Who makes passives for AI data centers

Passives are the smallest, most numerous parts on any board: MLCCs, inductors and tantalum or polymer capacitors sit beside the VRMs (4.08) to filter and stabilize the power rails feeding GPUs, CPUs and memory. A single high-power GPU board's power delivery network needs many more of these components than a conventional server board, because more numerous, more transient power rails each need their own local filtering. That volume increase is one of the more overlooked demand multipliers in an AI rack: it isn't a new part category, but sheer count reshapes procurement. High-specification MLCC supply has tightened as a result, with lead times reported to have lengthened well beyond what commodity passives see. For investors, this tier is normally viewed as undifferentiated and price-competitive, but the AI-driven volume and spec shift changes that calculus somewhat: capacity for high-spec, high-reliability grades is a narrower pool than capacity for commodity passives, and that gap is what determines who benefits from the volume increase rather than just absorbing more low-margin unit orders.