China gets Nvidia GPU for the AI race through Vietnam and Malaysia
Deepseek recently challenged the AI community and shook the US, revealing an escalating rivalry between the United States and China in the field of artificial intelligence. The competition has entered a new phase, with Washington tightening export controls on advanced semiconductor technologies since September 2022. Aimed squarely at curbing Beijing’s access to cutting-edge AI capabilities, the restrictions target high-performance graphics processing units (GPUs) produced by US chipmaker Nvidia, including the A100 and H100 models—critical components for training large-scale AI models. Despite these curbs, AEOW/INPACT has uncovered how China is succeeding in acquiring restricted chips through two intermediary countries in Asia. By analyzing export trade data, AEOW/INPACT has identified six recent shipments of NVIDIA A100 and H100 GPUs, valued at a combined $642,060, destined for two Chinese companies.
The Vietnamese trail to Suzhou Etron Technologies Co Ltd
The first shipment trail runs through Vietnam. The country gets its share of A100 GPU from Nvidia Singapore branch via Taiwan according to trade data. In July and October 2024, a Vietnamese company called Etron Vietnam Technologies company limited. According to IPC, a trade association for electronics, Etron Vietnam is linked to the website “etron-global.com” which belongs to Chinese company ETRON Global.
The company is shipping 8x NVIDIA A100 GPUs and 16x NVIDIA A100 GPUs to a company called Elb International in Hong Kong. The company has been incorporated in Hong Kong in 2019.
According to a press article, Elb International Ltd is actually a subsidiary of Suzhou Etron Technologies Co Ltd, listed on the Shanghai Stock Exchange (code: 603380 SH) and was set up to do procurement of electronics and expand overseas markets. The company which is a top electronics manufacturer located in Suzhou is also linked with the website etron.cn which has the same page as Etron Global.
The case underscores how unsanctioned Chinese firms are exploiting their regional footholds across Asia to funnel restricted AI components back into China, circumventing export controls.
The Malaysian trail to Chengdu AI hub
The second shipment trail runs through Malaysia. In February and March 2025, a Malaysian logistics firm, United Despatch Agent Sdn Bhd, exported two consignments labelled “A100 GPU baseboard” and “Geforce graphics NVIDIA H100.” The former refers to the full NVIDIA A100 GPU card—an advanced semiconductor device designed for AI training, inference, and high-performance computing. The latter, however, is a misnomer: no such product as a “GeForce NVIDIA H100” exists. GeForce is NVIDIA’s consumer-facing line, tailored for gaming and creative applications, while the H100 belongs to its data centre-grade Hopper architecture, developed for AI workloads. The suspicious labelling raises the prospect of deliberate obfuscation—possibly to disguise the shipment of restricted H100 GPUs amid tightening US export controls.
According to shipping records, the goods were flown to China via Malaysia Airlines. United Despatch Agent, which describes itself as a provider of end-to-end logistics solutions with a focus on electronics, appears to have acted as the freight forwarder. There is no information on the actual Malaysian supplier. The freight forwarder firm’s involvement adds a layer of plausible deniability to the transaction, potentially complicating enforcement efforts around export restrictions.
While the supplier cannot be identified, it has been possible to detect the import of A100 GPUs in Malaysia in 2024 to various Malaysian companies (Neview Technology Company Limited and M & S international trade technolog – Txi logistics M sdn bhd is a logistics company) from an Indian company based in Mumbai: Beehive Tech Solutions Private Limited
Looking at trade data, Indian companies are importing A100 GPUs from Nvidia Singapore branch via Taiwan.
The Chinese company buying GPUs is Chengdu Huo Feng Technology co. AEOW/INPACT has not been able to find the company but its address is in the High Tech Zone of Chengdu. Interestingly, Chengdu High Tech Zone has been chosen as a national hub for the development of artificial intelligence with several projects and an AI business ecosystem. On May 10th 2022, Huawei and several other Chinese companies signed contracts for the launch of the AI Computing Center.
The GPU could likely benefit to Huawei AI Computer Center and other local developments. The Center is the largest in southwestern China and will include platforms for AI computing, smart cities, and scientific innovation. It has the computing power of 300 petaflops.
These two circumvention schemes offer a clear illustration of how China continues to acquire critical AI components through indirect channels, underscoring its determination to stay competitive in the global race for artificial intelligence dominance.
It also highlights the limits of export control as they are right now and the porous nature of the semiconductor supply chain in the Indo-Pacific region.