Computer and Electronic Product Manufacturing (U.S.) — NAICS 334

A Histometrics rollup primer for public-market and private investors. This page sits one level above the six industry groups beneath it. Its job is the contrast across those six — their relative size, which way each is growing, who owns them, and how you actually buy exposure — plus this level's own ground-truth federal figures. For company-by-company depth, follow the links to each child primer.

("NAICS" is the North American Industry Classification System, the standard code framework U.S. statistical agencies use to define industries. 334 is a "subsector," a three-digit rollup; it contains six four-digit "industry groups," 3341 through 3346.)

---

1. Overview

NAICS 334 is where the physical hardware of the digital economy is built in the United States: chips and circuit boards, the instruments that sense and measure the world, the gear that carries voice and data, computers and their storage, consumer audio-video equipment, and the tape and discs data is archived on. It is one of the highest-value, most engineering-intensive slices of American manufacturing — the level pays about $105,000 per worker and ships about $432,000 of output per worker, both far above the manufacturing average, because the value here comes from design, patents, and certification rather than headcount.[1]

The name is misleading in a way worth fixing up front. Despite "Computer" leading the title, actual computer manufacturing is one of the smallest pieces of this subsector — under 6% of it. The subsector's center of gravity is two very different heavyweights: instruments (the picks-and-shovels sensors, medical devices, and defense electronics of NAICS 3345, ~48% of revenue) and semiconductors and electronic components (the chips, boards, and passives of NAICS 3344, ~33%). Everything else — communications gear, computers, audio-video, and recorded media — splits the remaining fifth.

For an investor, the single most useful thing this page does is lay the six children side by side, because they are genuinely different businesses. One is a diversified, defensive, recurring-revenue "instruments" basket; one is the deeply cyclical chip-and-component complex; one is three-markets-in-one communications gear; one is a small import-exposed consumer-electronics remnant that is shrinking in headcount while posting the fastest labor-productivity growth of any manufacturing industry the Bureau of Labor Statistics examined (11.2% a year, 2019–2024); and one is a $1.6 billion niche split between dying optical discs and booming archival tape.[4] They differ in size by more than 100-to-1, in concentration, and — most decisive for how you invest — in whether the leaders are U.S.-listed, foreign-listed, or private.

---

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

The six share a factory-floor DNA (design-heavy electronics, a thin domestic assembly slice, value migrating into chips and software) but sell into different worlds and behave differently as investments. This table is the heart of the page; it is ordered largest-first. ("CR4" is the combined revenue share of the four biggest firms; "HHI," the Herfindahl-Hirschman Index, sums every firm's squared market share — U.S. antitrust agencies treat anything below ~1,500 as "unconcentrated." "AI" is artificial intelligence; "PE" is private equity.)

The one-paragraph read. Two children are this subsector: instruments (3345) and components (3344) together are about four-fifths of the revenue and 85% of the jobs. They are near-opposites in temperament — 3345 is a diversified, certification-moated, recurring-revenue basket that rides many separate cycles and is therefore steady; 3344 is the classic boom-bust chip-and-board complex where timing the cycle matters as much as picking the company. Communications gear (3342) is a distant but real third, itself three unrelated markets under one label. The bottom three — computers (3341), audio-video (3343), and media (3346) — are small on U.S. soil not because the products are unimportant, but because the manufacturing largely happens offshore while the design, brand, and IP stay American (Section 3). Ownership is the sharpest divide for investors: public-market depth is concentrated in chips, connectors, contract assembly, defense electronics, medical devices, and computers, while the biggest exposures in wireless, media, and premium audio are foreign-listed or private.

One structural note — corrected in this pass. The revised children establish that five of the six (3341, 3343, 3344, 3345, and 3346) each contain exactly one five-digit industry, so their child primers double as the detailed pages; only communications equipment (3342) genuinely splits at that tier, into wired networking (33421, ~10% of that child), wireless and broadcast gear (33422, ~78%), and signaling (33429, ~11%).[2][3][4][5][6][7] For the other five, the interesting internal variation lives one tier further down at the six-digit level — most dramatically in 3345, whose nine sub-industries run from a $53.5 billion defense-electronics business to a ~$3.0 billion environmental-controls niche; in 3344, whose six run from semiconductors at ~54% of that child's receipts down to passives at ~3%; and in 3341, which divides into computers (~59%), storage (~24%), and peripherals (~17%).[2][5][6] Audio-video (3343) and media (3346) are single industries all the way down.[4][7]

---

3. Size (this level's rollup figures)

Our ground-truth federal statistics for NAICS 334, and — a useful check — how the six children reconcile up to them:[1]

Plants, workers, revenue, and payroll add up cleanly from the six children, and the revised pages confirm the same identity holds one tier further down inside each of them.[2][3][5][6] Firms are the one line that doesn't add — the children sum to 10,026, about 165 more than the subsector's 9,861 — because a firm counts once inside each industry group it operates in, but only once at the parent. That 165-firm gap is the fingerprint of diversified companies (a Honeywell, a Cisco, an HP) whose products straddle two or more children; every child reports the identical arithmetic against its own sub-industries.[2][3][5][6]

Which federal series you use now matters, and it is a level-wide caution. All six children surfaced alternative federal readings of their own scope, and the alternatives do not agree with the Economic Census / County Business Patterns basis used above — nor do they all point the same way. Audio-video is measured at $5.09 billion of Census receipts but $4.39 billion of BLS sectoral output, and at 10,977 payroll employees but ~19,700 workers once BLS adds the self-employed.[4] Communications shows the reverse for networking: BLS counts 13,506 jobs where County Business Patterns counts 8,893.[3] Semiconductors show it larger still — BLS counts roughly 181,000 jobs against CBP's 116,831 for the same sub-industry.[5] And the newer Annual Integrated Economic Survey runs above the 2022 Census in some places (electromedical $45.5 billion versus $39.6 billion; environmental controls $4.196 billion versus ~$3.0 billion) and below it in others (networking $3.92 billion versus $4.30 billion).[3][6] These are different instruments with different universes, not revisions. The $352.3 billion and 815,444 above hold only on a consistent EC-2022 / CBP-2023 basis; do not mix series into a new total.

