Benefiting from a surge in orders from major clients such as Nvidia, AMD, and Broadcom, TSMC is expected to see its AI-related revenue grow exponentially this year, poised to challenge the NT$1 trillion mark. The company is also projected to continue growing next year, setting a new record for the third consecutive year. This is expected to propel TSMC's USD revenue to surpass the US$100 billion mark in 2026, with order visibility extending to 2028.
Analysts say that the widespread adoption of AI startups by TSMC's customer base has kept TSMC fully booked, with AI revenue expected to exceed US$40 billion next year.
Industry sources indicate that TSMC's customer base continues to book capacity for advanced processes below 3nm, with not only Apple but also non-Apple companies showing strong interest. Although TSMC has announced plans to reflect sales value to clarify actual demand, this hasn't deterred the momentum of capacity bookings driven by the AI arms race, with orders extending into 2028.
In terms of technology applications, TSMC's most advanced 2nm process has entered mass production as scheduled. The company anticipates rapid growth in its 2nm business in 2026, driven by smartphones, high-speed computing, and AI applications. Furthermore, TSMC has already partnered with all global AI innovation partners. TSMC Chairman and President C.C. Wei previously mentioned at an earnings call that TSMC's server AI processor revenue will account for more than 20% of its total revenue by 2028.
However, *The Economist* points out that while the AI spending boom has significantly increased chip demand, it has also raised concerns about increasingly tight chip supply. TSMC's production increase rate is not meeting customer expectations, mainly because TSMC is taking a cautious approach to large-scale production increases based on its past experience with the boom-and-bust cycles of the semiconductor industry
*The Economist* notes that the chip industry's long-standing business cycle has been characterized by initial supply shortages, followed by expansion to meet demand, only to face overcapacity when demand cools. Moreover, expanding capacity is costly and slow; an advanced wafer fab costs approximately $20 billion and takes three to four years, with even higher costs in the US.
In particular, TSMC increased investment during the COVID-19 pandemic to alleviate shortages of mature process chips, but this capacity is now underutilized, raising concerns about a repeat of the past. Furthermore, TSMC may also be worried that if Intel and Samsung Electronics eventually resolve their production issues, the AI chip market could turn into a supply-demand imbalance. However, chip designers calling for increased capacity will not share the costs of building fabs, and while customers may be frustrated by TSMC's cautious approach, they should not expect TSMC to change its stance.
TSMC Dominates Advanced Process Technologies
Research firm TrendForce predicts that the semiconductor foundry industry will grow by approximately 20% in 2026. Advanced processes, benefiting from high-performance computing (HPC) demand, will lead the market with a 31% year-on-year growth rate, showing a polarized development trend with mature processes. The market expects TSMC, the world's leading semiconductor foundry, to dominate advanced processes globally. AI chip orders are driving supply shortages, and continued capacity expansion will boost Taiwan's semiconductor supply chain growth next year.
Regarding the semiconductor foundry market in 2026, TrendForce believes that geopolitics will lead to a two-sided landscape: advanced processes dominating while mature processes remain in a stalemate. The industry is projected to grow by approximately 20% in 2026, with advanced processes leading the market due to HPC demand. Advanced technologies, including front-end manufacturing and back-end packaging and testing, are becoming scarce resources in the short to medium term due to strong AI demand and a limited number of suppliers, leading to rising prices year by year.
TrendForce indicates that while demand for mature processes will increase as supply chains replenish inventory for various applications, overall growth momentum is limited due to a lack of strong innovative applications in the consumer market, coupled with multiple manufacturers opening new capacity. Furthermore, uneven regional resource allocation is emerging, leading to a continued price decline in 2026. Cost pressures and sales challenges are placing a double burden on foundries, making the effective allocation of regional capacity to meet localized demand a key challenge.
In addition, with the growth in AI computing power demand, advanced packaging areas are continuously expanding, extending from CoWoS to EMIB, CoWoP, and CoPoS. Competition and cooperation between foundries and packaging companies, and the corresponding supply chain coordination, are also crucial topics.
TrendForce further points out that driven by strong AI demand, the semiconductor IC industry landscape is undergoing significant changes, with the monopolistic position of leading semiconductor manufacturers becoming increasingly apparent.
In non-AI applications, the overall market demand growth is limited due to the sluggish global economy, and even the inventory adjustment cycle is getting longer. In particular, in the automotive and industrial semiconductor fields, market competition is becoming increasingly fierce. Chinese semiconductor manufacturers have broken the monopoly of the first-tier semiconductor giants in the mature process fields due to the increased localization rate.
Source: Content compiled from Semiconductor Industry Observer
