Computer and Peripheral Equipment Manufacturing (U.S.)

NAICS 2022 code 33411 — the five-digit industry that groups the three businesses that build finished computing hardware on American soil: the computers themselves, the drives that store their data, and the terminals and peripherals around them.

("NAICS" is the North American Industry Classification System, the standard code framework U.S. statistical agencies use to define industries. A five-digit code like 33411 is a rollup that contains more detailed six-digit industries — here, three of them.)

---

1. Overview

NAICS 33411 is where the physical machinery of computing is manufactured. It bundles three neighboring industries under one code:

• 334111 — Electronic Computer Manufacturing: the finished computers, from laptops to artificial-intelligence (AI) data-center servers to mainframes [4].
• 334112 — Computer Storage Device Manufacturing: the hard disk drives (HDDs), solid-state drives (SSDs), and tape units that store data [5].
• 334118 — Computer Terminal and Other Computer Peripheral Equipment Manufacturing: the printers, monitors, keyboards, mice, scanners, point-of-sale (POS) terminals, and automated teller machines (ATMs) around the computer [6].

Why an investor should care: this is the hardware layer directly beneath "the cloud" and "AI," and two of its three pieces are riding the largest capital-spending wave in technology's history. But the level carries a defining quirk that everything else in this primer builds on — the federal statistics count only what is physically assembled inside the United States, which is a small fraction of what U.S.-based companies actually sell, because the marquee brands (Apple, Dell, HP, Seagate, Zebra) design here and build abroad. Read the numbers in Section 3 as a measure of the domestic factory footprint, not of the American computer-hardware business, which is an order of magnitude larger.

The distinctive value of looking at 33411 as a whole is the contrast across its three children — they differ sharply in size, growth direction, ownership, and how the money is made. The newest evidence sharpens that contrast in an unexpected way: the same AI wave that is lifting revenue across the level is moving margins in opposite directions inside it — expanding them in storage, compressing them in computers, and arriving in peripherals as a pure cost increase with no demand offset [4][5][6]. Section 2 lays the contrast out first.

---

2. What's inside — the three child industries and how they differ

The three children are roughly the same vintage of "computing hardware," but they behave like three different businesses. The table compares them on the things that matter to an investor: how big each is within the level, which way it is heading, which way its margins are moving, who owns it, and how you would buy exposure.

(CR4 = the combined revenue share of the four largest firms — a standard market-concentration measure; CR8 and CR50 extend it to the top eight and top fifty. Tickers are defined in Section 4; ODM = original-design manufacturer, a contract factory that builds hardware to a brand's specification.)

The four ways to read that table:

1. Value is dense in computers and storage; bodies are dense in peripherals. Computers and storage together are ~83% of the level's receipts but only ~49% of its factory sites, and they pay the most per worker (roughly $138K and $129K average annual payroll, versus ~$99K in peripherals) [1]. Storage is the extreme case: ~24% of receipts on just ~11% of sites — a capital-intensive business run out of a handful of very large plants. Peripherals is the mirror image: ~51% of the sites but ~17% of the receipts, a fragmented trade of small design-and-assembly shops averaging about 39 employees per establishment [6].

2. Two children ride the AI wave as demand; the third meets it as a cost. Computers (AI servers) and storage (AI data-lake demand) are both pulled by hyperscaler capital spending [4][5]. Peripherals gets no demand lift — its buyers are retail, banking, logistics, and office refresh cycles — but it does get the AI wave's bill: Logitech reports that AI and data-center construction has raised demand for memory components and encouraged suppliers to redirect capacity away from it [6]. If you buy 33411 as a theme, know that one of its three children — about a sixth of measured receipts but half its factory sites — is not an AI trade at all.

3. The AI wave splits the level's margins. This is the newest and least obvious feature of the family. In storage, a shortage means the maker keeps the price increase: Seagate's operating income rose to $1.890 billion from $452 million and Western Digital's swung to $2.334 billion from a $403 million loss [5]. In computers, the expensive part is a bought-in chip that passes straight through, so revenue dollars explode while gross-margin percentage falls [4]. Memory inflation is, in effect, a transfer of profit from the computer and peripheral children to the storage child.

4. Each child is a concentrated oligopoly, but they are concentrated around different firms — which is why, as Section 3 shows, the combined level looks less concentrated than any of its parts.

