The DIGITIMES Research Center released a report on June 25th, stating that the global output value of server GPUs (including storage chips, boards, and subsystems) will exceed $100 billion for the first time in 2024, reaching $121.9 billion. Among them, the proportion of high-end server GPU output value will exceed 80%, reaching 102.2 billion US dollars, with a shipment volume of 4.82 million units, NVIDIA accounting for 92.5%, and AMD accounting for 7.3%.
DIGITIMES analysts say that generative artificial intelligence (AI) is still in its early stages of development, and cloud service providers (CSPs) are still actively reserving computing power. In addition to self-developed server ASIC accelerators, they are also purchasing a large number of high-end server GPUs to seize business opportunities. In addition to CSP, large global enterprises are also actively laying out generative AI services. Although their single order volume is not large, the large quantity will also drive up Nvidia's high-end AI server shipments in 2024. In addition, governments around the world are actively laying out computing infrastructure.
The institution stated that it is worth noting that Nvidia's Max GPU data center is less competitive than its competitors, resulting in lower order volumes. Intel has chosen to switch to the Gaudi acceleration chip.
In addition, Nvidia China's special edition AI GPU H20 is expected to obtain an export license from the US Department of Commerce, and it is expected that H20 will account for 97.9% of the high-end server GPU market in China.
