Meituan released its second-quarter and first-half performance report for 2026 on the afternoon of August 28, showcasing robust growth across its core operations. During the quarter, the company generated revenue of RMB 104.6 billion, marking a year-on-year increase of 14.4%. Benefiting from a more rational competitive landscape in the industry and the seasonal uptick typical of the second quarter, all business segments posted solid growth, with the company achieving a profit of RMB 2.2 billion for the period.

Meituan CEO Wang Xing commented: "We will steadfastly increase our investments in ecosystem development and technology, integrating AI into real-world business scenarios to enhance user and merchant experiences, as well as our operational efficiency, all while helping everyone eat better and live better."

In the second quarter, Meituan's instant retail business continued to solidify its advantages in order structure, user base, and operational efficiency, while core user loyalty strengthened further. The in-store, hotel, and travel segments also sustained high-quality growth. As a result, the core local commerce division generated revenue of RMB 71.5 billion, up 10.1% year-on-year, with operating profit turning positive sequentially to reach RMB 5.7 billion.

Meituan's grocery retail and international operations saw continued improvements in operational efficiency during the quarter. The new initiatives segment recorded revenue of RMB 33.1 billion, a 25% increase year-on-year, while narrowing its losses to RMB 1.7 billion. In the grocery retail arena, Meituan deepened its "online-plus-offline" retail channel strategy, with Xiaoxiang Supermarket expanding its footprint to 68 cities and Happy Monkey reaching 40 physical stores nationwide.

Additionally, Meituan strengthened its AI capabilities throughout the quarter, boosting quarterly research and development investment by 22.5% year-on-year to RMB 7.7 billion, which accounted for 7.3% of total revenue. In June, the company unveiled its next-generation proprietary large language model, LongCat 2.0, marking the industry's first trillion-parameter model to complete both training and inference entirely on a domestic computing cluster.