TSMC's 2nm Costs Surge: AI Chip Market Impact

As the year 2023 draws to a close, Taiwan Semiconductor Manufacturing Company (TSMC) faces an imminent rise in production costs for its cutting-edge semiconductor manufacturing processes. Currently at the forefront with its 3nm technology, a recent report underscores the anticipated significant cost escalation for the upcoming 3nm and 2nm nodes. Industry analysts posit that these potential cost increases may impact the profit margins of a range of technology products, including both high-end and entry-level devices, with a notable influence on Apple's offerings.

The 3nm process, currently the global pinnacle of semiconductor production, is predominantly utilized by TSMC, a major supplier for Apple's latest products. Reports from 2022 highlighted the estimated production cost of TSMC's 3nm wafers at approximately $20,000 per unit, with subsequent information suggesting a potential increase to $25,000 or a decrease below $20,000 per wafer. The emergence of artificial intelligence (AI) chips in 2023 has disrupted the semiconductor landscape, influencing cost calculations for products and shaping expectations for investments in high-cost chip manufacturing facilities.

While an IBS report forecasts a 50% increase in the price of 2nm products, reaching $30,000 per wafer, the consequential impact on Apple is expected to be significant. However, should the prevailing AI trend fully materialize over the next three years, companies such as AMD and NVIDIA may also encounter budgetary constraints.

NVIDIA, for instance, has already introduced AI products based on TSMC's 5nm architecture. Industry sources suggest that NVIDIA's forthcoming GPU, built on the 3nm process, will incur high chip costs during the initial production phase of 2nm, placing a direct burden on Apple. Despite the high wafer costs associated with the initial mass production of advanced chip technology, AMD and NVIDIA, due to the robust power of their processors and GPUs tailored for intensive workloads, are traditionally less susceptible to these cost fluctuations.

Nonetheless, the intricate supply and demand dynamics within the semiconductor market may drive an escalation in prices for AI chips. If demand surpasses the production capacity of foundries, the $20,000 cost per 3nm wafer could potentially rise further. Foundries such as TSMC and Samsung, boasting fixed production capacities, may consider passing on the initial costs of investing in new facilities to customers, presuming a confidence that demand will outstrip supply.

Especially in light of TSMC's strategic recalibration following a challenging 2022, a more measured approach to production expansion is anticipated. TSMC's periodic adjustments to revenue expectations for 2023, coupled with the need to manage optimistic customer expectations amid an overall industry-wide slowdown in chip production in 2022, underscores the company's cautious stance.