A note on ownership contrast, because the audience spans public and private investors: 334112 storage is almost entirely large public multinationals — just 38 U.S. firms, with essentially no small-operator segment; 334111 computers is a handful of public brands sitting atop mostly foreign, often-private contract manufacturers; and 334118 peripherals has by far the richest tail of small private and private-equity-owned firms (kiosk, receipt-printer, rugged-terminal, and payment-hardware specialists such as private-equity-held Verifone). A useful corrective to the receipts-weighted view: by the Small Business Administration's own thresholds, 86.5% of the computer child's firms and 96.5% of the peripheral child's firms are small businesses [3][4][6]. This is a level of a few giants and several hundred small shops, not a level of giants alone.

---

3. How big it is (and why the number looks small)

Our ground-truth federal figures for the level, NAICS 33411:

The children add up to the level — a good sign the data is clean. Establishments (212 + 61 + 284 = 557), employment (13,144 + 7,221 + 11,143 = 31,508), and payroll ($1.81B + $0.93B + $1.10B = $3.84B) all tie out, and receipts ($11.78B + $4.72B + $3.34B = $19.84B) match to the dollar [1][2][4][5][6]. The one small gap is firm count: the children sum to 387 but the level reports 384, because a firm that operates in more than one of the three industries is counted once at the level but can appear in each child — so firm counts are not perfectly additive.

The $19.84 billion is also a 2022 vintage, and it is a floor. Two of the three children now carry a later federal reading of their own domestic activity that runs above their Economic Census receipts: the computer child shows $14.07 billion of sales/shipments in the Census Annual Integrated Economic Survey for 2023 (against $11.78 billion of 2022 receipts) [4], and the peripheral child shows $4.2 billion of shipments in the Census M3 series for 2025, up from $3.7 billion in 2024 (against $3.34 billion of 2022 receipts) — though that M3 series may not be exactly coterminous with NAICS 334118 [6]. The storage child has no comparable later series. These are different surveys with different scopes and years and should not be added together into a new level total; read them as evidence that the domestic footprint has grown since the 2022 census benchmark, not as a replacement for it.

Now the caveat that dominates this level: the $19.84 billion is small on purpose. Federal manufacturing statistics count only production physically located in the United States. The U.S.-based firms that dominate global computing build offshore, and that revenue is booked under wholesale, retail, or corporate categories rather than domestic manufacturing. The scale of the undercount is hard to overstate: Apple's Mac line alone brought in about $33.7 billion in fiscal 2025 — roughly 1.7 times the entire level's measured receipts [4][7]. Seagate ($9.097 billion) and Western Digital's post-separation hard-drive business ($9.520 billion) each booked close to half the level again in a single fiscal year, almost none of it on U.S. soil [5]. Zebra alone reported about $5.4 billion of global revenue in 2025, more than the whole peripheral child's domestic shipments [6]. U.S. imports of computer hardware and semiconductors topped $450 billion in 2025 [4]. Treat the federal figure as "what is built here," and expect the true U.S. computer-hardware business to be roughly an order of magnitude larger once offshore production and imports are counted.

The undercount is not evenly spread, and it hides ownership differently in each child. In computers and storage, the missing mass is offshore factories owned by large public companies — so public investors can still reach it through the parent stocks. The storage child makes this starkest: Western Digital employs roughly 40,000 people with 88% of them in Asia-Pacific, and about 25,000 of Seagate's 30,000 are in Asia [5]. In peripherals, alongside the offshore effect, there is a genuine long tail of small private domestic firms (201 firms across 284 establishments averaging ~39 employees) whose economic weight is thin in receipts but real in count [6] — the place where individual and small-business ownership is most under-represented by a receipts-weighted lens.

A distinctive rollup insight — the level looks less concentrated than its parts. The level's HHI is 706.8 (the HHI sums the squared market shares of all firms; U.S. antitrust agencies generally treat anything below ~1,500 as "unconcentrated"), and its CR4 is only 43.9%. Yet each child is an oligopoly: computers CR4 60.5%, storage CR4 83.6%, and even the fragmented peripheral child reaches ~89% by its top 50 [4][5][6]. How can the whole be less concentrated than its pieces? Because the three children are dominated by different companies — Apple/Dell/HP lead computers, Western Digital/Seagate/Toshiba lead storage, Zebra/NCR/HP lead peripherals — so pooling them dilutes any single firm's share. Note the shape of the curves: the level's top-50 share (89.2%) is essentially the peripheral child's, but every child's concentration curve is steeper at the very top than the level's. (Notably, the level's HHI is published, whereas the Census suppresses the HHI for all three children individually [4][5][6]; the rollup is the only place we get an official concentration index for this family.)

