Nvidia has generated 1000% profit from its H100 GPUs

AI acceleration is poised to become one of the most profitable hardware industries in the coming months and years, and Nvidia is well positioned to capture a sizable piece of…

By Nivedita Bangari

August 19, 2023

2 min read

Comments

AI acceleration is poised to become one of the most profitable hardware industries in the coming months and years, and Nvidia is well positioned to capture a sizable piece of this market. The H100 data centre GPU is already generating significant income for the Santa Clara-based business.

If the forecasts are correct, this may be the start of an unprecedented golden era for Nvidia’s GPU business. According to Kim, the demand for H100 GPU units is so high that they are virtually sold out until 2024. Meanwhile, AI firms are racing to get enough GPU accelerators to run their generative models and AI-powered services.

According to Foxconn, the AI server industry will be worth $150 billion by 2027, and these modern AI servers rely significantly on the powerful computing capabilities of the latest Nvidia technology.

Inside, there’s a new steering wheel, a 10.25-inch touchscreen infotainment system, and a redesigned digital dashboard. Buttons appear to have been removed entirely from the dashboard. All controls are accomplished using a touch-sensitive console or touchscreen. Tata Motors may possibly provide a facelift for the Nexon with ADAS safety features, making it the brand’s third model after the Harrier and Safari’s Red Dark Edition.

The Hopper microarchitecture underpins the H100 GPU, which was created as the data centre counterpart to the Ada Lovelace architecture that drives the latest generation of GeForce RTX gaming GPUs. However, as evidenced by performance benchmarks, Hopper’s primary interest is not gaming. The H100 accelerator includes a modified GH100 GPU with 14,592 CUDA cores, as well as 80GB of HBM3 RAM with a 5,120-bit memory bus.

Nvidia is undoubtedly riding a high wave in the current AI boom, yet Kim’s recommended $3,320 estimated cost per H100 GPU warrants further elaboration. Creating a new GPU is an expensive and time-consuming endeavour that requires the participation of multiple hardware engineers and specialised specialists. These employees must be compensated, and current estimates show that the average wage for a hardware engineer at Nvidia is roughly $202,000 per year.

On the one hand, building and selling H100 GPUs could be substantially more expensive than the claimed $3,320 cost. Kim, on the other hand, has been following Nvidia for 30 years and is currently preparing a book that he describes as the “definitive history” of the firm.

Also Read:

Source