Taiwan’s Electronic Information Manufacturing Industry: 28% of Enterprises Implement AI
The Industrial Technology Research Institute (ITRI) recently announced the results of the “AI Adoption in Taiwan’s Electronic Information Manufacturing Industry” survey, revealing that 28% of enterprises in the sector have already implemented AI, while another 46% are in the planning stage. Among various sub-industries, the application of AI is more mature in the PCB, optoelectronic materials, and components sectors, while the development in other electronic components, computers, and peripheral devices is relatively slower.
The survey indicates that large enterprises are ahead of small and medium-sized enterprises (SMEs) in AI deployment progress. However, SMEs are accelerating their efforts, with an expected compound annual growth rate (CAGR) of 26% in AI investment from 2024 to 2026. ITRI industry analyst Zhang Jiafu stated that the main goals for manufacturers implementing AI are to improve performance and reduce costs, with key performance indicators including yield rates, production capacity, time-to-market, and costs.
Enterprises that have implemented AI are expected to invest an average of NT$2.09 million in 2024, increasing to NT$2.36 million in 2025, and reaching NT$2.61 million in 2026, resulting in a three-year CAGR of 11.5%. Among them, about 40% of enterprises are continuing to increase their AI investments, with 46% expected to raise their budgets in 2025 and 39% in 2026.
In terms of resource allocation, hardware expenditures in the AI sector will account for the highest proportion in 2025, reaching 46%, followed by software (42%), while services will account for the lowest proportion at 12%. This indicates that Taiwan’s automation industry possesses strong competitiveness in the hardware field, with promising future business opportunities.
Observation 1: Discriminative AI is Mainstream, Generative AI’s Potential Yet to be Developed
ITRI pointed out that the manufacturing industry continues to focus its AI technology investments on discriminative AI, with a budget allocation ratio of 73% for discriminative AI in 2025, significantly higher than the 27% for generative AI. In 2026, the investment share for generative AI is expected to slightly increase to 29%. Focusing on the current AI application status in manufacturing units, it is found that the number of enterprises using discriminative AI is 1.6 times that of those using generative AI.
Although current applications of generative AI mainly focus on “product development report generation,” with relatively low satisfaction, advancements in AI agents and human-machine collaboration technologies are expected to expand generative AI applications to more manufacturing and production processes in the future. Zhang Jiafu suggests that solution providers should continue developing related applications to seize emerging market opportunities.
Observation 2: Quality Inspection and Production are the Mainstream AI Applications
Among the top ten AI applications, half are related to manufacturing production, with the top three being defect detection, defect image labeling, and production process improvement. The demand for AI in the manufacturing production department remains the highest, followed by product development and quality inspection departments, indicating that these areas will further widen the gap in intelligence compared to other departments.
According to ITRI’s survey, the top ten AI applications are: defect detection, defect image labeling, production process improvement, product development report, root cause analysis of defects, production scheduling planning, design defect inspection, safety incident analysis, optimization of process parameters, and root cause analysis of production issues.
Observation 3: IT Department as the Core Driver of AI Implementation
In enterprises that have implemented AI, the IT department has the fastest AI development progress, with a practice rate of 60%, indicating that the IT department is generally the driver of digital transformation in enterprises. Following this are the manufacturing production and quality inspection departments, which are also the most concentrated areas for AI applications at present.
AI Implementation Satisfaction and Challenges: Data is the Core Issue
The survey shows that enterprises have varying degrees of satisfaction with the effectiveness of AI implementation, with the most significant improvements noted in increased revenue, alleviated labor shortages, and reduced costs. However, satisfaction regarding “improving the predictability of issues” is relatively low. Zhang Jiafu analyzes that the underwhelming predictive capabilities of AI may be influenced by factors such as market supply and demand fluctuations, political and economic environments, and the enterprises’ own data preparedness.
ITRI points out that data issues remain the biggest challenge in AI development for the manufacturing industry, with 80% of enterprises that have implemented AI facing data-related difficulties, particularly large enterprises where complex organizational structures make data management more challenging. Additionally, enterprises still planning to implement AI mainly face high costs and difficulties in assessing benefits.
Industry analyst Zhang Jiafu states that data readiness can only be verified through actual execution; insufficient data can lead to poor model performance, while excessive data without effective governance can also hinder the full potential of AI. Enterprises should start with the end in mind, confirming the AI application scenarios first, then planning the necessary data to ensure data volume, quality, and governance are in place to enhance the accuracy and effectiveness of AI applications.
This article is reprinted with permission from: Future Business