China is drafting a $295bn plan to build AI data centres, and to lock Nvidia out of them

China wants to win the AI race on its own hardware. A new plan shows just how much it is willing to spend, and how far it will go to cut American chips out of the picture.

Beijing is drafting a blueprint to spend around 2 trillion yuan ($295bn) over the next five years building a national network of AI data centres, according to Bloomberg, citing people familiar with the matter.

The plan, led by the powerful National Development and Reform Commission, would knit the country’s scattered computing facilities into a single, interconnected grid by 2028, mostly operated by state telecoms giants China Mobile and China Telecom.

The most pointed detail is what those data centres would run on. The blueprint calls for local suppliers, including Huawei, to provide at least 80 per cent of the core technology, AI chips included, effectively squeezing out Nvidia and AMD.

It is a deliberate echo of the campaigns that built national champions like Huawei in the past, now aimed at replacing US technology across the AI stack and closing the gap with American labs.

The 💜 of EU tech

The latest rumblings from the EU tech scene, a story from our wise ol' founder Boris, and some questionable AI art. It's free, every week, in your inbox. Sign up now!

The money would come largely from sovereign debt, including ultra-long-term special government bonds, plus state funds for strategic industries, with bank loans and private capital topping it up. The build is one prong of a broader “Six Networks” programme spanning water, power, and computing, and folding in the power grid could push the total past 5 trillion yuan.

“Elevating it to a national strategy ensures policy alignment and capital mobilisation,” said Charlie Dai, a principal analyst at Forrester.

For all the ambition, the numbers are smaller than they sound next to the West. The $295bn is spread over five years; US firms alone, led by Meta and Microsoft, are setting aside roughly $725bn for AI this year. The Chinese figure also excludes private spending by Alibaba and Tencent, and Chinese data centres are cheaper to build.

The point is less the raw sum than the coordination: a state marshalling debt, land, power, and chips behind a single national grid.

The timing reflects growing Chinese confidence in its own silicon. Washington has eased up, agreeing to let Nvidia sell its previous-generation H200 chips to Chinese buyers, but shipments have not started, and in May nine homegrown AI chips, from Huawei, Alibaba, Shanghai Biren, and Moore Threads, cleared a domestic security review, opening them to sensitive sectors.

Beijing increasingly believes it can fill the gap itself.

It is the same sovereignty logic now sweeping the West, from Britain’s sovereign-AI push to Europe’s scramble to cut its reliance on US clouds, only inverted: where Europe frets about depending on America, China is building to need it as little as possible. The plan is still early, the people cautioned, and details could change. But the direction is unmistakable.

The world’s two largest economies are each trying to wall off their own AI supply chains, and the era of a single global stack is ending.