---

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

There is no single "computer-hardware manufacturing" stock, and no U.S.-listed fund dedicated to NAICS 33411. Investors buy the level by choosing among its children, and value pools in different places in each. Tickers and figures below are global company-level numbers — only a slice of each is U.S. manufacturing.

A structural feature worth flagging first: several large public names span multiple children of this level at once. HP (HPQ) makes both PCs (Personal Systems $38.5B) and printers/peripherals (Printing $16.7B) [4][6]. Dell (DELL) makes both computers and storage — and is simultaneously the largest AI-server vendor by revenue growth and the largest external enterprise-storage vendor at 22.7% share [4][5]. HPE and IBM straddle computers and storage too. Buying one of these is buying a slice of the whole 33411 stack, not a single child.

334111 Computers — the AI-server cycle is the prize (largest child, ~59% of the level).

334112 Storage — a rare pricing-power window (~24% of the level).

334118 Peripherals — value is in recurring revenue, not the box (smallest child, ~17%).

Private-market exposure differs by child. Computers: mostly reachable only through the foreign ODM parents or private data-center-infrastructure funds building the new U.S. AI-server plants [4]. Storage: nearly closed — Toshiba delisted in December 2023 and Kioxia is Bain-consortium-controlled after its December 2024 Tokyo listing; private money goes into controller-chip and storage-software startups, not drive factories, and the incumbents' patent, head/media, and qualification advantages make a greenfield entrant improbable [5]. Peripherals: the most open private field — private equity is the natural owner of commoditized, cash-generative hardware (the Verifone template), with a deep bench of privately held kiosk, receipt-printer, and payment-terminal specialists [6].

One adjacency that shapes everything but sits outside this level: Nvidia (NVDA). It is classified as a semiconductor maker, not a computer maker, so it is not in 33411 — but it captures an estimated ~90% of AI-accelerator spending and increasingly sells rack-scale systems, setting the economics for the computer and storage children alike [4].

---

5. How the money works

All three children are manufacturing businesses, so the economics run on capacity, input costs, product mix, and cyclicality — not on utility rate base, real-estate occupancy, or asset-management fees. But the profit engine sits in a different place in each, and that is the level's central lesson: the metal box is rarely where the money is.

• Computers (334111): mix and scale, plus the rare proprietary platform. Assembling a PC or commodity server earns low-single-digit margins — Dell's Client Solutions ran a 5.6% operating margin on $51.0 billion, HP's Personal Systems 5.3% on $38.5 billion — so owners win on volume buying, high factory utilization, fast inventory turns, and thin working capital [4]. The big recent lever is mix: an AI server loaded with expensive graphics processing units (GPUs) can sell for 10–20× a plain server, so revenue dollars explode even as gross-margin percentage falls (the pricey GPU is a pass-through). Dell's consolidated gross margin fell to 20.0% principally on that mix shift [4]. Infrastructure still beats PCs — Dell's Infrastructure segment ran 11.7% — and the true exceptions are proprietary or integrated: IBM's mainframe (segment margins above 20%) and Apple's Mac (chip + operating system + brand) [4].

• Storage (334112): cost per terabyte, exabytes shipped, and capital discipline. Profit is the spread between what a terabyte (TB) costs to build and what it sells for, driven by packing more bits onto each platter or chip — heat-assisted magnetic recording (HAMR), which enables 30–50+ TB drives, is the current weapon [5]. The right unit of account is exabytes, not units: Western Digital shipped 696 exabytes in fiscal 2025 against 443 the year before, with average selling price per unit up 29% on capacity mix [5]. Heavy fixed costs mean margins swing hard with utilization; with only three hard-drive makers left, prices firm when the three hold capacity instead of flooding the market. That discipline drove a ~46% HDD contract-price jump between September 2025 and January 2026 and Seagate's move to 35% gross margin from 23% [5].

• Peripherals (334118): the installed base and what attaches to it. Because assembly is outsourced, factory metrics matter less than input costs and, above all, recurring revenue. The textbook case is printing's razor-and-blades model — HP's Printing segment turned $16.7 billion of revenue into $3.1 billion of operating earnings (18.7%), with supplies alone contributing $10.9 billion [6]. The modern version is software and services bolted onto lumpy hardware: NCR Atleos books 70.6% of revenue as recurring at 19.1% adjusted EBITDA margins [6]. Investors track average selling price, the attach rate of consumables/software/services, and refresh-cycle length. The market rewards the share of revenue that is recurring, not the hardware line.

