ASICs for AI on the rise at GPUs’ expense
Cette publication existe aussi en Français
Application-specific AI inference accelerators are gaining market share in the datacenter and this is benefiting chip vendors Broadcom and Marvell, rather than the dominant AI player Nvidia, according to The Information Network.
The semiconductor analyst reckons that in 2023 GPUs had 72 percent of the market for AI acceleration while ASICs had 22 percent. In 2025 the numbers will have moved significantly with GPUs losing 7 and ASICs gaining 8 percentage points. That trend will continue through 2027, the analyst estimates.
AI accelerator market share by chip type. Source: The Information Network.
With its GPUs able to attack lots of tasks in a fast-moving market Nvidia is the dominant vendor in the market for AI acceleration in the data center. However, these are primarily used for the development and training of models. Increasingly ASICs optimized for inference are making an impact said Robert Castellano, founder and principal analyst with The Information Network.
At the same time the most general purpose chips, CPUs and FPGAs are expected to occupy about 5 to 7 percent of the market.
Both Broadcom and Marvell announced strong earnings in December 2024 driven by ASIC sales into AI applications but there is nothing to prevent Nvidia competing in the ASICs market.
Related links and articles:
News articles:
TSMC, Broadcom mull deals to carve up Intel
Graphic shows Nvidia, US domination of chip industry
Broadcom first to add AI to network switch chip
Marvell partners with Micron, Samsung, SK Hynix on custom HBM
If you enjoyed this article, you will like the following ones: don't miss them by subscribing to :
