AMD beats Intel in the data center market for the first time - but NVIDIA earns many times that amount

For the first time in history, AMD has earned more from data center processors than Intel. However, NVIDIA has generated nearly four times more in this market than both processor giants combined.

For over two decades, Intel held a dominant position in the data center chip market. Intel Xeon processors were used in the vast majority of servers, while AMD only captured a small percentage a few years ago. However, the landscape has shifted dramatically: although Xeon still operates in most servers, the most expensive systems now utilize AMD EPYC processors. According to SemiAnalysis, AMD's data center division has outperformed Intel's equivalent division.

AMD's data center segment posted a revenue of $3.549 billion in the third quarter of this year, while Intel earned $3.3 billion in this sector. Just two years ago, Intel's DCAI group was earning $5-6 billion each quarter. As AMD EPYC gained competitive advantages over Xeon, Intel had to sell its server chips at significant discounts, which reduced profitability.

The flagship 128-core Intel Xeon 6980P Granite Rapids processor is priced at $17,800, making it the company’s most expensive standard processor. In contrast, AMD's most expensive 96-core processor, the EPYC 6979P, costs “only” $11,805. If demand for the 6900 series Xeon remains high and the company can supply these processors in large quantities, Intel’s data center profits may likely rebound and surpass AMD's sales. However, Intel still needs to ramp up production of its Granite Rapids products.

At the same time, NVIDIA is earning significantly more from data center graphics processors and network chips, which are essential for the integrated operation of AI processors. NVIDIA's network product sales reached $3.668 billion for the quarter, while sales for computing GPUs soared to $22.604 billion, greatly exceeding the combined revenues of Intel and AMD mentioned above. In total, NVIDIA sold nearly $42 billion in GPUs for AI and HPC in the first half of this year, and it is likely the company will sell even more in the latter half of the year.

Source: Tom's Hardware