Taiwan Semiconductor Manufacturing is making a record capital expenditure bet on AI chip demand, yet even the world's largest foundry concedes it cannot keep pace with the soaring requirements. During an earnings call this week, the company announced a projected capital expenditure of $56 billion for 2026, with all funds directed towards constructing new fabrication plants and upgrading existing production lines. However, CEO C.C. Wei explicitly stated that despite this massive investment, the company anticipates supply shortages will persist until 2027 and potentially beyond, rather than being confined to 2026.
This declaration starkly highlights a core contradiction within the semiconductor industry during the current AI super-cycle: the speed of demand expansion far outstrips the timeline for building new production capacity. The entire supply chain, from GPUs and CPUs to memory, and extending to supporting materials like voltage regulators, integrated circuits, and cables, is currently experiencing shortages. This is constraining the ability of major clients like Nvidia, AMD, and Apple to renew their wafer orders.
To address the relentlessly increasing demand for its 3-nanometer process technology, Taiwan Semiconductor Manufacturing is executing a parallel global capacity expansion strategy. In Taiwan, the company will add a new 3nm fab within the GIGAFAB cluster at the Tainan Science Park, with mass production expected in the first half of 2027. In Arizona, USA, a second fab also utilizing 3nm process technology has completed construction, with mass production scheduled for the second half of 2027. In Japan, the company plans to upgrade its second fab to 3nm capability, targeting mass production around 2028.
CEO C.C. Wei noted on the earnings call that beyond new fabs, the company is continuously converting existing 5nm equipment in Taiwan to support 3nm output. It is also implementing flexible capacity allocation mechanisms across its N7, N5, and N3 technology nodes to maximize support for all customers. He emphasized that Taiwan Semiconductor Manufacturing will not favor any single client over others amidst the capacity crunch.
The primary driver of this supply tightness is the comprehensive explosion in AI demand. The rise of agentic AI has significantly increased requirements for high-bandwidth memory and LPDDR. The consumption of computing infrastructure by AI applications has now spread from data centers to various end markets, including automotive and the Internet of Things.
Ongoing wafer order updates from tech giants like Nvidia, AMD, and Apple are further intensifying the capacity pressure on Taiwan Semiconductor Manufacturing. Mr. Wei pointed out that once existing production lines reach peak capacity, upgrade projects will be initiated to supplement incremental demand, but this process requires significant time.
These supply bottlenecks are also prompting some manufacturers to diversify their foundry partnerships. Reports suggest Tesla is collaborating with both Taiwan Semiconductor Manufacturing and Samsung on its next-generation AI chip and is advancing a Terafab cooperation plan with Intel. Intel is also rumored to be poised to win a major client this year with its 14A process technology. Meanwhile, Samsung's foundry business is fielding more customer inquiries, though its current focus remains concentrated on memory chip production like HBM.
Taiwan Semiconductor Manufacturing's $56 billion capital expenditure plan is substantial, but the timeline from fab construction to mass production typically spans several years. This means that even with funding secured, a rapid increase in output is unlikely. The new fabs in Taiwan and the US will not achieve mass production until the second half of 2027 at the earliest, with the Japan facility following in 2028.
This timeline indicates that the tight supply situation for AI chips will most likely continue for the next two to three years. Pressures on delivery lead times and procurement costs across the related supply chain are unlikely to ease quickly. For technology companies reliant on advanced process chips, capabilities in supply chain management and capacity securing will become critical variables influencing their competitive standing.
Comments