TSMC AI Revenue to Hit NT$1T, Orders Through 2028

According to reports from Taiwan, TSMC is seeing a surge in high-value orders from major AI players such as NVIDIA, AMD, Broadcom, and several ASIC innovators, driving its AI-related revenue to multiply this year and push toward NT$1 trillion. Momentum is expected to continue into next year, setting TSMC up for three straight years of record highs and keeping its USD revenue above US$100 billion through 2026, with solid order visibility already extending to 2028.

In 2024, when global AI demand took off, TSMC crossed a major milestone as its AI-related revenue exceeded US$10 billion for the first time. This year, with NVIDIA leading the wave—followed closely by AMD, Broadcom, and other large ASIC customers—TSMC's advanced-process capacity has remained fully loaded. AI revenue is projected to reach US$28–33 billion, more than doubling year-over-year, and surpass NT$1 trillion—a new record for the company. Broadly defined AI revenue (including AI switching, networking ASICs, and AI-edge devices) is expected to account for 23–28% of total annual revenue, exceeding earlier estimates.

The rapid expansion of AI applications across TSMC's customer ecosystem continues to push demand sky-high. Analysts expect TSMC's AI-related revenue to climb past US$40 billion in 2026, reaching roughly NT$1.24 trillion, again setting a new all-time high.

Industry sources note that customers are locking in sub-3nm advanced-process capacity at an aggressive pace as part of a global AI buildup. Even though TSMC has emphasized aligning orders with real demand, the capacity race remains intense, giving the company clear visibility on orders through 2028.

TSMC has long worked closely with global AI innovators. Chairman C.C. Wei previously mentioned that AI server processors will account for over 20% of revenue by 2028.

The company highlighted that it collaborates closely with both customers and their downstream partners when planning capacity. With over 500 active customers and rising process complexity, project engagement now needs to begin two to three years in advance.

TSMC defines AI-related demand as including AI GPUs, AI ASICs, and—newly added—HBM controllers, all key components powering AI training and inference in data centers. The company expects AI-accelerator revenue to double again in 2025.

On the technology front, TSMC confirmed that its 2nm process has entered mass production on schedule. Fueled by smartphones, high-performance computing, and expanding AI use cases, the company expects its 2nm business to scale rapidly in 2026.