During the 2026 Spring Festival, China’s tech giants threw real money into the AI battlefield and kicked up a haze of smoke. Tencent Yuanbao splurged RMB 1.0 billion in cash red packets (USD 144.12 million), Alibaba’s Qwen followed with RMB 3.0 billion in “free orders for daily life” (USD 432.35 million), while ByteDance’s Doubao took a different route and secured the exclusive AI cloud partnership for the 2026 CCTV Spring Festival Gala—together, the three giants invested more than RMB 4.5 billion (USD 648.53 million). While ordinary users were immersed in the revelry of “drinking milk tea for RMB 0.01 (USD 0.0014)” and the excitement of “shake to claim red packets,” this seemingly simple marketing-and-subsidy battle is, in fact, a critical turning point as AI applications shift from a “technology race” to a fight over “entrances.” Behind the red-packet war are three bets by Yuanbao, Doubao, and Qwen on the ultimate form of AI: subsidies in exchange for time, mindshare in exchange for space, and scenarios in exchange for habits. When the fireworks fade, who will be able to leave an irreplaceable spot on users’ phones?
Yuanbao’s Blitzkrieg: The Paradox of Social Viral Growth and Subsidy Dependence
Tencent Yuanbao chose the most familiar and most direct path—replicating the classic 2015 WeChat red-packet surprise attack on Alipay. RMB 1.0 billion in cash red packets (USD 144.12 million) was deeply embedded in the WeChat social relationship graph: to receive red packets, users needed to share and needed friends to assist; every click reinforced the so-called “moat of social viral growth.” Data shows that during the Spring Festival, Yuanbao’s peak daily volume of generated share links exceeded 200 million times; the WeChat Index surged 50.12% day-over-day on February 1, and on February 2 downloads jumped from 210,000 to around 300,000. This strategy of “using capital density in exchange for time acceleration” did indeed push users from “knowing” to “downloading” in the shortest possible time.
Yet the ghost of history still hovers. In 2019, Baidu’s Spring Festival Gala red-packet campaign once set a record with daily active users soaring from 160 million to 300 million, but after the carnival, the 7-day user retention rate was under 2%. Red packets attract “price-sensitive tourists,” but AI survival requires “value-dependent residents.” Even more subtly, when Yuanbao’s red-packet links and sharing passcodes were restricted inside WeChat, this “left-hand-versus-right-hand” dynamic exposed Tencent’s internal strategic tension—WeChat both needs to funnel traffic to Yuanbao and must also guard against excessive marketing eroding the user experience. Subsidies can buy an entrance, but they cannot buy staying; social viral growth can create noise, but it cannot guarantee retention. Yuanbao’s all-in bet lies in whether its content ecosystem (Channels, Official Accounts) can catch this wave of traffic and complete the transformation from “transaction” to “relationship,” but that road is destined to come with an astonishing loss rate.
Qwen’s Infiltration War: The Ambition to Go from a Tool to an Operating System
Alibaba Qwen’s approach is clearly more intricate. RMB 3.0 billion in “free orders for daily life” (USD 432.35 million) is not simply handing out money; it deeply binds AI agents to local lifestyle services—users can order takeout and book hotels through the Qwen App while paying only RMB 0.01 (USD 0.0014). Within 9 hours of the campaign going live, order volume surpassed 10 million orders, directly causing app crashes and overwhelming milk-tea shops with orders. This gameplay of “AI orders with one sentence” is, in essence, cultivating the shortest-path habit from “intent” to “satisfaction.”
Qwen’s open strategy is to become the “conversational operating system” between users and Alibaba’s digital economy ecosystem. When you say “order milk tea,” it mobilizes Ele.me; say “buy New Year goods,” it connects to Taobao; plan travel in the future, and it can seamlessly coordinate Fliggy, Amap, and Alipay. Compared with Yuanbao’s purely acquisition-driven logic, Qwen’s “service equals retention” strategy builds higher switching costs—once users are accustomed to using AI to manage eating, drinking, entertainment, and travel, migrating means giving up an entire set of already-validated efficient habits. But this path is also the most perilous: Qwen’s share links were quickly blocked inside WeChat, showing that when AI goes deep into transactions and lifestyle services, it has already touched the “core territory” of super-platform defense. Qwen is carrying out a difficult “Normandy landing” within the “national-level app” matrices of Tencent and ByteDance.
Doubao’s Positioning War: The Spring Festival Gala IP and Long-Term Mindshare Bonds
Unlike the “money offensive” of the other two, ByteDance Doubao chose zero cash red packets, but it secured the status of exclusive AI cloud partner for the 2026 CCTV Spring Festival Gala. The brilliance of this move lies in upgrading “attention economics”—red packets buy short-lived “click behavior,” while binding with the Spring Festival Gala buys the “collective unconscious memory” of prime time on Lunar New Year’s Eve. When AI-generated custom couplets appear on the Gala screen, or when the whole family participates in AI riddle-guessing interactions, “Doubao” becomes deeply tied to high-emotion symbols like “New Year” and “reunion.” Data shows that in the week after the Spring Festival, searches associated with “Doubao + Spring Festival Gala” surged 650% week-over-week.
