4 Trillion New Goals Set! Shanghai Releases a Major Plan for the 14th Five-Year Plan in Software: What Key Signals Are Being Released?

Recently, the Shanghai Municipal Commission of Economy and Information Technology officially issued the "Shanghai Plan for the Development of Software and Information Services during the 14th Five-Year Period." This important document not only outlines an ambitious blueprint for future industrial development but also sends a strong signal to promote technological innovation and the deep integration of industries.

According to the plan, by 2030, Shanghai will position the software and information services industry as a "power source" for economic growth, the main battlefield for AI applications, and a stronghold for global competition. In specific development goals, the industry scale is expected to reach 4 trillion yuan, and the added value of the industry will exceed 1.1 trillion yuan. At the same time, the quality and efficiency of the industry will be further improved, with the number of enterprises with revenue exceeding 10 billion yuan increasing to 35, and a group of emerging enterprises with industry ecosystem leadership and potential leading power will be cultivated.

To achieve this grand goal, Shanghai has made comprehensive arrangements in technological innovation and industrial layout. In terms of technical routes, the plan proposes to focus on next-generation model architectures, and actively explore multiple technical routes based on non-Transformer architectures such as state space models, recurrent neural network variants, and liquid neural networks. It also aims to accelerate the layout of cutting-edge basic model technologies such as physical intelligence, world models, quantum intelligence, and brain-like intelligence.

In terms of intelligent computing hardware supply, Shanghai will focus on overcoming the networking technology of ultra-large-scale intelligent computing clusters. It will comprehensively enhance the supply capacity of core areas such as high-performance computing chips, quantum chips, high-speed optical interconnects, high-bandwidth memory, and heterogeneous servers, and vigorously promote the deep integration between self-developed chips and mainstream large models. In addition, focusing on new storage retrieval and data-model collaboration, it will break through key technologies such as high-precision heterogeneous processing, native multi-modal fusion, and dynamic value alignment, and fully build automated complex reasoning capabilities covering the entire lifecycle of training data.