Generated by Codex with GPT-5

What this article is about

This summary covers The Economist’s March 28th, 2026 China column on page 44, headlined The state of AI and described in the contents as Chaguan AI diffusion.

The article’s core argument is that the most important AI contest may not be who builds the flashiest model, but who is best at spreading the technology through the real economy. On that measure China looks formidable. It has cheap open-source models, a huge manufacturing base, plenty of researchers and, above all, a state willing to push capital and institutions in the same direction. But the article also argues that this same state introduces theatre, caution and bureaucratic drag. China is not simply unleashing AI. It is trying to manage, choreograph and contain it at the same time.

China’s advantage is organised diffusion

The piece starts with a deliberately exaggerated picture of AI showing up everywhere in Chinese life, from shopping apps and ride-hailing to face-scanned yoghurt orders and toilet seats that promise health monitoring. That scene-setting makes the article’s larger point: China wants AI to become ordinary infrastructure, not just a frontier lab exercise.

The turning point, in the article’s telling, came well before last year’s DeepSeek moment. After AlphaGo’s victory in 2017, Beijing published a national AI-development plan and began treating the technology as a strategic priority. More recently it launched an AI Plus initiative to push adoption across industries such as health care and energy. The scale of state involvement is striking. Of China’s 700bn yuan of AI capital spending last year, the government is estimated to have supplied about 60%, or roughly \$100bn. The article thinks that kind of co-ordination helps explain why China may prove especially strong at applying AI in factories, services and consumer products.

The state speeds things up, but also turns them into performances

The article is sceptical of glossy official claims. It notes that government statistics say 42.8% of China had already “adopted” generative AI by 2025, but treats that number less as a trustworthy measure than as evidence of how eager the state is to display momentum.

That performative quality shows up in the reporting from Yizhuang, a Beijing tech hub focused on “embodied AI”. Visitors can watch robot waiters, robot shows and even a robo-triceratops gliding through a restaurant. Yet one of the animals was being steered by a human with a game controller, and staff admitted the kitchen still relied on people because the robots could not cook well enough. The article’s point is not that China is faking AI progress altogether. It is that state-backed targets encourage showcase demos and public-relations wins, which can make adoption look deeper and more mature than it really is.

Caution is part of the model

The article’s most interesting twist is that China’s state is not only an accelerator. It is also a brake. Officials have moved quickly to restrict tools such as autonomous AI assistants in government work and state-owned firms because of fears about leaking sensitive information. They are also wary of labour-market disruption. According to the article, senior officials have pushed companies to direct AI toward jobs that are dangerous or hard to staff, rather than toward indiscriminate automation that could trigger mass layoffs.

There is another limit, too: much of the Chinese workforce sits inside the public sector and state-owned enterprises, which are famously slow adopters of new digital tools. The article describes how some organisations responded to orders to use DeepSeek by installing isolated machines for display purposes instead of integrating the model into actual workflows. In other words, the same command structure that can mobilise adoption fast can also produce shallow compliance.

The takeaway

The Economist’s view is that China may be better positioned than many outsiders assume to weave AI into everyday economic life, because the state can marshal money, data and institutions behind the effort. Yet that does not mean China is pursuing a freewheeling AI revolution. Its leaders want the productivity gains and strategic edge, but they also want control over information, employment and the public narrative.

That tension is the real theme of the article. China could become very good at practical AI deployment precisely because the state is so powerful. But that power also makes the rollout more cautious, more performative and less spontaneous than the hype suggests.