Doubao is issuing a “long-term mindshare bond”: it does not seek an immediate spike in users, but rather aims to establish a priority akin to a “default option” in users’ future AI choice lists. The long-term value of building this kind of “cognitive convenience” may far exceed one-time user acquisition. But the challenge lies in the “high-risk leap from cognition to habit”—the Spring Festival Gala can make hundreds of millions of people “know” Doubao, but it cannot force them to “use” Doubao. Turning brand awareness into daily tool dependence is separated by a massive “behavior-change chasm.” Doubao must prove that it is not only a “smart toy” for the Spring Festival Gala, but can also become an “efficiency co-pilot” in users’ work.
The Talent Shadow War
While the Spring Festival AI red-packet war sparked a nationwide participation craze, another war without smoke was unfolding in parallel. On February 5, job postings on the workplace social platform Maimai showed that Alibaba’s “Qwen App User-Growth Algorithm Engineer” role offered an annual salary up to RMB 1.28 million (USD 184,470). Tencent Yuanbao’s user-operations and campaign-operations roles offered annual salaries above RMB 750,000 (USD 108,088). ByteDance’s “Doubao AI Application Engineer” role offered an annual salary close to RMB 1.0 million (USD 144,117). The AI job types posted by internet giants on Maimai span product, operations, growth, R&D, algorithms, and more. On one side: tens of billions of RMB in cash fighting for users; on the other: million-RMB annual salaries fighting for talent.
This talent shadow war reveals the essence of AI competition: technical strength and product experience are the decisive keys to victory. No matter how much money is spent, without excellent talent to support it, even more users may only be a flash in the pan. Tencent is sparing no effort in its talent investment for Yuanbao, which is internally positioned as a “battle that must be won.” Alibaba is mobilizing group-wide resources to support Qwen’s development, tilting comprehensively from technical R&D to marketing promotion. ByteDance is leveraging its advantages in short video to build a unique growth path for Doubao. Doubao’s current UGC marketing is also a form of groundwork; in the long run, it still needs to gradually connect with Douyin’s consumption, transactions, and lifestyle services.
The Hidden Costs of War: The Compute Black Hole and the Homogenization Trap
The price of this red-packet war goes far beyond the RMB 4.5 billion on the books (USD 648.53 million). While the giants are frantically subsidizing on the front end, back-end compute is burning real money. By industry estimates, a chatbot with 100 million daily active users consumes tokens on the order of 10 trillion per day; to stabilize daily active users at 200 million, at least 150,000 additional GPU cards would still be needed. Google once estimated that the average cost per large-model question is about RMB 0.5 (USD 0.0721)—assuming 100 million people claim red packets and do tasks every day, with each person asking an average of one question, compute cost alone would reach RMB 50 million per day (USD 7.21 million/day). During the period when Yuanbao was distributing red packets, congestion prompts such as “compute has been pushed to the point of smoking” already appeared, exposing infrastructure bottlenecks.
A deeper anxiety comes from homogenization. Qwen emphasizes the tool attribute of “being able to get things done,” Yuanbao stresses a “social” positioning, and Doubao leans toward a general assistant—but these differences are blurring at the user-experience level. Sources say Yuanbao is negotiating cooperation with JD.com and Meituan, and Doubao is also integrating with Trip.com and DiDi—in the end, everyone will converge toward a hybrid form of “ads + e-commerce + local lifestyle.” Just as Warren Buffett commented 20 years ago on the competition between Freddie Mac and Fannie Mae: “Even if there are only two companies, they will fight to the death and do some foolish things.” When AI assistants cannot form differentiation, price wars will become the norm, and user retention will be the ultimate test.
Endgame Projection: Tiered Coexistence and Ecosystem Moats
When the smoke clears, this “Three Kingdoms battle” may not produce a single winner, but instead move toward a steady state of “tiered coexistence.” Yuanbao, by virtue of WeChat’s social advantages, will occupy the entrance for light entertainment and simple tasks among acquaintances; Doubao, relying on ByteDance’s content ecosystem, will become an “inspiration partner” in work and study; Qwen, backed by Alibaba’s commercial closed loop, will become a “get-things-done steward” for lifestyle consumption.
The real dividing line lies in the breadth and depth of “ecosystem callability.” The binding power of red packets and mindshare is relatively easy to break when competitors offer larger subsidies or more creative content. The traffic carnival of red packets and milk tea will eventually fade; users pulled in by subsidies will ultimately return to rationality; the true winner will only become clear after the Spring Festival. The ultimate lesson of this war is that in the AI era, the strongest form of capital is no longer the cash on the balance sheet, but the model’s ability to solve real problems—and the efficiency of calling an ecosystem to complete services. When the fireworks of subsidies are gone, the AI that remains on users’ phones and is used repeatedly will be the real winner—it may not be the loudest one, but it will certainly be the one that understands users best, helps the most, and is hardest to live without.
[Disclaimer]: The above content reflects analysis of publicly available information, expert insights, and BCC research. It does not constitute investment advice. BCC is not responsible for any losses resulting from reliance on the views expressed herein. Investors should exercise caution.