China’s AI industry shifts from models to agents, state report says

China's AI sector is undergoing a structural shift away from building ever-larger foundation models and toward deploying AI agents at scale, according to a report from the China Telecom Research Institute carried by state broadcaster CCTV on Saturday. The report projects that inference — the computing cost of running models in real time — will account for 80% of China's computing-power market by 2029, overtaking the training workloads that have dominated demand until now.thenextweb

Chinese technology companies are expected to spend close to 600 billion yuan, roughly $89 billion, on AI this year — more than a tenth of all investment in the country, according to the report. Agents are projected to drive close to tenfold annual growth in computing demand over the next two to three years.thenextweb

From Capital Expense to Operating Expense

The pivot carries broad economic implications. Training a model is a one-time capital expenditure; inference is an ongoing operating cost that scales with every user query, every automated workflow, every agent interaction. As Chinese firms move from showcasing model benchmarks to monetizing agent-based services, the economics of the industry tilt toward companies that can optimize inference efficiency rather than those that simply build the largest models.thenextweb

The shift aligns with a broader government push. On September 2, the Ministry of Industry and Information Technology released a dedicated action plan for integrating AI into the software and information technology services sector. The plan sets a two-stage roadmap: by 2028, China aims to extend AI-powered coding tools and development platforms to more than 20,000 major software enterprises and build 100 benchmark agent-based applications across key industries, while incubating at least five open-source projects. By 2030, the ministry envisions full intelligent upgrades across critical software systems, with AI agents and intelligent services emerging as new growth engines.sohu+1

Europe's Parallel Bet

Europe is preparing for the same inference-heavy future on a different timeline. The European Commission opened bidding in July for up to seven AI gigafactories, a program combining up to €10 billion in public funding with at least €20 billion in expected private investment. Applications close on November 12, with awards expected in early 2027 and machines due to run by mid-2028.digital-strategy.europa+2

That schedule has already slipped — bidding moved from May to July, and interest narrowed from about 70 companies to roughly ten expected bidders. Most of the public funding depends on an EU budget for 2028 to 2035 that member states have not yet agreed upon.thenextweb

A Race Over Running, Not Building

Chinese authorities predict AI-related sectors will surpass 10 trillion yuan in combined scale by 2030. The government work report has deployed the "AI Plus" initiative for three consecutive years, this year calling for "creating new forms of smart economy" for the first time. The inference pivot suggests the next phase of the global AI competition will be defined less by who can train the biggest model and more by who can run agents cheaply and reliably at scale.china