Register D. FY = calendar year (Dec). All figures JPY unless noted. Q1 FY2026 results filed 2026-05-12.
Register D. FY = calendar year (Dec). All figures JPY unless noted. Q1 FY2026 results filed 2026-05-12.
Thesis (cautious bull, valuation-flagged): MEC sells the CZ-series copper microetchant — the specialty chemistry that lets HDI PCB and IC substrate fabs hit adhesion and signal-integrity specs that no incumbent alternative reliably matches. CZ-8101 is qualified into chiplet packaging including CoWoS and is now riding AI server / general-server / PC / smartphone substrate recovery simultaneously. The business throws off 28% operating margin and >¥4B annual FCF on a clean net-cash balance sheet, with capital return policy (35%+ payout + 4%+ DoE + opportunistic buybacks) explicitly committing to return cash. The Kitakyushu plant Dec-2026 start is the only material near-term EPS overhang. The catch: the stock has re-rated to ¥11,190 (~115% above the December 2025 Nippon-IBR research-update reference price of ¥5,200 in five months) and now trades at 41x TTM P/E vs an analyst mean PT of ¥8,900 — sell-side is materially behind the price and conviction here depends on whether the AI-PCB layer-count thesis can absorb a 26-30% OPM trajectory through FY2027.
MEC Company Ltd. (4971.T, TSE Prime, Specialty Chemicals) was founded 1969, headquartered Amagasaki Hyogo, 508 employees. The business is a single-platform specialty chemicals company: copper surface-treatment chemistries for printed circuit boards and IC substrates, sold as consumables to global PCB and substrate fabs. Web: https://www.mec-co.com. Latest IR materials: 2030 Vision Phase 2 medium-term plan (Feb 13, 2026); Nippon-IBR sponsored research update (Dec 8, 2025); Q1 FY2026 Tanshin (May 12, 2026).
Business lines (FY2025 chemical-segment 9M run, % of chemical sales):
| Line | FY25 9M sales (¥mil) | % of chem sales | What it does |
|---|---|---|---|
| Adhesive Enhancers — CZ series total | 9,749 | 66.9% | Inner-layer copper roughening to enable next-layer dielectric adhesion at controlled ~0.1µm roughness |
| └ CZ-8101 (core product) | 5,217 | 35.8% | PC / smartphone / server packaging incl. chiplet / CoWoS |
| └ CZ-8100 | 1,016 | 7.0% | Legacy (launched 1995), now EV/ADAS automotive packaging |
| └ Other CZ (incl. CZ-8201, CZ-8401) | 3,516 | 24.1% | Next-gen ultra-fine, low signal-loss variants |
| V-Bond (adhesive enhancer) | 618 | 4.2% | Multilayer substrates for auto, smartphone, satellite comms |
| Etching chemicals — EXE | 1,054 | 7.2% | Chip-on-film (COF) for TVs |
| Etching chemicals — SF | 342 | 2.3% | Touch panel etch |
| Other surface-treatment | 650 | 4.5% | Anti-tarnish, DFR pre-treat, laser-direct-drill prep, etc. |
Plus a small equipment/process-line business not separately broken out in chemical-segment table; total revenue includes equipment + chemical.
Business model. Recurring consumable chemistry sold by kg/litre into qualified production lines at PCB / substrate fabs. Each board recipe is qualified at the OEM level (Intel / Nvidia / AMD / Apple). Re-qualifying with a substitute chemistry is a 6-18 month cycle and a yield risk customers refuse to take. Result: ~62% gross margin and 27-28% operating margin in a chemicals business. The Dec-2025 research update shows chemical-segment 9M FY2025 GPM of 62.4%, +1.0ppt YoY, driven entirely by CZ-8101 mix shift.
Geographic mix. Overseas sales 65.0% of total (FY25 9M, +3.8ppt YoY). Sensitivities: ¥0.1/NTD = ¥59M sales / ¥40M OP; ¥0.1/RMB = ¥23M sales / ¥11M OP. No formal geographic-segment percentages publicly broken out in English, but commentary confirms Suzhou strong on smartphones + displays, Zhuhai strong on general demand recovery, Taiwan supports the Taiwan PCB cluster.
| Site | Function | Status |
|---|---|---|
| Amagasaki HQ (Hyogo, Japan) | R&D + production | Operating |
| Taiwan plant | Production for Taiwan PCB cluster | Operating |
| Suzhou (China) | China east-coast production | Operating; +30% capacity expansion targeting Dec-2026 |
| Zhuhai (China) | China south production | Operating |
| Thailand | SEA production | Operating |
| Europe | EU service base | Operating |
| Kitakyushu (Japan, new) | New advanced CZ + future AP-series plant | Under construction; start Dec-2026. ~¥4B of the ¥8B 3yr capex |
None disclosed. MEC operates wholly-owned subsidiaries in each geography. Customer relationships are pure supplier–buyer.
