Frost & Sullivan has released the China AI4S Research Intelligent Computing Infrastructure Market Research, 2026. The report defines AI4S research intelligent computing infrastructure as the integrated computing foundation that supports scientific research, engineering simulation, and the training and inference of scientific foundation models at universities, research institutes, and scientific platforms, spanning intelligent computing clusters, high-speed networks, parallel storage, software platforms, scheduling systems, and research toolchains. As scientific research moves into petabyte-scale data, tens of millions of samples, and simulations of hundreds of millions of atoms, the industry is shifting from pilot projects to large-scale deployment. The potential market for AI4S infrastructure at Chinese research institutions is expected to grow from about RMB 42 billion in 2025 to about RMB 76 billion in 2027, and to reach about RMB 148 billion by 2030.
The report identifies three development trends. First, infrastructure is moving from separately built supercomputing and AI computing centers toward converged HPC-AI platforms that support both high-precision scientific computing and AI training and inference. Second, the focus is shifting from accumulating computing power to releasing effective computing power, with parallel efficiency, storage throughput, job scheduling, and long-term stable operation becoming key measures. Third, competition is moving from construction scale to application outcomes, with model accuracy, R&D cycles, and research output becoming the main evaluation criteria. In terms of adoption, use by individual research groups accounts for the largest share at 34% of universities and research institutions, while only 15% have built institution-level platforms, leaving substantial room for platform development.
On the competitive landscape, the report finds that competitive barriers are evolving from single-hardware performance toward system-level capabilities spanning customer projects, effective computing power, system delivery, platform software, research scenarios, and security and controllability. In the near term, vendors with strong capabilities in complex system delivery, platform productization, and foundational software adaptation are better positioned to win large-scale projects. Over the longer term, vendors that can turn computing resources into capabilities that are usable for research, sustainable in operation, and replicable at scale are expected to gain greater room for growth as AI4S moves from pilots to large-scale adoption.
