“We are seeing that TSMC is hitting [production capacity] limits,” Natarajan Ramachandran, director of product marketing in Broadcom’s Physical Layer Products division, told reporters on Tuesday. He noted that just a few years ago, TSMC’s capacity would have been considered “infinite.” “They will be increasing the capacity to 2027, but that has become a bottleneck, or that has kind of choked the supply chain in 2026,” he added. TSMC, the world’s leading producer of advanced AI chips, did not immediately respond to a request for comment.
The Taiwanese firm, whose customers include Nvidia and Apple, acknowledged in January that its advanced production lines were fully stretched due to the rapid expansion of AI infrastructure. It has been working to narrow the gap between supply and demand.
Ramachandran said supply constraints extend beyond semiconductors, affecting multiple adjacent industries. “Even though there are multiple suppliers in the industry today… there is definitely a supply constraint in the laser space,” he said, adding that printed circuit boards (PCBs) have also become an “unexpected” bottleneck. Lead times for PCBs used in optical transceivers, he noted, have lengthened from approximately six weeks to six months, with both Taiwanese and Chinese suppliers facing capacity limits.
To mitigate these risks, many customers are entering long-term agreements with suppliers, locking in capacity commitments for three to four years. The trend was echoed last week by memory chipmaker Samsung Electronics, which announced efforts to shift major customers toward contracts spanning three to five years. These moves reflect the industry’s desire for greater supply security and suppliers’ need to manage volatility in demand.