Semiconductor and Other Electronic Component Manufacturing (U.S.) — NAICS 33441

A Histometrics rollup primer for public- and private-market investors. NAICS = North American Industry Classification System, the federal scheme used to define and count industries. This is a 5-digit "industry" that contains six more detailed child industries.

---

1. Overview

This is where electronic hardware is physically made in the United States: the chips that compute, the bare boards that carry them, the resistors and capacitors that condition power, the connectors that join circuits, the loaded assemblies that become a finished product, and a long tail of specialty parts (crystals, sensors, microdisplays, high-frequency laminates). NAICS 33441 groups all of it into one industry with six children — and the single most useful thing to understand is that they are wildly unequal in size, ownership, economics, and how you invest in them.[1]

One five-digit line therefore spans the two economic extremes of manufacturing at once: the most capital-hungry business in the entire economy (a leading-edge chip factory runs roughly $10 billion for the plant plus about $5 billion of equipment, with the largest complexes quoted at $15–20 billion) and one of the thinnest-margin (contract board assembly earns high-single-digit to low-teens gross margins — 7.0% to 10.1% across the listed bench in fiscal 2025).[4][5] What ties the six together is a common demand engine — artificial intelligence (AI) data centers, electric vehicles (EVs), defense, and medical devices are pulling on every one of them at the same time — and a common vulnerability: most of the world's electronic-hardware production migrated to Asia, so U.S. factory output is a fraction of what the U.S. economy actually consumes.[6][7][8]

Why an investor should care: this level is the physical backbone of the digital economy, and it offers unusually different entry points. Two children (semiconductors, connectors) hold large, liquid U.S.-listed champions; two (bare boards, passive components) are overwhelmingly private or foreign-listed; and the rest sit in between. Picking the child is most of the investment decision. The divergence is not subtle: while semiconductors run a subsidised build-out and global chip sales rose 25.6% to $791.7 billion in 2025, bare-board fabrication has shed 80.6% of its U.S. workforce since 2000 — 137,501 jobs down to 26,676.[6][9][10] Same five-digit code, opposite trajectories.

---

2. What's inside — the six children and how they differ

The children are not variations on one theme; they are different businesses with different owners and different ways in. The table below is the heart of this primer. Shares are of this level's domestic receipts ($115.9 billion total, 2022); "jobs" is share of the level's 279,841 U.S. workers (2023).[2][3]

How to read the contrasts:

• Size is lopsided. Semiconductors alone are ~54% of the level's revenue and ~42% of its jobs; the two smallest children (bare boards + passives) together are under 7%. This is one giant child, one large child (EMS), and four smaller ones.
• Receipts and jobs don't line up — that's the economics showing through. Semiconductors earn 54% of revenue on 42% of the workforce (capital-heavy, high pay); "other components" (334419) is the mirror image — 20% of the jobs but 12% of revenue (labor- and engineering-heavy, lower revenue per head).
• Pay tracks capital intensity, and the full ladder is now visible. Average pay (payroll ÷ employment) runs roughly $137,000 in semiconductors, $85,000 in connectors, $73,000 in both assembly and other components, $65,000 in bare boards, and $60,500 in passives — against about $99,000 across the level.[2][3] The more automated and capital-intensive the process, the higher the pay per remaining worker.
• Ownership is the real fork in the road. Semiconductors are almost entirely public/large-corporate; connectors are a barbell where the largest franchises are as likely to be private (Molex, Samtec, Yazaki) as public (Amphenol, TE); passives are dominated by Asian firms you can only buy on foreign exchanges; bare boards are overwhelmingly private. The same "electronics manufacturing" headline hides completely different cap tables.
• Even the firm counts mean different things. Census counts employer establishments classified by principal activity; trade bodies count operating businesses. In bare boards the federal file shows 333 firms while the industry association (PCBAA) counts roughly 145 remaining U.S. fabricators — not an error, a definitional gap that recurs whenever you compare census data to an industry census.[3][18]
• Firms straddle the codes. Vishay spans semiconductors and passives; TTM spans bare boards and assembly; Sanmina runs board-fab lines inside an assembly business; Bel Fuse spans passives, connectors and assembly; Knowles now appears in both passives (specialty capacitors, Cornell Dubilier) and other components (MEMS acoustics). No child is a clean corporate boundary.[29][31]
• Direction of travel diverges. Every child is being pulled by AI, but bare-board fabrication is shrinking on U.S. soil (import competition), passives are riding a polarised price cycle, connectors compound on rising content-per-device, and semiconductors are in a subsidised demand supercycle. "The electronics cycle" lifts them together and cuts them together, but their structural trajectories differ.

---

3. How big it is (this level's rollup figures + the undercount caveat)

Our ground-truth federal statistics for NAICS 33441:

The children exhaust the parent — a clean rollup, now verified line by line. The six children's receipts sum to $115.9 billion, their establishments to exactly 3,686, their employment to exactly 279,841, and their payroll to $27.8 billion — matching the level totals.[2][3] That is a useful check: 33441 is fully accounted for by its six parts, with semiconductors and EMS supplying roughly three-quarters of the whole. The SBA thresholds span the same spread as the economics: 550 employees in passives, 750 in boards, assembly and other components, 1,000 in connectors, 1,250 in semiconductors.[12]

One firm-count wrinkle worth knowing. The children's firm counts add to 3,470, but the level reports 3,376 unique firms.[2] The 94-firm gap is real economics, not error: some companies operate in more than one child industry and are counted once at this level but in each child below (see the straddlers named in Section 2). The electronics supply chain is stitched together by firms that cross codes.

