The Spring Festival AI battle is just a display? JPMorgan: Token consumption is about to enter a period of rapid growth, possibly increasing 370 times in five years!

``` JPMorgan Chase pointed out that although major technology giants launched fierce AI application promotion campaigns during the Spring Festival, this is only a surface phenomenon. Beyond the short-term contest of "who will win the AI application battle," the market should focus on a deeper trend: consumers' ways of obtaining information and consuming content are undergoing a structural change, which will drive token consumption into a multi-year period of rapid growth. By the 2026 Spring Festival, Tencent, Alibaba, and Baidu will compete for users' minds with subsidies totaling billions of RMB. According to Wind Trading Desk, on February 6, JPMorgan Chase analyzed in its latest China Internet and AI Industry Report that these large Internet platforms are spending real money to accelerate users' migration from traditional search to chatbots and to develop intelligent commercial habits. This process is actually driving up the volume of inference, thus speeding up token consumption. The current GPU computing power shortage or supply-demand imbalance is expected to persist in the coming years. The bank predicts that from 2025 to 2030, the compound annual growth rate of token consumption in China will be as high as about 330%. Based on this, JPMorgan believes it is too early to judge the winners and losers in the AI application sector, and that the more certain investment opportunities lie with the "second-order winners." These include direct beneficiaries of China's AI infrastructure (such as Alibaba and Baidu) and companies with exposure to online advertising businesses (such as Kuaishou). For investors, China's AI infrastructure has become one of the most attractive investment themes in the TMT sector. Headline Offensive by Tech Giants During Spring Festival: Fighting for User Habits, Not Just Traffic Technology companies will adopt distinctly different strategies during the Spring Festival, with the core goal being to seize user entry points and mindshare in the AI era. According to JPMorgan's observations, Tencent has adopted a classic "growth first" strategy, investing 1 billion RMB in red envelope campaigns. By leveraging WeChat's social channels to lower the user threshold, its goal is to rapidly increase the installation and activation volume of "Yuanbao." However, this campaign also exposed the boundaries of platform governance, as some induced sharing behaviors led to link restrictions. Analysts believe that in the future, Tencent may need to transition from link-driven viral spreading to product-driven retention cycles, integrating AI into the WeChat ecosystem in a smoother, user-experience-first manner. Alibaba's strategy shows structural differences. Alibaba announced a 3 billion RMB "Spring Festival Treat" plan, integrating Taobao, Alipay, Fliggy, and Amap. JPMorgan points out that Alibaba's goal is not only to increase the adoption rate of "Tongyi Qianwen" but also to foster commercial habits of initiating transactions from AI interfaces. The strategic goal is to shorten the consumer funnel and improve conversion efficiency. If these effects persist after the holiday, it will help Alibaba gain upstream advantage in high-friction areas such as travel planning. By contrast, Baidu invested about 500 million RMB, continuing to adhere to a "AI Assistant + Search Distribution" integrated strategy, seeking to embed AI in more intent-driven search sessions. ByteDance, while not announcing any large-scale cash subsidies, relies on product iteration and the distribution capabilities of Douyin and other ecosystem entry points; its Doubao series of models reportedly has reached 100 million daily active users. JPMorgan believes that even though the monetization model is still uncertain, the cost of losing mindshare at the traffic entry layer is extremely high. This is the key reason why tech giants are willing to "buy time" through marketing. Token Consumption Outlook: A Five-Year Infrastructure Feast with 370x Growth JPMorgan gave a staggering growth forecast in its report: China's AI inference token consumption is expected to grow from about 1,000 trillion in 2025 to about 39,000 trillion in 2030, an increase of about 370 times over five years. This explosive growth is mainly driven by two factors: first, increased AI penetration in consumer and enterprise workloads; second, an expansion of application scenarios from simple conversational AI to complex intelligent agents and multimodal outputs. Data shows that as of December 2025, ByteDance's Doubao series models' daily token consumption has exceeded 50 trillion, a year-on-year increase of more than tenfold. Structurally, the makeup of inference demand will undergo significant changes. Conversational AI's share of total consumption is expected to drop from nearly half in 2025 to a high single-digit percentage in 2030. Instead, "knowledge worker AI agents" and "multimodal AI" will take over. The former, involving multi-step reasoning and document workflows, is expected to account for one-third of total consumption by 2030; the latter, with the increase of image, audio, and video workloads, is expected to grow from the current 15% to nearly 40%. It is worth noting that JPMorgan reminds investors that token growth is not linearly equivalent to chip demand growth. With improvements in inference performance brought by the new generation of hardware such as NVIDIA's Blackwell (e.g., 4 times Hopper), and optimizations in the software stack, tokens processed per GPU hour are also significantly improving. Risk Warning and Disclaimer The market carries risks, and investment should be approached with caution. This article does not constitute individual investment advice, nor does it take individual users' specific investment objectives, financial status, or needs into account. Users should consider whether any opinions, views, or conclusions in this article fit their particular circumstances. Investment decisions made based on this are at your own risk. ```