en.Wedoany.com Report on Mar 25th, The Shenzhen Municipal Industry and Information Technology Bureau recently issued the "Shenzhen City Action Plan for Accelerating the High-Quality Development of the Artificial Intelligence Server Industry Chain (2026-2028)". It clearly outlines a series of technological breakthroughs and industrial upgrading measures centered around memory chips, a key link in the AI server industry chain.
The plan proposes supporting the R&D and application of memory chips, accelerating breakthroughs in advanced packaging technologies for memory chips, and prioritizing the development of high-end storage products such as enterprise-grade SSDs and enterprise-grade memory modules. Simultaneously, it emphasizes strengthening R&D for novel technologies like near-memory packaging and compute-in-memory, enhancing the supply capacity for high-end storage, and building a supply system for high-end storage products suitable for large model training and supercomputing centers.
With the explosive growth in computing power demand for AI large model training and inference, storage systems are becoming one of the key bottlenecks constraining the overall performance of AI servers. Large model training requires high-speed reading and writing of massive data, posing challenges for traditional storage architectures in terms of bandwidth, latency, and capacity. As core storage components of AI servers, the performance of enterprise-grade SSDs and enterprise-grade memory modules directly impacts the efficiency and cost of model training. Advanced packaging technologies further improve storage bandwidth and energy efficiency by optimizing interconnect density and signal transmission efficiency between chips.
As a major hub for China's electronic information industry, Shenzhen has formed a relatively complete industrial chain in areas such as memory chip design, packaging and testing, and module manufacturing. This newly issued action plan aims to seize the window of opportunity presented by the expanding AI server market to promote the high-end upgrade of the local storage industry. By tackling advanced packaging technologies and developing enterprise-grade high-end products, Shenzhen strives to establish core competitiveness in the niche track of AI storage.
The plan also proposes strengthening R&D for novel technologies like near-memory packaging and compute-in-memory. Near-memory packaging reduces data movement latency by shortening the physical distance between processors and memory; compute-in-memory attempts to perform calculations directly within the memory cells, fundamentally breaking through the "memory wall" bottleneck. The layout of these cutting-edge technologies will build up technical capabilities for next-generation AI computing architectures.
As AI computing power demand extends from training to inference, and model parameter counts continue to grow, market demand for high-end storage products will continue to be released. This three-year action plan issued by Shenzhen not only points out the direction for technological breakthroughs for local storage enterprises but also provides important support for the autonomous and controllable development of the national AI server industry chain.
