【重磅干货】AI芯片生存指南，2025TrendForce集邦咨询半导体产业高层论精华分享

Core Viewpoint - The TSS Semiconductor Industry Forum highlighted the significant impact of AI on the global semiconductor market, emphasizing the need for strategic adjustments across various sectors within the industry [1][3]. Group 1: AI's Influence on Semiconductor Demand - The demand for high-performance computing chips driven by AI applications is expected to grow significantly, with advanced processes and packaging technologies being the primary growth drivers in the foundry sector, projected to grow by 19.1% in 2025 [6]. - Advanced process technology, particularly 2nm, will begin mass production in the second half of this year, while advanced packaging capacity is expected to expand with an annual growth rate of 76% [6]. - Despite global geopolitical pressures, the resilience of AI development is evident, with transformations in AI servers and models driving future industry growth [6]. Group 2: Opportunities and Challenges in IC Design - The global IC design market is projected to reach $647.3 billion in 2024, reflecting a year-on-year growth of 25.6%, largely fueled by strong AI demand [9]. - While AI applications are thriving, non-AI sectors are experiencing limited growth due to a sluggish global economy, leading to prolonged inventory adjustments, particularly in automotive and industrial semiconductor markets [9]. - Chinese semiconductor manufacturers are gaining ground in mature process technologies, breaking the previous monopolies held by leading firms [9]. Group 3: Storage Demand in the AI Era - AI's high requirements for flash memory performance, capacity, and energy efficiency are pushing manufacturers to invest heavily in R&D for advanced NAND Flash technologies [12]. - The intense competition in the flash memory market is pressuring manufacturers to reduce costs while meeting AI's high-performance demands [12]. - The explosive growth in AI demand may lead to short-term supply shortages and price increases, but potential market corrections could result in oversupply and price declines [12]. Group 4: Trends in the Robotics Industry - The AI wave is driving the continuous upgrade of smart terminal devices, with humanoid robots expected to become a key export in the next generation of computing, potentially reaching a global market value of $4 billion by 2028 [15]. - Chips are critical components in humanoid robots, responsible for data processing, AI inference, and motion control, which will determine their intelligence and application depth [15]. - The evolution of AI models and the increasing demand for edge computing will push humanoid robot chips towards higher efficiency, lower power consumption, and greater integration [15]. Group 5: HBM Market Dynamics - HBM, characterized by TSV and stacking technologies, is expected to see significant advancements in bandwidth and capacity due to process node upgrades [18]. - The HBM market is projected to see the HBM/e generation account for over 90% of shipments by 2025, with suppliers needing collaborative development capabilities to enhance their technological barriers [18]. - NVIDIA continues to dominate the HBM consumption market, and the balance between supply and demand remains crucial as HBM's share of overall memory capacity increases [18]. Group 6: AI Server Market Insights - The AI server market is evolving, with major cloud service providers expanding their in-house ASIC chip development capabilities [21]. - NVIDIA's recent introduction of the NVLink Fusion solution indicates its strategic move into the ASIC domain to enhance its technology ecosystem [21]. - TrendForce provides in-depth observations on the development opportunities in the AI market and the long-term trends for AI servers [21]. Group 7: Wide Bandgap Semiconductor Trends - SiC is establishing a leadership position in high-voltage applications, particularly in electric vehicles and industrial sectors, despite short-term pressures in the EV market [24]. - GaN is approaching large-scale application, transitioning from low to high-power applications, with significant potential in automotive and AI data centers [24]. - The shift towards 8-inch wafers is expected to dominate the GaN market, with 12-inch wafers anticipated to enter mass production within the next decade [24].