Alibaba Cloud Leads China's AI Infrastructure Market with 23% Share

Alibaba Cloud has emerged as the leader in China's AI infrastructure market for 2024, capturing 23% of the market share, according to the latest report from IDC. This positions Alibaba Cloud ahead of its competitors, with its market share exceeding the combined total of the second and third-ranked companies. The company's dominance is particularly pronounced in the generative AI infrastructure sector, where it leads in both model training and model inference markets.

AI Infrastructure as a Service (AI IaaS) relies on specialized AI chips like GPUs for computational support. This infrastructure is divided into two categories: generative AI IaaS (GenAI IaaS), which is centered around large models, and traditional non-generative AI IaaS. With the rapid advancement of large models, generative AI IaaS has become the dominant force. IDC's report highlights that in the second half of 2024, China's GenAI IaaS market experienced a year-over-year surge of 165%, reaching 8.74 billion yuan, and for the first time, it accounted for more than 70% of the overall AI IaaS market.

The report also provides a detailed breakdown of the market shares for 2024. Alibaba Cloud holds 23% of the AI IaaS market, followed by Huawei Cloud with 10% and ByteDance's Volcano Engine with 9%. IDC attributes Alibaba Cloud's leadership to its extensive cloud resources and the capabilities of its PAI and Lingjun products, which have positioned it at the forefront of both large model training and inference IaaS markets. The report predicts that leading companies will continue to increase their computational investments to enhance model performance and establish technological and ecological dominance.

Furthermore, the report indicates that in 2024, the generative AI IaaS resources consumed for model training were approximately 3.25 times those used for model inference. As "blockbuster" large models emerge, IDC anticipates that AI computational demand will transition from training-driven to inference-driven by 2025. This shift is expected to create more market opportunities for domestic computational power, particularly in sectors such as the internet, telecommunications, and autonomous driving. Major computational clients in these fields are actively validating, testing, purchasing, and deploying domestic computational power to address the computational gaps in both cloud and edge environments.