Concentration looks trivial at this level — and that reading is a trap. An HHI of 93.9 and a CR4 of just 15.2% say "highly fragmented, competitive." Yet every child is more concentrated than the whole (child HHIs run from 163 up to 707), and the revised children show the same dilution repeating at every tier: 3345's HHI of 163.2 is lower than any of its nine sub-industries', and its CR4 of 19.0% is below the lowest of them (23.5%); 3342's HHI of 506.4 sits below its largest child's 795; 3344's CR4 of 29.3% sits below its three largest sub-industries individually; and 3341's CR4 of 43.9% masks computers at 60.5% and storage at 83.6%.[2][3][5][6] The real markets underneath are outright oligopolies — semiconductor test is a two-firm ~80% duopoly, high-end imaging's top five vendors hold ~96% of the U.S. market, hearing aids six firms ~80%, hard drives three makers shipping more than 95%, and LTO data tape a two-company world.[2][6][7] The subsector looks unconcentrated only because it pools children dominated by different firms; stacking them dilutes any single company's share. Concentration is real one or two levels down, and it washes out on the way up.

Undercount caveat — essential here. Read $352 billion as value physically manufactured inside U.S. factories coded to 334, not as the size of the electronics economy Americans design, sell, or consume. Two structural forces push the true figure far higher, and both run through most children:

1. Production moved offshore. The marquee U.S. brands design at home and build abroad. Most consumer electronics, roughly 90% of bare circuit boards (the U.S. makes about 4% of global board output and under 1% of advanced substrates), most passive components, and the leading-edge chips America designs are made in Asia and imported — TSMC alone held about 72% of foundry revenue, and the U.S. imported roughly $115 billion of telephones in 2024 and more than $450 billion of computer hardware and semiconductors in 2025, almost none of it counted here.[2][3][5]
2. Value moved to design and services, counted in other codes. America's most valuable chip names — Nvidia, AMD, Qualcomm, Broadcom — are "fabless" (they design and own no factories), so their revenue sits elsewhere; U.S.-headquartered firms captured about $318 billion, ~50.4% of global chip revenue, in 2024.[5] Apple's Mac line alone earned about $33.7 billion in fiscal 2025, nearly none of it domestic manufacturing, and Nvidia — classified as a semiconductor maker — captures an estimated ~90% of AI-accelerator spending while setting the economics for the computer and storage businesses.[2] The giants that dominate instruments are booked under whichever code is their primary product, with recurring service and software often classified as services: GE HealthCare split 2025 into $13.7 billion of product and $7.0 billion of service, and its service remaining performance obligations ($10.7 billion) are more than double its product obligations.[6]

The children now disagree on the direction of the trade gap, and that is worth stating plainly. Five of the six describe an import-swamped picture in which domestic shipments badly understate domestic consumption. Instruments does not: the electromedical sub-industry runs the other way, with U.S. receipts of $45.5 billion (AIES 2023) against an estimated U.S. domestic market of about $26.3 billion in 2024, because America is a net exporter of high-end implants and devices — even though the broader medical-instruments trade line showed $35.8 billion of exports against $41.3 billion of imports in 2024.[6] Read 334 as a mix of net-export and net-import pieces, not one import-swamped whole.

Crucially, this is not the kind of undercount caused by cash-only micro-operators slipping through the data. The manufacturing census is mandatory and thorough, and at ~$36 million of receipts per firm on average, this is a subsector of substantial incorporated companies, not hidden proprietors. One nuance the revised children add: "small business" means something unusual here, because SBA size standards across these codes run from 550 to 1,350 employees — so 86.5% of the computer sub-industry's firms, 96.5% of the peripheral sub-industry's, 94.84% of the signaling industry's, and an estimated 99.1% of media firms qualify as small.[2][3][5][6][7] The receipts-weighted view of a few giants and the firm-weighted view of several thousand modest plants are both true; the modest tail is real in headcount and thin in receipts, concentrated in computer peripherals, custom audio, specialty components, building controls, and vinyl pressing.[2][4][5][6][7] The gap is the offshoring-and-value-migration kind: the true economic weight of what these firms design and sell globally is several times the made-in-America figure.

---

4. Investable universe (where value concentrates across the children)

Because the subsector is really six markets, the investable map is six maps, and public depth is unevenly distributed. (A "ticker" is a stock's exchange symbol; "ETF" is an exchange-traded fund, which trades like a stock; "EMS" is electronics-manufacturing services, i.e. contract factories; "ADR" is an American Depositary Receipt, a foreign share traded in U.S. markets.)

