AI is taking China by storm.

Chinese tech companies are entering a new competitive landscape centered around AI agents, with platforms like OpenClaw gaining significant user attention.

On March 6th, many customers, including older adults, brought their MacBooks or laptops to the new system rollout events, creating a scene reminiscent of Android community gatherings a decade ago.

This trend is rapidly spreading across the industry. Xiaomi is internally testing MiclawAgent to integrate AI into its "person, car, home" ecosystem, connecting phones, cars, TVs, and home appliances into execution points. Meanwhile, several cloud service providers are also actively participating in this race by directly integrating AI into their operating systems and end devices.

An elderly person brought their computer to Tencent's headquarters in Shenzhen, China, on March 6th to have OpenClaw installed. Photo: Tencent AI.

The new AI race

According to experts, the explosion of AI is this year's automation competition. It's also being likened to a race to control the next "hyper-connected gateway" of the Internet.

One of the main reasons driving this trend is the business viability of AI models. Over the past two years, major Chinese tech companies have invested heavily in computing infrastructure. ByteDance, Alibaba, and Tencent are expected to spend more than $60 billion on data centers and GPUs this year.

However, a purely chatbot model does not generate enough revenue to offset the enormous operating costs.

AI will play a crucial role in 2026. Photo: Bloomberg .

Common interactions like writing emails or creating diagrams consume very few tokens, leaving computing resources easily idle. In this context, some AI agents like OpenClaw have taken full advantage of this redundancy.

When performing a complex task, the system can break down the work, search for information, and self-correct errors. Each step generates requests to cloud data, resulting in token consumption hundreds of times higher than a typical query.

“OpenClaw adopts an open-source model primarily due to its low-cost advantage. This allows for more frequent API usage and generates revenue for cloud providers,” an AI analyst told Wall Street Insights .

Data gold mine

Beyond revenue, tech companies also aim for long-term goals when collecting training data for the next generation of AI. During the process of users assigning tasks to AI agents, the entire process from information retrieval and synthesis to query completion is recorded as a sequence of actions, also known as "trajectory data".

According to Alan Feng, an OpenClaw community manager in China, the system's value lies not only in automation. "Orbital data feedback allows the model to continuously optimize, improving the capabilities of AI agents," he said.

Data is the lifeblood of future AI systems. Photo: Bloomberg .

This type of data is a crucial resource as high-quality text sources on the internet are almost exhausted. A source from Alibaba's Qwen project said that collecting orbital data could help Chinese AI models improve faster and close the gap with international competitors.

In this context, AI is seen by many companies as the new "interface layer" of the Internet. Instead of opening each application individually, users simply make a request and the AI will decide which service to use, from ordering food to making payments.

This shift could diminish the role of traditional super apps. Therefore, OpenClaw's recent popularity is an early sign of a major change. As AI moves from a "chat tool" to an automated system, the internet's architecture could enter a completely new competitive phase.