Shenzhen: Apoiar produtos de chips nucleares domésticos como GPU, NPU, CPU, DPU para acelerar a aplicação e iteração

Shenzhen Municipal Bureau of Industry and Information Technology recently issued the “Action Plan for Accelerating the High-Quality Development of the Artificial Intelligence Server Industry Chain (2026–2028)”, which proposes supporting domestic GPU, NPU, CPU, DPU and other core chip products to accelerate application and iteration, forming a complete industrial ecosystem. It encourages R&D and industrialization of high-performance computing chips based on RISC-V architecture, development of high-speed Ethernet switch chips, PCIe/CXL switch chips, etc., to promote technological iteration and performance upgrades. It supports the development and application of third-generation semiconductor power devices, the development of scenario-specific power supply solutions, and forward-looking layout of next-generation power semiconductor technologies.

Shenzhen City’s Action Plan for Accelerating the High-Quality Development of the Artificial Intelligence Server Industry Chain (2026-2028)

In order to seize the commanding heights of artificial intelligence industry technological development, accelerate the high-quality development of the AI server (hereinafter referred to as AI server) industry chain, and pioneer the creation of a new form of intelligent economy, Shenzhen is committed to building a global AI pioneer city. In accordance with the spirit of documents such as the “Opinions of the State Council on Deepening the Implementation of the ‘Artificial Intelligence+’ Action,” and based on the actual situation of our city, this action plan has been formulated.

1. Development Goals

Leverage Shenzhen’s advantages in new-generation electronic information, artificial intelligence, new energy, and other industries. Adhere to innovation-driven, software-hardware synergy, ecological cultivation, and open cooperation. Promote innovation clustering across the entire AI server industry chain, including R&D design, core components, complete machine manufacturing, system integration, and supporting services. Facilitate collaboration among large, medium, and small enterprises, close cooperation between upstream and downstream companies, and integrated production, supply, and sales links. Build the most complete industry support system, the most efficient resource allocation, and the best innovation ecology for AI servers. Establish a large-scale intelligent computing cluster R&D and manufacturing hub, providing solid support for Shenzhen to become a leading global advanced manufacturing center and an influential industry technology innovation center.

By 2028, the capacity and shipment volume of products across the entire AI server industry chain will achieve leapfrog growth. The global market share of key products such as core chips, storage, printed circuit boards (PCBs), power storage, optical modules, and passive components will significantly increase. Breakthroughs in core chip and key product technologies will be achieved, cultivating a batch of technically proficient and market-advantaged “specialized, refined, distinctive, and innovative” (“specialized and innovative”) enterprises and “single champion” companies. An industrial echelon of large, medium, and small enterprises working in synergy will be formed.

2. Key Areas

(1) Core Chips. Support domestic GPU, NPU, CPU, DPU and other core chip products to accelerate application and iteration, forming a complete industrial ecosystem. Encourage R&D and industrialization of high-performance computing chips based on RISC-V architecture. Develop high-speed Ethernet switch chips, PCIe/CXL switch chips, etc., to promote technological iteration and performance upgrades. Support third-generation semiconductor power device R&D and application, develop scenario-specific power supply solutions, and proactively layout next-generation power semiconductor technologies.

(2) Storage Products. Support R&D and application of storage chips, accelerate advanced packaging technology breakthroughs for storage chips, and focus on high-end storage products such as enterprise SSDs and memory modules. Strengthen research on near-memory packaging and integrated compute-storage technologies, improve high-end storage supply capacity, and build a supply system for high-end storage suitable for large model training and supercomputing centers.

(3) Printed Circuit Boards. Relying on Shenzhen’s global PCB manufacturing core base advantages, focus on developing multi-layer high-speed boards for AI servers, rigid-flex combination boards, flexible boards, advanced packaging substrates, 6-layer and above ultra-high-density HDI boards, and ABF substrates. Promote large-scale application of cutting-edge low-dielectric high-end substrates. Strengthen R&D and capacity layout for high-end carrier boards and flexible circuit boards, expand small-scale and customized PCB services, and support enterprise R&D and pilot testing. Accelerate applications of special PCBs (for high-speed high-frequency communications), FCBGA/BT packaging substrates, etc., to support efficient interconnection of data centers and backbone networks.

(4) Power Supply and Energy Storage. Accelerate the empowerment of “lithium battery+” advantages, continuously improve UPS, 800V high-voltage DC power supplies, lithium/sodium energy storage systems, and other key products. Strengthen “power electronics+” capabilities, promote iteration of new HVDC power supply solutions, and accelerate R&D of high-voltage SiC MOSFETs, high-frequency GaN power devices, multi-phase controllers, intelligent power modules (DrMOS), and highly integrated power modules. Tailor solutions to domestic and international market demands, create benchmark zero-carbon data centers with “photovoltaic/offshore wind + energy storage + green power direct connection,” and support local consumption of computing power and efficient use of green electricity. Innovate in power supply energy storage participation in virtual power plants and explore integrated source-grid-load-storage business models.

(5) Optical Modules. Promote optical modules from 800G to 1.6T/3.2T generations, support mass production projects for 800G and above optical modules. Focus on developing high-speed, low-power silicon photonic modules, CPO/LPO/NPO packaged optical modules, and breakthroughs in core technologies such as high-end thin-film lithium niobate and high-end indium phosphide. Promote evolution and industrialization of all-optical switching technology, and enhance自主研发能力 of core materials, optical chips, and optical devices.