In any multilayer printed circuit board or IC substrate, copper traces are laminated between layers of resin dielectric. If the copper-to-resin bond fails, the board delaminates under thermal cycling, signal integrity collapses, and yield craters. Historically, fabs got adhesion by chemically oxidising the copper to a microscopically rough black-oxide surface — the resin then mechanically keys into the roughness.
Two problems killed brown/black oxide as PCB tech advanced:
The industry needed copper roughening that (a) doesn’t dissolve in acidic dielectrics, (b) gives controlled sub-µm roughness for low-loss signalling, (c) is uniform enough to hit 50ppm-class yield, and (d) plays well with photoresist and direct laser drilling. That’s CZ chemistry.
CZ chemistry is a controlled microetch — an aqueous oxidant solution (typically based on cupric salts with organic complexing agents) that selectively dissolves grain boundaries in copper foil. The result is a controlled roughness topology at the sub-µm scale, with an organic-passivated copper surface that bonds chemically (not just mechanically) to the next dielectric.
Three things make CZ hard to copy:
Fundamental limit: roughness can theoretically go to zero with a perfectly smooth copper surface, but bond strength at zero roughness depends entirely on the chemical-bond layer rather than mechanical interlock. That’s the next platform — MEC’s AP series targets exactly this no-roughening regime for next-generation PCB substrate use in data centres (per the Dec-2025 update).
A typical inner-layer copper roughening pass at a PCB fab:
The fail modes: roughness too low (delamination), roughness too high (signal loss), uneven across panel (pattern-level yield loss), residue (resin adhesion fails), tarnish bleed-through (visible discolouration on finished product). Process engineers obsess over CZ-bath chemistry uniformity because every one of these fails recurs in PCB fabs that try to switch suppliers.
What investors should track: Q-over-Q chemical shipment volume (tonnes), CZ-8101 share of CZ-series, OPM trajectory through Kitakyushu start-up, and any disclosure on the AP series roadmap.
CZ-series is now ~67% of chemical revenue and the single line that matters for the equity. Within CZ:
[Cu sulfate / H2SO4 / additives] → [MEC: CZ chemistry] ★ → [PCB / substrate fab] → [OSAT / EMS] → [End OEM Intel/Nvidia/Apple]
↑
Spec / qualification gate
happens here — and lasts
a product generationMEC sits at the chemistry layer, but its leverage comes from the spec lock at the next-stage customer (PCB / substrate fab). The fab cannot swap MEC for a competitor without re-qualifying the recipe with the end-OEM — a multi-quarter exercise. MEC effectively captures more value than its tiny revenue footprint implies, because the substitution cost downstream is several orders of magnitude larger than MEC’s own annual spend.
Suppliers. Sulfuric acid, hydrogen peroxide, copper salts, organic complexing agents — commodity inputs. No supplier concentration risk; no specialty-supplier bottleneck.
| Supplier | Layer | Bypass-ability | Supplier MC vs 4971 | Market-pricing |
|---|---|---|---|---|
| Bulk H₂SO₄ / H₂O₂ | Commodity chemicals | Yes (multiple) | n/a | Priced-in |
| Cupric salts | Commodity | Yes | n/a | Priced-in |
| Specialty organic ligands | Speciality chem | Partial — likely internal IP / contract synth | n/a | Priced-in |
Bottleneck verdict: None upstream. MEC’s leverage is downstream (spec lock) not upstream. No supplier alpha candidates here.
