TechSoda

Taiwan Semiconductor Manufacturing Co (TSMC), the world's leading contract chipmaker, announced on May 27 the establishment of a new chip design center in Munich, Germany. This strategic move, confirmed at TSMC's 2025 Technology Symposium, signals a significant shift in the semiconductor industry towards AI-driven chip creation and is expected to bolster innovation, particularly within the European market.

The Munich Design Center is slated to open in the third quarter of 2025. Paul de Bot, President of TSMC Europe, stated that the center's focus will be on supporting European customers in designing high-density, high-performance, and energy-efficient chips, with key applications in automotive, industrial, AI, and IoT sectors.

This announcement comes as AI becomes an increasingly integral and essential part of chip design, especially for chips at 28 nanometers and below. The smaller nodes are crucial for the high transistor density and low power consumption required by demanding AI applications. According to Bob O'Donnell of Technalysis, over 50% of advanced silicon designs are now being realized with significant AI assistance.

The benefits of integrating AI into chip design are substantial, including significant gains in efficiency and time savings, potentially cutting weeks off project timelines. AI excels at tasks like verification and physical layout, rapidly crunching possibilities and exploring more options than humans can in the same timeframe. This not only boosts ROI and competitiveness but also leads to dramatic productivity jumps, sometimes 5x or even 10x improvements, enabling design teams to handle more projects and foster faster innovation cycles, resulting in better-performing chips.

AI is also enabling the acceleration of hardware needed for generative AI, high-performance computing, and data centers, while also making customized designs at scale more feasible. This means the creation of specialized chips for specific AI tasks is becoming more practical and cost-effective. The industry is witnessing a "virtuous cycle" where current AI is used to design tomorrow's AI chips, a feedback loop projected to drive the semiconductor market to a trillion dollars by 2029, largely fueled by Gen AI and HPC needs.

Looking ahead, AI methods are expanding beyond just optimizing chip components to designing and analyzing entire systems. There's also discussion of "agentic AI," which could manage entire design workflows more autonomously, freeing engineers for path-finding research. These AI platforms are also expected to drive breakthroughs in infrastructure AI, physical AI, and scientific AI, along with enhancing energy efficiency in chip design for applications like IoT devices and AI robots.

The Munich design center's potential involvement in developing AI-optimized chips, possibly for the European market or specific AI tasks, aligns with Europe's strengths in automotive, IoT, and industrial applications. TSMC's advanced packaging technologies, such as CoWoS and InFO, are crucial for building the complex multi-die designs required for demanding AI workloads, suggesting the Munich center could leverage this expertise for powerful integrated AI chip solutions.

This new center is also expected to strengthen collaborations with European companies on AI-driven designs, potentially shortening the path from AI concepts to working silicon for local tech companies like NXP. TSMC is also involved in a joint venture, European Semiconductor Manufacturing Co (ESMC), to build a microchip manufacturing plant in Dresden, Germany, with partners Infineon, NXP, and Robert Bosch. TSMC executives noted that both the ESMC plant and the design center could assist in meeting Europe's AI chip ambitions.

The establishment of the Munich design center is not merely a strategic expansion for TSMC but a profound embedding into the European ecosystem to drive the next wave of AI hardware innovation. This geographical spread of design expertise could foster new partnerships and unexpected innovations in a world increasingly reliant on tailored AI hardware.

資料來源