Image source: NVIDIA
According to JPR, there are currently seven GPU architecture developers in the market, as well as 20 companies creating discrete, integrated and embedded GPUs. Most of these solutions are entry-level integrated GPUs, with only a few companies developing discrete GPUs for gaming graphics cards and compute accelerators.
Most of the revenue in the global GPU market comes from solutions that are not used for graphics processing: in quantitative terms, sales of GPUs for AI and high-performance computing (HPC) workloads are only a few million units per year, but since they cost tens of thousands of dollars per unit, the sale of such products brings tens of billions of dollars to Nvidia, and billions to AMD.
Image Source: AMD
In just two quarters of the current fiscal year, Nvidia’s revenue from GPU shipments for AI and HPC was $42 billion, and for the full year these figures could increase to $90 billion. AMD expects revenue from the sale of its AI accelerators to exceed $3 billion. Speaking of other market participants, such as Biren or MetaX, they still lag significantly behind the leaders in terms of income.
«GPUs have become ubiquitous and can be found in virtually every industrial, scientific, commercial and consumer product manufactured today,” said Dr. Jon Peddie, president of Jon Peddie Research. “Some market segments, such as AI, have made headlines for their rapid growth and high average selling price (ASP), but they are small in volume compared to other market segments.”