TSMC won a large number of orders for AI chips by virtue of sub-7nm technology, and Nvidia A100 and H100 emergency orders were added.

Thanks to CTOnews.com netizens for the delivery of clues on the way! CTOnews.com May 23 news, with the global popularity of ChatGPT, the demand for a variety of generative artificial intelligence applications has also ushered in a rapid growth. According to the Electronic Times, TSMC has won a large number of orders for artificial intelligence chips with its sub-7nm process.

TSMC has seen an increase in orders for AI chips from Nvidia in the past few months, according to sources. Relevant media also quoted market sources as saying that orders for the two high-performance GPU models Nvidia A100 and H100 for data centers are increasing, and they have also increased the number of films in TSMC.

The source also pointed out that the demand for Nvidia AI chips by major companies around the world has been very strong, such as China's Aliyun, Baidu AICloud and Tencent Cloud and other big companies are actively acquiring artificial intelligence chips; and cloud service providers in other parts of the world also have strong demand for Nvidia AI chips.

According to reports, Nvidia H100 GPU uses TSMC 4nm process, A100 is 7nm process, and Nvidia's top H100 Tensor Core GPU uses Hopper architecture, based on TSMC 4nm process manufacturing, it is equipped with the fourth generation Tensor Core and Transformer engine, single card is equipped with 80GB memory, compared to the A100 GPU per kilowatt computing efficiency is increased by two times.

Nvidia's DGX H100 system carries eight H100 GPU, which can be formed as a whole through NVIDIA NVLink connections. As far as CTOnews.com knows, Nvidia H100 system has been used by many well-known universities and technology companies in the world.

TSMC currently sees no real competitors in supporting the sub-7nm process for AI chip manufacturing. Intel has returned to foundry, but has not made any meaningful progress.