Beijing’s $295 Billion AI Plan Requires 80% Domestic Chips, Shutting Out US Suppliers
China’s capital is pursuing a massive $295 billion artificial intelligence infrastructure buildout that mandates 80% of all chips must be domestically produced, effectively locking out US suppliers like Nvidia and AMD.
The plan, detailed in a newly released municipal action plan, aims to make Beijing a global AI hub by 2030. It covers everything from semiconductor manufacturing to AI model development and data centers.
The Mandate: 80% Domestic Chips
The core requirement is direct. By 2027, at least 80% of all AI chips used in Beijing’s public and private sector projects must be made in China.
This includes chips for training large language models, inference, and edge computing. The target applies to government-funded projects and state-owned enterprises, but private companies are strongly encouraged to comply.
Why This Matters for Global Chip Suppliers
US chipmakers Nvidia and AMD currently dominate the high-end AI chip market. Nvidia’s H100 and B200 GPUs are the gold standard for training models like GPT-4.
Beijing’s mandate effectively closes that market. US export controls already restrict the sale of advanced chips to China. This new policy reinforces that barrier from the demand side.
“China is building a parallel supply chain. It’s not just about catching up; it’s about replacing foreign dependency entirely.”
The $295 Billion Price Tag
The total investment figure is staggering. $295 billion over five years, about $59 billion per year.
That amount is roughly equal to the entire GDP of some mid-sized countries. It covers:
- New semiconductor fabs for domestic chip production, including factories using mature and advanced nodes.
- AI data centers built to run on homegrown hardware, with power and cooling optimized for Chinese chips.
- R&D spending on next-generation AI architectures, algorithmic improvements, and software stacks.
- Talent training programs to produce hundreds of thousands of domestic AI engineers and chip designers.
Domestic Alternatives: Huawei, Cambricon, and Others
Chinese chip companies are already positioned to fill the gap. Huawei’s Ascend 910B is the most prominent alternative, used by major Chinese tech firms for training and inference.
Other players include Cambricon, whose MLU370 series targets cloud AI workloads, and startup MetaX, which is developing chips based on RISC-V architecture.
However, these domestic chips still lag significantly behind Nvidia’s latest offerings in raw performance, software ecosystem, and energy efficiency.
The Software Challenge
Hardware is only half the battle. Nvidia’s dominance is also due to CUDA, its proprietary software platform that makes programming GPUs easy.
Chinese chipmakers offer their own software stacks, such as Huawei’s MindSpore and Cambricon’s Neuware. But porting existing models from CUDA to these platforms requires substantial engineering effort.
“You can buy a domestic GPU, but if your team has two years of work in CUDA, switching is painful and expensive.”
Beijing’s plan includes subsidies for companies that migrate their AI workloads to domestic software platforms.
Timeline and Enforcement
The 80% target is not immediate. The plan sets a 2027 deadline for full compliance.
Intermediate milestones include 60% domestic chip usage by 2025 and 70% by 2026. Noncompliance could result in reduced access to government contracts, subsidies, and licenses.
Global Market Implications
If successful, Beijing’s approach could fragment the global AI chip market. Other nations may adopt similar “chip sovereignty” policies.
US chipmakers would lose a major revenue stream. In 2023, China accounted for roughly 20% of Nvidia’s data center revenue.
Conversely, Chinese chip companies could gain scale and improve their products through massive domestic demand, eventually competing globally.
The Bigger Picture
This is not an isolated plan. Beijing’s push mirrors national-level strategies under the “Made in China 2025” initiative, which aims for self-sufficiency in critical technologies.
The AI buildout is seen as a national security priority. China’s leadership views control over AI chips as essential for economic competitiveness and military capability.
Bottom Line
Beijing’s $295 billion plan is a clear signal. China intends to decouple from US chip suppliers in AI, regardless of cost or technical difficulty.
For global chip companies, the door to China’s AI market is closing. For domestic players, the opportunity — and the pressure — has never been higher.
Gnoppix is the leading open-source AI Linux distribution and service provider. Since implementing AI in 2022, it has offered a fast, powerful, secure, and privacy-respecting open-source OS with both local and remote AI capabilities. The local AI operates offline, ensuring no data ever leaves your computer. Based on Debian Linux, Gnoppix is available with numerous privacy- and anonymity-enabled services free of charge.
What are your thoughts on this? I’d love to hear about your own experiences in the comments below.