However, the details are crucial. Because the year-ago quarter was still affected by inventory corrections, a comparison between the prior year's quarter and the current quarter is not entirely fair, points out Dean McCarron, president of Mercury.
"The corrections were largely completed by Q3 2023, so the current quarter results are likely to be consistent with actual CPU sales, and from Q3 2024 most comparisons will be valid - but the annual results are certainly not " he wrote in his report.
"After a year of focusing on clearing out existing inventory, the current quarter's results are likely more reflective of actual demand and market share, rather than inventory changes - there isn't that much inventory left to suppress Intel's sales, so what we're seeing is what CPU customers are really buying, and for Intel they're buying fewer units than a year ago, and for AMD they're buying more," he added in an email.
There are some interesting nuances to Intel's decline. Since the market has shifted to chips with a higher number of cores - up to 64 cores in a CPU - fewer chips are being sold. A two-socket server with 32-core chips can do the work of four 16-core chips, which means fewer servers. It appears that companies prefer fewer physical servers with high core counts than many physical servers with low core counts.
"Intel, which has the majority of the market share and thus the highest volume of low-core devices, had more to lose from a device dispatch perspective as the market converts than AMD does - AMD is basically growing with the move to high-core, and although Intel's new products also does, it loses legacy products and it's not a 1:1 replacement, so the migration to high core counts is more negative for Intel's device and market share than for AMD," McCarron said.