TCL TECH.(SZ:000100) Nan Fang Du Shi Bao·2025-12-13 13:59Core Insights - The focus of the manufacturing industry is shifting towards the practical application of AI, aiming to bridge the "last mile" of AI implementation [1][2] - TCL anticipates that its AI applications will generate comprehensive benefits exceeding 1 billion yuan by 2025, with R&D expenses projected to reach 15 billion yuan for the year [1] - The company emphasizes the importance of investing in core competencies despite the reduced costs of AI training due to open-source models [2] Group 1: AI Implementation and Investment - TCL's CTO, Yan Xiaolin, highlighted that while open-source models like DeepSeek have significantly lowered training costs, the company plans to substantially increase its budget for large model training next year [2] - The company aims to develop "domain-specific multimodal large models," with a focus on in-house training for critical processes such as domain training and reinforcement learning [2] - TCL's R&D center in Poland is taking on significant AI development tasks, with plans to establish a 50-person team in Eastern Europe to collaborate with the Shenzhen team [2] Group 2: AI Applications and Efficiency Gains - TCL's "Xingzhi X-Intelligence 3.0" model has been successfully applied in various industrial scenarios, surpassing the capabilities of DeepSeek R1-671B in the semiconductor display domain [3] - The integration of AI in automated design has drastically reduced circuit board design cycles from 15 days to 3 days, and layout design from 3 days to 3 hours, achieving quality comparable to senior engineers [3] - The transition from Automatic Defect Classification (ADC) to Automatic Defect Repair (ADR) has enabled full automation in inspection, judgment, and repair, yielding over 50 million yuan in annual economic benefits [3] Group 3: Future of AI and Industry Perspectives - Experts indicate that current AI models have reached a "graduate level" in knowledge acquisition and understanding, with the next competitive frontier being embodied intelligence [4] - TCL maintains a cautious stance regarding the implementation of embodied intelligence in industrial settings, acknowledging that while progress has been made, there is still a significant gap to achieve the required reliability of 99.9% [4] - The industry is expected to see a pivotal year ahead as leading manufacturing companies shift their focus from technical metrics to tangible economic benefits from AI [4]