One economic feature the level shares end to end is customer concentration on the buy side, and it is now quantified: cloud customers account for 88% of Western Digital's revenue, its top ten customers 68%, with three individual customers at 17%, 12%, and 10%; Seagate books $7.3 billion of its $9.1 billion through OEMs with one customer at roughly 10% [5]. The same handful of hyperscalers sets the order book for the computer child. The unifying thread across all three: the box is a low-margin, cyclical commodity, and durable profit lives in mix, scale, proprietary platforms, and attached software, services, and consumables.

---

6. What drives demand

• AI data-center build-out (dominant for two of three children). Hyperscalers plan on the order of $600–725 billion of capital spending in 2026, up from ~$410 billion in 2025, roughly half of it on servers and chips, with cumulative 2025–2027 capex estimated near $1.15 trillion; the generative-AI server market alone was put at ~$104 billion in 2025 [4]. That single wave lifts the computer child (AI servers) and the storage child, where nearline data-center demand is growing roughly 25% or more a year [4][5]. It supplies no demand to peripherals.
• PC and enterprise refresh cycles. Multi-year replacement waves set the baseline for the computer child — 2025 PC shipments grew ~9% to more than 270 million units, helped by Windows 10's end of support on October 14, 2025, with "AI PCs" reaching ~31% of shipments (~78 million units) — and for consumer peripherals [4][6]. Some of that 2025 strength was demand pulled forward rather than a new growth regime, and Gartner forecasts a ~10% shipment decline in 2026 as DRAM and SSD prices rise ~130% [4].
• The data explosion, independent of AI. Video, backups, and regulatory archives grow relentlessly, underpinning storage even without the AI surge. The likely outcome is tiering rather than a single winner — flash for performance, hard drives for bulk (enterprise SSDs still cost roughly 16× per TB), and tape for archive, where 176.5 exabytes of compressed capacity shipped in 2024, up 15.4% and the highest on record [5].
• Retail, banking, and logistics capital budgets (the peripherals clock). Self-checkout, ATM/branch transformation, and warehouse/parcel automation drive the peripherals child on an enterprise-refresh cycle that is largely disconnected from AI [6].
• Government and defense procurement. A policy-sensitive slice of demand across all three, and an incentive for U.S.-based assembly.

The contrast is the point: computers and storage share one demand engine (AI capex); peripherals runs on a separate, more mature one and feels AI only through input costs. A shock to hyperscaler spending would hit the ~83% of the level's receipts that sit in its two AI-exposed children hard, and the last child barely at all.

---

7. Regulation

None of the three children is a heavily licensed industry, but each meets policy at a different edge:

• Trade and national-security policy (heaviest on computers and storage). The U.S. Bureau of Industry and Security (BIS) restricts exports of advanced AI chips and computers, chiefly aimed at China; after tightening through 2025, BIS moved in January 2026 to case-by-case review for specified products including Nvidia H200- and AMD MI325X-class chips [4]. A 25% Section 232 (national-security) tariff on certain advanced semiconductors — with carve-outs including data-center use — flows straight through to finished-system prices, because an AI computer is mostly chip cost [4]. Because storage assembly is concentrated in Thailand, Malaysia, and China, the same trade regime reshapes its supply chain [5].
• Enforcement is not theoretical. BIS imposed a $300 million civil penalty on Seagate over 429 transactions involving 7,420,496 hard drives valued at $1.1 billion sold to Huawei, together with audit obligations and a suspended denial order — the level's clearest illustration that export-control exposure is a live financial risk, not a compliance formality [5].
• Industrial subsidy — upstream, not here. The CHIPS and Science Act subsidizes domestic semiconductor fabrication (Micron secured up to $6.1 billion and has announced roughly $150 billion of planned U.S. memory-manufacturing investment), not final computer, drive, or peripheral assembly — but it reshapes the chip supply chain all three depend on [5].
• Payment, accessibility, and product regimes (heaviest on peripherals). POS terminals and ATM PIN pads must meet Payment Card Industry (PCI) point-of-interaction security and EMV (Europay/Mastercard/Visa) chip standards; ATMs must meet Americans with Disabilities Act (ADA) requirements, and federal purchases of kiosks and transaction machines are governed by Section 508 accessibility standards; monitors and printers face FCC emissions, UL safety, ENERGY STAR, and RoHS/WEEE e-waste rules (CRT displays are hazardous waste under RCRA), plus a growing wave of state right-to-repair laws [6].
• Environmental compliance across the level. Electronics manufacturing facilities meeting relevant thresholds fall under EPA greenhouse-gas reporting, and end-of-life hardware creates metals and e-waste exposure in all three children; drives add rare-earth magnet and helium handling [4][5].
• Government procurement. Buy-American provisions and the Trade Agreements Act govern which hardware federal agencies may buy across all three children, adding an incentive for U.S.-based assembly.