Customer names are not disclosed in MEC IR. The Nippon-IBR Dec-2025 update mentions end-uses (CoWoS chiplet packages, AI server packages, PC/smartphone, EV/ADAS, satellite comms) but not customer names. Inference table below.
| # | Likely Customer | Ticker | Likely Revenue Share | Relationship |
|---|---|---|---|---|
| 1 | Ibiden | 4062.T | undisclosed (large) | ABF IC substrate (CoWoS underlay) |
| 2 | Unimicron | 3037.TW | undisclosed (large) | ABF IC substrate + HDI PCB |
| 3 | Shinko Electric | 6967.T | undisclosed | IC substrate |
| 4 | Compeq / Tripod / Nan Ya PCB | 2313.TW / 3044.TW / 8046.TW | undisclosed | AI-server HDI PCB |
| 5 | AT&S | ATS.VI | undisclosed | European ABF substrate (Kulim ramp) |
| 6 | Kinsus / Simmtech | 3189.TW / KQ:222800 | undisclosed | Memory + IC substrate |
For each: undisclosed financial-health detail beyond the public profiles. Switching costs are uniformly high — every one of these fabs runs CZ in their qualified inner-layer roughening line and would need OEM re-qualification to switch.
Concentration risk. Top-1 and top-5 not disclosed. The ~100% PC-CPU substrate market-share claim implies that Intel- and AMD-side substrate suppliers collectively are a large block. This is the single biggest disclosure gap in the public file on MEC — it would meaningfully tighten the thesis to know what % of revenue is concentrated in the top 3 customers.
Strategic partnerships. None disclosed publicly.
Why it matters. Every advanced multilayer PCB and IC substrate built today needs a controlled copper-roughening step. As line/space shrinks (HDI → mSAP → ABF) and signal-loss budgets tighten (M9-grade CCL, 224G SerDes, PCIe 6/7), the alternatives to CZ-class chemistry get progressively worse. MEC’s product replaces brown-oxide chemistry that no longer works at advanced nodes, which is why share has held through 20 years of PCB technology transitions.
End-use applications:
TAM. Wet-process chemistry inside PCB + IC substrate is approximately $2-3B addressable globally (order of magnitude, no clean IDC/Prismark line). CZ-class copper roughening is a high-value sliver within that — call it $400-600M of premium-segment TAM. MEC at ¥14.5B FY25 9M chemical sales = ~$130M USD run-rate → mid-20% to 30% of premium segment, dominant at the highest end.
Secular tailwinds:
ai-server-pcb-primer.md).Headwinds:
AI-server substrate demand has accelerated CZ-8101 from ~¥4.5B FY23 sales to a >¥7B FY25 run-rate, with Q3 FY25 alone hitting ¥2.02B (37.5% YoY). General-server, PC, and smartphone recovery layered on AI is happening simultaneously rather than sequentially. MEC’s Kitakyushu plant doesn’t open until Dec-2026, so the supply side is tight through 2026 — which means pricing power is intact and mix shift to high-value CZ-8101 is uncontested. The competitive set (Atotech / MKS, Uyemura) has not delivered a CZ-equivalent at advanced nodes. Sell-side coverage is thin (3 analysts, mean PT below spot) and the new shareholder-return regime (35%+ payout + 4%+ DoE + active buybacks) hasn’t been fully priced.
Japanese small-cap; English disclosure on individual executives is limited. Detailed forensic work requires the Yuho (Japanese annual securities report) and TDnet filings. The /mgmt-dd skill will do the deeper pass — this section is the framework view.
| Name | Title | Tenure | Background |
|---|---|---|---|
| Maezawa Yasushi | Representative Director, President & CEO | multi-year | Career MEC executive |
| (CFO/COO/CTO) | Not individually disclosed in English IR | — | — |
The English IR file does not name the CFO, COO, or CTO publicly. Full board / NEO disclosure is in the Yuho — out of scope for this section, flagged for /mgmt-dd.
This level of management/founder ownership is high for a Japanese specialty-chemicals mid-cap and a clear positive alignment signal. Combined with the buyback policy + 35%+ payout commitment, the equity holder is sharing the upside with insiders symmetrically rather than being diluted by them.
Cannot construct from English filings. Requires Japanese DEF 14A equivalent (Yuho 大株主) — flagged for /mgmt-dd. The 18.45% aggregate suggests insider holdings are dominantly in MEC itself, not parked elsewhere, but this needs primary-source confirmation.
No red flags surfaced at this level of disclosure. Deeper Yuho 関連当事者取引 (related-party transactions) read needed for the /mgmt-dd skill.
Capital allocation grade: A-. Genuinely shareholder-friendly capital return alongside disciplined, capacity-aligned capex. The minus is the absence of a clearly-articulated long-term ROIC target in English disclosure.
