You are currently viewing SemiWiki as a guest which gives you limited access to the site. To view blog comments and experience other SemiWiki features you must be a registered member. Registration is fast, simple, and absolutely free so please, join our community today!
With NVIDIA's recent blowout quarter, there's no doubt that they will have nearly all of the AI accelerator market for the medium term.
Will there be cascades down into AMD's MI250, MI300 lines and Intel's Gaudi and Ponte Vecchio products as H100 lead times are nearing one year?
Pat Gelsinger mentions "Out pipeline of opportunities through 2024 is rapidly increasing and is now over $1billion and continuing to expand with Gaudi driving the lion's share."
Super Micro also mentioned in their most recent earnings that "We are also getting good traction on Intel's Guadi 2 and PVC, as well as AMD's MI250 accelerator solutions"
The near term AI boom are all for LLM. Gaudi performs very poor in LLM training though it does well in other AI/ML task. Since all CapEx are allocated to LLM training, other AI/ML market could actually suffer.
A lot of capex is going for inference as well. Things like Microsoft co-pilot seem aimed to support several hundred million customers of Office, and the models are too big to download and keep updated even if client machines had the processing. That points to a large fleet in the cloud for the inferencing. Google and Meta have similar needs. It is not all just training.
