▶ AI server demand tightens passive component supply, lifting Taiwanese suppliers' share

• Passive component distributor Nichidenbo noted that rising demand for AI servers and their associated power systems is rapidly extending lead times for high-spec MLCCs, large-format electrolytic capacitors, and hybrid aluminum electrolytic capacitors.

• Typical lead times have stretched from the previous 1.5–2 months to 3–4 months, with some part numbers facing waits considerably longer than that.

• Nichidenbo further projects that, as next-generation AI server products approach volume production, shortages of AI-related passive components are likely to intensify sharply from 2H26 through 2027.

• According to Nichidenbo, leading Japanese and Korean passive component makers still hold priority supply status in the AI server market, but the surge in AI orders is already exceeding their total production capacity.

• As a result, lead times for some AI-related components have reportedly extended to six months, or even more than a year.

• Major Japanese and Korean suppliers have begun halting new orders for certain MLCC and electrolytic capacitor part numbers, and their relatively conservative capacity-expansion stance is further accelerating the tightness in the mid- to high-spec passive component market.

• Based on Nichidenbo's supply chain estimates, AI server demand is expected to lift 2026 MLCC demand by 11% YoY, with overall capacitor demand potentially expanding by a further 30%.

• In particular, next-generation AI servers adopting a vertical power delivery architecture require tens of thousands of additional passive components per rack, which is set to absorb still more of global suppliers' capacity.

• Citing Nvidia's VR300 as an example, Nichidenbo explained that the compute node requires roughly 220,000 passive components and the power distribution section roughly 110,000, bringing total per-rack demand to around 330,000 units.

• For Google's TPU v8, compute node demand is estimated at about 400,000 units and power distribution usage at about 200,000, pushing total per-rack passive component demand above 600,000 units.

• Nichidenbo emphasized that, because both per-rack usage and capacitance requirements are rising simultaneously, AI servers consume far more passive components than automobiles do.

likes: 167 | retweets: 15 | replies: 10 | views: 26812