Shenzhen's government work report this year outlined plans to strengthen the development of a fully independent and controllable AI software and hardware ecosystem. The city will concentrate efforts on making breakthroughs in core technologies such as algorithm theory, model architecture, AI chips, foundational software, and intelligent robotics. It aims to effectively utilize major national AI training infrastructure like Pengcheng Cloud Brain III and the second phase of the National Supercomputing Center in Shenzhen, striving to create a world-leading new paradigm for the integration of AI and advanced manufacturing.

Shenzhen is moving beyond the simple addition of "manufacturing plus AI" as a tool. Instead, it is adopting a development logic for its AI industry centered on "large AI models + advanced manufacturing + application scenario innovation + ecosystem development." This approach aims to empower high-quality manufacturing development and forge new productive forces, charting an unconventional path through comprehensive exploration to establish a world-class model for AI-manufacturing fusion.

In traditional manufacturing workshops, scenarios where experienced masters manually adjust parameters and conduct visual inspections for defects are being replaced by AI-driven intelligent production modes. In Shenzhen, AI is no longer just an "add-on" to manufacturing processes but is making comprehensive inroads into core areas such as R&D design, production scheduling, and supply chain management. This shift enables a transition from "post-facto remediation" to "preemptive prediction and real-time control."

At the lithium battery production facility of leading new energy manufacturer Sunwoda, an intelligent optimization system built on deep reinforcement learning operates 24/7, directing robotic arms. This system has replaced the traditional production line's reliance on manual shutdowns for parameter testing.

Yuan Mengfei, Deputy General Manager of the AI Research Institute at Sunwoda Electronic Co., Ltd., stated that traditional parameter adjustment was not only time-consuming but also led to quality fluctuations. The AI system has reduced optimization time by 66% and improved the process capability index (CPK) by 29%, effectively resolving production challenges caused by environmental variations and equipment restarts.

The deep integration of AI and advanced manufacturing relies on comprehensive ecosystem support. Leveraging its trillion-yuan industrial clusters in next-generation information communication and new energy vehicles, Shenzhen possesses a massive base of real industrial application scenarios and rich data. Combined with the underlying large models, embodied intelligence software/hardware, and computing power provided by leading local tech firms, Shenzhen's foundation for developing "AI + Advanced Manufacturing" is solid.

The "Shenzhen Action Plan for 'AI+' Advanced Manufacturing (2026-2027)," released in February, sets a clear goal: by 2027, establish a development framework characterized by "one base, one center, one alliance, one hundred scenarios, and multiple applications" in the "AI+Advanced Manufacturing" sector. This not only outlines a clear roadmap but also directly addresses common challenges in manufacturing's intelligent transformation.

Li Enhan, Director of the Token Digital Economy Research Center at the China Development Institute (Shenzhen), commented that the core logic of this layout is targeting pervasive challenges in manufacturing's smart transformation, such as "difficulty understanding the technology, high implementation costs, and difficulty finding suitable application scenarios." It aims to bridge the gap between technological R&D and industrial application through institutional innovation. The National AI Application Pilot Base is intended to facilitate the pilot testing phase for moving technology from the lab to the production line. The Industrial Intelligent Body Innovation Center is tasked with breakthroughs in digital employee technology, providing core technical support. The Industrial Software and Industrial Knowledge Alliance will transform tacit manufacturing knowledge and experience into replicable software and models, significantly lowering the barrier to entry for small and medium-sized enterprises (SMEs).

Currently, many manufacturing enterprises in Shenzhen, especially SMEs, face three common pain points in AI implementation: significant data silos, high retrofitting costs, and a shortage of interdisciplinary talent.

Li Enhan suggests that SMEs should not blindly pursue building their own general-purpose large models. The key solution lies in utilizing platform-based services and applying large and small models collaboratively. By leveraging open industry ecosystems and adopting modular solutions that combine vertical industry-specific large models with specialized small models or intelligent agents, and with guidance from government and leading chain enterprises, trial-and-error and deployment costs can be reduced, enabling SMEs to dare to and be able to use AI.

From Yuan Mengfei's perspective, the development of "AI+Advanced Manufacturing" requires breakthroughs in three key dimensions: data, computing power/algorithms, and application scenarios. On the data front, multi-modal data fusion is essential. Regarding computing power and algorithms, the challenge of excessive computational consumption by large models must be addressed by compressing them into lightweight models suitable for the real-time requirements of industrial settings. Collaboration between industrial manufacturing, universities, and professional research institutions is poised to become a major trend in industrial AI transformation. Leading enterprises can provide the industrial validation and implementation support needed for technological achievements from universities and startups, while also accelerating their own development by leveraging cutting-edge technologies.

Li Enhan believes that while appropriately increasing the openness of industrial application scenarios, Shenzhen should guide enterprises towards differentiated development to avoid excessive intra-industry competition. Furthermore, the government could take the lead in establishing credible industrial data pilot platforms and third-party safety and ethics evaluation platforms.