The net effect: a maker's addressable market, cost structure, and factory location are now partly set in Washington — most acutely for the computer and storage children, where trade and export policy is now a first-order variable.

---

8. Consolidation

All three children are consolidation stories, but of different kinds:

• Storage (334112) is the textbook case of reduction, then focus. Dozens of hard-drive makers in the 1990s collapsed into three (Western Digital ~47%, Seagate ~42%, Toshiba ~11% of capacity shipped), which now ship more than 95% of drives; NAND flash is nearly as tight among five suppliers [5]. The 2025 twist was separation: Western Digital completed its split from SanDisk on February 21, 2025, leaving each a pure-play on the bet that disciplined, focused firms hold pricing power — a bet that has so far paid [5].
• Computers (334111) is consolidation by restructuring and reshoring. Hewlett-Packard split into HP Inc. and HPE (2015); Dell bought EMC (2016). The live shift is the ODM-direct threat — hyperscalers buying "white-box" servers straight from contract manufacturers, squeezing the branded vendors, with Nvidia also centralizing AI-server assembly among selected partners — and the reshoring of AI-server assembly onto U.S. soil (Foxconn, Quanta, Wistron, Pegatron, Inventec plants in Texas, Wisconsin, and California). That reshoring is one of the few places domestic 33411 employment could actually grow [4].
• Peripherals (334118) is consolidation by survival and scale, and it is still moving. NCR split into Voyix (retail) and Atleos (banking) in 2023; Diebold merged with Wincor Nixdorf, then passed through Chapter 11 and re-emerged; Xerox completed its $1.5 billion Lexmark acquisition in July 2025; Verifone and Ingenico went private/absorbed. The newest move removes a listed name: Brink's agreed in February 2026 to acquire NCR Atleos for approximately $6.6 billion including assumed debt, with both shareholder bases approving in June 2026 [6].

A rollup point the children make only separately: the three are moving in opposite directions on domestic manufacturing itself. The computer child is pulling final assembly onshore; the peripheral child is pushing it further out — NCR Voyix completed the transfer of its self-checkout and POS hardware design and manufacturing to an outsourced model in March 2026 [6]; and the storage child builds essentially no new drive plants anywhere, putting its capital into chip fabs and software instead [5]. Since the federal statistics in Section 3 count only domestic assembly, these three trajectories, not the demand cycle, are what will move this level's measured size over the next several years. Across the level, expect further vertical moves — into software, services, cooling/power, and supply — more than simple horizontal mergers. Concentration remains high inside each child even as the pooled level reads as moderate (Section 3).

---

9. Risks

• AI-capex "digestion" risk (concentrated in two children). The computer and storage boom rests on a handful of hyperscalers spending unprecedented, increasingly debt-funded sums. If AI returns disappoint or budgets pause, orders can fall fast — and ~83% of the level's receipts sit on that single bet [4][5].
• Memory-cost inflation cuts across the level in both directions. The same shortage that hands the storage child pricing power is a direct cost shock to the other two: Gartner projects DRAM and SSD prices up ~130% by the end of 2026 and forecasts a ~10% drop in worldwide PC shipments as a result, while Logitech reports memory, microcontroller, and optical-sensor shortages as suppliers redirect capacity toward AI [4][6]. A single-name or single-child view will misread this; at the level it is a transfer, not a uniform headwind.
• Cyclicality and commoditization (the whole level). These are boom-bust, thin-margin, substitutable-product businesses; a maker that loses its cost, scale, or platform edge has little to fall back on. Storage's history is oversupply → price crash → losses, and Seagate notes its vertically integrated cost base cannot be reduced quickly when demand falls; peripherals' post-2020 boom-then-glut whipsawed results within two years [5][6].
• Customer concentration. Server and drive makers depend on a few cloud giants — cloud is 88% of Western Digital's revenue, its top ten customers 68% — and those same customers can bypass the brand via ODM-direct [4][5].
• Supply-chain concentration. The level leans on Taiwan-centered chips and Southeast-Asian assembly; high-bandwidth memory and advanced packaging were reported sold out through 2026, Thailand's 2011 floods once crippled global drive supply, and China originates the majority of global printed-circuit-board exports [4][5][6].
• Trade and export-control whiplash. Sudden tariff or licensing changes can strand inventory, close markets, or reroute production overnight — heaviest on computers and storage, with the $300 million Seagate settlement as the concrete precedent [4][5].
• Fixed-cost and balance-sheet risk. Storage carries obligations that persist through downturns — SanDisk owns 49.9% of each Kioxia joint venture and must fund roughly half its fixed costs regardless of output taken, and Seagate spent $724 million on product development in fiscal 2025 alone. In peripherals, commoditized hardware plus leverage produced Diebold's bankruptcy [5][6].
• Technology substitution and obsolescence. Flash could erode hard drives' cost edge (Pure Storage's Meta design win is an early shot, and Seagate itself acknowledges SSD encroachment); soft-POS phones threaten payment terminals; product generations turn over fast and unsold hardware loses value weekly [5][6].
• Secular decline in parts of peripherals. Office print volume and cash usage are both structurally falling — HP explicitly tied its fiscal 2025 supplies decline to lower installed-base usage — a headwind unique to the smallest child [6].

