What happened
In its 2026 interim report released on August 28, Meituan posted second-quarter revenue of 104.6 billion yuan, up 14.4% year on year, and said operating profit swung from negative to positive on a quarterly basis.
At the earnings call, CEO Wang Xing outlined Meituan's AI strategy for the first time, saying the company does not intend to become a Token factory and will instead use models and AI products to support its core business. He stressed that AI's long-term value lies in entering real workflows and solving real problems.
Meituan also revealed that its self-developed trillion-parameter model LongCat-2.0, known externally as Owl Alpha, ranked among the world's top three in monthly API calls on multiple platforms in the first half of the year.
Why it matters
Meituan is positioning AI not as a separate business but as a way to make its existing local-services engine more efficient. Its argument is that local life is hard for AI because each order depends on a tangle of real-world factors such as foot traffic, weather, time slots, inventory and delivery distance.
The company argues that the industry no longer lacks models, but rather proven ways to embed AI into real business operations. By leveraging its accumulated merchant, user, fulfillment and review data, Meituan believes it can bring AI into actual transaction chains rather than just text and search scenarios.
Wang Xing said AI represents a full upgrade of Meituan's organization, products and workflows, pointing to three strategic directions: building LLMs, AI at work, and AI in products. This suggests Meituan is deliberately skipping the model arms race to focus on street-level applications.
Key facts
Meituan's second-quarter revenue was 104.6 billion yuan, up 14.4% year on year, with operating profit turning positive quarter on quarter.
Wang Xing said Meituan's AI strategy covers Building LLM, AI at Work and AI in Products.
Meituan's Owl Alpha model is actually its self-developed trillion-parameter large model LongCat-2.0, which ranked global top three in monthly calls on several platforms in the first half of the year.
Meituan's AI agent platform CatPaw helps merchants automatically aggregate sales, review and marketing data to generate daily reports.
During an early AI training phase, Meituan had 100,000 employees raise 80,000 digital shrimp and spent 10 million yuan per day on model tokens.
What to watch next
Whether Meituan can keep turning AI experiments into tools merchants actually use, such as CatPaw and the upgraded assistant Xiaotuan, and whether these tools improve the efficiency of its thin-margin food delivery business.
How Meituan balances its stated focus on core business efficiency with its continued investment in building large models and AI capabilities.
Whether rivals follow Meituan's path of applying AI to real-world local commerce rather than competing purely on model scale.