With the global explosion of ChatGPT, the demand for various generative artificial intelligence applications has also ushered in rapid growth. According to DigiTimes, TSMC has won a large number of artificial intelligence chip orders with its sub-7nm process.
TSMC has seen an increase in Nvidia’s AI chip orders over the past few months, the sources said. Relevant media also quoted market sources as saying that orders for Nvidia A100 and H100, two high-performance GPUs for data centers, are increasing, and they have also increased their investment in TSMC.
The source also pointed out that the demand for Nvidia AI chips from major companies around the world has been strong. For example, major companies such as Alibaba Cloud, Baidu AICloud and Tencent Cloud in China are actively acquiring artificial intelligence chips; Demand for Nvidia’s AI chips was also strong.
According to reports, Nvidia’s H100 GPU uses TSMC’s 4nm process and A100’s 7nm process. Nvidia’s current top H100 Tensor Core GPU adopts the Hopper architecture and is manufactured based on TSMC’s 4nm process. It is equipped with the fourth-generation Tensor Core and Transformer engines. The card is equipped with 80GB of video memory, which doubles the computing efficiency per kilowatt compared to the A100 GPU.
Nvidia’s DGX H100 system features eight H100 GPUs connected via NVIDIA NVLink to form a single unit. The NVIDIA H100 system has been used by many well-known universities and technology companies in the world.
TSMC currently sees no real competition in the sub-7nm process that supports AI chip manufacturing. Intel has returned to the foundry space but has yet to make any meaningful progress.