Only one employment series rolls up cleanly — use CBP. The Bureau of Labor Statistics measures these industries on different boundaries, and the children now make the mismatch explicit. BLS's payroll survey counts roughly 181,000 semiconductor-manufacturing jobs (April 2026) against County Business Patterns' 116,831, because the two draw different universes.[3][16] Since its 2025 benchmark BLS publishes bare boards (334412) and assembly (334418) as a single combined series, so neither can be pulled out alone.[17] And for other components (334419) BLS typically publishes only the broader 3344 group. Comparing a BLS figure for one child against a Census figure for another will produce nonsense; the CBP column above is the only set that reconciles across all six.

Two vintages of revenue now exist. The 2022 Economic Census — the basis for the shares in Section 2 — is the only source that covers all six children on one footing. A later federal survey (the 2023 Annual Integrated Economic Survey) puts semiconductors at $61.0 billion and printed circuit assembly at $27.2 billion, versus the census figures of $62.4 billion and $24.5 billion.[2][15] These are different instruments, not a revision; do not mix them into one total.

The undercount caveat — read this before comparing to any headline. The $115.9 billion measures only hardware physically manufactured on U.S. soil. It understates the industry's true economic weight in two structural ways that repeat across every child:

1. The value migrated offshore. Roughly 90% of bare boards are imported and the U.S. makes about 4% of global board output (and under 1% of advanced IC substrates); most passive components are made in Asia, with North America accounting for roughly 36% of consumption; the leading-edge chips America designs are built by TSMC in Taiwan, which held ~72% of foundry revenue (~$132.9 billion); most contract assembly is done in Mexico and Asia.[7][8][20][21] The corporate footprints show the same thing: Amphenol employed about 15,000 people in the United States out of roughly 170,000 worldwide at the end of 2025, with 79% of its long-lived assets outside the U.S.; Jabil booked 75% of fiscal-2025 revenue as foreign-source; Flex produced $6.9 billion in Mexico against $4.3 billion in China.[5][19] The census counts the U.S. factory, not the enterprise.
2. The value migrated to design (counted elsewhere). America's most valuable chip names — Nvidia (~$206 billion of 2025 revenue), Broadcom (~$40 billion of chips), Qualcomm (~$39 billion), AMD (~$35 billion) — are "fabless": they design chips and own no factories, so their revenue is not in this manufacturing code at all. U.S.-headquartered firms captured roughly half of global chip revenue (~$318 billion in 2024), many multiples of the "made-in-America" census figure.[9][24]

The scale of the gap is easiest to see child by child against the relevant global market — not as a single sum, since these are separately scoped studies that overlap and cannot be added: chips ~$791.7 billion (2025);[6] all electronic components ~$701 billion (2025);[23] EMS ~$648 billion (2025), heading toward ~$850 billion by 2030;[22] connectors $86.5 billion worldwide in 2024, of which North American sales were $20.1 billion — nearly three times this level's $7.05 billion of U.S. connector shipments;[11] bare boards ~$80–92 billion;[20] and passives ~$45 billion, forecast toward ~$63 billion by 2031.[21]

Crucially, this is not the kind of undercount caused by tiny cash operators slipping through the data — the manufacturing census is mandatory and thorough, and the most fragmented child (334419, 1,189 firms) is well captured.[2] The gap here is the opposite: the industry is undercounted because its value moved to design (booked in other codes) and to foreign fabs (booked nowhere in U.S. data). Two further points: the receipts are a 2022 snapshot, so the AI-driven surge since — global chip sales alone rose 25.6% to $791.7 billion in 2025, on track past $1 trillion in 2026 — is not captured;[6] and this level is statistically unconcentrated (HHI 317.5), even though several of its children are global oligopolies whose scale sits abroad.

---

4. The investable universe — where value concentrates across the children

Because the six children have such different cap tables, there is no single "electronics hardware" trade. Value concentrates unevenly:

• Public depth lives in three children. Semiconductors (334413), EMS (334418), and connectors (334417) hold nearly all the liquid, large-cap U.S. exposure. The clearest names: chip makers Intel, Micron, Texas Instruments, Analog Devices, GlobalFoundries, ON Semiconductor, Microchip, SkyWater (plus fabless designers Nvidia, AMD, Broadcom, Qualcomm, Marvell that define the sector from outside the code);[24] connector compounders Amphenol (APH, $23.1 billion of 2025 sales) and TE Connectivity (TEL, $17.3 billion in fiscal 2025), each individually larger than the entire domestic connector industry, with Aptiv, ITT and Methode as sector-tilted alternatives;[25][26] and the EMS bench Jabil (JBL), Flex (FLEX), Celestica (CLS), Sanmina, Plexus, Benchmark, Fabrinet, Kimball.[5] Tickers, yields, and multiples belong in company-level diligence; these are cyclical industrials and trade like it.
• Private/foreign depth lives in the other three. Bare-board fabrication (334412) has one U.S.-listed pure-play — TTM Technologies (TTMI), ~$2.9 billion of company-wide revenue — plus a smaller, more defense-focused Canadian alternative, Firan Technology Group (TSX: FTG), whose Circuits segment generated C$125.4 million of revenue and C$23.4 million of adjusted EBITDA in fiscal 2025, atop a private, PE-consolidating base.[27][28] Passive components (334416) offer one U.S. name of scale — Vishay (VSH), ~$3.07 billion of FY2025 revenue of which passives are ~$1.63 billion (53%) — with Knowles (KN), Bel Fuse, Vishay Precision Group, Standex and CTS as partial angles, while the true producers (Murata, TDK, Yageo, Kyocera, Samsung Electro-Mechanics, Panasonic) trade only on Asian exchanges and the best U.S. brands (Coilcraft, Bourns, KEMET, Kyocera AVX) are private or foreign-owned.[29][46] "Other components" (334419) is a pick-your-slice bench — SiTime, Knowles, Rogers, CTS, M-tron, Interlink, Kopin, Richardson Electronics — with the global leaders again foreign or private.[30][31]
• The codes leak, so verify what you are buying. The marquee "other components" name, SiTime, makes MEMS-silicon timing devices that may well be classified under semiconductors rather than 334419; neither Amphenol nor TE is a pure connector company; TTM is neither U.S.-only nor purely bare-board.[19][27][31] Exposure should be measured from product and segment disclosure, not from a database industry label.
• There is no dedicated fund for this level. Broad semiconductor/electronics ETFs (for example iShares Semiconductor, ticker SOXX; VanEck Semiconductor, ticker SMH) capture mostly the chip child — and even there they blend fabless designers, equipment vendors and foreign foundries — giving only diluted exposure to boards, passives, connectors, and assembly. An investor wanting the non-chip children generally has to buy individual names or go private.
• The private markets are where the tails are. Across bare boards, EMS, and other components, the domestic industry is a long tail of small, certified, family- or PE-owned shops — the field where private equity roll-ups (defense/medical/quick-turn specialists) and strategic acquirers do their buying (Section 8).

