Nvidia and AMD Banned From Supplying China and Russia

The U.S. Securities and Exchange Commission has notified Nvidia that it cannot export its latest data centre GPUs to Russia and China without a license.

According to the latest notice from the U.S. Securities and Exchange Commission (SEC), Nvidia will need a license to export any systems such as A100 GPUs or DGX/HGX based on the latest architecture to China and Russia. The newly specified requirements also apply to the upcoming H100 (Hopper) GPU or products that achieve similar performance.

AMD received a similar request to cut off high-end GPU chips from Chinese customers. A company spokesperson confirmed that the licensing requirements also apply to AMD. However, the company does not believe these requirements will have significant impacts on the business.

How powerful are the A100 and H100?

The A100 offers a 20x performance boost over Nvidia’s previous-generation Volta GPU, making it ideal for artificial intelligence, data analytics, scientific computing, and cloud graphics workloads. Composed of 54 billion transistors, the chip packs a third-generation Tensor core with acceleration for sparse matrix operations, especially useful for AI inference and training. In addition, each GPU can be divided into multiple instances to perform different inference tasks, and multiple A100 GPUs can be used for larger AI inference workloads using Nvidia NVLink interconnection technology.