Not disclosed in English IR at NEO level. Yuho will contain individual NEO comp for the top 5 executives. Flagged for /mgmt-dd.
TSE Prime listed → requires ≥⅓ independent outside directors. Specific names not disclosed in English IR. No dual-class structure. No poison pill. No staggered board.
| Dimension | Rating | Key finding |
|---|---|---|
| Skin in the Game | Green | 18.45% insider ownership — high for Japan mid-cap |
| Holdings Concentration | Yellow | Not verified at NEO level (English disclosure gap) |
| Shell / Cross-Holdings | Green | No red flags surfaced |
| Capital Allocation | Green | 35%+ payout, 4%+ DoE, opportunistic buybacks, 10% revenue → R&D, disciplined capex |
| Compensation Alignment | Yellow | Not verified — Yuho needed |
| Governance Quality | Green | TSE Prime; ≥⅓ independent board; no anti-takeover devices |
| Litigation / Enforcement | Green | None surfaced in public record |
| Overall (preliminary) | Green-Yellow | High-quality governance pending NEO-level Yuho verification |
| Company | Ticker | Segment(s) | Approx revenue | Share | Moat | Pure-play? |
|---|---|---|---|---|---|---|
| MEC | 4971.T | CZ copper microetch — high-end | ¥20.3B FY25E | ~100% PC CPU substrate adhesion; HSD overall | Spec lock + 20yr process IP | Yes |
| Atotech / MKS Instruments | MKSI | Specialty plating + surface chem | $3.6B (Atotech only) | Large in general PCB chemistry; lost high-end share | Global scale | No |
| Uyemura International | private (JP) | Plating + surface treatment | private | Asia plating focus | Asia footprint | No |
| Shipley / Rohm and Haas (Dow) | DOW | Legacy PCB chemistry | mixed | Mid-tier PCB chemistry | Legacy scale | No |
| Resonac (formerly Showa Denko) | 4004.T | Substrate films + chemistry | ¥1.3T (group) | Different layer — substrate film, not microetch | Materials breadth | No |
MEC is the most concentrated pure-play CZ-class microetch supplier listed. Its competitive moat is the combination of (a) spec lock at the OEM end (PCB fabs can’t unilaterally swap chemistry), (b) 20+ years of bath chemistry data + process knowhow, and (c) qualified install base across the world’s top-10 PCB / substrate fabs. The 62%+ gross margin in a chemicals business is the financial signal of moat depth — if a credible substitute existed, pricing would have collapsed.
Quality verdict: high-quality / durable.
PCB / IC substrate consumable chemistry is moderately consolidated at the high end (3-4 names global) and fragmented at the low end. Industry has been cyclical historically (PC cycle, smartphone cycle) but is now layered with the AI capex cycle — which has different dynamics.
| Q1 FY25 | Q2 FY25 | Q3 FY25 | Q1 FY26 | |
|---|---|---|---|---|
| Chemical sales YoY % | n/a (similar Q1 FY24 base) | n/a | +18.8% | n/a (full FY26 Q1 split TBD) |
| Group revenue YoY % | n/a | n/a | n/a | +35% (est) |
The Q3 FY25 chemical print at +18.8% YoY with +15.6% QoQ is a meaningful acceleration vs the +9.5% 9M YoY pace. Q1 FY26 group revenue at ~+35% YoY is another acceleration. Implied: FY26 will materially beat the 26-30% OPM band that management put forward in Feb-2026 — unless Kitakyushu start-up costs land heavier than guided in H2 FY26.
| Metric | Value |
|---|---|
| Share price | ¥11,190 |
| Market cap | ¥204.3B |
| Enterprise value | ¥200.4B |
| P/E (TTM) | 41.1x |
| Forward P/E (FY+1) | ~36-41x (depends on whether consensus updates post Q1 FY26 beat) |
| EV/EBITDA (FY25E) | ~31x (EV ¥200.4B / EBITDA ¥6.34B FY25E per Nippon-IBR) |
| P/FCF (FY25E) | ~48x (cap ¥204B / FCF ¥4.27B FY25E) |
| EV/Revenue (FY25E) | ~9.9x |
| FCF yield | ~2.1% |
| Dividend yield | ~0.76% (¥85 / ¥11,190) |
| 52-week range | ¥2,385 - ¥11,780 |
The stock has gone from ¥5,200 on Dec 4, 2025 (Nippon-IBR reference) to ¥11,190 today (May 12, 2026) — a 115% move in five months, on top of an 822% 52-week range. Q1 FY26 beat is real but the valuation absorbs that and more.
