Nvidia's AI GPU Orders Surge, TSMC Production Ramps Up

According to a report by Digitimes, market sources indicate that orders for Nvidia's A100 and H100 AI GPUs have increased, leading to an increase in the wafer production at TSMC (Taiwan Semiconductor Manufacturing Company).

Nvidia recently announced a partnership with ServiceNow to develop powerful enterprise-level AI capabilities. These capabilities will transform business processes by leveraging automated workflows to make them faster and smarter.

ServiceNow is utilizing Nvidia's software, services, and accelerated infrastructure to develop customized Large Language Models (LLMs) for their intelligent end-to-end digital transformation platform. Additionally, ServiceNow is utilizing Nvidia's data customization and running foundational models on their hybrid cloud infrastructure to assist Nvidia in streamlining their IT operations. This infrastructure comprises Nvidia DGX Cloud and on-premises Nvidia DGX SuperPOD AI.

The DGX H100 series from Nvidia started shipping in May and has been receiving significant orders. The DGX H100 is a component of the Tokyo-1 supercomputer in Japan, which employs simulation and AI to accelerate the drug discovery process. The average performance of each Nvidia H100 Tensor Core GPU in the DGX H100 series is six times higher than previous GPUs. The DGX H100 is equipped with 8 GPUs, each featuring a transformer engine to accelerate the generation of AI models. Compared to the DGX A100, the DGX H100 has twice the energy efficiency per exaflop.

The H100 GPU is based on the Hopper architecture and utilizes TSMC's 4nm process, while the A100 is based on TSMC's 7nm process. Nvidia has modified both models to H800 and A800 for export to the Chinese market. Nvidia is reportedly closely following TSMC, and its order backlog is said to be full. Nvidia's orders will help TSMC fill the capacity gap caused by reduced orders from MediaTek.