Chinese artificial intelligence companies poured an estimated $1.2 billion into user subsidies during the Lunar New Year holiday, racing to win traction for “agentic” chatbots that can complete tasks like ordering meals. The spending spree, centered in China’s biggest annual shopping season, signals a push to turn chatbots from chat companions into digital helpers that transact on a user’s behalf.
The surge in giveaways and coupons, rolled out across apps and mini-programs, aimed to lock in daily usage at a moment when millions of people were off work and on their phones. Companies sought to prove not only that their tools could talk, but that they could act, integrating with food delivery, travel booking, and e-commerce.
“Chinese AI companies spent $1.2bn on user subsidies during Lunar New Year as they compete to develop agentic chatbots that can perform tasks like ordering meals.”
Why Holiday Subsidies Matter
Lunar New Year is China’s prime season for user acquisition. Big platforms have long used splashy promotions to grab attention, from ride-hailing discounts to digital cash “red packets.” AI firms are adopting the same playbook. The goal is simple: get users to try a new feature when habits are in flux, then keep them coming back.
Agentic chatbots shift the focus from conversation to completion. Instead of suggesting a noodle shop, the bot places the order, applies a coupon, and shares the delivery time. That leap demands reliable integrations with services such as food delivery, maps, payments, and customer support.
From Talk to Transactions
China’s tech giants already run many of the services these bots need. That gives them an edge in wiring chatbots to real commerce. Popular models like Ernie, Qianwen, Hunyuan, Doubao, and others have been layering tools for search, shopping, and bookings. The subsidy push is a stress test for those connections.
Consumers win first. They get cheaper meals, faster service, and less screen tapping. Merchants get traffic and bigger baskets as bots bundle add-ons with a single tap. But the bill lands with platforms, which hope that scale and stickiness will outlast the discounts.
Echoes of Past Subsidy Wars
China has seen this movie before. Ride-hailing, food delivery, and community group buying burned through cash to gain share, then pulled back once habits formed. Some categories stabilized; others saw retrenchment, consolidation, or tighter rules on promotions.
AI agents add a twist. The subsidy is not just to win a price-sensitive order. It is to teach users a new behavior: trust a bot to act. That takes repetition and low friction. The $1.2 billion figure suggests companies are willing to pay handsomely for that training period.
- Short-term effect: cheaper orders and rapid user trials.
- Medium-term test: retention once coupons fade.
- Long-term bet: fees from transactions and premium plans.
Costs, Constraints, and Competition
Building capable agents is expensive. Models need frequent updates, tool-use reliability, and guardrails for payments and refunds. Compute costs and chip supply add pressure, especially under export controls that limit high-end hardware. That pushes firms to show near-term traction to justify heavy investment.
Global competition looms. U.S. labs are also moving agents from demos to daily tasks. Chinese platforms counter with local integrations and super-app reach. If agents become the default interface for ordering, whoever controls the agent can steer demand and collect a fee on each action.
What Users and Regulators Will Watch
Trust will make or break these tools. Users need clear permissions, accurate billing, and easy dispute resolution. Merchants will look for fair ranking, transparent fees, and ways to keep their brand visible when a bot stands between buyer and seller.
Regulators have experience managing subsidy battles and platform behavior. Issues include predatory pricing, data protection, and disclosure when an automated system places an order. Clear rules could help the market mature without repeating past excesses.
Signals to Track Next
Investors and operators will monitor a few markers after the holiday surge. Daily active users, repeat agent-triggered orders, and average order value will show whether behavior is sticking. Merchant churn and customer support cases will reveal friction. Most telling will be how fast platforms taper subsidies without losing usage.
The holiday blitz proves one thing: chatbots that act are moving from concept to checkout. Whether the $1.2 billion spend becomes a smart down payment or a costly tradition depends on retention, trust, and unit economics once the party ends.