---

5. How the money works

Every child is a manufacturing business, so the economics are about volume, capacity utilization, yield, mix, and the cycle — not fees, rents, or same-store sales. The shared levers:

• Operating leverage is the whole game. Plants carry heavy fixed costs (cleanrooms, plating and assembly lines, test). Fill them and incremental volume drops disproportionately to profit; run them half-empty and margins collapse. Two children now quantify it: bare-board lines have run near 65% utilization in soft periods against roughly 85% needed to hit return-on-capital targets;[33] and at Kimball Electronics a 13% fiscal-2025 sales decline produced a 26% gross-profit decline, partly from lost absorption.[5] This is why every child is cyclical and why utilization is the profit switch.
• Mix and complexity, not commodity volume, is where domestic margin lives. Across all six, the money is in the high-reliability, hard-to-second-source tier — defense/aerospace, medical, automotive-qualified, high-layer-count, high-speed. Commodity work has migrated offshore; U.S. survivors specialize. Vishay reports that only 17% of its resistor revenue was commodity product in 2025, against 52% certified and 31% custom.[29]
• Design-win stickiness is the moat. A part (a connector, a crystal, a specific board, an analog chip) that is "designed in" and qualified into a car platform or a defense program ships for that product's multi-year life with limited price erosion. Qualification is slow and costly — which is exactly what makes the revenue durable. The counterweight, visible in the connector and other-component filings, is that orders remain cancellable or reschedulable and customers keep asking for price reductions.[19][30]
• Input costs and yield. Copper and gold (boards, connectors), tantalum, palladium and nickel (passives), quartz and specialty ceramics (other components), plus energy, water and skilled labor. Scrap and rework destroy margins, so process yield is a core metric.
• Book-to-bill and backlog (orders ÷ shipments; above 1.0 signals a growing backlog) are the shared near-term demand gauges to watch — North American bare-board book-to-bill ran 1.24 in March 2025, and North American EMS book-to-bill ran above 1.0 entering 2024.[5][27]

The defining contrast is capital intensity and margin, and the children now let us lay the whole ladder out — bottom to top, within one five-digit code:

• EMS/assembly (334418) is the floor. Customers own the design and can re-bid the work, and much of reported revenue is pass-through of customer-specified components. Fiscal-2025 gross margins were 8.9% at Jabil, 8.4% at Flex, 8.8% at Sanmina, 10.1% at Plexus and 7.0% at Kimball; operating margins were 4.0%, 4.5%, 5.0% and 3.1% respectively. Double-digit operating margins are almost unheard of.[5]
• Bare boards (334412) sit just above, and were once worse. A federal survey of the industry found average net margin of only 2.7% across 2012–2015, with a third of respondent facilities reporting negative net income in the average year — and a steep scale gradient (6.6% above $40 million of sales, 1.6% below $10 million).[32] Today's survivors, concentrated in advanced and defense work, do better: TTM's company-wide adjusted EBITDA margin was 15.7% in 2025, up from 14.4%.[27]
• Passives and other components occupy the middle. Vishay's combined resistor, inductor and capacitor segments earned a 22.7% gross margin and 17.8% segment operating margin in fiscal 2025; Bel Fuse's magnetics business reached 27.6%.[29] In specialty components, Rogers' advanced-electronics segment earned 29.6% and CTS 38.4%.[30]
• Connectors are the best of the non-chip children. TE Connectivity earned a 35.2% gross margin and 18.6% operating margin on $17.3 billion of fiscal-2025 sales; Amphenol's operating margin rose to 25.4% in 2025 from 20.7%, on volume leverage and cost control.[19][26]
• Semiconductors (334413) are the capital-heavy, high-margin — and most violent — extreme. Industry capital spending reached about $166 billion in 2025 and is projected near $200 billion in 2026, running roughly 35–50% of revenue at foundries and 15–25% at integrated makers.[4] The reward is pricing power and high pay; the cost is whiplash. Micron's revenue rose about 49% in fiscal 2025 and its gross margin swung to 39.8% from 22.4% — while in the same window Microchip's revenue fell about 42% on an inventory correction.[34] Those are two different segments of one child, not one number: memory and analog/embedded run separate cycles.

For an owner, the durable strategy is identical across all six: win pricing power through complexity or certification, run the plant as full as possible, keep yields high, and time capacity to arrive into a shortage rather than a glut.

