TAIPEI (Taiwan News) — TSMC is considering raising the production price of its 4nm chips made in Arizona by 30%.
The chipmaker is also reportedly preparing to raise its average wafer foundry prices for clients by about 10%, per Commercial Times. Nvidia CEO Jensen Huang (黃仁勳) backed the price adjustment on Friday, saying that while TSMC’s advanced process technologies are expensive, they are “very worthwhile.”
Industry sources said the planned price hikes stem from multiple factors, including higher fab construction costs in the US, global inflation, exchange rate fluctuations, and rising investments in advanced technologies. TSMC’s Arizona fab is reportedly mulling the 30% increase to reflect the higher production costs associated with “Made in USA” chips.
TSMC Chair and CEO C.C. Wei (魏哲家) recently said during an earnings call that revenue from AI processors—including CPUs, AI accelerators, and GPUs—is expected to double this year. From 2024 to 2029, the compound annual growth rate is projected to approach 45%.
Wei added that strong market demand is driving the growth. To sustain high capital expenditures while maintaining healthy profit margins, TSMC sees a pricing adjustment as necessary.
Supply chain sources noted that Nvidia is already among the clients lined up for TSMC’s 2nm process, which it plans to use for its next-generation AI chips. As process nodes continue to shrink, complexity and the risk of failure increase, making higher wafer prices increasingly common.
According to electronic design automation firms, the success rate for initial tape-outs—handoffs of chip designs for manufacturing—is rapidly declining. Once around 30%, the rate dropped to 24% in 2023–2024 and is expected to fall to 14% in 2025.
The decline is attributed to more customized chip designs, longer design verification cycles, and growing R&D and financial risks.
Industry insiders said that as process thresholds rise, foundries and customers are becoming more tightly integrated. Switching foundries entails significant time, cost, and manpower, leading many IC design companies to favor long-term partnerships with trusted suppliers to remain competitive in the AI innovation race.
TSMC reportedly continues to apply a “buy more, save more” pricing model. While the price hike presents some pressure for clients like Nvidia, they can still benefit from their large-scale collaboration through technology access and priority production scheduling.