• Instruments (3345) — the biggest child — offers its cleanest public franchises in defense electronics (through aerospace-and-defense primes) and health care (medical-device and imaging large-caps). The industrial-measurement middle is dominated by diversified serial acquirers — Danaher, Fortive, Roper, AMETEK, Emerson, Honeywell, Thermo Fisher, GE HealthCare — where instruments are one segment, and the scale mismatch against the federal data is startling: GE HealthCare's Imaging segment alone (~$9.2 billion in 2025) is about the size of the entire irradiation sub-industry's $9.18 billion of U.S. shipments, Honeywell's Building Automation segment booked ~$7.4 billion against environmental controls' ~$3.0 billion, and Thermo Fisher's Analytical Instruments segment was $7.55 billion against the whole analytical sub-industry's $19.35 billion.[3][6] The middle is getting less shallow, though — the conglomerates are now splitting these niches back out (Section 8). Several category leaders in process, test, and metering remain private or foreign and cannot be bought publicly.[6]
• Semiconductors & components (3344) concentrates public liquidity in three of its six sub-industries — chips, contract assembly (EMS), and connectors (Amphenol at $23.1 billion of 2025 sales and TE Connectivity at $17.3 billion in fiscal 2025, each several times the entire domestic connector industry's ~$7 billion of shipments; Jabil, Flex, Celestica, Sanmina in EMS). Bare boards, passives, and other components are largely private, foreign-listed, or a lone U.S. pure-play each (TTM, Vishay, Rogers, CTS, Bel Fuse). A warning the child makes explicit: the codes leak — neither Amphenol nor TE is a pure connector company, and exposure should be measured from segment disclosure rather than a database industry label.[5]
• Communications (3342) puts its most investable names in the two smaller of its three markets: Cisco, Arista, Ciena and HPE (now the owner of Juniper) in wired networking, with small caps Calix, Adtran, Extreme, NETGEAR and Aviat below them; and, in signaling, three focused manufacturers — NAPCO (~$182 million), Daktronics (~$839 million), and Federal Signal's Safety & Security segment (~$343 million, under half that company) — beneath diversified Honeywell, Johnson Controls, and Siemens. In wireless, Motorola Solutions (~$11.7 billion of 2025 revenue) is the clearest single exposure and L3Harris carries defense communications (~$5.5 billion in 2024), but the biggest exposures are the hardest to buy: no U.S. company leads mass-market cellular base stations, where the top five suppliers hold roughly 94–96% and none is American, so the routes are foreign-listed (Ericsson, Nokia ADRs), chip-layer (Qualcomm), or private (SpaceX/Starlink).[3]
• Computers (3341) is the most public of the small children: Dell, HPE, Super Micro, IBM, HP Inc., the integrated Apple, storage names Seagate, Western Digital, Micron, SanDisk, NetApp, Pure Storage, Quantum, and a mature peripherals bench (Zebra, Logitech, NCR Voyix, Diebold Nixdorf, Xerox, Corsair) — several straddling more than one business in a single ticker. The listed pure-play ATM exposure is on its way out with Brink's pending purchase of NCR Atleos.[2]
• Audio-video (3343) is a short, mostly small-cap list — Sonos, GoPro, Turtle Beach, Universal Electronics, Koss — with Dolby as the highest-margin adjacent name, taking about 93% of FY24 revenue as royalties at roughly 89% gross margin. Diversified giants (Apple, Sony, Samsung) bury AV inside far larger businesses, and so does Gentex after its VOXX purchase, where premium audio is only about 6% of sales. Much value is locked in private or subsidiary brands (Bose, Shure, Fender, and the Samsung-owned Harman empire).[4]
• Media (3346) has no U.S.-listed pure-play at all; the data-tape duopoly sits inside foreign parents (Sony, Fujifilm), platters inside Japan's Resonac, and optical discs inside Taiwan's CMC Magnetics. A real share of the activity is captive — Seagate and Western Digital both state they make substantially all or a significant portion of their own recording media internally — so most public investors reach the same AI-archival tailwind through those adjacent drive-makers, or through Quantum (tape systems and LTO royalties) and Alliance Entertainment as a downstream read on physical-media sell-through, all in other codes.[7]

The through-line: there is no single "electronics hardware" stock and no ETF for NAICS 334. Passive investors get the subsector bundled across broad semiconductor, technology-hardware, and aerospace-and-defense funds — which also fold in the fabless chip designers (Nvidia-style) that capture the fattest margins but sit outside this manufacturing code. Named tickers, revenues, and scale live in each child primer, which is where an actual portfolio decision gets made.

---

5. How the money works

Every part of 334 is a manufacturing business, so the economics run on capacity utilization, yield, product mix, and the cycle — not utility rate base, real-estate occupancy, or asset-management fees. Three lessons carry across all six children:

• Operating leverage is the game. Plants carry heavy fixed costs, so filling them drops incremental volume straight to profit while running them half-empty in a downturn collapses margins — at contract assembler Kimball Electronics a 13% fiscal-2025 sales decline produced a 26% gross-profit decline.[5] This is most violent in the chip child, where Micron's revenue rose about 49% in fiscal 2025 with gross margin swinging to 39.8% from 22.4% while, in the same window, Microchip's fell about 42% on an inventory correction — two companies in one sub-industry, opposite directions.[5] It is gentlest in the diversified instruments child.[6]
• Value keeps migrating out of the metal box. Assembling a commodity computer, board, or speaker earns low-single-digit margins; durable profit lives in mix (an AI server loaded with graphics processing units sells for many times a plain one), proprietary platforms, silicon and software content, and — above all — the recurring annuity attached to the hardware. The strategic template is identical everywhere: convert a one-time equipment sale into subscriptions, consumables, calibration, service, and licensing. Recurring revenue is roughly 83% of sales at Thermo Fisher and 81% at Danaher; a security-alarm maker's cellular-monitoring line runs ~91% gross margin at 48% of revenue; Dolby collects a royalty on every device shipping its format; a tape library pulls steady cartridge purchases for years.[3][4][6][7]
• The same input shock can move margin between children rather than up or down. The revised computer child makes this concrete: memory-cost inflation is expanding storage margins (where the maker keeps the price increase), compressing computer margins (where the expensive part is a bought-in chip that passes straight through, so revenue rises while gross-margin percentage falls), and hitting peripherals as a pure cost with no demand offset.[2] Read parts of this subsector as a profit transfer, not a uniform tailwind.