---

6. What drives demand

A striking feature of this level is that one demand stack powers all six children at once — which is why they rise and fall together:

• AI data centers — the current supercycle, and the dominant swing factor for every child. AI servers need more advanced logic and memory (334413), denser high-layer-count boards (334412 — AI-server layer counts have gone from ~16–20 to 28–36, lifting board value per server by more than 30%), tens of thousands of multilayer ceramic capacitors and high-Q inductors per rack (334416 — one next-generation rack can absorb 40,000–60,000 MLCCs), thousands of high-speed and optical connectors (334417), specialist assembly (334418), and low-jitter precision timing (334419). No child is untouched, and the numbers are showing up in results: TE's digital-data-networks sales rose from $1.3 billion to $2.2 billion in fiscal 2025, and Jabil's Intelligent Infrastructure revenue rose 34%.[5][6][13][14][26]
• Automotive electrification. EVs and advanced driver-assistance systems multiply electronic content per vehicle — more chips, more boards, 3–5× the passive content (on the order of 15,000 multilayer ceramic capacitors versus about 3,000 in a legacy car), more connectors, more sensors — a durable multi-year tailwind across the level.[21][30]
• Defense and aerospace. Radar, avionics, missiles, satellites and secure communications require trusted, U.S.-made, qualified hardware. This is the anchor demand where domestic production is protected and defended across every child — and where the export-control regime (ITAR) actively keeps the work onshore.
• Medical, industrial, 5G/telecom, and IoT. Steady, higher-reliability, diversified demand that favors domestic and specialty suppliers.
• Reshoring — but read it as nearshoring. Tariffs and supply-chain policy are pushing production out of China; the larger flow is to Mexico under the USMCA rather than to U.S. soil. Flex's fiscal-2025 geography — $6.9 billion produced in Mexico against $4.3 billion in China — is the clearest illustration. U.S. capacity wins where intellectual-property protection, export controls or trusted-source requirements outweigh the domestic cost premium.[5]
• The broad electronics inventory cycle sets the volume tone underneath all of it, amplified by distributor destock/restock swings (the 2023–24 correction cut across the group).

The near-term caveat that also applies to all six: a large share of current strength is AI capital spending, which could digest — a shared, forward-looking risk (Section 9).

---

7. Regulation

None of these children is rate-regulated like a utility; the binding rules are trade, materials, quality certification, and industrial policy — and here the level shows an important policy asymmetry, which the children now let us size:

• Industrial subsidy is overwhelmingly concentrated on one child. The 2022 CHIPS and Science Act put $52.7 billion — roughly $39 billion in manufacturing incentives plus an investment tax credit — almost entirely behind semiconductor fabrication (334413); by late 2025 Commerce had allocated over $36 billion across roughly 19 firms, with headline awards to Intel (~$7.86B), TSMC (~$6.6B), Micron (~$6.2B) and Samsung (~$4.75B).[36] The credit is 25% on qualifying equipment placed in service through 2025, rising to 35% thereafter, but is scheduled to expire at the end of 2026 — a live policy risk.[36][37] What the rest of the supply chain received is smaller by orders of magnitude: a $39.9 million Defense Production Act award to Calumet Electronics for bare-board and substrate capacity, an $11.7 million DPA award to expand printed-circuit-assembly capacity for hypersonic systems, and about $300 million of Commerce funding for advanced-packaging substrate research.[39][40][41] The proposed Protecting Circuit Boards and Substrates (PCBS) Act would extend CHIPS-style support to boards and substrates — roughly $3 billion plus a 25% credit for buyers of U.S.-made boards — but as of mid-2026 it had not been enacted.[7] Passives, connectors, and assembly have no equivalent program. America is subsidizing the chip but barely the board, passive, or assembly it must be built onto.
• Subsidy is now edging into ownership. In August 2025 the U.S. government converted part of Intel's support into a roughly 10% equity stake, with reports it could seek stakes in other CHIPS recipients — a departure from grant-only support that equity investors in this child now have to price.[38]
• Export controls apply across the level, hardest at the top. ITAR, the Export Administration Regulations (EAR), and — for advanced chips and tools sold to China — the Commerce Department's Bureau of Industry and Security (BIS) rules govern defense- and high-end product. BIS policy has whipsawed repeatedly: after restricting advanced AI chips outright, by late 2025 it moved to case-by-case licensing of some advanced parts into China in exchange for the U.S. government taking 25% of the revenue, effective January 2026.[42][43] That is a live business risk for the semiconductor child specifically, and the whiplash itself is the risk.
• Materials and hazardous-substance rules are universal. The EU's RoHS (ten restricted substances) and REACH regimes, U.S. conflict-minerals disclosure under Dodd-Frank (the "3TG" minerals — tin, tungsten, tantalum, gold, with tantalum especially relevant to passives and gold plating to connectors), EPA's TSCA PFAS reporting, EPA electroplating effluent rules and OSHA beryllium limits apply across essentially every child and add real supply-chain reporting and process cost.
• Critical-materials dependence is federally documented. The United States was 100% net-import reliant for both gallium (with China supplying 99% of primary low-purity production) and tantalum (roughly $190 million of consumption) in 2025 — one input binding on compound semiconductors, the other on tantalum capacitors.[44]
• Quality certification is the real gatekeeper. AS9100 (aerospace), ISO 13485 and the FDA's Quality Management System Regulation (effective February 2026, medical), IPC workmanship standards including J-STD-001 and IPC-A-610 Class 3, IPC-1791 "trusted" fabricator status and DMEA accreditation for defense boards, AEC-Q qualification for automotive, and MIL-PRF/MIL-DTL specifications decide which lucrative work a plant can win. These are barriers to entry more than compliance line items — and the moat that keeps high-value work onshore.
• Trade policy is a live, two-way variable. Section 301 tariffs on Chinese components and 2025–26 tariff actions raise input costs but also push reshoring.[45] The direction is not uniformly favourable to domestic producers: a U.S. fabricator can pay duty on full-value copper, laminate and chemical inputs while an imported finished board is dutied only once, and components routinely cross the U.S., Canadian and Mexican borders several times.
• Environmental permitting (metal-finishing effluent, hazardous waste, PFAS, water and energy use) is a genuine cost and siting barrier, heaviest for board fabrication and chip fabs.

---

8. Consolidation

The level is statistically unconcentrated (HHI 317.5; top-4 firms 29.3% of receipts),[2] but that average hides a wide spread — and is partly an artifact of aggregation. The level's CR4 sits below that of semiconductors, EMS and connectors individually, because combining six heterogeneous product lines dilutes measured concentration.

