As the renowned American investment bank Raymond James told the tech journalist Tae Kim, who writes for Bloomberg, Dow Jones & Company and Barron’s, among others, has reported that Nvidia’s profits on each (G)H100 series AI accelerator sold are estimated to be almost 1,000 percent. The company could soon do without other sectors, such as the graphics cards from the Geforce series, without any problems.
Estimated 1,000 percent gain per AI accelerator
The internationally operating financial services company estimates that the AI accelerator Nvidia H100 (“Hopper”), which is currently in demand worldwide, costs the manufacturer around 3,300 US dollars to produce, and the professional accelerator with 80 GiByte HBM3 is sold at prices of 25,000 to 30,000 US dollars -Dollar, while in Germany 30,000 euros are called for. In 2023 alone, Nvidia is expected to generate $16.5 billion in sales in the sector.
But the average price achieved by Nvidia for the AI accelerator is significantly higher, because the H800, a trimmed version of the H100, which was specially designed for the Chinese market, is currently sold for around 70,000 US dollars.
Nvidia is fully focused on AI computing
At SIGGRAPH 2023, Nvidia once again made it clear that the company will focus more and more on the business area with professional GPGPUs (“General Purpose Computation on Graphics Processing Units”) for accelerating artificial intelligence. New products of all sizes were presented here.
New AI superchips and AI supercomputers for 2024
The GH200 superchip (“Grace Hopper”) with 144 GiByte HBM3e for AI computing and the DGX supercomputer are also aimed at this market segment, which has meanwhile demoted the Geforce graphics cards addressed to gamers to secondary actors.
Sources: Raymond James via Tae Kim via Twitter
