China Responds to the Chip Embargo with $295 Billion — And 80% Must Be Domestic

Geopolitical Analysis — China is preparing to spend $295 billion (2 trillion yuan) over five years on AI data centers — with at least 80% of the technology sourced domestically. Nvidia and AMD are explicitly locked out. This is the direct response to Taiwan’s chip export controls and the US embargo.

The Plan

Key details:

• Government-led: National Development and Reform Commission drafting the blueprint
• State-operated: China Mobile and China Telecom run the data centers
• Huawei as chip supplier: Huawei’s Ascend AI chips replace Nvidia/AMD
• 80% domestic requirement: Explicit policy to squeeze out US suppliers
• Part of “Six Networks” program: Broader infrastructure initiative
• Private sector excluded: Alibaba, Tencent, ByteDance spend is on top of this

The AI Fracture Deepens

Yesterday we mapped the AI Fracture — four jurisdictions splitting the Map of AI by geography. China’s $295B plan is the most concrete evidence yet:

The Scale Comparison

How does $295B compare?

$295B sounds massive. In context, it is not. Meta alone is spending $145B this year — half of China’s entire 5-year plan in one year. Goldman’s $7.6T projection for the US-allied stack is 25x China’s public budget.

But China’s plan excludes private sector spend (Alibaba, Tencent, ByteDance, Baidu). Add that and the real number is likely $400-500B. Still a fraction of the US-allied total — but enough to build a parallel, self-sufficient AI infrastructure.

The 80% Rule Changes Everything

The 80% domestic requirement is the structural story. It means:

• Huawei’s Ascend chips become the default — not because they’re better, but because they’re mandated. Performance gap vs Nvidia is real but narrows with scale.
• An entire domestic supply chain must be built — from chip fabrication (SMIC, trailing TSMC by 2-3 generations) to networking equipment to cooling systems
• Nvidia and AMD lose a $236B market — 80% of $295B that cannot go to US suppliers = $236B locked out
• Open-source models become even more strategic — China weaponizes open source because efficient models reduce the hardware disadvantage

The Product Overhang Read

China is building its overhang deliberately. $295B in infrastructure investment creates compute capacity that doesn’t produce value until models are trained and deployed on it. The overhang builds over 5 years. When it releases — when Huawei’s chips reach good-enough performance and China’s domestic models mature — it will be the largest coordinated overhang release in AI history.

The US chip embargo was designed to prevent this. China’s response: build the entire stack domestically, accept a performance gap, and bet that scale + efficiency + open source closes it.

The AI Fracture is no longer theoretical. It has a budget: $295 billion.

Related:
The AI Fracture: Four Jurisdictions
Taiwan Criminalizes Chip Exports to China
Goldman Sachs: Where $7.6 Trillion Goes
Map of AI

Sources: Bloomberg, Yahoo Finance, The Decoder, Tech Startups, MarketScreener (June 9-10, 2026)