Surge in Orders! TSMC’s 3nm Capacity Ramps Up Ahead of Schedule

Three Major Chipmakers Place Rush Orders, Capacity Booked Through 2027

The core driving force behind this round of accelerated capacity expansion is the explosive demand for AI high‑performance computing chips. Wedbush analysts pointed out that the three major AI chipmakers—Nvidia, AMD, and Broadcom—have continued to place additional orders, directly pushing TSMC’s 3nm production lines to operate at full capacity.

Supply chain data shows that TSMC’s 3nm capacity is already fully booked through 2027, and even with accelerated expansion, it remains difficult to fully absorb market orders. To match customer demand, TSMC is not only building new 3nm wafer production lines but also converting some 5nm equipment for 3nm production, in order to maximise available capacity.

2nm Ramps Up in Tandem, Capital Expenditure Hits Record High

Meanwhile, the ramp‑up of the next‑generation 2nm process is also accelerating. In the first half of 2026, TSMC’s monthly 2nm capacity had already reached 50,000–60,000 wafers, and the company is targeting 100,000 wafers per month by the end of the year, indicating that the commercialisation of TSMC’s next‑generation advanced process is progressing at a faster pace, laying the manufacturing foundation for subsequent iterations of AI chips. 

Faced with the surge in AI orders, TSMC has significantly raised its capital expenditure guidance, lifting its 2026 capex to US$52–56 billion, a record high. The massive capital outlays are mainly directed towards advanced 3nm and 2nm wafer production lines and supporting advanced packaging equipment, reflecting TSMC’s optimistic outlook on the long‑term prosperity of the AI computing power sector.

Tight production capacity triggers changes in the economic landscape

However, the capacity bottleneck has also brought about subtle changes in the industry landscape. Due to the tight supply of 3nm capacity, six core clients—including Apple and Nvidia—have had to actively seek alternative foundry solutions, turning to Samsung and Intel for some of their orders. 

Market research firms project that TSMC's market share in advanced process foundry services may slightly decline from 95% to 90%. Although Samsung and Intel are unlikely to challenge TSMC's leading position in the short term, the annual transfer of approximately $5 billion in orders has already opened up a growth window for competitors, signaling that the advanced process competition is no longer a one‑horse race for TSMC.

Supply-Demand Logic Reshaped, Competition Enters a New Stage

From an industry perspective, TSMC’s 3nm capacity coming online ahead of schedule is a direct validation of AI computing power demand exceeding expectations. The surging demand for AI chips is reshaping the supply‑demand dynamics of the global wafer foundry industry: advanced‑process capacity has become a scarce resource, and clients are no longer solely focused on cost—securing supply reliability has become their top priority. 

On the one hand, TSMC continues to dominate the advanced foundry space, leveraging its technological and scale advantages. On the other hand, capacity constraints are forcing top customers to diversify their supply chains, opening up opportunities for Samsung and Intel to win orders.

Of course, uncertainties remain in the industry. If the growth rate of AI computing power demand were to slow in the future, the supply‑demand balance in advanced processes could reverse. Meanwhile, whether Samsung and Intel can continue to improve their yield rates and expand capacity will also determine the competitive trajectory of the advanced foundry market going forward. Overall, this round of capacity expansion signals once again confirms that AI has become the most important growth engine for the semiconductor industry. 

In the short term, the tightness in advanced‑process capacity will not be quickly alleviated, and 2027 is still expected to see a state of tight supply‑demand equilibrium. Moreover, as 2nm capacity gradually comes online, the manufacturing competition for next‑generation AI chips will also enter a brand‑new phase.