THE COMPONENT SIGNAL

Issue #011 · Monday, March 24, 2026 · DRAM Crisis + Foundry Watch + Automotive EMC

DRAM — Q1 closes at +95%. The numbers are final: DRAM contract prices rose 95% in Q1 2026. HBM4 is sold out through 2026 with zero spot availability. Standard DDR5 is up 60–70% depending on density. Samsung and SK hynix are both prioritizing HBM4 production over standard DRAM — which means the shortage on commodity memory will persist even as total wafer output increases.

What this means for your BOM: If your product shipped in 2025 with $8 DRAM, it now ships with $15.60 DRAM. For a product with 16GB, that's a $120 BOM increase — on memory alone. Some teams are redesigning to reduce memory density. Others are locking 6-month supply agreements at current prices, betting that Q2 goes higher. TrendForce says they're right: Q2 forecast is another 58–63%.

Samsung foundry wins Qualcomm 3nm. Samsung's foundry division (not to be confused with their memory division) secured a Qualcomm order for the Snapdragon 8 Gen 4 on Samsung's SF3E (3nm GAA) process. This matters beyond the mobile market: Qualcomm's automotive Snapdragon Ride platform uses derivatives of the same silicon. New process nodes = new EMI signatures that existing shielding may not cover.

Passive components — stable for now. MLCCs, resistors, and inductors are in a holding pattern. The April 1 price hikes (Murata, Samsung Electro-Mechanics) haven't triggered panic buying yet. Distributor inventories for standard sizes remain above 12 weeks. Automotive and high-cap grades are tighter — see Issue #001 for the breakdown.

We've reviewed 14 automotive EMC failures in the last 6 months. Eight of them had the same root cause: shield can apertures sized for the previous-generation design, not the current one.

Here's the pattern: a design team reuses a PCB layout template from a previous project. The shield can footprint comes along for the ride. We see this constantly — the shield looks right in the layout review because it's the same part number. But the board inside it changed. But the new design has a different component placement — taller capacitors, relocated connectors, rerouted flex cables. The shield can aperture (the gap where components or cables exit the shielded zone) grows by 2–5 mm to accommodate.

At 1 GHz, a 5 mm aperture increase changes the slot antenna resonance from above the test band to inside it. The shield goes from 40 dB of attenuation to 25 dB — right at the CISPR 25 limit.

The fix is simple but has to happen at layout, not at test:

• • Re-derive aperture dimensions for the new design
• • Maximum aperture: λ/20 at the highest test frequency
• • At 2.5 GHz (CISPR 25 current limit): max aperture = 6 mm
• • At 6 GHz (CISPR 25 proposed extension): max aperture = 2.5 mm

If 2.5 mm is too tight for your cable exit: Use a waveguide-below-cutoff approach — extend the aperture into a tunnel (shield wall folded inward) at least 3x the aperture width. This adds 15–20 dB of attenuation without shrinking the opening.

Metals

If you're reusing a shield can footprint from a previous design, re-derive your aperture dimensions against the new layout. Especially if component placement or cable routing changed. This takes 20 minutes in layout review and prevents a 6-week retest cycle.