Equipment Component Lead Times Double to 40 Months as AI Capex Meets a Tool-Chain Bottleneck
Lead times for critical semiconductor equipment components have more than doubled, with deposition parts stretching from four to ten months and one laser toolmaker waiting up to 40 months for Japanese components, a bottleneck that now threatens to delay memory and foundry expansion while advanced packaging stays dominated by CoWoS-L.
The AI buildout has reached the part of the supply chain that makes the tools. ETNews reports that lead times for essential semiconductor equipment components have more than doubled: parts for deposition equipment that once took four months now take as long as ten, a laser processing equipment maker faces up to 40 months for key Japanese components, and a packaging equipment maker says four-month parts now need at least six, about 50% longer than usual. Suppliers cannot accelerate deliveries even at a premium because capacity is full. TrendForce relayed the report on September 18.
Equipment itself is already stretched. ETNews reported in July that lead times for major tools from Applied Materials, ASML, Lam Research, Tokyo Electron and KLA had risen by 1.5 to two times, with roughly six-month deliveries becoming as long as a year. The root cause is underinvestment: component suppliers did not add capacity ahead of the AI boom, and uncertainty over how long the current semiconductor supercycle lasts is making Japanese component makers in particular reluctant to expand. US-China trade tensions remove the obvious fallback, because equipment makers are wary of qualifying Chinese-made components in tools used for advanced production.
Scarcity is turning into price. Citing a China Securities research report, Guandian describes a broad-based wave of price increases across the global semiconductor equipment component market, with lengthening lead times at overseas suppliers of valves and piping, ceramic components, RF power supplies and gas boxes. China Securities estimates that expanding a component production line takes 12 to 18 months, which means even a decision taken today does not add supply until 2028. For chipmakers, that converts a cost problem into a schedule problem.
The schedule risk is concrete. The reports specifically flag that prolonged equipment deliveries and component shortages could delay semiconductor manufacturing investment worldwide, including Samsung Electronics' and SK hynix's expansion projects in Pyeongtaek and Yongin - the fabs that are supposed to relieve the memory shortage. Any slip there pushes out the bit supply that the entire AI hardware and memory cost curve depends on.
Packaging is the other half of the constraint. TrendForce's September 11 report on advanced packaging finds that as cloud providers scale custom ASICs alongside GPUs, CoWoS-L is expected to remain the mainstream AI chip packaging technology through 2028 because it fits larger dies and more HBM, with Intel's EMIB-T emerging as a CSP alternative and FOPLP and hybrid bonding still needing until around 2028 to mature. With TSMC previously guiding CoWoS capacity up by more than 60% by 2027, equipment and component availability is now one of the binding limits on how fast that capacity can actually be installed.