For years, Nvidia’s powerful GPUs have been the backbone of China’s AI revolution — running everything from search engines and smartphones to autonomous vehicles and advanced generative AI systems.
But the U.S. decision to ban the export of Nvidia’s most advanced chips to China, along with President Donald Trump’s declaration that the new Blackwell chips are “only for the United States,” has forced China to accelerate its own semiconductor development.
Why the US Blocked Nvidia Chips
Washington imposed the restrictions due to concerns that Nvidia’s high-end GPUs could enhance China’s military capabilities. Trump emphasized this stance in an interview on 60 Minutes, making it clear that the newest Blackwell chips would not be shared internationally.
Some US lawmakers even compared the idea of easing restrictions to “handing Iran weapons-grade uranium.”
Nvidia CEO Jensen Huang, however, warned that the ban may backfire. He argued that cutting China off from U.S. technology could push Chinese engineers — who make up nearly half of the world’s AI talent — to innovate even faster and build strong domestic alternatives.
Huawei Leads China’s Mission for AI Chip Independence
Huawei, which has been under U.S. sanctions since 2019, has become the driving force behind China’s self-reliance in AI hardware.
Huawei’s major chip developments:
- Ascend 910B: Comparable to Nvidia’s older A100.
- Ascend 910C: A chiplet-based design aiming to reach Nvidia H100-level performance.
- Ascend 950/960/970: Huawei’s upcoming next-gen chips aimed at supercomputing workloads.
Huawei is also building massive AI supercomputers like the Atlas 900 A3, with even larger Atlas 950 and 960 clusters in development.
To break Nvidia’s dominance in software, Huawei has developed:
- MindSpore (alternative to PyTorch)
- CANN (alternative to CUDA)
A growing coalition of major Chinese companies — including Baidu, ByteDance, iFlytek, SenseTime, and 360 — are already testing and adopting Huawei hardware.
Alibaba and Baidu Roll Out Their Own Nvidia Replacements
Alibaba:
Alibaba’s semiconductor arm, T-Head, is now a critical national asset.
Its latest PPU chip, equipped with 96GB HBM and PCIe 5.0, is already running in China Unicom data centers.
More than 16,000 PPU chips are currently deployed, making the chip a strong alternative to Nvidia’s H20.
Alibaba is also expanding entire AI server ecosystems under its Panjiu Supernode architecture.
Baidu:
Baidu has made a major shift by building a 30,000-chip Kunlun P800 cluster, capable of training large DeepSeek-style AI models.
The Kunlun P800 offers performance similar to:
- Nvidia’s A100
- Huawei’s 910B
Baidu has also secured $139 million in chip orders from China Mobile and promises to release a new AI chip every year until 2030.
China’s New AI Chip Players Are Rising Rapidly
Beyond tech giants, a new wave of AI chip startups is emerging, including:
- Cambricon
- Moore Threads
- Biren
- MetaX
- Enflame
- Hygon
DeepSeek’s announcement that its next big AI model will run on Chinese-made chips has further boosted investor confidence, pushing Cambricon’s valuation to the top of the Chinese market.
While these startups still face challenges — from US blacklisting to limited fabrication resources — analysts believe two or three will eventually become national champions.
Is China Overtaking Nvidia?
Instead of slowing China down, the US Nvidia ban has triggered one of the fastest and most unified semiconductor pushes in China’s history.
Companies like Huawei, Alibaba, and Baidu are racing to rebuild a complete AI hardware ecosystem that once depended heavily on American technology.
Whether China will surpass Nvidia remains uncertain — but one thing is clear: China is determined to control its AI future without relying on US chips.
