中国的 AI 竞争已从主要把资金投入算力与人才，转向在农历新年期间投入约 $720 million 的用户补贴；Alibaba、Tencent、ByteDance 与 Baidu 正试图推动聊天机器人服务的大众采用。这一策略瞄准大规模的消费者行为：在过去 1 year 内 AI 能力明显提升，但除了一小部分技术领先型工作者外，日常使用仍处于早期阶段。各公司正利用红包式激励，推动用户从尝试性使用走向在 App 生态内的常规交易。

Alibaba 宣布了规模最大的计划，为 3 billion yuan ($433 million)，透过 Qwen 补贴购物、外送与票务预订；国内报导称奶茶商家被 Qwen 带动的订单淹没。ByteDance 把促销与一场全国性晚会绑定，提供包含装置在内的奖励，另有最高 8,888 yuan 的红包；Tencent 则提供最高 10,000 yuan 的 Yuanbao 现金激励。Baidu 启动一项 500 million yuan 的活动，以在 AI 助手成为推荐、下单与支付入口之际，捍卫搜寻时代流量，让分发演变为由补贴主导的拉新竞赛。

这些数字显示出典型的中国网际网路打法：前期重补贴、快速抢流量、再做生态锁定；但 AI 带来持续且高昂的推理成本与不明确的变现路径，增加了利润率长期承压的风险。历史类比包括网约车多年补贴战，以及之后在智慧手机与外送领域的竞争；然而 AI 竞争似乎把这一周期提前加速了，在稳定的重复使用模式尚未完全成熟之前就已发生。较可能的结果是，围绕已整合多种服务的超级 App 生态出现更强集中，赢家获得长期用户黏性，而若产品效用不足以支撑补贴强度，输家将面临高烧钱率。

China’s AI competition has shifted from spending primarily on compute and talent to spending about $720 million on user subsidies during Lunar New Year, as Alibaba, Tencent, ByteDance, and Baidu try to drive mass adoption of chatbot services. The strategy targets consumer behavior at scale in a market where AI capabilities improved markedly over the past 1 year, but everyday usage is still early-stage outside a small set of tech-forward workers. Firms are using red-packet style incentives to push users from experimentation toward routine transactions inside app ecosystems.

Alibaba announced the largest program at 3 billion yuan ($433 million), subsidizing shopping, food delivery, and ticket booking via Qwen; domestic reports said milk-tea merchants were flooded with Qwen-driven orders. ByteDance tied promotions to a national gala and offered rewards including devices plus red packets up to 8,888 yuan, while Tencent offered direct Yuanbao cash incentives up to 10,000 yuan. Baidu launched a 500 million yuan campaign to defend search-era traffic as AI assistants become entry points for recommendations, ordering, and payments, turning distribution into a subsidy-led user acquisition race.

The numbers indicate a classic China internet playbook: heavy upfront subsidies, rapid traffic capture, then ecosystem lock-in, but AI adds high ongoing inference costs and unclear monetization, increasing the risk of prolonged margin pressure. Historical analogs include multi-year subsidy wars in ride-hailing and later battles in smartphones and food delivery, yet AI competition appears to be accelerating that cycle earlier, before stable repeat-use patterns fully mature. The likely outcome is stronger concentration around super-app ecosystems that already bundle many services, with winners gaining long-term user stickiness and losers facing high burn rates if product utility fails to justify subsidy intensity.