How China’s AI infrastructure push supports its drive for tech self-sufficiency

Even as US-China trade dynamics command the global spotlight, Chinese leaders are focused on “doing [their] own work well.”

• That goes double when it comes to their push toward tech self-sufficiency and AI development.

After weeks of dissecting the trade war drama, it’s nice to be able to turn back to analyzing China’s domestic regulatory environment – aka “the Trivium wheelhouse.”

Three specific developments this week highlight Beijing’s determination to press ahead with its overarching technological ambitions – even as external distractions mount:

• Policymakers have banned foreign AI chips in state-funded data centers.
• Local governments are increasingly subsidizing the electricity bills of AI data centers.
• And officials are rolling out a new “AI Plus Manufacturing” plan, under the broader AI Plus initiative.

Let’s take these one by one.

First, on November 5, Reuters reported that authorities have ordered state-funded data centers to stop using foreign AI chips in new projects or those still early in construction:

• “The Chinese government has issued guidance requiring new data center projects that have received any state funds to only use domestically-made artificial intelligence chips.”
• “In recent weeks, Chinese regulatory authorities have ordered such data centers that are less than 30% complete to remove all installed foreign chips, or cancel plans to purchase them, while projects in a more advanced stage will be decided on a case-by-case basis.”

The vast majority of data centers take at least some state funds, so under these requirements, only a small subsection of private companies that build their own data centers would be exempted from the rules.

• That said, the fact that this guidance is not a publicly-released regulation means Beijing still has the flexibility to allow in some foreign chips later – if this turns out to be an ill-conceived self-sufficiency gun-jump that runs up against market realities.

This all matters because Washington is debating whether to allow Nvidia to sell a downgraded version of its Blackwell chips to China.

• But that debate is moot if Beijing won’t approve their import or use.
• And for now, every indication is that China is all in on AI chip self-sufficiency.

Second, on November 4, the FT reported that China’s local governments are subsidizing electricity costs for AI data centers:

• “China has increased subsidies that cut energy bills by up to half for some of the country’s largest data centers.”
• “The new subsidies come after several tech groups complained to regulators about the increased costs of using domestic semiconductors from companies such as Huawei and Cambricon, most of which are less energy-efficient than Nvidia’s.”

The key point here is that Beijing is trying to capitalize on one of its few AI infrastructure advantages over the US – cheap and abundant power – to offset the cost of switching to domestic chips.

• The US has the most advanced AI accelerators, but currently lacks the power to drive the expected data processing boom.
• China is in the opposite situation – it has plenty of power but weaker chips.

To get more concrete, Huawei’s AI server system – CloudMatrix 384 – is less energy-efficient than Nvidia’s NVL72 system.

• Since it can’t beat the US on per-chip performance, Huawei’s strategy is to raise aggregate performance by linking up more chips.
• But bigger systems consume more power – which is fine if the government covers the bill.

That’s why China’s plentiful supply of cheap energy is a boon for its AI ambitions.

• It gives Huawei, and potentially other companies, room to compensate for lower per-chip performance by scaling horizontally.
• And it also means China’s AI infrastructure buildout will go exponential once its chip supply chain issues are resolved.

This is a moment that Chinese policymakers are trying to force: By tightening access to even the most advanced foreign chips in data centers, as discussed above, Beijing is tacitly promising a protected market to domestic chip innovators that come close to – or meet – foreign chip performance.

The key point: If China’s chipmaking ramp happens before the US resolves its power bottlenecks, Beijing will slingshot ahead in the AI race.

And third, on November 3, Li Lecheng, Minister of Industry and Information Technology (MIIT), revealed that MIIT will soon issue an “AI Plus Manufacturing” plan, under the aegis of the broader AI Plus program.

The plan won’t be short on ambition, with Li indicating it will focus on:

• Rolling out AI upgrades in key industries, and promoting them across the entire manufacturing sector
• Expanding use cases like intelligent assisted design, virtual simulation, and early defect detection
• Promoting the replacement of last-generation consumer devices with products like AI-enabled mobile phones and computers
• Accelerating R&D for next-gen intelligent devices such as humanoid robots and brain-computer interfaces

Why it matters: This AI Plus Manufacturing plan will offer one of the clearest roadmaps yet for how Beijing intends to diffuse AI across the economy – which is a cornerstone of its broader innovation and industrial-upgrading goals.

Putting it all together: Just when you might have expected Xi Jinping and co. to sit back and breathe a sigh of relief after sealing a trade deal with the US, officials instead pushed forward on three major AI policy fronts in a single week.

• It’s clear that, when it comes to AI, Chinese policymakers are going full-throttle as they simultaneously seek to 1) force self-sufficiency in chips, 2) underpin the data center buildout with cheap power, and 3) rapidly diffuse AI across the manufacturing sector.

The bottom line: China’s leaders will stay laser-focused on the innovation agenda – undeterred by any distractions from the US, or other Western governments, on the trade front.