China is preparing to spend roughly 2 trillion yuan, about $295 billion, over the next five years to build a nationwide network of interconnected AI data centers, according to reporting by Bloomberg. The plan, still in draft form, would knit together computing hubs across the country into a single state-operated grid, and it is designed to lean on domestic suppliers like Huawei for at least 80% of core technology, including AI chips, effectively squeezing Nvidia and AMD out of the buildout.
The reporting traces the blueprint to key government agencies including the National Development and Reform Commission (NDRC), China's top economic planning body. State carriers China Mobile and China Telecom would operate the bulk of the facilities and ensure they are linked, with the broad goal of connecting scattered data centers into a cohesive national grid by 2028. That expands the telecom giants' role well beyond connectivity into the provision of sovereign compute at national scale.
What the money covers, and what it doesn't
The 2 trillion yuan headline figure is for data center construction. It would be funded mainly through sovereign debt, including ultra-long-term special government bonds typically of more than 10 years' tenure, along with state strategic-industry funds, bank loans, and supplementary private capital. Folding in the power grid upgrades needed to feed those facilities could push total capital requirements past 5 trillion yuan, according to the reporting.
Crucially, the $295 billion does not include the AI capital spending of China's private tech champions. Alibaba and Tencent are running their own large capex programs in parallel, so the state plan represents a floor on Chinese AI infrastructure investment rather than the full national total. The buildout is framed as part of Beijing's latest five-year plan, in which the government pledged to prioritize data infrastructure construction.
A caveat worth keeping front of mind: as of this reporting, the plan is described as a draft blueprint rather than a finalized, formally announced policy. The specifics come from Bloomberg's sourcing and secondary aggregation rather than a published government document, so figures and timelines could shift before any official rollout.
The chips problem at the center of it
The 80% domestic-sourcing target is, in many ways, the most consequential detail. It marks a decisive push away from Western silicon in Chinese state infrastructure and accelerates a decoupling that US export controls have been driving from the other direction. Washington has spent years tightening restrictions on advanced AI chips and the equipment to make them, aiming to slow China's access to frontier compute. A national grid built around Huawei's Ascend accelerators and other domestic parts is Beijing's answer: turn the constraint into an industrial-policy mandate.
The challenge is supply. Analysts cited in coverage from outlets such as Tom's Hardware flagged that a 2028 timeline could collide with the limits of domestic chip production. China's leading-edge fabrication still trails the most advanced foreign processes, and producing enough high-performance accelerators to fill a nationwide grid, while also serving private-sector demand, is a tall order. The plan's ambition may outrun the country's near-term ability to manufacture the silicon it requires.
How it stacks up against the US
The scale is enormous by any measure, but it sits in a different bracket from US private-sector spending. Industry coverage noted that the $295 billion five-year total compares with the roughly $725 billion that US companies including Meta and Microsoft are setting aside for AI investment in 2026 alone. Chinese data centers generally cost less to build and operate, thanks to lower labor, component, and construction costs plus local-government incentives, so the yuan stretches further. Still, the contrast underscores a structural difference: America's AI buildout is led by deep-pocketed hyperscalers, while China's is increasingly state-directed, debt-financed, and routed through national champions.
Why it matters
This is industrial policy at the scale of national infrastructure. By financing compute through sovereign bonds and routing operations through state carriers, Beijing is treating AI capacity the way it once treated highways, high-speed rail, and 5G: as strategic public infrastructure to be planned and built top-down. The 80% domestic-chip rule turns US export controls into a forcing function for a fully sovereign AI stack. If it works, China reduces a key vulnerability and hands Huawei a guaranteed domestic market large enough to fund further chip development. If the chip supply can't keep pace, the plan becomes a stress test of just how far state direction can substitute for access to the world's best silicon.
What to watch
Official confirmation: whether the NDRC or State Council formally announces the plan, and whether the $295 billion figure, the 2028 grid target, and the 80% domestic threshold survive into final policy. Chip supply reality: signs that Huawei and other domestic suppliers can produce enough advanced accelerators, or evidence of bottlenecks that force compromises on the homegrown-silicon mandate. Funding mechanics: how much ultra-long special bond issuance is actually allocated, and whether power-grid upgrades push the real bill toward 5 trillion yuan. And the US response: any tightening of export controls or countermeasures aimed at China's domestic chip ecosystem as Beijing moves to lock out Nvidia and AMD.
“China is preparing one of the largest state-directed digital infrastructure programmes in history, committing $295 billion over five years to build a network of AI data centres operated by its state telecoms carriers.”— Capacity, Telecom & data-center trade publication