Meituan Wang Xing talks about AI investment again, stating that they will not blindly invest beyond financial capabilities.

Author: Wang Xiaojuan

At the recently held Meituan shareholders meeting, Meituan CEO Wang Xing clearly outlined the company's AI development strategy for the next stage. Wang Xing stated that Meituan will consider its own cash flow situation and will not blindly invest beyond its capital capacity.

He emphasized that AI is a positive productivity tool, and Meituan will invest within its reasonable range, but will not make irrational investments beyond its financial capabilities.

This statement indicates to a certain extent that Meituan's strategy in the AI field is shifting from early-stage capital deployment and technology trials to a stage driven by capital efficiency, business synergy, and ROI.

Over the past two years, Meituan's external investments in the AI field have mainly focused on two core tracks: large model underlying architecture and embodied intelligence. As related target companies enter the listing process, Meituan's early strategic investments are gradually transforming into financial assets with realization dividends.

In the embodied intelligence field, Meituan participated in multiple rounds of financing for Unitree Robotics.

Unitree Robotics passed its IPO hearing on the STAR Market on June 1, with a post-issuance valuation of approximately 42 billion yuan. Public data shows that Meituan-related funds hold a 9.65% stake in Unitree Robotics, making it the largest external institutional shareholder. Beyond financial returns, this investment mainly points to Meituan's long-term reserves for an automated delivery network on the technology front.

In the large model track, Meituan holds a 3.86% stake in Zhipu AI. After Zhipu AI went public on the Hong Kong Stock Exchange in January, its market value once reached 650 billion Hong Kong dollars. Against the backdrop of Meituan's management emphasizing improved capital utilization efficiency, the book value appreciation of this equity asset is becoming a potential source of non-recurring gains for the company.

At the level of independent R&D and product application, Meituan's investments show clear boundary demarcation: avoiding participation in the high-energy-consuming underlying large language model arms race, and concentrating R&D resources on the application layer and automation tool chain.

On June 9 this year, the AI-native browser "Tabbit 1.0," developed by the Meituan Guangnian Zhiwai team, officially launched. This product does not adopt a single path of pure self-developed models. Instead, it integrates Meituan's self-developed LongCat model as well as third-party mainstream models such as DeepSeek, Zhipu GLM, and Kimi through a multi-model access approach, positioning itself as a cross-software, cross-webpage task execution entry point.

Data shows that Tabbit's Agent task execution success rate has increased from 53.1% during the public beta in March this year to the current 91.8%. In terms of operational cost control, the product adopts a tiered pricing model: the standard version is free, while advanced automation functions and scenario customization are monetized through a professional version (9.9 yuan/week), thereby alleviating computing power cost pressure.

Additionally, in its core local services business, Meituan is advancing low-cost AI grayscale testing, including launching a delivery Skill that interfaces with various third-party AI assistants, and deploying the "Xiao Mei" agent on platforms such as Tencent Yuanbao.

Furthermore, Wang Xing introduced that Meituan has been grayscale testing the "Xiao Tuan" feature within its main app, serving as a direct AI entry point for local life scenarios.

Regarding the initial data of this feature, Wang Xing gave a rational assessment, admitting, "We haven't seen explosive results yet." However, he also made a prediction about the evolution of terminal interaction methods: "I believe typing will decrease, while voice usage will increase more and more."

This judgment indicates that Meituan's future iteration focus for C-end AI products will likely tilt towards a deep integration of voice commands and LBS scenarios, attempting to find the most natural entry point within the existing order transaction chain.

For Meituan, its core business model is inherently built on a high-frequency, low-margin offline transaction and fulfillment network. Unlike pure online content distribution platforms, AI large models are unlikely to generate disruptive revenue growth for its main businesses such as food delivery and in-store services in the short term.

In the current situation where the local services track faces multi-line competition and core businesses have entered a stock game, Wang Xing's restraint in AI investment essentially clarifies the company's strategic focus. This capital discipline, which strictly confines AI to an efficiency tool level and takes cash flow safety as the bottom line, also reflects to some extent the realistic choices of domestic internet giants when facing technological cycles.

Risk Warning and Disclaimer

        Markets are risky, and investment requires caution. This article does not constitute personal investment advice, nor does it consider the specific investment objectives, financial situation, or needs of individual users. Users should consider whether any opinions, views, or conclusions in this article are suitable for their particular circumstances. Investing based on this is at your own risk.