Gross margin, in short, tracks how much intellectual property and software ride on the box, and the revised children now give the full ladder in one code: from 7.0%–10.1% at contract board assemblers (operating margins 3.1%–5.0%) and a historical 2.7% average net margin in bare boards, through ~27% for a commodity scoreboard and ~28% for satellite terminal product, to 35.2% at TE Connectivity, 42.0% at an optical-transport vendor, ~55.6% at a code-certified safety panel maker, 63.7% on Cisco's products, and ~89% where the product is a license rather than a device.[3][4][5][7] Instruments spans roughly 32% to 69% on its own, with R&D running 6–19% of sales.[6] Two cautions travel with those numbers: gross margin is not profit (Sonos posted a 43.7% gross margin alongside a $50.5 million operating loss; Nokia's Mobile Networks segment earned a 37% gross margin but only 2.8% operating, with R&D absorbing 27% of segment sales), and asset-light is not risk-light (Sonos ended fiscal 2025 with roughly $173 million of open finished-goods purchase orders, so a demand miss lands as inventory).[3][4] The leading indicators for the capital-goods parts are order backlog and book-to-bill (new orders versus shipments; above 1.0 means backlog is building) — Motorola Solutions carries roughly $14 billion of it — though the instruments child warns that backlog is not revenue where customers can cancel or reschedule without penalty.[3][6] For anything sold to carriers or cloud buyers, the customer capital-spending cycle rules, and concentration is disclosed and high: cloud customers are 88% of Western Digital's revenue, Jabil's top five are 36% (one at 16%), Sanmina's top ten 52%, and one optical vendor's top five 49.7%.[2][3][5]

---

6. Demand drivers

The subsector's demand is the union of its children's, and the drivers pull partly independently — which is exactly why 334 as a whole is steadier than its most cyclical parts. The dominant force cutting across the level is one wave:

• AI data-center buildout — the loudest driver, and it touches almost everything. Hyperscalers plan on the order of $600–725 billion of capital spending in 2026, up from roughly $410 billion in 2025, about half of it on servers and chips.[2] That single wave pulls the chip child (logic and memory, with industry capital spending itself around $166 billion in 2025 and projected near $200 billion in 2026), the components child (AI-server board layer counts up from 16–20 to 28–36, lifting board value per server by more than 30%; 40,000–60,000 ceramic capacitors in a next-generation rack; high-speed connectors, where TE's digital-data-networks sales rose from $1.3 billion to $2.2 billion in fiscal 2025), the computer child (AI servers and the data those models retain), the networking child (high-speed switching and optical), the instruments child (semiconductor test, where the category leader's revenue rose 18.8% in 2025 — the loudest driver within 3345), and even the media child (cheap archival tape for cold AI training data, at a record 176.5 exabytes of LTO capacity shipped in 2024, up 15.4%).[2][3][5][6][7]
• Defense, public safety, and the threat environment — about $900.6 billion for national defense in FY2026, of which roughly $295 billion is procurement plus research and development, the buckets that buy radar, guidance, tactical radios, and first-responder systems; the anchor demand where U.S. domestic production is actually protected.[3][6]
• Health care and demographics — devices, imaging, and radiation therapy driven by 61.2 million Americans aged 65 and over in 2024 (18.0% of the population) and an estimated 2.1 million new U.S. cancer cases in 2026; a defensive, non-cyclical base. Pharma and research funding (NIH's FY2025 appropriation was $48.5 billion) drive analytical labs.[6]
• Automotive electrification — a driver the parent previously lacked: EVs and driver-assistance multiply electronic content per vehicle, including roughly 3–5× the passive content (on the order of 15,000 ceramic capacitors versus about 3,000 in a legacy car), and car audio and infotainment feed the audio-video child.[4][5]
• Industrial and energy capital spending, plus utility infrastructure — process industries account for 55–60% of instrumentation demand, and North American water advanced-metering penetration was still only about one-third at end-2022 against an EPA-assessed $625 billion of 20-year drinking-water need. Newer: the nuclear revival tied to AI data-center power, with technology companies committing over $10 billion to nuclear partnerships, and the regulatory shift away from ethylene-oxide sterilization toward electron-beam and X-ray machines.[6]
• Wireless carrier cycles — 5G, its 5G-Advanced upgrade, and eventually 6G, plus the low-Earth-orbit satellite terminal boom; U.S. providers invested about $29 billion in networks in 2024 serving 579 million connections, though traffic growth does not translate one-for-one into equipment revenue.[3]
• Building codes, construction, and infrastructure subsidy — fire/life-safety mandates (NFPA 72, school-safety laws), traffic systems, and programs like the $42.45 billion Broadband Equity, Access, and Deployment (BEAD) fiber build — which, both communications children now caution, reaches equipment makers indirectly and late.[3]
• Consumer replacement and secular niches — PC refresh (2025 shipments grew on Windows 10's end of support, but Gartner forecasts a ~10% decline in 2026 as DRAM and SSD prices rise ~130%), TV and soundbar cycles mostly filled offshore, streaming's 106.5 million U.S. paid music subscriptions supporting networked speakers, the vinyl revival ($1.043 billion on 46.8 million units in 2025, up 9.3%, again outselling CDs at $312.4 million and 29.5 million units, down 7.8%), a collector-driven 4K Blu-ray counter-trend up 12%, and the streaming-driven decline of optical discs.[2][4][7]

The contrast is the point. A shock to hyperscaler budgets would hit the chip, component, computer, and networking children hard and near-simultaneously, while defense, health-care, metering, and building-safety demand would barely flinch — and inside the computer child, peripherals would feel the AI wave only as an input cost either way. The instruments child's internal diversity is the subsector's shock absorber.

---

7. Regulation

None of 334 is rate-regulated like a utility; the binding rules are trade, industrial subsidy, quality certification, and product/materials standards — and across the subsector regulation is as often a moat and a demand creator as a cost.

