Taiwan Semiconductor Manufacturing Co. is signaling that the race to build AI infrastructure is likely to keep its factories under strain for the foreseeable future. Chief executive C.C. Wei told shareholders that the company expects it will take years before chip output can fully catch up with customer demand, especially for the most advanced AI processors.
The comments underline how quickly demand has grown from cloud providers and other large buyers that are pouring money into AI systems. Even as TSMC expands manufacturing capacity in multiple regions, including its Arizona operations, Wei said the extra production coming online is still not enough to satisfy all customers.
Bloomberg reported that the shortage is being driven in part by hyperscalers, the large cloud companies that are building out data center capacity at a rapid pace. According to the report, that demand is increasing faster than TSMC can bring new output online.
TSMC has spent heavily to widen its global footprint, but Wei’s remarks suggest those investments will not resolve the bottleneck quickly. The company is already one of the key suppliers behind the AI hardware surge, producing advanced chips used in accelerated computing systems.
Wei said it will be a long time before the company can meet customer needs in full. That is notable because TSMC has opened several fabs around the world in recent years, yet still sees demand running ahead of available capacity. The company’s U.S. expansion has been closely watched as governments and customers look for more geographically diverse chip production, but even that move does not appear to be enough to eliminate the shortage.
Despite the supply tightness, TSMC is still projecting solid growth. The company expects sales to rise 30% this year, according to the report. That forecast indicates that demand remains strong enough to support substantial revenue growth even while output remains constrained.
Wei also told shareholders that TSMC does not intend to take advantage of the shortage by sharply raising prices. He said the company plans to keep pricing stable and avoid abrupt increases similar to those seen in the memory and storage chip markets.
That approach suggests TSMC wants to avoid creating additional volatility for customers already struggling with tight availability and heavy AI spending commitments. Stable pricing could also help preserve long-term relationships with major buyers, many of whom depend on TSMC for the most advanced chips.
The comments reflect a wider shift in the semiconductor industry, where AI demand has become the dominant growth driver. Major technology firms continue to expand their budgets for data centers and AI hardware, leaving foundries like TSMC with the challenge of scaling production fast enough to keep pace.
For now, TSMC’s message is that the imbalance is not temporary. Even with more fabs coming online, the company expects supply to remain behind demand for years, making advanced chip capacity one of the most closely watched constraints in the AI boom.