TSMC says AI chip demand remains above current supply as data-center power constraints shape capacity planning

TSMC Chairman C.C. Wei said during the company’s third-quarter earnings conference that AI demand remains very strong, with demand for AI chips materially exceeding the company’s current ability to supply them. He said all AI companies are working with TSMC, spanning demand for AI accelerators, AI CPUs and networking silicon.

The company is actively expanding manufacturing capacity, but its planning extends beyond semiconductor demand. TSMC is incorporating data-center construction progress, electricity availability, financing and the supply of other components into its capacity decisions. This reflects the interdependence between chip availability and the physical infrastructure required to deploy AI systems at scale.

TSMC also sent teams to the United States to assess data-center construction progress and power conditions. According to Wei, feedback on cloud service provider readiness was positive, although TSMC did not identify specific operators, sites or supporting readiness metrics.

The company’s HPC platform revenue rose 22% quarter on quarter in the third quarter and accounted for 68% of total TSMC revenue. TSMC said it has not seen signs of a slowdown in AI demand.

TSMC had previously projected AI-accelerator revenue to grow at a CAGR in the mid-to-high 50% range over 2024–2029, and indicated that actual growth could be higher. That projection remains forward-looking rather than an achieved result.

For data-center developers and infrastructure investors, the comments suggest that AI capacity deployment could be governed by synchronized delivery across silicon, grid interconnection and power supply, site construction, financing and networking equipment. This could increase the value of early power-capacity reservation and closer coordination across the infrastructure supply chain.