• Trade and national security is the heaviest and most cross-cutting lever: export controls on advanced AI chips and computers (Commerce Department Bureau of Industry and Security rules, plus ITAR/EAR on defense gear), which have whipsawed — after outright restrictions, BIS moved by late 2025 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, and to case-by-case review for specified computing products the same month. A 25% Section 232 duty on certain advanced semiconductors, also effective January 2026, plus Section 301 tariffs, raise the cost of anything assembled abroad — cutting both ways by pressuring margins while nudging reshoring, and not always in domestic producers' favor, since a U.S. fabricator can pay duty on full-value inputs while an imported finished board is dutied once.[2][3][5][6] Enforcement is not theoretical: BIS imposed a $300 million civil penalty on Seagate over hard-drive sales to Huawei, with audit obligations and a suspended denial order.[2]
• Industrial subsidy is concentrated on semiconductors: the 2022 CHIPS and Science Act put $52.7 billion (about $39 billion in incentives, with over $36 billion allocated across roughly 19 firms by late 2025, plus an investment tax credit of 25% on qualifying equipment placed in service through 2025 rising to 35% thereafter but scheduled to expire at the end of 2026) almost entirely behind chip fabrication. What the rest of the chain received is smaller by orders of magnitude — a $39.9 million Defense Production Act award for bare-board and substrate capacity is representative — and the analogous bare-board (PCBS) bill, which would add roughly $3 billion plus a 25% buyer credit, had not been enacted as of mid-2026.[5] Subsidy is also edging into ownership: in August 2025 the U.S. government converted part of Intel's support into a roughly 10% equity stake, a departure from grant-only support that equity holders now have to price.[5]
• Certification is the real gatekeeper that keeps high-value work onshore and locks out competitors: FDA device pathways and CMS reimbursement in health care — now including the FDA's Quality Management System Regulation, in effect since February 2, 2026, and section 524B cybersecurity requirements for "cyber devices"; NIST-traceable calibration in measurement; AS9100/ISO 13485/AEC-Q/MIL-PRF/IPC quality regimes and DFARS/CMMC flow-downs to defense subcontractors; FCC equipment authorization, the December 2025 "Covered List" ban on specified Chinese network gear and the roughly $5 billion "rip and replace" reimbursement program (only about 42% of projects finished by mid-2026); UL 864 and NFPA 72 — whose 2025 edition adds a dedicated cybersecurity chapter — and the MUTCD 11th Edition for traffic devices; and PCI/EMV/ADA/Section 508 rules on payment and accessibility hardware.[2][3][5][6]
• Domestic-content rules are tightening on a schedule. Build America, Buy America preferences attach to federally funded projects: Commerce's BEAD waiver requires certain manufacturing steps to happen domestically for covered equipment, while federal-aid highway projects obligated on or after October 1, 2025 require U.S. final assembly of covered manufactured products and those obligated on or after October 1, 2026 require U.S.-produced components exceeding 55% of component cost.[3]
• Materials and product rules are universal — RoHS/REACH restrictions, conflict-minerals disclosure, energy-efficiency and e-waste standards, and a rising wave of state right-to-repair laws — with critical-input dependence federally documented: the United States was 100% net-import reliant for both gallium and tantalum in 2025.[2][4][5]
• Fiscal policy now cuts both ways with dated cliffs in the instruments child: the 25C residential efficiency credit expired for equipment placed in service after December 31, 2025, the 179D commercial deduction terminates for construction beginning after June 30, 2026, and the Buy America waiver keeping federally funded advanced water-metering projects buildable runs only to December 19, 2027.[6]

---

8. Consolidation

The subsector is statistically unconcentrated (HHI 93.9), but it is restructuring steadily — and the revised children establish that there are now three engines at work, not two:

• Strategic M&A at the top. HPE's $13.6 billion purchase of Juniper Networks (closed July 2025, revised down from the parent's earlier ~$14 billion) and Nokia's ~$2.3 billion buy of Infinera in networking; Amphenol's serial connector acquisitions (including a ~$10.5 billion CommScope connectivity deal) and Asian roll-ups of Western passive brands (Yageo–KEMET at ~$1.6–1.8 billion, Yageo–Pulse at ~$740 million); in instruments the Waters–BD Biosciences combination (~$17.5 billion, closed February 9, 2026), Siemens–Varian (~$16 billion), Blackstone–Copeland (~$14 billion), J&J–Shockwave (~$13.1 billion), and Emerson–National Instruments (~$7.8 billion); Dell's earlier absorption of EMC, Xerox's Lexmark acquisition, and Brink's pending ~$6.6 billion purchase of NCR Atleos in computers and peripherals; and, at the small end, brand consolidation in audio (Samsung's Harman buying the Sound United portfolio for $350 million — against roughly $1 billion Masimo paid for the same brands in 2022 — and Gentex buying VOXX for ~$196 million) and optical media (CMC Magnetics buying Verbatim for $32 million).[2][3][4][5][6][7] It is not frictionless: DuPont's ~$5.2 billion agreement to buy Rogers Corporation was terminated after failing to clear Chinese antitrust review, with a $162.5 million break fee — cross-border review now shadows any large deal here.[5]
• Private-equity roll-ups of the fragmented tail. PE aggregates small certified board and assembly shops into regional platforms (Summit Interconnect's sale to Lindsay Goldberg; the APCT/Advanced Circuits combination into AdvancedPCB at roughly $220 million); takes signaling specialists private — Carrier's 2024 restructuring was a three-way breakup, sending its access business to Honeywell (~$5.0 billion), its commercial and residential fire business to Lone Star Funds ($3 billion, forming Kidde Global Solutions), and its industrial fire business to Sentinel Capital ($1.425 billion, forming Spectrum Safety Solutions), alongside Cubic (Veritas) and Iteris (Almaviva, ~$335 million); and buys sticky, code-protected, recurring-revenue safety, traffic, and audio assets (Bose Professional to Transom Capital).[3][4][5][6]
• De-conglomeration — the update this pass adds. The same companies that spent decades rolling niches up are now splitting them back out, and each break-up manufactures a new focused public company: Honeywell is separating into three, Fortive spun out Ralliant in June 2025, Danaher spun Veralto in 2023, GE spun GE HealthCare in 2023, Johnson Controls sold residential and light-commercial HVAC to Bosch for ~$8.1 billion, and Xylem sold its non-North-American metering arm; in computers, Hewlett-Packard's earlier split into HP Inc. and HPE, Western Digital's completed SanDisk separation in February 2025, and NCR's split into Voyix and Atleos follow the same logic.[2][6] The caveat is real — a newly focused pure-play concentrates the downside as well as the upside, and Ralliant took a $1.44 billion goodwill impairment in its first full year.[6]

The direction of travel is vertical and toward scale — into software, services, silicon, cooling, power, and supply — more than simple horizontal merger. For public investors the net effect is now two-sided: the listed opportunity set keeps narrowing in several children (signaling, audio, media, bare boards), which raises the scarcity value of the few remaining pure-plays, while widening in instruments and storage as conglomerates surface pure businesses. Capital intensity and certification are the moats, so scale keeps concentrating even as the headline statistics stay fragmented.

