9,800 EFLOPS by 2030: China's Five-Year Plan Commits $532 Billion to Quadrupling AI Compute

China has put a hard number on its AI infrastructure ambitions. On Monday, the Ministry of Industry and Information Technology (MIIT) released its five-year plan for the information and communications industry, setting a target of 9,800 EFLOPS of intelligent computing capacity by 2030 and calling for 3.8 trillion yuan (US$532 billion) in cumulative information infrastructure investment over the 2026–2030 period.

The scale of the target is easiest to grasp against the baseline. According to MIIT, China’s intelligent computing capacity reached 2,185 EFLOPS at the end of June 2026 — up 177 per cent from a year earlier. By the end of July, the National Data Administration put the figure at roughly 2,450 EFLOPS. Reaching 9,800 EFLOPS by 2030 means more than quadrupling the June level in under four and a half years, and a six-fold expansion from the 1,590 EFLOPS recorded at the end of 2025.

What the plan actually says

The document — formally the “15th Five-Year Plan for the Information and Communications Industry” — lays out 13 headline indicators for 2030. Beyond the compute target, they include:

• 4.1 trillion yuan in information and communications industry revenue by 2030
• 7 per cent average annual growth in total telecom business volume
• 50 5G and 5G-Advanced base stations per 10,000 people
• 95 per cent penetration of 5G (including 5G-A) subscribers

On computing specifically, the plan calls for the “orderly deployment” of intelligent computing clusters built around 10,000 accelerator cards, alongside larger facilities using 100,000 cards or more, plus inference computing facilities tailored to different application domains. It also directs greater effort toward adapting infrastructure to home-grown computing chips — a line that reads as a direct response to the US export-control regime that has cut Chinese AI labs off from Nvidia’s most advanced accelerators.

Two further strands stand out in the full text. The plan encourages green power direct-connection schemes and compute-power coordination (绿电直连、算电协同) — pairing data centres directly with renewable generation rather than drawing on coal-heavy grids. And it commissions forward-looking research into space-based computing, an early signal that Beijing is studying the same orbital data-centre concepts that have surfaced in US and Japanese proposals.

Building on “East Data, West Computing”

The targets extend an expansion already under way. China had built 52 intelligent computing facilities, each equipped with more than 10,000 accelerator cards, as of mid-2026, according to MIIT. The new plan layers onto the “East Data, West Computing” project launched in 2022, which shifts power-intensive workloads from the densely populated eastern seaboard toward western regions with cheaper land and abundant energy. That project has since crystallised into a network of eight national computing hubs, ten data-centre clusters, and three regions dedicated to coordinating computing facilities with power supplies.

In July, officials went further, describing the ambition to turn the national computing grid into a “public utility” — computing priced and delivered like electricity or water, available on demand to enterprises and researchers nationwide. The 9,800 EFLOPS target is, in effect, the supply-side commitment behind that vision.

The context that matters

Three forces make this plan more than a routine planning document.

First, export controls have made compute a strategic resource. With Washington restricting access to advanced AI chips, Beijing’s ability to hit 9,800 EFLOPS depends heavily on domestic accelerators from Huawei, Cambricon, and a growing cohort of startups — which is precisely why the plan explicitly ties infrastructure deployment to home-grown silicon. The target doubles as a demand guarantee for China’s chip industry: someone will need to buy, install, and operate those millions of accelerator cards.

Second, the money is enormous but not unmatched. The 3.8 trillion yuan commitment follows a separately reported plan, surfaced in June, to spend roughly 2 trillion yuan (US$295 billion) over five years on nationwide data-centre construction. For comparison, US hyperscalers are collectively committing hundreds of billions of dollars to their own buildouts. China’s figure covers broader information infrastructure — including 5G networks — not just AI data centres, but it signals that state-directed capital will underwrite capacity at a scale no single Western company can match alone.

Third, utilization is the open question. Analysts have already flagged that parts of China’s AI computing buildout are underused — capacity built ahead of demand, in regions far from the customers who need it. A six-fold expansion compounds that risk. The plan’s emphasis on “orderly deployment” and on automated monitoring of supply-demand matching suggests MIIT is aware of the problem, but hitting both the capacity number and healthy utilization will require inference workloads — not just training runs — to migrate onto the national grid in volume.

What to watch

The plan’s credibility will be tested by a few concrete markers over the coming years: whether the 100,000-card clusters multiply from today’s handful; whether domestic accelerators can deliver the efficiency to make the EFLOPS arithmetic work without breaking provincial energy budgets; and whether inference demand from enterprises, cities, and the “AI for science” programme absorbs the capacity being built. The 177 per cent year-on-year growth recorded through June shows the expansion is real and accelerating. The question the 2030 target leaves open is not whether China can build the machines — it has demonstrated that it can — but whether the economy that grows up around them justifies the bill.

For the global AI industry, the message is unambiguous: the compute race is now explicitly a state-planned contest, measured in EFLOPS and five-year increments, and China has just published its score to beat.