuan, a year-on-year increase of 2735.7%, accounting for 86.5% of total revenue; under the current situation that most model companies still rely on financing for blood transfusion, Token revenue can support the financial statements of a company. The actions of the industrial chain are also in the same direction. In March this year, Alibaba established the Token Hub Business Group directly responsible by the group CEO. Tencent Cloud upgraded the MaaS platform to Token Hub in the same month. In May, the three major telecom operators collectively launched Token packages. From cloud vendors to telecom operators to independent model manufacturers, the entire industrial chain is operating the same thing as standardized commodities, and Double 11 is just a scene visible to the public for this operation. 

The significance also extends to hardware. 

This Double 11, AI PCs, AI glasses, and AI headphones are all unveiled intensively, but hardware without model capabilities is just a concept. Si Dengyu said, "Now some brands are already launching AI PCs and AI mobile phones. If these are separated from Tokens or large language models, they are just concepts... We connect this to hardware products, and in the future, consumers can do hybrid deployment, which leaves huge room for imagination." 

When consumers buy both hardware and Token Plan, it means that besides the profit from purchase and sales price difference, hardware brands will have a second growth curve of subscription revenue. 

As for the significance of buying Tokens on Tmall for consumers, we need to analyze it calmly at different levels. Geeks and heavy users will not be locked by packages. Wang Zimin, who makes AI videos in Beijing, spends 30,000 to 40,000 yuan on Tokens every month, and chooses official channels or e-commerce channels depending on who offers more "discounts". 

E-commerce packages actually serve medium and low-frequency C-end users, family users, learning and research users, and one-person companies. According to Si Dengyu's judgment, the covered population will start from geeks, then to office white-collar workers, then to ordinary consumers who are passionate about technology, and finally to the general public. The current user structure is exactly at the second link of this expansion chain. 

In other words, e-commerce is not a substitute for official websites, but adds a retail layer to the Token market, reaching people who would never open the manufacturer's official website in the first place. 

The wholesale market is billed by volume, and the retail market adopts package subscription, and the two layers solve their own problems respectively. 

Future packages will be more detailed. According to the vision of platforms and manufacturers, Token packages will be like beauty brands matching sets for different crowds, combining products around the needs of study, family, one-person companies and other scenarios, turning AI usage needs into products that are easier to choose. But at the moment, there are also real difficulties. The measurement standards are not unified, and it is difficult to convert between different packages. 

Wang Jiandong, Deputy Director of the Big Data Development Department of the State Information Center, has suggested that the Token measurement standard should be unified as soon as possible and third