---

9. Risks

Because so much of the level leans on one demand stack, the biggest risks are shared and correlated:

• Cyclicality and AI-capex digestion. The chip, component, computer, and networking children ride customer capital budgets, and a large share of current strength is hyperscaler AI spending funded by a handful of increasingly debt-financed buyers — a pause would hit much of the subsector at once. The recent template is on the record: the 2023-24 inventory correction, global radio-access-network sales falling roughly 22% from ~$45 billion in 2022 to ~$35 billion in 2024, and Microchip's ~42% revenue decline in the same window Micron's rose ~49%.[2][3][5]
• Memory-cost inflation as an internal transfer. DRAM and SSD prices are projected up ~130% by the end of 2026 with a forecast ~10% drop in PC shipments — a pricing-power windfall for drive makers and a direct cost shock to computer and peripheral makers within the same child.[2]
• The offshore cost wall and permanent value migration. For commodity tiers (bare boards, passives, consumer audio, low-end assembly) the structural cost gap versus Asia is existential without policy support — U.S. bare-board employment fell 80.6% between 2000 and 2024, from 137,501 to 26,676, which also means the labor pool for any reshoring push is thinner than policy assumes. The highest-value activity — chip design, brand, software — is booked in other codes, permanently capping domestic manufacturing economics.[4][5]
• Reshoring may be nearshoring. Trade policy is pushing production out of China, but the larger flow is to Mexico under the USMCA rather than onto U.S. soil: Flex produced $6.9 billion in Mexico against $4.3 billion in China in fiscal 2025. Within computers, the three six-digit businesses are moving in opposite directions on domestic assembly — AI-server assembly is being pulled onshore while self-checkout and point-of-sale hardware manufacturing was outsourced away and no new drive plants are being built anywhere. Those location decisions, more than the demand cycle, are what will move this subsector's measured size.[2][5]
• Geopolitics, China, and tariff/export-control whiplash. Leading-edge chipmaking concentrated in Taiwan is a flashpoint; China is simultaneously a major buyer, a major input source, and — via antitrust review — a gate on Western consolidation; and tariffs are now a quantified margin item rather than a hypothetical, with GE HealthCare estimating tariffs cut its 2025 operating income by approximately $245 million and cash flow by approximately $285 million, Gentex reporting a ~110-basis-point hit to consolidated gross margin, and GoPro's Thailand- and Malaysia-made cameras moving from a 10% to a 19% U.S. tariff in August 2025 — a reminder that diversifying out of China moves the exposure rather than eliminating it.[4][5][6]
• Customer and supplier concentration. A few cloud hyperscalers, carriers, and government buyers drive demand (cloud is 88% of Western Digital's revenue; Jabil's top five customers 36%; Sanmina's top ten 52%; one optical vendor's top five 49.7%), while Taiwan-centered chips and Southeast-Asian assembly drive supply, with documented 100% U.S. net-import reliance for gallium and tantalum and hard-to-substitute pinch points in helium, iodine, cobalt-60, and sole-source detector materials.[2][3][5][6]
• Policy dependence. Much of the reshoring case rests on subsidies that are partly expiring (the CHIPS tax credit at end-2026) or unlegislated (the bare-board bill), on reimbursement and budget decisions across distinct public payers (the Pentagon, CMS, utility rate cases, research funding) with several dated 2026–2027 cliffs, on a rip-and-replace program only ~42% complete, and now on a novel governance question created by the U.S. government taking equity in a subsidy recipient.[3][5][6]
• Technology substitution and secular decline. Flash could erode hard drives; Open RAN and merchant silicon can obsolete a product line; office print, cash usage, and optical discs are structurally falling — media's sectoral output is $1.458 billion in 2024 against $9.984 billion in 1998 — and even inside instruments, environmental-controls employment fell 63.5% between 2000 and 2024.[2][3][6][7]

The offset is diversification: defensive instruments, health care, metering, and building-safety demand cushion the cyclical chip-and-computer complex — which is why the aggregate is steadier than any single child.

---

10. How to invest & outlook

How to invest — pick the child first, because the subsector is not one trade. There is no ETF or single stock for NAICS 334; you assemble exposure, and the practical route depends on which of the six markets you want.