(6) Cooling Products. Consolidate China’s leading position in cold plate liquid cooling, accelerate development of immersion cooling, chip-level direct cooling, embedded microchannel liquid cooling, and other advanced cooling technologies. Improve the energy efficiency of high-end air cooling systems, layout phase-change energy storage cooling, thermoelectric cooling, and other frontier directions. Focus on developing liquid cooling cabinets, cold plates, liquid cooling plates, high thermal conductivity interface materials, and cooling modules. Strengthen supply of high-end core components such as VC uniform heat plates, 3D VC, and high-efficiency thermal gels.

(7) Passive Components. Focus on developing high-speed connectors, large current power inductors, high-capacitance MLCC, low ESR capacitors at high frequencies, 3D silicon-based capacitors, and other core products for AI servers. Promote iterative upgrades of high-speed connectors, power inductors, and magnetic core materials, and improve mass production capacity of integrated inductors and solid-state polymer capacitors.

(8) Complete AI Servers. Relying on Shenzhen’s advantages in complete machine R&D, manufacturing, and ecological support, develop core products such as all-liquid cooling AI super nodes, high-end AI training servers, AI inference servers, and large model training and inference integrated machines. Actively layout lightweight AI integrated machines and mini hosts. Support the development of liquid-cooled CDU server clusters and connection systems, strengthen and expand the complete machine integration industry, and enhance ODM/OEM/JDM manufacturing services. Build a full-chain service system covering design, manufacturing, testing, and delivery, and cultivate globally influential AI server brands.

3. Key Tasks

(1) Support Major Projects. Strengthen city and district coordination, utilize national policies such as “new infrastructure,” “double innovation,” and financial tools, to support technological transformation and major project planning. Deeply explore industry chain enterprise needs, encourage capital investment and expansion. Provide full-process services for high-end enterprise expansion projects, and accelerate project implementation.

(2) Guarantee Industrial Space by Tier and Category. Based on enterprise needs, ensure industrial space through tiered and categorized approaches. City and district coordination to select high-end R&D and manufacturing projects, and strengthen industrial land supply. Promote models like “industrial building upwards” to quickly provide space for lightweight manufacturing, R&D testing, and flexible production.

(3) Deepen Industry Chain Collaboration and Innovation. Explore “chain leader sets questions, upstream and downstream answer” collaborative innovation models, leverage chain leader enterprises to drive joint R&D and technological breakthroughs. Focus on core technologies for new demands in AI servers, and explore industrialization. Support upstream enterprises to join leading supply chains through policies like first sets, first batches, and first editions, to accelerate domestic equipment, components, materials, and software application.

(4) Enhance Resilience and Security of Industry and Supply Chains. Implement national strategies for high-quality development of key industry chains, guide enterprises to strengthen and stabilize supply chains. Utilize international trading centers for electronic components and integrated circuits for multi-source reserves, trading, procurement, and cross-border services to ensure supply stability. Attract upstream core materials, components, and enterprise headquarters, R&D centers, and manufacturing bases in Shenzhen, to optimize the industry layout.

(5) Build a Healthy Industry Ecosystem. Promote regular cooperation mechanisms between chain leader enterprises, cloud service providers (CSP), and large model companies. Accelerate deep adaptation of AI computing infrastructure technology systems, support the healthy development of domestic AI computing ecology. Promote large-scale application of domestic computing chips, attract supporting enterprises, and foster a virtuous cycle of “application-driven technological iteration.” Support deep integration of open-source Hongmeng, open-source Euler, and RISC-V architecture, creating a new ecosystem of operating systems and RISC-V in the AI server field.

(6) Promote Manufacturing and Service Industry Collaboration. Support the development of clusters in complete machine manufacturing and computing infrastructure, including system integration, computing leasing, model training, and operation management. Build integrated solutions of “hardware + algorithms + services” that are full-stack, green, and reliable. Support OEMs to transform into system service providers, offering full lifecycle services from planning and deployment to intelligent operation and maintenance. Encourage cloud service providers and AI companies to participate deeply in hardware customization, creating a positive cycle of demand-driven supply and supply-driven demand. Leverage Shenzhen’s software and information service advantages to promote the coordinated development of AI servers, large industry models, and intelligent computing platforms, building a globally leading AI computing infrastructure ecosystem.

(7) Optimize Full-Cycle Investment and Financing Services. Strengthen support for SMEs, focus on early and mid-stage industry chain startups, and promote government investment funds to target high-quality AI server projects. Promote deep cooperation between government funds and industry venture capital (CVC), providing funding and supply chain support. Establish platforms for bank-enterprise and investor-enterprise matchmaking, reduce financing costs, and assist companies in accessing capital markets and listing.

(8) Improve Talent Development System. Encourage local universities to add relevant majors and expand enrollment. Deepen industry-education integration, jointly build industry colleges and training bases with top enterprises and universities, and train skilled talents. Use talent policies to improve housing, healthcare, and other supporting services to attract and retain technical professionals, stabilizing the workforce.

4. Support Measures

(1) Focus on Task Implementation. City-level coordination to develop overall strategies, proactively serve enterprises, and strengthen financial and investment support; districts to clarify responsibilities, implement measures, and ensure industrial land, facilities, and financing needs are met. Use models like “industrial building upwards” to quickly match high-quality industrial space and ensure project efficiency.

(2) Improve Work Mechanisms. Under the “20+8” strategic emerging industry cluster framework, promote close cooperation among industry authorities, planning, natural resources, environmental protection, water departments, and districts. Fully align with national and provincial AI and infrastructure plans, mobilize resources, and support the development of AI server industry chain enterprises and major projects, strengthening policy and resource coordination.

(3) Strengthen Support Systems. Support the construction of public technology pilot platforms, integrating R&D, testing, IP, compliance, and enterprise services to provide full-cycle, inclusive professional services. Leverage industry associations and research institutions to enhance industry research, exchanges, and talent training, empowering industry development.

（Source: People’s Financial News）