| Metric | FY22 | FY23 | FY24 | FY25E (Nippon-IBR Dec25) | FY26E (consensus) |
|---|---|---|---|---|---|
| Revenue (¥mil) | 16,329 | 14,020 | 18,234 | 20,300 | 23,807 |
| Rev YoY % | — | -14.1% | +30.1% | +11.3% | +13.8% |
| Gross profit | 9,804 | n/a | 11,101 | n/a | n/a |
| Gross margin % | 60.0% | n/a | 60.9% | n/a (62.4% 9M actual) | n/a |
| Operating profit | 4,004 | 2,492 | 4,562 | 5,500 | n/a |
| Operating margin % | 24.5% | 17.8% | 25.0% | 27.1% | n/a |
| Net income | 3,065 | 2,304 | 2,291 | 4,300 | n/a |
| Net margin % | 18.8% | 16.4% | 12.6% | 21.2% | n/a |
| EPS (¥) | 161 | 122 | 122 | 233 | 275 (consensus) |
FY23 net income (¥2,304M) and FY24 net income (¥2,291M) are flat despite FY24 OP growing 83%. Likely tax + one-off impact; flagged for /mgmt-dd. FY25E NI ¥4,300M is an +87.6% jump — biggest single-year EPS step in the modern history of the business.
| Metric | FY25 |
|---|---|
| Operating cash flow | ¥3.98B |
| Capex | ¥2.77B |
| Free cash flow | ¥1.21B (yfinance) / ¥4.27B (Nippon-IBR FY25E) |
| Cash | ¥10.35B |
| Total debt | ¥1.25B |
| Net debt | -¥9.1B (net cash) |
| ROE | 12.8% (yfinance) / 15.3% (Nippon-IBR FY25E) |
| Net D/E | -0.4x (Nippon-IBR) |
Note: yfinance FCF (¥1.21B) and Nippon-IBR FY25E FCF (¥4.27B) diverge because yfinance includes the buyback as a financing outflow netted in trailing FCF while Nippon-IBR estimates pre-buyback FCF. The Nippon-IBR ¥4.27B is the cleaner read on underlying cash generation.
Using chemical-segment YoY 9M (FY25 vs FY24) as the cleanest read:
What this tells us:
Current multiples (TTM/FY25E):
Historical context: stock 5x’d off 52w low. Pre-rally (Dec-2025) Nippon-IBR had FWD P/E at 22.3x — that’s a more “normal” multiple for this business. At the current price the multiple has nearly doubled in 5 months while FY25E earnings revised up by only +20-30%.
Implied expectations at ¥11,190:
Where consensus is behind:
Downside scenario:
Upside scenario:
The asymmetry is poor at this price. A DCF is essential before sizing.
| Tailwind | Mechanism | Magnitude | Duration |
|---|---|---|---|
| AI server PCB layer count expansion | More inner-layer surfaces → more CZ per board | Direct multiplier on volume | 3-5yr |
| ABF substrate area expansion (Hopper→Blackwell→Rubin) | Bigger panels → more CZ per package | Direct multiplier | 3-5yr |
| 224G SerDes / PCIe 7 / low-loss specs | Tighter roughness tolerance favours CZ over alternatives | Mix shift to high-ASP CZ-8101 / 8201 / 8401 | 5-10yr |
| AP-series no-roughening | Optionality — captures next platform transition | TBD; commercial timing not disclosed | 5-10yr |
| Smartphone + general server recovery | Demand recovery off cyclical lows | +10-15% volume | 1-3yr |
| Headwind | Likelihood | Impact |
|---|---|---|
| Etching chemicals SF contraction | Continuing | Drag on chemicals total growth ~1-2ppt |
| Kitakyushu plant start-up costs | High | OPM compression FY26-27 (already in guide) |
| Glass substrate / hybrid bonding long-term displacement | Medium 5-10yr | Erodes part of CZ TAM; AP-series is hedge |
None individually disclosed; MEC sells consumables under long-running spec qualifications rather than discrete contracts. The “contract” structure is implicit in the OEM qualification process at PCB / substrate fabs.