• Public-market investors get the deepest, cleanest choices in instruments and components: aerospace-and-defense funds and medical-device/imaging large-caps for 3345, plus the diversified serial acquirers (Danaher, Fortive, Roper, AMETEK, Emerson, Honeywell, Thermo Fisher, GE HealthCare) and the growing set of spun-out pure-plays; broad semiconductor ETFs, connector leaders Amphenol and TE Connectivity, and EMS contractors (Jabil, Flex, Celestica, Sanmina) for 3344; Cisco, Arista, Ciena, HPE, and Motorola Solutions — plus NAPCO, Daktronics, and Federal Signal in signaling — for 3342; Dell, HPE, Super Micro, IBM, Apple, Seagate, Western Digital, Micron, and the peripherals bench (Zebra, Logitech, HP Inc.) for 3341; and small caps Sonos, GoPro, and Turtle Beach with licensing name Dolby for 3343. The biggest wireless and media exposures require going foreign-listed (Ericsson/Nokia ADRs; Sony/Fujifilm) — and broad exposure comes only through diversified technology-hardware, semiconductor, or defense funds that also fold in the fabless chip designers sitting outside this code. Verify what you are buying from segment disclosure rather than an industry label: the codes leak in every child.[2][3][4][5][6][7]
• Private-market investors reach the parts public markets can't: SpaceX/Starlink satellite terminals; private-5G, Open-RAN, and specialty-component startups; private-equity roll-ups of certified board, assembly, calibration, and test-lab businesses; the U.S. contract plants being stood up for domestic-content rules; venture bets on next-generation instruments (which mostly exit by strategic acquisition rather than IPO, so timing tracks the majors' M&A appetite); the private premium-audio brands (Bose, Shure, Fender); vinyl pressing — the most accessible live opportunity in media, since data-tape manufacturing is effectively closed to new entrants by duopoly, patents, and capital; and the code-protected, recurring-revenue safety and traffic assets that private equity keeps rolling up.[3][4][5][6][7]

A caution on comparing numbers across this tree, carried up from the children because it applies to every figure above: Census statistics measure domestic establishments by primary production activity; company filings consolidate worldwide operations, most of them offshore; analyst share estimates measure global units or capacity; and BLS and AIES series measure yet other universes. Those are different populations and should never be combined into a single "market size."[2][3][4][5][6][7]

Outlook. The tailwinds are unusually aligned: an AI capital-spending wave lifting demand and per-unit content across most of the subsector, automotive electrification raising electronic content per vehicle, defense budgets steering dollars to trusted domestic suppliers, health-care and metering demand providing a defensive floor, and trade policy nudging production out of China. The counterweights are equally shared: deep cyclicality, a structural cost gap versus Asia in the commodity tiers, a thinned-out domestic skills base, dependence on subsidy that is partly expiring or unpassed, and the fact that much current strength is AI capex that could normalize. Two through-lines hold at every level of this tree: value keeps migrating out of the metal box and into chips, software, services, and recurring revenue, and the plausible path for U.S. production is a smaller-but-higher-value footprint concentrated in defense, AI, health care, and specialty work rather than a return to commodity-scale dominance — with the honest caveat that the production genuinely leaving China is going to Mexico at least as often as to the United States. For the full company-by-company map, per-child economics, and detailed risk and regulatory treatment, read the six child primers: 3341, 3342, 3343, 3344, 3345, and 3346.

---

Sources

Level figures (§3) are our ground-truth federal reference statistics for NAICS 334 [1]. All other figures and company detail are synthesized from the six child primers [2]–[7] and the sources they cite.

