China Demands Alibaba, ByteDance Explain Nvidia AI Chip Purchases
▼ Summary
– Beijing is requiring tech firms like Alibaba and ByteDance to justify their orders of Nvidia’s H20 AI chips, complicating Nvidia’s business in China.
– Regulators, including MIIT, are asking companies to explain why they need Nvidia’s H20 chips over domestic alternatives.
– Some tech firms plan to reduce H20 chip orders due to regulatory scrutiny, according to sources.
– Purchasing Nvidia’s H20 chips is seen as politically sensitive, though not officially banned, per a data center operator.
– Chinese regulators discourage using Nvidia’s chips for government or security-related projects, as reported by Bloomberg.
Chinese authorities are pressing major tech firms like Alibaba and ByteDance to clarify their purchases of Nvidia’s H20 AI chips, signaling tighter scrutiny over foreign semiconductor imports. The move creates fresh challenges for Nvidia’s operations in China despite previous export agreements with U.S. officials.
Industry sources reveal that regulators, including the Ministry of Industry and Information Technology (MIIT), have formally requested justifications for opting for Nvidia’s H20 processors instead of domestically produced alternatives. Several companies that previously relied on these chips are reportedly scaling back orders amid the heightened regulatory pressure.
One executive from a Chinese data center described the situation: “While not officially prohibited, buying these chips now carries political sensitivity.” The stance reflects Beijing’s broader push to reduce dependence on foreign technology, especially for projects tied to government or national security.
Recent reports indicate that Chinese agencies have circulated advisories discouraging the use of Nvidia’s H20 chips in state-linked initiatives. Neither Alibaba, ByteDance, nor MIIT provided immediate comments on the matter. The development underscores the escalating tensions between global tech supply chains and China’s drive for self-sufficiency in critical hardware.
(Source: Ars Technica)