Concentration varies sharply by child. Semiconductors are the most top-heavy (CR4 48%, HHI 897), followed by EMS (CR4 43.6%, HHI suppressed) and connectors (CR4 39.6%, HHI 582.5); bare boards sit in the middle (CR4 28.7%, HHI suppressed), while passives (CR4 24.1%, HHI 226.3) and especially "other components" (CR4 ~11%, HHI 84.6) are among the most fragmented corners of all manufacturing.[2] Read the level as a barbell: a few giants at the top of some children, a long fragmented tail beneath all of them. And read the domestic HHIs with care — connectors measure "unconcentrated" in the United States while the global market is led by a stable top tier, and passives measure unconcentrated only because the oligopolists' plants are abroad.[11][21]

Two consolidation engines run across the level:

1. Strategic M&A at the top. Connectors show the archetype in Amphenol, a decentralized serial acquirer that rolled up dozens of businesses and, in 2024–26, closed a $10.5 billion CommScope connectivity deal.[25] Passives show the other pattern — Asian groups buying Western brands: Yageo acquired KEMET (~$1.6–1.8B), Pulse Electronics (~$740M) and Chilisin; Kyocera absorbed AVX.[46] In semiconductors, leading-edge fabs are usually built, not bought, so M&A clusters in analog, design tools, and packaging.
2. Private-equity roll-ups of the tail. Across bare boards, EMS, and other components, PE firms are aggregating small certified shops into regional platforms — Summit Interconnect (sold by HCI Equity Partners to Lindsay Goldberg in 2024) and AdvancedPCB (APCT plus Advanced Circuits, built through more than ten acquisitions, with the APCT transaction at ~$220 million enterprise value) in boards, and steady roll-ups of defense/medical job shops in assembly and specialty components.[47] Qualification barriers and fragmentation make this the most active private thesis at this level.

Consolidation is not frictionless. The clearest recent counterexample sits in "other components": DuPont's ~$5.2 billion agreement to buy Rogers Corporation was terminated in late 2022 after failing to clear Chinese antitrust review, with Rogers collecting a $162.5 million break fee and continuing as an independent public company.[48] Cross-border review — antitrust abroad, CFIUS at home — now shadows any large deal in this level.

The through-line: capital and certification are the moats, so scale keeps concentrating at the top of each child even as the statistical industry stays fragmented at the trailing edge and in the specialty tail.

---

9. Risks

Because one demand stack drives all six children, the risks are largely shared — and correlated:

• Cyclicality. The defining risk. Utilization-driven margins swing hard with the electronics cycle; the industry regularly overbuilds into shortages, then suffers gluts. Memory is the most extreme (Micron up ~49% in fiscal 2025), but the swings are not confined to it — Microchip fell ~42% on an analog/embedded inventory correction in the same window, and every child breathes in and out.[34]
• AI-capex concentration. A large share of current growth leans on a handful of hyperscale AI buyers. A pause or "digestion" of AI data-center spending would hit all six children at once — the correlated downside of a shared driver.
• The offshore cost wall and import competition. For boards and passives especially, the structural cost gap versus Asia is existential; commodity work will not return without policy support.[7][20][21]
• Geopolitics and China. Leading-edge chipmaking concentrated in Taiwan is a flashpoint; China is a major buyer, a major source of inputs and low-end supply, and — via its antitrust review — a gate on Western consolidation; export-control and tariff whiplash can strand revenue or disrupt inputs overnight.[8][42][43][48]
• Input-cost, materials and currency swings. Copper, gold, tantalum, palladium, nickel, quartz and specialty materials feed straight into margins, amplified by foreign-exchange moves because so much production is offshore. Two inputs carry documented 100% U.S. import reliance — gallium and tantalum.[44]
• Customer concentration. Defense-heavy and hyperscaler-heavy players depend on a few programs or buyers; one program's timing can move a supplier's quarter. The disclosed levels are real — Jabil's five largest customers were 36% of fiscal-2025 revenue with one at 16%; Sanmina's top ten were 52%; M-tron's top four were 61%.[5][31]
• Policy dependence. Much of the reshoring thesis rests on subsidies — the CHIPS tax credit expires end-2026, and the PCBS Act for boards has not passed — while the new government-equity model adds a novel governance risk.[7][36][38]
• An eroded domestic skills base. The 80.6% collapse in bare-board employment since 2000 means the labor pool for a reshoring push is far thinner than the policy assumes, and federal surveys repeatedly identify skilled-worker availability as the binding limit on adding shifts and raising utilization.[10][32] The same shortage appears at the top of the level, where the industry association projects tens of thousands of unfilled semiconductor positions by 2030.
• Capital intensity and technology risk. New advanced capacity costs a great deal and takes 18–36 months; missing a process node, a packaging transition, or a substitution (MEMS timing displacing quartz; optical displacing copper; integrated passives removing discrete parts) can be financially punishing.
• Valuation risk for public holders. The best franchises — the leading chip names, Amphenol, and the fastest-growing specialty names — trade at premium multiples, so much good news is already priced in.

---

10. How to invest & outlook

Public-market routes — choose the child first.

