NVIDIA faces significant challenges due to new export regulations by the United States. These regulations not only affect custom chips like the A800 and H800 but also impact consumer-grade graphics cards like the RTX 4090. In this article, we explore the specifics of the new regulations, how they differ from previous ones, and the implications for NVIDIA and the technology industry.
Previous Export Regulations
Before the new licensing rules, U.S. export restrictions focused on the on-device connection speed. NVIDIA adapted to these regulations by developing custom products based on Ampere and Hopper architectures, namely the A100 and H100 series. These products were initially designed to comply with the previous set of U.S. export regulations.
The New U.S. Export Licensing Rules
On October 17, 2023, the U.S. Government announced a shift in the criteria for export restrictions, moving from connection speed to performance thresholds. The newly introduced Interim Final Rule specifically amends Export Control Classification Numbers (ECCN) 3A090 and 4A090, affecting the export of certain NVIDIA integrated circuits to several countries, most notably China.
Regulatory Details: The Interim Final Rule
The Interim Final Rule imposes additional licensing requirements for exporting to China and other Country Groups like D1, D4, and D5. The restrictions encompass not just chips but also any system incorporating one or more of the specified integrated circuits. This means that even NVIDIA DGX and HGX systems fall under the new export control purview.
Implications for NVIDIA
NVIDIA faces a two-fold challenge. First, there is a considerable number of companies in China involved in the final packaging of consumer-grade RTX 4090 series graphics cards. Second, many board partners operate exclusively in China. The new licensing requirements make it difficult for NVIDIA to engage in these critical aspects of their supply chain.