BEIJING, Nov. 9 (Xinhua) -- A livestream studio once synonymous with lipstick and beauty products has added a very different item to its Singles' Day shopping list: artificial intelligence (AI) services.
During a recent Singles' Day livestream, a Chinese livestreaming star sold 100,000 monthly membership cards for QwenWork, an AI-powered productivity service, in just half an hour at a promotional price of 0.01 yuan (about 0.0015 U.S. dollars) each; the subscriptions carried a combined allowance of about 2 trillion tokens.
The sale marked an unusual moment for China's annual shopping festival. The discounted monthly membership gave consumers access to AI-powered tools for everyday tasks. In the livestream, the familiar shopping slogan "Everybody, buy it" gave way to a new pitch: "321, tokens up."
The livestream was part of a wider push to bring AI services onto mainstream e-commerce platforms. Since launching its token recharge center on Sept. 3, Tmall has brought in AI service providers including DeepSeek, Alibaba Cloud, Zhipu AI and Kimi. MiniMax has opened its first official flagship store on the platform, while QwenWork and StepFun have also joined.
The trend has gained traction during this year's Singles' Day campaign. Transaction value in Tmall's token category soared 450 percent month on month as the platform's sales campaign kicked off on Oct. 20, according to a Tmall representative.
Tokens are the basic units for measuring large language model inference, and providers use the number consumed to calculate charges. Until recently, token consumption was billed through AI providers' own systems according to usage, leaving users with little information about unit prices and few ways to compare costs. Putting tokens and AI subscriptions on e-commerce shelves, notably, is changing how consumers access and pay for these services.
AI ENTERS EVERYDAY LIFE
Who is buying these services? According to a Tmall representative, the main customers are individuals and small businesses with clear AI needs, including households, people using AI for learning and research, and one-person businesses. Many are already regular users of large language models but previously had to pay different providers separately based on usage. E-commerce platforms now allow them to compare prices in one place and purchase tokens according to their needs.
Among the buyers was Zhang Ji, an office worker at a tobacco processing company in east China's Shandong Province, who purchased a QwenWork membership during the livestream. She has used Qwen to prepare work-related documents since May and has also tried Doubao and DeepSeek. "I thought it was normal to pay for access to more AI capabilities, but never imagined seeing tokens sold in a livestream. The price was so low that I bought one right away," she said.
Her use of AI goes beyond work. Before traveling to Singapore with her parents and young child in August, Zhang turned to Qwen for help planning the trip, including transportation from Singapore Changi Airport to the hotel, attractions suitable for a five- or six-year-old, reservations and routes between destinations. The tool offered several options based on her family's needs.
"You tell AI what you need, and it offers different options, but the final decision is still yours," she said. Instead of spending half a day planning the trip herself, Zhang saved time to spend with her family.
Other consumers are putting AI to work in their hobbies and businesses. Su Guoqing, a short-video enthusiast in Xining, northwest China's Qinghai Province, stocked up on tokens during the shopping festival to improve his videos. He said more complex scene creation and action storyboarding required more tokens than simpler productions.
Businesses are also taking advantage of the discounts. Wang Daoxing, executive director of an AI technology company based in Chongqing Municipality in southwest China, said his firm spends more than 2,000 yuan a month on tokens for software development, presentations and promotional images. During the sales campaign, it purchased a year's worth of monthly memberships for Qwen's standard version at a lower price than on the provider's official website.
Fan Yubo, deputy director of the institute of economics at the Shandong Academy of Social Sciences, said consumers previously had to navigate technical jargon and complicated billing rules to access large language models. By packaging AI capabilities into monthly subscriptions for everyday work and personal use, providers are now making these services easier to understand and buy, much like topping up mobile phone credit or buying mobile data packages.
The growing range of token packages on e-commerce platforms also suggests that some consumers are beginning to include AI services in their regular spending budgets. While users once tended to rely on free versions, some are now willing to pay for more stable usage quotas and stronger multimodal capabilities. In some applications, AI is moving beyond novelty value to become a productive tool, Fan said.
MAKING AI A CONSUMER PRODUCT
Fan said low or even nominal prices are intended to encourage consumers who have never paid for AI services to try them for the first time. Whether they renew their subscriptions after using up an initial package, however, will depend on whether AI can continue to deliver stable and tangible value in everyday applications. Building lasting usage habits will be key to opening up the broader consumer market, he said.
Token retail is still at an early stage of market development. A survey released by Tencent Research Institute in May found that only 9.8 percent of AI users surveyed had paid for AI tools, while more than 90 percent still relied on free services. The 0.01-yuan membership offered during the livestream was a promotional price designed to encourage consumers to buy the membership and use its included quota, rather than the actual market price of the computing power consumed.
Fan noted that tokens are intangible, which can leave consumers with limited information about how much they have consumed and whether the service is worth the price. Making AI services easier to buy must therefore go hand in hand with making their usage and costs easier to understand.
For ordinary consumers, tokens are generally available through two purchasing options. One is prepaid credits charged according to usage, much like mobile phone credit, which users can top up as needed. The other is a monthly or annual subscription with a fixed quota. Fixed-quota subscriptions, however, may not match actual needs: consumers who buy too much may end up with unused credits, while those who buy too little may run out before the subscription period ends.
However, token services cannot simply follow the same sales model as mobile data, said Ding Ding, deputy general manager of China Unicom Data Intelligence Co., Ltd. Mobile data services benefit from greater economies of scale as their user base grows. Tokens, by contrast, represent AI-generated output, with each unit involving costs related to graphics processing units, electricity and equipment depreciation. Token services cannot follow the same low-marginal-cost model as mobile data, and overselling can lead to queues and throttling, Ding said.
Putting tokens on e-commerce shelves is only the first step. Several industry insiders said that for token services to become mainstream consumer products, challenges remain in measurement and accounting, pricing mechanisms, security controls and compliance.
Ma Shenglong, a solutions manager at China Mobile Qinghai's computing power business operations center, said token consumption for the same text could vary by 20 to 50 percent across different models. Billing rules also vary widely among providers, making it difficult to compare different packages on an equivalent basis.
Wang Jiandong, deputy director of the big data development department at the State Information Center, suggested establishing unified standards for token measurement and conversion, improving billing disclosures and introducing third-party verification.
Some consumers also want e-commerce platforms to strengthen after-sales services for tokens, with clear rules for handling unused quotas and purchase disputes, so that shoppers know what they are paying for and can buy with confidence. ■