• Broadest, most liquid exposure is the semiconductor child: chip makers (Intel, Micron, Texas Instruments, Analog Devices, GlobalFoundries, ON Semiconductor, Microchip, SkyWater) for direct U.S.-manufacturing exposure; fabless designers (Nvidia, AMD, Broadcom, Qualcomm, Marvell) for the highest-margin way to own AI compute; equipment and design-software names (Applied Materials, Lam Research, KLA, Synopsys, Cadence) as a "picks-and-shovels" play; and semiconductor ETFs (SOXX, SMH) to spread single-stock and cycle-timing risk — noting those funds blend all of the above rather than tracking the manufacturing code.[24]
• Best-in-class industrial compounders are in connectors: Amphenol (APH) and TE Connectivity (TEL), with content-per-device as the structural edge, plus Aptiv (auto tilt) and ITT (aerospace/defense).[25][26][35]
• Direct AI-hardware and reshoring leverage is in EMS: Celestica (CLS) and Fabrinet (FN) for AI/optics; Jabil (JBL) and Flex (FLEX) for diversified scale; Sanmina, Plexus, Benchmark for the higher-margin regulated tier.[5]
• Narrow, specialized bets are in bare boards (TTM/TTMI, with FTG on the TSX as a smaller defense-focused alternative),[27][28] passives (Vishay/VSH, Knowles/KN, Bel Fuse, or foreign listings for the real producers),[29] and other components (SiTime, Knowles, Rogers, CTS, M-tron, Kopin, Richardson — treat as individual small/mid-cap industrials, not a sector).[30][31]

Reserve valuation judgments (multiples, dividend policy) for company-level diligence; dividends across the group are modest, with capital more often returned via buybacks, and the cyclical names trade on where you are in the cycle.

Private-market routes. This is where the non-chip children truly live. The active theses: PE roll-ups of certified bare-board and EMS job shops (defense/ITAR, medical, quick-turn); direct ownership of specialty passives and other-component makers serving defense and medical programs; and backing the fragmented specialty tail that the public consolidators covet. In semiconductors, private capital is better deployed in the supply chain around the fabs (packaging/test, materials, real assets, construction) than in financing a new leading-edge fab outright. The diligence traps recur across all six children and are worth naming: reconcile distributor sell-in to actual point-of-sale consumption; separate commodity, certified and custom revenue; test qualification life, sole-source status and change-control clauses; examine yield and utilization line by line; quantify environmental and plating liabilities; and normalize working capital across the inventory cycle. The largest single underwriting error at this level is capitalizing peak shortage-era earnings as structural profitability.

Outlook (forward-looking). The tailwinds are aligned and, unusually, point the same way for all six children: an AI capital-spending wave lifting demand and per-unit content; automotive electrification raising content per vehicle; firm defense budgets steering dollars to trusted domestic suppliers; and reshoring/trade policy nudging production out of China — though more often to Mexico than to the United States.[5][6][11][14] The counterweights are equally shared: deep cyclicality, a structural cost gap versus Asia for the commodity tiers, a thinned-out domestic skills base, dependence on policy that is partly unlegislated or expiring, and the fact that much of the current strength is AI capex that could normalize. The most durable investment truth at this level is that timing the cycle matters as much as picking the company — and that which child you buy determines almost everything about the risk, the ownership, the margin structure, and the way in. The plausible path is a smaller-but-higher-value U.S. footprint concentrated in defense, AI, and specialty work, rather than a return to commodity-scale dominance.

