Taiwan Semiconductor Manufacturing Co. (TSMC) announced on April 21, 2026, that it will increase its capital expenditure to a record $56 billion for the 2026 fiscal year. Despite this massive investment aimed at expanding fabrication capacity, CEO C.C. Wei stated during the company’s annual technology symposium that the global supply of high-end artificial intelligence chips is expected to remain constrained beyond 2027. The announcement highlights the persistent challenges in scaling production to match the rapid growth of the AI sector.
The $56 billion budget represents a significant escalation from previous spending cycles, focusing primarily on the expansion of 2-nanometer (2nm) and 3-nanometer (3nm) production lines. TSMC confirmed that a substantial portion of this capital will be directed toward the construction of new GigaFabs in Taiwan and the acceleration of equipment procurement for its international sites. Specifically, the company is fast-tracking the installation of High-NA Extreme Ultraviolet (EUV) lithography machines, which are essential for the next generation of sub-2nm transistors. TSMC aims to increase its 3nm capacity by 50% by the end of 2026 to meet the surging demand from major clients including Nvidia, Apple, and AMD.
CEO C.C. Wei addressed the supply-demand gap by identifying advanced packaging as the primary bottleneck. While wafer fabrication capacity is expanding, the complexity of Chip-on-Wafer-on-Substrate (CoWoS) technology continues to limit the final output of AI accelerators. Wei explained that the lead times for specialized manufacturing equipment and the intricate nature of multi-chip integration mean that supply parity is unlikely to be achieved before the end of 2027. He noted that even with the company’s efforts to triple its advanced packaging capacity over the next two years, the demand from data center operators continues to outpace facility readiness.
To address these constraints, TSMC is allocating approximately $12 billion of the 2026 CapEx specifically to advanced packaging facilities. This includes the expansion of its AP6 and AP7 plants in Taiwan and the integration of advanced testing services. Senior Vice President of Operations, Y.P. Chin, noted that the transition to 2nm mass production, scheduled to ramp significantly throughout 2026, requires unprecedented levels of precision and cleanroom space. The company is also investing in dedicated power infrastructure at its Science Park locations to support the high energy requirements of these advanced nodes.
TSMC’s financial outlook remains tied to the AI transition, with AI server processors now accounting for approximately 25% of the company’s total revenue as of the first quarter of 2026. This is a marked increase from the 15% reported in the previous year. The company reiterated its long-term revenue growth target of 15% to 20% in U.S. dollar terms, citing the structural shift toward high-performance computing. Despite the projected shortages, TSMC officials emphasized that the company remains on track with its global expansion plans, including the development of its third fab in Arizona and its first facility in Dresden, Germany.