Despite the US government’s stringent export restrictions, top Chinese universities and research institutes have quietly obtained high-end Nvidia artificial intelligence (AI) chips. A recent review by LA News Center has uncovered a concerning pattern.
**Trail of Tenders Reveals Forbidden Purchases**
An analysis of hundreds of tender documents reveals that at least 10 Chinese entities have acquired advanced Nvidia chips in recent months. These chips were embedded in server products manufactured by reputable companies like Super Micro Computer Inc., Dell Technologies Inc., and Gigabyte Technology Co. Ltd.
Nvidia’s high-end AI chips remain in high demand in China despite export restrictions. (Image: Wong Yu Liang | Moment | Getty Images)
The purchases occurred after the US expanded its embargo on November 17th, 2022. The ban prohibits the sale of certain chips to China, including Nvidia’s most advanced technology. However, it’s important to note that the purchase and possession of these chips are not illegal within China.
**Cloud of Secrecy Surrounds Chip Sellers**
The 11 sellers who supplied these chips to Chinese entities are relatively unknown Chinese retailers. LA News Center was unable to determine whether they had stockpiled the chips prior to the US export restrictions or had acquired them through other means.
**Nvidia: No Violation Evidence**
Nvidia has stated that the tenders in question specify products that were exported and widely available before the restrictions. The company maintains that none of its partners violated the export control rules, with the chips representing a negligible fraction of worldwide sales.
**Supply Chain Challenges: A Loophole for Diversion?**
Daniel Gerkin, a partner at Kirkland & Ellis, believes that Nvidia chips could have been diverted to China without the manufacturer’s knowledge due to limited visibility into downstream supply chains. He suggests that thorough due diligence on the manufacturer’s part could make it difficult for the US government to enforce penalties.
**US Monitoring Diversion Attempts**
The US Commerce Department emphasized its commitment to monitoring the diversion of restricted chips, including through end-use checks and investigations into potential breaches. The department stated that credible allegations of violations would be thoroughly investigated.
**Chinese Buyers and Sellers Remain Silent**
The Chinese buyers and retail sellers involved in the tender transactions have chosen to remain silent and have not responded to inquiries from LA News Center.
**Experts: Chips Valuable for AI Training and Research**
Analysts and industry executives indicate that the acquired chips, though limited in quantity, could prove useful for training AI models and conducting advanced research. The transactions raise concerns about China’s continued access to advanced AI technology and its potential military applications.
**Companies Respond: Compliance and Investigations**
Super Micro Computer Inc. claims to adhere to US requirements on the sale and export of GPU systems and will investigate any reported violations. Dell Technologies denies shipping products to the listed entities with the restricted chips but will continue to investigate. Gigabyte Technology Co. Ltd. asserts compliance with Taiwanese and international regulations.
**Limited Snapshot, Wider Implications**
The transactions exposed by the tenders only represent a small fraction of China’s chip purchases. However, they serve as a reminder that advanced technology may still be flowing into the country despite export restrictions. This raises concerns about the potential development of AI-powered military applications in China.
**US Commitment to Enforcement**
Companies and individuals found guilty of violating US export controls face substantial penalties, including fines and prison sentences. The US government remains vigilant in its efforts to enforce these restrictions and prevent the unauthorized transfer of sensitive technology to countries perceived as a threat to national security.
Data sourced from: cnbc.com