---

Sources

1. NAICS Association / U.S. Census Bureau. "NAICS 33441 — Semiconductor and Other Electronic Component Manufacturing" and its six child definitions (334412, 334413, 334416, 334417, 334418, 334419). 2022. https://www.naics.com/naics-code-description/?code=33441
2. U.S. Census Bureau. 2022 Economic Census — receipts, firm counts, and concentration ratios for NAICS 33441 and each child (level: receipts $115.9B, 3,376 firms, CR4 29.3%, CR8 40.2%, CR20 53.2%, CR50 65.8%, HHI 317.5; children's CR4/HHI as cited). 2022. https://www.census.gov/programs-surveys/economic-census.html
3. U.S. Census Bureau. County Business Patterns — establishments (3,686), employment (279,841), annual payroll ($27.8B) for NAICS 33441 and each child. 2023. https://www.census.gov/programs-surveys/cbp.html
4. Semiconductor Intelligence / Deloitte. "CapEx Up for Foundry, Memory" (fab build cost ~$10B plus ~$5B equipment; industry capex ~$166B in 2025, ~$200B in 2026; capex intensity by business model). 2026. https://www.semiconductorintelligence.com/capex-up-for-foundry-memory/
5. Jabil Inc. / Flex Ltd. / Plexus Corp. / Kimball Electronics FY2025 Forms 10-K (EMS gross margins 8.9% / 8.4% / 10.1% / 7.0% and operating margins 4.0% / 4.5% / 5.0% / 3.1%; Jabil 75% foreign revenue and top-five customers 36%; Flex $6.9B produced in Mexico vs $4.3B in China; Kimball sales −13% with gross profit −26%; Sanmina FY2025 8.8% gross margin and top-ten customers 52%; North American EMS book-to-bill above 1.0). 2025. https://www.sec.gov/Archives/edgar/data/898293/000162828025045293/jbl-20250831.htm
6. Semiconductor Industry Association. "Global Annual Semiconductor Sales Increase 25.6% to $791.7 Billion in 2025" (on track past $1 trillion in 2026). 2026. https://www.semiconductors.org/global-annual-semiconductor-sales-increase-25-6-to-791-7-billion-in-2025/
7. Summit Interconnect / Printed Circuit Board Association of America. "Bipartisan PCBS Act" (U.S. ~4% of PCB output; Asia ~90%; under 1% of advanced substrates; ~$3B authorization plus 25% buyer credit; not yet enacted). 2025. https://summitinterconnect.com/blog/article/bipartisan-pcbs-act/
8. Dataconomy. "TSMC Dominates Foundry Market With 72% Share" (full-year foundry revenue ~$132.9B; most leading-edge U.S.-designed logic built in Taiwan). 2025. https://dataconomy.com/2025/12/23/tsmc-dominates-foundry-market-with-72-share-in-q3-2025/
9. Semiconductor Industry Association. "2025 SIA Factbook" (U.S.-headquartered firms ~$318B / 50.4% of global chip revenue, 2024; end-market mix). 2025. https://www.semiconductors.org/wp-content/uploads/2025/05/2025-SIA-Factbook-FINAL-1.pdf
10. U.S. Bureau of Labor Statistics. "Industries with employment decreases from 2000 to 2024" (NAICS 334412: 137,501 in 2000 to 26,676 in 2024; −80.6%). 2025. https://www.bls.gov/opub/ted/2025/industries-with-employment-decreases-from-2000-to-2024.htm
11. Bishop & Associates / TTI. "2024 World Connector Market Review" (worldwide connector shipments $86.5B, +5.6%; North American sales $20.1B). 2025. https://www.ttieurope.com/content/ttieurope/en/resources/marketeye/categories/connectors/me-bishop-20250228.html
12. U.S. Small Business Administration. Table of Small Business Size Standards (NAICS 33441 children: 550 employees for 334416; 750 for 334412, 334418, 334419; 1,000 for 334417; 1,250 for 334413). 2023. https://www.sba.gov/document/support-table-size-standards
13. TrendForce. "2026 MLCC Market Update" (polarized demand: AI and advanced applications tightening, consumer electronics weak). 2026. https://www.trendforce.com/presscenter/news/20260205-12914.html
14. UGPCB. "AI Server PCB: High-Frequency Materials & Market Trends" (AI-server layer counts from 16–20 to 28–36; PCB value per server +30%). 2026. https://www.ugpcb.com/news/trade-news/ai-server-pcb/
15. U.S. Census Bureau. Annual Integrated Economic Survey — NAICS 334413 sales $61.0B and NAICS 334418 revenue $27.2B. 2023. https://data.census.gov/table/AIESBASICTIMESERIES.AIES31BASIC01?codeset=naics~334413&g=010XX00US
16. U.S. Bureau of Labor Statistics. Current Employment Statistics, NAICS 334413 (~181,000 employees, April 2026). 2026. https://www.bls.gov/ces/data/employment-and-earnings/2026/table1a_202605.htm
17. U.S. Bureau of Labor Statistics. "2025 Benchmark Series Changes" (NAICS 334412 and 334418 published as one combined employment series). 2025. https://www.bls.gov/ces/notices/2025/2025-bmk-series-changes.htm
18. Printed Circuit Board Association of America. "PCB 101" (approximately 145 remaining U.S. bare-board fabricators). March 2025. https://www.pcbaa.org/wp-content/uploads/2025/04/PCB101_WEB-Doc_MAR2025_v6.pdf
19. Amphenol Corporation. Form 10-K FY2025 (operating margin 25.4% vs 20.7%; ~15,000 U.S. of ~170,000 worldwide employees; 79% of long-lived assets outside the U.S., 37% in China; ~65% of sales outside the U.S.; distributors ~19% of sales; order and pricing terms). 2026. https://www.sec.gov/Archives/edgar/data/820313/000110465926013549/aph-20251231x10k.htm
20. HiL Electronic. "Global PCB Market Hits $100B on AI Demand" (2025 market ~$80–92B; Asia ~90% of supply). 2025. https://hilelectronic.com/global-ai-pcb-market/
21. SNS Insider. "Passive Electronic Components Market Size" (global ~$45B in 2025 toward ~$63B by 2031; North America ~35.8% of consumption; EV passive content 3–5×). 2026. https://finance.yahoo.com/sectors/technology/articles/passive-electronic-components-market-size-073000399.html
22. Fortune Business Insights. "Electronic Manufacturing Services (EMS) Market" (global EMS ~$648B in 2025 toward ~$850B by 2030). 2025. https://www.fortunebusinessinsights.com/electronic-manufacturing-services-ems-market-105519
23. Mordor Intelligence. "Electronic Components Market Size & Share" (global component output ~$701B, 2025). 2025. https://www.mordorintelligence.com/industry-reports/electronic-components-market
24. TechPowerUp / AnySilicon. "Top 10 Global Fabless IC Firms 2025" (Nvidia ~$206B; Broadcom ~$40B chips; Qualcomm ~$39B; AMD ~$35B — designed in the U.S., built abroad, outside 334413). 2026. https://www.techpowerup.com/347923/ai-compute-demand-drives-44-yoy-growth-for-top-10-global-fabless-ic-firms-in-2025
25. Amphenol Corporation / StockTitan. "Amphenol Reports Record Fourth Quarter and Full Year 2025 Results" ($23.1B FY2025 sales; $10.5B CommScope connectivity acquisition). 2026. https://www.stocktitan.net/news/APH/amphenol-reports-record-fourth-quarter-and-full-year-2025-nkeprvqejb11.html
