DIGITIMES Research has upwardly revised the 2024 shipment forecasts for high-end AI servers and general AI servers to increase year-on-year by 182.6% and 54.2%, respectively, according to the latest data from the firm's AI server report.
The global generative AI boom that started in the first half of 2023 has since created numerous applications. Demand for multimodal model applications, on-device small language models (SLM), retrieval augmented generation (RAG), and large language model agents (LLM agents) has been rising, generating better-than-expected growth in demand for AI servers that are at the heart of generative AI computing.
In 2024, the top-5 players will account for more than 75% of the high-end AI server shipments from major global cloud service providers (CSP) and brands. The concentration will be lower than the 85% recorded in 2023. Among them, Google will be one of the most active players in deploying AI servers in 2024, with its shipments surpassing Microsoft to take the top place in the ranking, the report's figures show.
For Microsoft, in addition to deepening its cooperation with OpenAI in advanced LLM in 2024 to maintain its leading edge in AI cloud services, it will also actively train on-device SLM that can be run on both PCs and cloud AI for hybrid computing, enabling comprehensive AI services. It has slowed down the procurement of AI servers compared to 2023, but it is faster than competitors in adopting new accelerator solutions.
Benefiting from expanded server procurement orders from first and second-tier CSP customers and its more comprehensive liquid-cooling solutions, Supermicro is expected to surpass Amazon in shipments to rank in third place. Meanwhile, Dell will have the highest shipment growth among all vendors in 2024.
In terms of accelerators used in high-end AI servers, the proportion of high-end AI servers using Nvidia GPUs is expected to drop about 10pp in 2024. HGX H100 will still see the highest share among Nvidia GPUs in 2024, while shipments for GB200 will not pick up until 2025.
Google's heavy investment in AI computing resources will drive its TPU server shipment share to account for almost a quarter of the total volume. AMD MI300X, launched at the end of 2023, has attracted much attention from CSPs and brands due to its high price-performance ratio. It will enjoy a major increase in shipment share in high-end AI servers in 2024, surpassing Amazon's in-house-developed ASIC AI server shipments.
In terms of high-end AI server shipments from manufacturers in 2024, Inventec, which had strong results in 2023, is expected to lead in L6 (server motherboards) shipments in 2024, but its share will decrease significantly. In terms of full systems and racks (L10-12), the ranking will change significantly.