| Risk | Likelihood | Existing Mitigants | Mgmt De-risk Plan | Closable? |
|---|---|---|---|---|
| Valuation re-rate — stock priced for continued AI growth at 40x FWD P/E | High | Strong fundamentals; capital return policy supports floor | Capital return + buybacks defend valuation in correction | No — closes only via earnings catch-up or multiple compression |
| AI capex pause / hyperscaler reset in 2027 | Medium | Diversified end-use; general-server + smartphone + auto recovery layered | Capacity sizing paced to qualified demand | No — structural to customer base |
| Customer concentration in ABF substrate cluster (Ibiden / Unimicron / Shinko) | Medium | 100% PC CPU substrate spec lock; multi-year qualification moat | Active Chinese substrate qualification | Partially — closes if MEC qualifies into Chinese substrate entrants pre-share-shift |
| Kitakyushu start-up cost drag (FY26-27) | High (near-certain, in plan) | 26-30% OPM band absorbs it | Phased ramp; revenue pre-booked | Yes — closes Q3/Q4 FY27 at budgeted utilisation |
| AP-series substitution risk (long-tail platform displacement) | Low near-term, Medium 5-10yr | MEC owns AP-series IP | R&D allocation | Yes — if AP-series captures the displacement |
| FX (JPY appreciation) | Medium | Geographic production hedges some | No hedge programme disclosed | No — structural |
| Etching SF revenue decline | High (continuing) | Small absolute drag (¥220M annualised) | Mix shift to CZ already in progress | No — managed not closed |
Effectively zero. Share count is flat-to-declining (2.49% reduction via buyback Aug 2025). No convertibles. No ATM. No warrant overhang. Net cash + ¥4B+ FCF self-funds Kitakyushu plus capital return.
Moderate. CEO is a career insider. CZ chemistry know-how is held tacitly across the R&D team (not single-person dependent). No public succession plan disclosed in English — flagged for /mgmt-dd.
What invalidates the thesis: (a) CZ-8101 share loss in advanced packaging — Atotech or Chinese entrant qualifies a substitute at Ibiden or Unimicron; (b) AP-series fails technically; (c) AI capex pauses materially in 2027.
(Detail in 4971-profile.md; summary here.)
The PT-below-spot is the single most important valuation signal. Either consensus catches up (revisions upward in coming weeks following Q1 FY26 beat) or the stock drifts back to the PT zone (¥8,000-10,000) over coming months.
Conviction: Medium on the business; Low-Medium on the entry. Valuation is the binding constraint.
Suggested entry framework:
Stop-loss triggers (if owned):
Re-evaluation triggers (after Q2 FY26 print Aug 2026):
Searched ~/Dropbox/Wafflebun/KB/wiki/semianalysis/ on
2026-05-12. No direct SA coverage of MEC Company. Two false-positive
hits (“MEC” in Meta-Superintelligence and Google Apollo posts) —
unrelated. No SA contradiction or supporting view to flag.
Internal wiki cross-reference:
~/Dropbox/Wafflebun/KB/wiki/ai-server-pcb-primer.md lists
MEC as an “indirect play” with ~$2B market cap (note: the MC reference
is stale — current is ~$1.36B USD, ~₹200B), ~15% supply-chain share, and
“Specialty CCL chemical + small fab.” This deep-dive is consistent with
the primer view but materially deeper on the chemistry and product
specifics.
Verdict: PASS at ¥11,190. Re-engage at ¥7,000-9,000.
The business quality is real and the moat is among the deepest in Japanese specialty chemicals. The AI-server / ABF substrate tailwind is in-progress and visible in the Q3 FY25 and Q1 FY26 prints. Capital return policy is best-in-class. But the entry economics are poor: 41x TTM P/E on a chemicals business with consensus PT below spot, after a 5x move in 12 months, is not where to start a new position. The Bear case (-61%) and Bull case (+18% over 12 months) give negative expected value at this price.
Set price alert at ¥9,000 and ¥7,000. Re-evaluate after Q2 FY26 print (Aug 2026).
~/Dropbox/Wafflebun/KB/wiki/ai-server-pcb-primer.md (AI
server PCB supply chain primer, May 2026);
~/Dropbox/Wafflebun/KB/wiki/packaging-glass-substrate-primer.md;
~/claude/output/profile/4971-profile.md.