1. U.S. Census Bureau, 2022 Economic Census — Industry Statistics & Concentration, and 2023 County Business Patterns, NAICS 334 (receipts $352.3B; 9,861 firms; CR4 15.2% / CR8 22.1% / CR20 34.7% / CR50 49.8%; HHI 93.9; 11,262 establishments; 815,444 employees; $85.8B annual payroll; $22.7B Q1 payroll). Histometrics ingested federal statistics. https://www.census.gov/programs-surveys/economic-census.html | https://www.census.gov/programs-surveys/cbp.html
2. Histometrics primer, 3341 Computer and Peripheral Equipment Manufacturing (a one-to-one rollup of 33411) — and its cited sources (six-digit split: computers ~59% / storage ~24% / peripherals ~17%, computers CR4 60.5%, storage CR4 83.6%; Apple FY2025 Mac revenue ~$33.7B; Dell, HPE, HP, Super Micro, IBM filings; Seagate $9.097B and Western Digital $9.520B, cloud 88% of WDC revenue; three HDD makers >95%; Brink's–NCR Atleos ~$6.6B and the Xerox–Lexmark deal; Gartner PC and memory-price forecasts (DRAM/SSD ~+130%, PC shipments ~−10%); hyperscaler capex $600–725B in 2026 vs ~$410B in 2025; Nvidia ~90% of AI-accelerator spending; BIS $300M Seagate penalty and January 2026 licensing review; SBA small-business shares 86.5% / 96.5%; Census EC 2022 / CBP 2023 for 3341: receipts $19.84B, 557 establishments, 31,508 employees, $3.84B payroll, 384 firms, CR4 43.9%, HHI 706.8).
3. Histometrics primer, 3342 Communications Equipment Manufacturing — and its cited sources (three children: 33421 ~10% / 33422 ~78% / 33429 ~11% of revenue; Cisco FY2025 10-K, ~$28.3B networking revenue and 63.7% product gross margin; Motorola Solutions 2025 10-K, ~$11.7B revenue, ~52% gross / ~26% operating, ~$14B backlog; Ciena 42.0% gross margin and top-five customers 49.7%; Nokia Mobile Networks 37% gross / 2.8% operating; NAPCO ~$182M with $86.3M recurring at ~91% gross margin; Daktronics ~$839M at 27.3% gross / 7.3% operating; Federal Signal Safety & Security ~$343M; Honeywell Building Automation ~$7.4B; Dell'Oro/Omdia RAN data (top five 94–96%; $45B 2022 → $35B 2024); Qualcomm FY2024 ~$38.9B; L3Harris communications ~$5.5B; U.S. telephone imports ~$115B in 2024; networking employment 104,129 in 2000 → 13,506 in 2024; HPE–Juniper $13.6B and Nokia–Infinera ~$2.3B; the Carrier three-way breakup, Cubic and Iteris take-privates; FCC Covered List (December 2025) and ~$5B rip-and-replace at ~42% complete; BEAD $42.45B; Buy America highway thresholds; NFPA 72 (2025), UL 864, MUTCD 11th Edition; Census EC 2022 / CBP 2023 for 3342: receipts ~$41.3B, 1,138 establishments, 73,884 employees, ~$9.0B payroll, 1,062 firms, CR4 40.8%, HHI 506.4).
4. Histometrics primer, 3343 Audio and Video Equipment Manufacturing (a pass-through to 33431) — and its cited sources (BLS productivity: fastest labor-productivity growth of manufacturing/mining industries examined, 11.2%/yr 2019–2024, output +8.6%/yr, hours −2.4%/yr; BLS sectoral output $4.39B and employment ~19,700 in 2025 vs 60,300 in 1987; Sonos FY2025 revenue $1,443.4M, 43.7% gross margin, $50.5M operating loss, ~$173M open purchase orders; GoPro FY2025 revenue $651.5M, 31% in Q4, Thailand/Malaysia tariffs 10%→19%; Dolby FY24 licensing 93% of revenue at ~89% gross margin; Gentex–VOXX ~$196M with premium audio ~6% of sales and ~110bps of tariff margin impact; Harman–Sound United $350M vs Masimo's ~$1B in 2022, and Samsung–Harman ~$8.02B in 2017; RIAA 2025 streaming data; Census EC 2022 / CBP 2023 for 3343: receipts ~$5.09B, 539 establishments, 10,977 employees, ~$861M payroll, 506 firms, CR4 43.6%, HHI 682.8).
5. Histometrics primer, 3344 Semiconductor and Other Electronic Component Manufacturing (a pass-through to 33441) — and its cited sources (six sub-industries: semiconductors ~54% down to passives ~3%, with CR4s of 48% / 43.6% / 39.6% / 28.7% / 24.1% / ~11%; SIA global chip sales $791.7B in 2025, +25.6%, and U.S.-HQ firms ~$318B / 50.4% of 2024 global revenue; TSMC ~72% of foundry revenue; ~90% of bare boards imported, U.S. ~4% of global output; BLS bare-board employment 137,501 → 26,676, −80.6%; EMS gross margins 7.0–10.1% and operating 3.1–5.0%; Kimball −13% sales / −26% gross profit; Micron +49% with gross margin 39.8% vs 22.4%, Microchip −42%; Amphenol $23.1B and 25.4% operating margin, TE Connectivity $17.3B at 35.2% gross / 18.6% operating; connector market $86.5B worldwide / $20.1B North America; Flex $6.9B Mexico vs $4.3B China; Jabil and Sanmina customer concentration; AI-server PCB layer counts and MLCC content; CHIPS Act $52.7B and the 25%→35% credit expiring end-2026; the PCBS Act and the $39.9M DPA award; the ~10% U.S. equity stake in Intel; BIS case-by-case licensing with 25% revenue share from January 2026; USGS 100% import reliance for gallium and tantalum; DuPont–Rogers termination and $162.5M break fee; SBA standards 550–1,250 employees; Census EC 2022 / CBP 2023 for 3344: receipts $115.9B, 3,686 establishments, 279,841 employees, $27.8B payroll, 3,376 firms, CR4 29.3%, HHI 317.5).
6. Histometrics primer, 3345 Navigational, Measuring, Electromedical, and Control Instruments Manufacturing (a pass-through to 33451) — and its cited sources (nine 6-digit sub-industries from $53.5B defense electronics to ~$3.0B environmental controls; gross margins ~32–69% and R&D 6–19% of sales; recurring revenue ~83% at Thermo Fisher and ~81% at Danaher; GE HealthCare 2025 product $13.7B vs service $7.0B, service RPO $10.7B vs product $5.0B, Imaging ~$9.2B, and ~$245M operating-income / ~$285M cash-flow tariff impact; Honeywell Building Automation $7.367B, Thermo Fisher Analytical Instruments $7.55B, Keysight $5.4B; electromedical AIES 2023 receipts $45.5B vs ~$26.3B U.S. market and HS 9018 trade data; FY2026 national defense ~$900.6B with ~$295B procurement + RDT&E; 61.2M Americans 65+; NIH FY2025 $48.5B; semiconductor-test leader +18.8%; water AMI penetration and EPA's $625B need; >$10B of tech-company nuclear commitments; FDA QMSR (February 2, 2026) and §524B, DFARS/CMMC, IEC 62443; the 25C, 179D, and BABA cliffs; the acquire-and-split record — Waters–BD ~$17.5B, Siemens–Varian ~$16B, Blackstone–Copeland ~$14B, J&J–Shockwave ~$13.1B, Emerson–NI ~$7.8B, Honeywell's three-way separation, Fortive–Ralliant and its $1.44B impairment, Danaher–Veralto, GE–GE HealthCare, JCI–Bosch ~$8.1B; environmental-controls employment −63.5% 2000–2024; SBA standards 600–1,350 employees; Census EC 2022 / CBP 2023 for 3345: receipts $168.6B, 5,050 establishments, 414,694 employees, $44.0B payroll, 4,427 firms, CR4 19.0%, HHI 163.2).
7. Histometrics primer, 3346 Manufacturing and Reproducing Magnetic and Optical Media (a pass-through to 33461) — and its cited sources (Sony/Fujifilm LTO tape duopoly and the 2016–2019 ITC dispute; CMC Magnetics–Verbatim $32M; Resonac platters; Seagate and Western Digital captive media manufacturing; Quantum LTO royalty revenue −9% in fiscal 2025; Alliance Entertainment FY2025 mix (vinyl ~32%); LTO 176.5 EB shipped in 2024, +15.4%; RIAA 2025 vinyl $1.043B / 46.8M units and CD $312.4M / 29.5M units; 4K Blu-ray +12%; tape-storage market ~$6.8B in 2025 to ~$12.4B by 2034; BLS sectoral output $1.458B in 2024 vs $9.984B in 1998 and employment 10,500 in 2025 vs 65,900 in 2000; SBA 1,250-employee standard with 99.1% of firms small; Census EC 2022 / CBP 2023 for 3346: receipts $1.61B, 292 establishments, 4,540 employees, $320.5M payroll, 271 firms, CR4 54.6% / CR8 65.1% / CR20 79.1% / CR50 91.3%, HHI suppressed).