Major Taiwan passive-component vendors sent October price letters on MLCCs and power inductors this week as original equipment manufacturers finalized AI server power budgets, according to distributors and brokerage field checks.
Where hikes land
Increases are steepest on automotive-grade and high-temperature MLCCs feeding voltage-regulator modules, less so on legacy consumer sizes where fabs still have slack capacity. Power inductors tied to GPU rail designs saw mid-single-digit percentage list moves, with spot markets asking premiums for expedited lots.
OEMs building edge-AI appliances are pulling the same families as hyperscale server buyers, squeezing allocation lists that were already stretched after earlier memory supply scares.
AI power stages drive mix
Each additional kilowatt of AI compute requires more bulk capacitance and shielded inductors to keep ripple within spec. Contract manufacturers said they are double-ordering passives on new platform spins because a single missing reel can idle an SMT line costing six figures per hour.
Copper and ferrite raw materials factor into the letters, but vendors also cite rework costs when customers cancel orders after allocation locks. October resets give legal cover to pass those penalties through list prices.
Vendor competition
Japanese and Korean passive giants are raising quotes in parallel, reducing the discount Taiwan shops could offer by switching geography. Local champions benefit when buyers want shorter logistics lanes ahead of typhoon season, yet they cannot infinitely expand capacity on mature MLCC lines.
Some buyers are qualifying second sources on lower-margin SKUs to preserve budget for AI-only parts—a strategy that helps procurement but complicates quality audits.
Signal for tech investors
Passive hikes corroborate restocking notes from PCB laminates and packaging names on the same trading day. It is a micro confirmation that AI capex is feeding the supply chain below the headline chip names, even when the TAIEX retreats on foreign selling.
If lead times stretch further, expect OSAT and EMS firms to mention component frictions on earnings calls, not just advanced silicon shortages.
Buyer guidance
Split forecasts by platform: legacy consumer programs can defer pulls, but AI power rails need committed forecasts with penalty clauses understood. Document approved alternates before October letters expire so SMT planners are not guessing under holiday staffing gaps.
Holiday staffing
EMS plants in Taoyuan said they will run skeleton receiving crews through the Double-Ten closure but will not schedule major line changeovers until foreign fund flows clarify after the break. That pause gives passive buyers a narrow window to renegotiate October letters before November consumption data arrives.