September 3, 2026 / SemiMedia / — TSMC Senior Vice President and Deputy Co-Chief Operating Officer Cliff Hou said AI-driven semiconductor capacity demand is growing at a pace and frequency not seen in the industry over the past 30 years, with the company's equipment requirements nearly doubling in just six months.

Hou said that if TSMC's equipment purchasing estimate at the end of 2025 for meeting 2026 customer demand is taken as a baseline of 1x, the requirement increased to 1.5x after only one quarter and reached 1.9x by July, representing an increase of about 90% within six months.

TSMC is currently advancing construction of nearly 20 fabs worldwide, including 13 in Taiwan and another five to six overseas. That compares with only four to five fabs typically under construction at the same time in the past, but Hou said customers are still reporting insufficient capacity.

The challenge is no longer limited to whether an individual company is willing to expand. Hou said the broader semiconductor ecosystem must also support the rapid increase in demand, while construction labor, equipment installation and supply chain resources all face practical limits.

AI is also increasing requirements for advanced process technologies, advanced packaging, energy efficiency and overall system performance. Semiconductor manufacturing is therefore moving away from a largely linear production model toward closer collaboration across wafer fabrication, packaging, testing and system integration.

TSMC has also introduced AI into multiple manufacturing processes to improve productivity and increase wafer output from existing capacity. Hou said the longer-term opportunity is to use AI to redesign semiconductor manufacturing itself rather than simply adding AI tools to existing production processes.

The 90% increase in TSMC's equipment requirements and the simultaneous construction of nearly 20 fabs show how quickly AI infrastructure demand is outpacing the semiconductor industry's traditional capacity expansion cycle.