The Component Signal · Issue #17
The Component Signal #017 — Intel's Terafab Anchor Model and the 0402 Ferrite Bead Squeeze
Intel's anchor-customer foundry model rewrites allocation logic, the Supreme Court tariff ruling opens a refund window, and we dissect 0402 ferrite-bead shortages and the EMC fix when your filter part is gone.
By Mike Kwak, Director · POCONS USA · How we report
The Signal
A general-purpose fab sells capacity. An anchor-customer fab sells a relationship. The difference rewrites how you secure allocation.
Intel's Terafab-style partnership model marks a shift away from general-purpose contract manufacturing toward anchor-customer facilities — fabs built around one or two committed lead customers who underwrite the capex. For everyone outside the anchor tier, this changes allocation logic fundamentally: capacity is no longer a market you bid into, it is a relationship you must already be inside.
Supply Chain
Supreme Court tariff ruling. The Court has blocked executive tariff authority (May 2026) — a structural ruling. Importers who paid 2025 assessments under the now-invalidated authority now have a refund pathway. The cash-flow opportunity is real, but it is retroactive and contingent on clean duty-paid records.
Passives. MLCC and inductor lead times stretched to 8–12 weeks as automotive demand rebounds — and within that, the ferrite bead is the acute pain (see Technical).
Korean Intel
A leading Korean foundry is reallocating 3 nm capacity toward AI accelerators and deprioritizing traditional mobile SoCs. Korean semiconductor exports jumped 19.4% month-over-month, reflecting both AI-chip demand and the memory-pricing recovery.
Technical
The 0402 ferrite bead shortage — and the EMC fix. Ferrite beads in the 0402 case size are on 8–12 week lead times, and for some impedance grades, longer. This is a genuine design problem, not just a procurement one, because a ferrite bead is not a generic part — it is a tuned component.
A ferrite bead is a frequency-dependent impedance: at low frequency it is a small inductor (passes DC and signal), and across a target band it becomes resistive, dissipating high-frequency noise as heat. It is specified by its impedance at 100 MHz (e.g., 600 Ω, 1000 Ω) and its DC resistance and rated current. You cannot substitute freely — swapping a 600 Ω bead for a 1000 Ω bead shifts the filter corner and can create a resonance with the downstream decoupling capacitor.
A ferrite bead plus a bypass cap is a low-pass LC filter — and an under-damped one can peak, amplifying noise at its resonant frequency f₀ = 1/(2π√(LC)) before it rolls off. If you must substitute, re-simulate the L-C interaction and check for peaking. Better: where the bead's job is purely EMI containment rather than rail filtering, solve it at the enclosure — a board-level shield can over the noisy block removes the radiated-coupling path entirely and takes the bead off the critical-shortage BOM line.
One Thing
Intel's anchor-customer model and the ferrite bead shortage have the same lesson: capacity flows to whoever has the relationship, not whoever has the order. The teams with locked supplier relationships and documented second-source qualifications are already paying the price they quoted. Everyone else is repricing at today's rate.
POCONS USA — EMI shields + components. Products manufactured in Korea (IATF 16949). San Diego, CA.