Despite US restrictions on the export of advanced artificial intelligence chips, Chinese buyers are resorting to unofficial channels to obtain banned latest Nvidia processors, such as Blackwell and Hopper. This indicates the difficulty of enforcing restrictions amid a fierce global tech race.
Article contents:
- Introduction
- Overview of US restrictions and their impact
- Smuggling methods and exorbitant prices
- Can this workaround be curbed?
- Frequently Asked Questions
Introduction
Amid the global race to develop artificial intelligence technologies, advanced chips have become a strategic element that countries seek to control supply for and restrict access to competitors, creating a parallel market to bypass these restrictions.
Overview of US restrictions and their impact
Since 2022, the US administration has imposed restrictions on the export of advanced chips necessary to run sophisticated artificial intelligence models, aiming to hinder China’s access to this sensitive technology. Despite tightening those rules, an underground commercial network has emerged to smuggle equipment associated with companies like Nvidia into the Chinese market.
Blackwell and Hopper processors are considered among the most advanced processors in the field of artificial intelligence, used to train massive language models and run giant cloud infrastructures. Since the restrictions were activated, demand for these chips in China has increased, and intermediaries in East Asia have stepped in to facilitate smuggling, either by passing parts individually or assembling them into complete systems before shipping.
Smuggling methods and exorbitant prices
Reports indicated that Chinese buyers are bypassing restrictions by using companies registered outside China (in regions such as Malaysia, Vietnam, and Taiwan) that obtain chips and systems from authorized distributors. Afterwards, these systems are sold in the Chinese market at inflated prices; a server containing 8 Blackwell chips can cost more than $600,000, compared to a lower global price.
One example of this: universities in Shenzhen and Wuhan have purchased 6 servers equipped with Hopper chips since December, according to official contracts. Some sellers have also published photos of warehouses full of servers ready for immediate delivery. As Chinese cities head toward a global competition in artificial intelligence, institutions find themselves forced to search for these chips by any means, boosting the prosperity of this parallel market.
Can this workaround be curbed?
US authorities face major challenges in controlling smuggling and bypassing restrictions. Even if one loophole is closed, another opens amid the expansion of the intermediary network and multiple countries and trading platforms. Although some regional governments have tightened oversight, it remains difficult to prevent all flows, especially in small transactions or shipments passed through cooperating intermediaries.
In addition, reports indicate that Chinese companies are striving to develop domestic alternatives, but these efforts still need time to compete with the performance of Nvidia chips. Future results remain hostage to how keen Washington is to tighten and enforce restrictions, and whether China will succeed in making chips that match their American counterparts. Ultimately, the global artificial intelligence landscape remains in a delicate balance between technology, politics, and commerce.
Frequently Asked Questions
1. Why is America blocking the export of Nvidia chips?
It seeks to limit China’s ability to develop advanced artificial intelligence models, given the importance of these chips in training and computing.
2. What are the most prominent processors banned from export to China?
The Hopper family and Blackwell series from Nvidia are the most prominent, given their high performance in artificial intelligence applications.
3. How do Chinese traders bypass the ban?
By using intermediary companies outside China, and transferring products to Chinese entities at a higher price.
4. Are these quantities sufficient to run major projects?
They may not reach the hundreds of thousands of units used by American companies, but they are enough to support some ambitious Chinese projects.
5. What is the role of universities in this race?
Some Chinese universities are purchasing advanced servers to conduct research or train artificial intelligence models on a wider scale.
6. Are prices rising as a result of this demand?
Yes, as sellers exploit the limited market and sell advanced servers with a large profit margin.
7. Is there a local alternative to American chips?
Chinese development in this field is still less advanced than Nvidia, and it is expected to take time to provide alternatives that match current performance.
