The global semiconductor supply chain is under increasing strain due to sustained demand for AI chips. According to Broadcom, manufacturing partners are reaching their limits. This is said to apply particularly to TSMC.
According to Reuters, a Broadcom executive stated that TSMC’s production capacity is now visibly under pressure. While the manufacturer was viewed as having virtually unlimited capacity just a few years ago, that situation has since changed. Although expansions are planned through 2027, Broadcom says the current constraints are causing delays that will noticeably impact the supply chain in 2026.
Demand for AI infrastructure is the primary driver of this trend. TSMC previously reported that its advanced production lines are operating at full capacity. Major clients such as Nvidia and Apple rely on this capacity, further intensifying the pressure. At the same time, the chip manufacturer is working to narrow the gap between supply and demand, though this remains a challenge in the short term.
Bottlenecks Beyond Chip Production
According to Broadcom, the problems extend beyond chip production alone. Bottlenecks are also emerging in related sectors. For example, there is limited availability of certain optical components, including lasers. In addition, printed circuit boards are unexpectedly proving to be a weak link, as suppliers in both Taiwan and China are reaching their capacity limits. This leads to longer delivery times and greater planning uncertainty.
The situation is forcing customers to adjust their strategies. Multi-year contracts are increasingly being signed to secure production capacity. While such agreements were less common in the past, companies are now opting for contracts spanning several years. Other players in the sector, such as Samsung, are also moving in this direction and working with customers to secure longer contract terms.
The shift toward long-term contracts underscores the importance of supply security in a market heavily dependent on a limited number of manufacturing partners. At the same time, suppliers are using this approach to better protect themselves against demand fluctuations, which are difficult to predict due to the rapid rise of AI.