26. TE Connectivity. Form 10-K FY2025 ($17.3B net sales; 35.2% gross margin; 18.6% operating margin; digital-data-networks sales from $1.3B to $2.2B). 2025. https://www.sec.gov/Archives/edgar/data/1385157/000110465925109150/tel-20250926x10k.htm
27. TTM Technologies, Inc. "Fourth Quarter and Fiscal Year 2025 Results" (~$2.9B company-wide revenue; adjusted EBITDA margin 15.7% vs 14.4%; North American PCB book-to-bill 1.24, March 2025). 2026. https://investors.ttm.com/news-events/press-releases/detail/400/ttm-technologies-inc-reports-fourth-quarter-fiscal-year
28. Firan Technology Group. 2025 Annual Report (Circuits segment: C$125.4M revenue, C$23.4M adjusted EBITDA). 2026. https://www.ftgcorp.com/wp-content/uploads/2026/02/Annual-Report_FINAL-FTG-2025.pdf
29. Vishay Intertechnology, Inc. Form 10-K FY2025 and segment financial data (~$3.07B revenue; resistors, inductors and capacitors ~$1.63B / 53%; passive segments 22.7% gross margin and 17.8% operating margin; 17% commodity / 52% certified / 31% custom resistor mix; Bel Fuse magnetics 27.6% gross margin). 2026. https://www.sec.gov/Archives/edgar/data/103730/000010373026000019/form10k.htm
30. CTS Corporation Form 10-K FY2025 ($541.3M sales; 38.4% gross margin) and Rogers Corporation Form 10-K FY2025 (Advanced Electronics Solutions 29.6% gross margin; materials and order terms). 2026. https://www.sec.gov/Archives/edgar/data/26058/000119312526067039/cts-20251231.htm
31. SiTime Corporation / Knowles / Rogers / CTS / M-tron results (precision timing, acoustics, high-frequency materials, frequency control — the "other components" public bench; M-tron top-four customers 61% of revenue; SiTime's MEMS timing straddles the semiconductor boundary). 2025–26. https://www.nasdaq.com/articles/sitime-corporation-reports-fourth-quarter-and-fiscal-year-2024-financial-results-revenue
32. U.S. Bureau of Industry and Security. "U.S. Bare Printed Circuit Board Industry Assessment" (2012–2015: average net margin 2.7%; 6.6% above $40M of sales vs 1.6% below $10M; 32% of respondents reporting negative net income; skilled-labor constraints). 2017. https://media.bis.gov/media/documents/u.s.-bare-printed-circuit-board-industry-assessment-2017.pdf
33. OpenPR. "Institutional Capital Targets PCB Exposure Amid Substrate and Advanced Packaging Boom" (utilization ~65% in soft periods vs ~85% needed for return targets). 2025. https://www.openpr.com/news/4453427/institutional-capital-targets-pcb-exposure-tooling-amid
34. Micron Technology FY2025 Form 10-K (revenue ~$37.4B, +49%; gross margin 39.8% vs 22.4%) and Microchip Technology FY2025 results (revenue ~$4.4B, −42% on inventory correction). 2025. https://www.sec.gov/Archives/edgar/data/723125/000072312525000038/mu-20251125.htm
35. GM Insights. "Connector Market Size & Share, 2026–2035" (content-per-device growth; global market structure; end-market drivers). 2026. https://www.gminsights.com/industry-analysis/connector-market
36. GovConFeed / Manufacturing Dive. "CHIPS and Science Act — $52.7B; ~$39B manufacturing incentives; over $36B allocated across ~19 firms; tax credit expiring December 2026." 2025. https://govconfeed.com/article/chips-act-30-billion-awarded-tax-credit-december-2026
37. U.S. Internal Revenue Service. "Advanced Manufacturing Investment Credit" (Section 48D: 25% through 2025, 35% thereafter). 2025. https://www.irs.gov/credits-deductions/advanced-manufacturing-investment-credit
38. TrendForce. "U.S. Reportedly Mulls Stakes in CHIPS Act Recipients After Intel" (~10% U.S. government equity stake in Intel). 2025. https://www.trendforce.com/news/2025/08/20/news-u-s-reportedly-mulls-stakes-in-chips-act-recipients-after-intel-raising-risks-for-tsmc-samsung/
39. U.S. Department of Defense. "DoD Awards $39.9 Million to Strengthen U.S. Supply Chains for Printed Circuit Boards" (Defense Production Act award to Calumet Electronics). 2023. https://www.defense.gov/News/Releases/Release/article/3589418/dod-awards-399-million-to-strengthen-us-supply-chains-for-printed-circuit-boards/
40. U.S. Department of Defense. "Department of Defense to Expand Manufacturing of Printed Circuit Board Assemblies" ($11.7M Defense Production Act award for hypersonic-systems PCBA capacity). 2024. https://www.defense.gov/News/Releases/Release/Article/3701760/department-of-defense-to-expand-manufacturing-of-printed-circuit-board-assembli/
41. U.S. Department of Commerce. "CHIPS for America Announces Funding Opportunity to Expand U.S. Semiconductor Advanced Packaging Capabilities" (~$300M for advanced-packaging substrate R&D). 2024. https://www.commerce.gov/news/press-releases/2024/02/chips-america-announces-funding-opportunity-expand-us-semiconductor
42. Congressional Research Service. "U.S. Export Controls and China: Advanced Semiconductors" (R48642; BIS rules, ITAR/EAR framework). 2025. https://www.congress.gov/crs-product/R48642
43. Baker McKenzie. "BIS Revises License Review Policy for Advanced Computing Commodities / AI Semiconductors to China and Macau" (case-by-case licensing with 25% revenue share, effective January 2026). 2026. https://sanctionsnews.bakermckenzie.com/bis-revises-license-review-policy-for-advanced-computing-commodities-ai-semiconductors-to-china-and-macau-when-exported-from-the-united-states/
44. U.S. Geological Survey. Mineral Commodity Summaries 2026 (U.S. 100% net-import reliant for gallium, China 99% of primary production; 100% net-import reliant for tantalum, ~$190M consumption). 2026. https://pubs.usgs.gov/periodicals/mcs2026/mcs2026.pdf
45. Accuris. "New Electronic Component Tariffs: How to Prevent Supply Chain Disruptions" (Section 301 and 2025–26 tariff actions; sourcing shifts). 2026. https://accuristech.com/blog/electronic-component-tariffs/
46. Wikipedia / EEPower. "KEMET Corporation" (Yageo acquisition ~$1.6–1.8B) and "Yageo to Acquire Pulse Electronics for US$740 Million" / Kyocera–AVX and Chilisin (Asian roll-up of Western passive brands). 2021–25. https://en.wikipedia.org/wiki/KEMET_Corporation
47. PR Newswire / Lincoln International. "HCI Equity Partners Completes Sale of Summit Interconnect" (to Lindsay Goldberg, 2024) and the APCT / Advanced Circuits combination into AdvancedPCB (~$220M enterprise value, 10+ acquisitions). 2023–24. https://www.prnewswire.com/news-releases/hci-equity-partners-completes-sale-of-summit-interconnect-301385097.html
48. Rogers Corporation. "Rogers Announces Termination of Merger Agreement with DuPont" (~$5.2B deal terminated after failing Chinese antitrust review; $162.5M break fee). 2022. https://www.rogerscorp.com/news/2022/rogers-announces-termination-of-merger-agreement-with-dupont