• Driven by strong initial demand, TSMC’s 3nm node is the fastest to reach full utilization.
• Resilient AI-related products also help TSMC’s 5/4nm node regain its momentum after the inventory adjustment period of Q3 2022 to Q1 2023.
• TSMC’s US investments could represent up to 30% of the total 2nm capacity, thereby sustaining its high utilization rate.

TSMC, the foundry market’s preeminent player, has reinforced its dominance since the inventory adjustments of late 2022, with its leading-edge utilization rates reflecting sustained strength.

According to Counterpoint’s data, the 3nm process saw the strongest initial demand, driven by robust adoption of the Apple A17 Pro/A18 Pro, x86 PC CPUs and other AP SoCs to reach full utilization by the fifth quarter after mass production (Q+5).

Looking ahead, the introduction of NVIDIA’s Rubin GPUs, along with other AI Asics such as Google’s TPU v8 and Aws’s Trainium 3, is expected to sustain this high utilization, fueled by rising demand for AI and high-performance computing (HPC) applications.

In contrast, older nodes like 7/6nm and 5/4nm experienced slower initial ramps, primarily due to their focus on the smartphone market. While 7/6nm’s utilization peaked around 2020, driven by strong mobile demand at that time, 5/4nm began to regain momentum in mid-2023 and has shown steady recovery since then. This recovery was largely driven by surging demand for AI accelerators, particularly from NVIDIA’sH100, B100, B200 and GB200, which are fueling the expansion of AI data centers and contributing to a notable increase in overall utilization rates for the 5/4nm node.

When discussing the potential of the 2nm node, we believe that it is on track to reach full utilization by the fourth quarter (Q+4), faster than any previous node. This is due to strong demand from both smartphone and AI-related applications, echoing TSMC’s statement during the Q1 2025 earnings call: “We expect the number of new tape-outs for 2nm technology in the first two years to be higher than both 3nm and 5/4 nm in their first two years, fueled by both smartphone and HPC applications.”

Beyond Apple, other potential adopters of 2nm technology include Qualcomm, MediaTek, Intel and AMD. This broader adoption is expected to contribute meaningfully to maintaining high utilization rates for the 2nm node.

To mitigate geopolitical risks while meeting rising US consumer demand, TSMC is investing up to $165 billion in its Arizona fabs, which will not only span 4nm and 3nm but also 2nm and advanced nodes.

While core R&D and process development will remain in Taiwan, the US expansion could eventually represent up to 30% of TSMC’s 2nm and beyond capacity. This dual-location strategy strengthens TSMC’s geopolitical resilience while aligning capacity with customer needs, especially in AI and HPC, and positioning the company to sustain high utilization rates across its most advanced nodes through 2030 and beyond.