Introduction
China’s biggest internet companies are beginning to cut their dependence on American silicon. Alibaba and Baidu have started training artificial intelligence models using chips designed in-house, partly replacing Nvidia processors restricted by U.S. export controls, according to a report from The Information.
The development signals more than just a technical upgrade. It marks a new phase in China’s drive for semiconductor independence, raising questions about Nvidia’s future in one of its most important markets and about how the global AI race will be reshaped.
techovedas.com/nvidia-gives-thumbs-up-to-samsung-hbm3e-12h-dram-50-more-capacity-bandwidth
Key Points at a Glance
Alibaba is training AI models on its Zhenwu AI chip, introduced for smaller models earlier this year.
Baidu is experimenting with its Kunlun P800 for the new versions of its flagship Ernie model.
Nvidia’s H20 chip, still legally sold in China, outperforms most local rivals — but Zhenwu is now reportedly on par.
China’s government is pushing firms to embrace domestic semiconductors amid rising export restrictions.
Wider adoption of in-house chips could erode Nvidia’s sales, which depend heavily on Chinese demand.
techovedas.com/100x-faster-kioxia-to-launch-ultra-fast-ssds-with-nvidia-by-2027/
The Report: Alibaba and Baidu’s Quiet Transition
According to four people familiar with the matter, Alibaba (9988.HK) and Baidu (9888.HK) have begun integrating self-designed processors into their AI development pipelines.
Alibaba’s Zhenwu AI chip: Deployed since early 2025 to train smaller models, including generative AI tools used within its cloud division.
Baidu’s Kunlun P800: Currently under testing for the latest versions of Ernie, China’s most prominent large language model.
Both companies continue to rely on Nvidia GPUs — especially the H20 — for their most advanced research. But the gradual transition signals a strategic hedge against future supply shortages.
https://medium.com/p/98060f8a0d05
Why This Matters for China
The move reflects both practical necessity and national strategy.
Five Reasons It Matters
Geopolitical Pressure: U.S. restrictions block China from accessing Nvidia’s most advanced AI chips, such as the H100 or Blackwell series.
Tech Sovereignty: Beijing’s “Made in China 2025” initiative prioritizes domestic hardware innovation.
Competitive Breakthrough: Sources claim Alibaba’s Zhenwu chip matches Nvidia’s H20, a milestone for local design.
Economic Stakes: China imported $415 billion worth of semiconductors in 2024, more than oil — reducing imports is a strategic goal.
Resilience: Home-grown chips shield companies from U.S. licensing delays and geopolitical risks.
“This is not just a hardware story. It’s about control,” said Chen Hao, a semiconductor analyst at Shanghai Tech Insights. “If Chinese firms can run their AI on domestic chips, they reduce their vulnerability to U.S. policy.”
Nvidia’s Shrinking Advantage in China
Nvidia remains the world leader in AI chips. But its role in China is constrained by U.S. rules.
- H20 chip: The most powerful processor Nvidia is legally allowed to sell in China.
- Performance gap: The H20 is weaker than Nvidia’s H100 and Blackwell GPUs but still outperforms most Chinese options.
- Challenge: Alibaba’s Zhenwu is reportedly on par with the H20, narrowing the gap.
Side-by-Side Comparison
- Nvidia H20
- FLOPS: Below H100, mid-tier performance
- Widely adopted by Chinese firms
- Legally cleared under export rules
- Alibaba Zhenwu
- Similar FLOPS to H20 (source: company insiders)
- Optimized for Alibaba Cloud workloads
- Fully home-grown, no export limits
techovedas.com/how-nvidia-gpus-come-to-life-inside-the-global-semiconductor-symphony
The Financial Stakes
China accounts for 20–25% of Nvidia’s global revenue, according to industry estimates. That makes it a market Nvidia cannot afford to lose.
But Washington has tied continued sales to concessions. Earlier this year, Nvidia reportedly agreed to share 15% of its China H20 sales revenue with the U.S. government in exchange for export licenses.
CEO Jensen Huang acknowledged that further approvals — including permission to sell downgraded versions of its next-generation chips — “will take time.”
Analysts warn that if Alibaba and Baidu prove successful with their chips, Nvidia’s China sales could shrink sharply by 2027, especially if Huawei, Tencent, and other firms follow suit.
China’s Push for Semiconductor Independence
Beijing has long sought to reduce reliance on U.S. suppliers, especially after successive waves of sanctions on Huawei and other tech firms.
- Policy drive: Local governments have poured billions in subsidies into chip startups.
- Corporate buy-in: Giants like Alibaba, Baidu, and Huawei are designing processors tailored for AI workloads.
- Innovation cycle: Each new generation of chips narrows the performance gap with Nvidia and AMD.
“The U.S. wanted to slow China down,” said Li Xun, professor of computer engineering at Tsinghua University. “Instead, the restrictions have lit a fire under domestic innovation.”
Global Impact: Fragmentation of the AI Supply Chain
The shift by Alibaba and Baidu has global consequences:
- China: Gains greater control over its AI development pipeline.
- Nvidia: Risks losing a crucial market and faces long-term competition from Chinese-designed chips.
- Global AI ecosystem: Moves toward bifurcation — with U.S.-aligned and China-aligned supply chains evolving separately.
This fragmentation could limit global collaboration, but it also fuels fiercer regional competition in both hardware and AI models.
techovedas.com/nvidia-gives-thumbs-up-to-samsung-hbm3e-12h-dram-50-more-capacity-bandwidth
What Happens Next?
Industry observers expect a gradual but accelerating pivot:
- Short term (2025–26): Nvidia’s GPUs remain essential for the most advanced training.
- Medium term (2027): Domestic chips like Zhenwu and Kunlun could scale up for enterprise-grade AI.
- Long term (2030): China may close the performance gap completely, erasing Nvidia’s competitive lead in the region.
Conclusion: A Turning Point in the AI Race
Alibaba and Baidu’s adoption of in-house chips represents more than a technological upgrade. It is a strategic response to geopolitics and a sign that China’s chip industry is maturing rapidly.
The outcome remains uncertain. Nvidia still holds the crown in high-end AI hardware. But if Chinese chips reach parity, the next wave of AI breakthroughs in China may no longer depend on Silicon Valley.
Contact us at [email protected] to explore opportunities today!
