Cloud AI Grabs Capacity, Chip Lead Times Extend Across the Board
N.R. Finch
Cloud-AI demand is squeezing wafer and packaging capacity so hard that longer lead times are spreading from AI accelerators into analog, power, and high-speed interconnect chips. Multiple majors have raised prices two to three times in one year — even if you have nothing to do with AI, parts are getting harder to secure.
How much have lead times actually stretched?
ADI has notified customers that some analog-chip lead times have extended to six months, urging them to place orders early.
STMicroelectronics MCU — microcontroller, the small chip that handles a device's basic computing — lead times have reached 52 weeks; distributors are already asking customers for full-year 2027 demand.
This means → the supply chain is no longer stocking for next quarter — it is locking capacity a year and a half out.
Why are power chips hit first?
Texas Instruments confirmed a July 1 price increase on PMICs — power-management ICs that distribute electricity to other chips — and MOSFETs. This is the third hike in one year.
STMicroelectronics enacted its second 2026 price increase on June 28; Infineon and NXP followed with their own second rounds.
In plain terms = AI servers are power-hungry. AI orders grab power-chip capacity first; everyone else waits longer and pays more.
Why are high-speed interconnect chips stuck too?
One U.S. chipmaker's PCIe Gen5 retimer — a chip that restores signal quality on high-speed data links — now has a 52-week lead time.
The reason: these products are migrating to 7 nm and 6 nm process nodes, which are simultaneously packed with AI accelerators and wireless chips.
This means → demand did not spike — the queue is simply full. AI chips have higher priority, so interconnect chips get pushed back.
When might the TSMC capacity bottleneck ease?
A Broadcom physical-layer executive warned publicly as early as late March 2026 that TSMC capacity was approaching its ceiling. Expansion capacity is not expected to come online until 2027.
Taiwanese IC-design houses broadly report that lead times for non-cloud-AI applications have also lengthened in recent weeks.
In plain terms = it is not just AI chips that cannot get in line — consumer and industrial chips with zero AI exposure are harder to source because capacity is being crowded out.
Who stands to benefit from this shortage?
Market observers note that with major vendors' lead times stretching and quotes rising, Taiwanese analog-chip makers may see relief from two years of pricing pressure.
PMICs, motor-driver ICs, USB PD controller ICs, and server-power and industrial applications are expected to benefit most.
This means → customers accelerating second-source qualification to cut single-vendor risk. Vendors that have already been qualified stand to capture emergency orders and redirected volume.
Is there a ceiling on the Taiwanese opportunity?
Taiwanese vendors face the same queuing pressure for wafer starts and packaging capacity.
Whether redirected orders translate into actual shipment growth ultimately depends on each vendor's own capacity allocation.
In plain terms = orders may shift your way, but if you cannot get on TSMC's schedule either, the opportunity stays on paper.
Content is for reference only, not financial advice.