---

10. How to invest, and the outlook

Match the child to the thesis — the level is not one trade.

• Direct AI-hardware torque: the computer child's AI-server names — Dell (DELL), HPE (HPE), and especially Super Micro (SMCI) — give the most direct, highest-beta exposure to the capex wave, at the cost of thin and currently compressing margins [4].
• A rare pricing-power window: the storage child's hard-drive makers, Western Digital (WDC) and Seagate (STX), are the cleanest bets on the 2025–26 shortage and the only place in the level where the AI wave is expanding margins; Micron (MU) and SanDisk (SNDK) add (more cyclical) memory exposure [5].
• Quality and recurring revenue: IBM's mainframe (IBM), Apple's integrated Mac (AAPL), and the storage-systems names NetApp (NTAP) and Pure Storage (PSTG) trade less like commodities and more like software-attached franchises [4][5].
• The mature, cash-generative corner: the peripherals child — Zebra (ZBRA), HP printing (HPQ), Logitech (LOGI), NCR Voyix (VYX) — is where to look past the hardware line to the recurring-revenue share, not to ride AI. Note that NCR Atleos (NATL) is pending acquisition by Brink's, so the listed pure-play ATM exposure is on its way out [6].
• Span-the-stack names: HP, Dell, HPE, and IBM each touch more than one child at once, giving diversified 33411 exposure in a single ticker (Section 4).
• Private-market routes differ sharply by child: the U.S. AI-server reshoring build-out and its power/cooling adjacencies (via ODM parents and infrastructure funds) for computers; storage-software, controller-chip, and archival-tape startups plus the memory-fab supply chain for storage; and private-equity-owned payment, kiosk, and rugged-hardware specialists for peripherals [4][5][6].

There is no single U.S.-listed fund for this level; passive investors get it bundled inside broad technology-hardware and semiconductor funds — which also fold in the Nvidia-style chip layer that captures the fattest margins in the value chain but sits outside 33411 [4].

A caution on comparing numbers across this family, carried up from the storage child because it applies to the whole level: Census figures 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 systems vendors mix devices with software and services. A Census shipment total, a global share estimate, and a public company's consolidated revenue are three different universes and should not be combined into a single "market size" [5].

Near-term outlook (forward-looking). The base case is strong dollar demand for two of the three children: hyperscaler capex is guided sharply higher into 2026, hard drives are effectively sold out through the year with lead times stretching toward two years and pricing power projected into 2027 against a $25–$30 per-terabyte target for 2027–28, and reshoring should add some domestic assembly [4][5]. The peripherals child stays cyclical and, in printing and cash, secularly soft, with reshoring of security-sensitive gear a modest offset [6]. The through-line for the whole level — and the reason the box-maker is rarely the best investment in it — is that value keeps migrating out of the metal and into mix, proprietary platforms, and attached software, services, and consumables. The watch-items now differ by child rather than running in parallel: for computers, whether GPU-heavy revenue and memory-cost inflation keep compressing margins even as shipments grow; for storage, whether three-maker capital discipline holds and the HAMR ramp qualifies with hyperscalers; for peripherals, whether recurring revenue grows fast enough to outrun the decline in print and cash. Volumes look set to grow, but 33411 remains a cyclical, thin-margin, capital-disciplined family of industries where the winners manage cost, mix, and working capital best — not simply the ones who ship the most hardware.

---

Sources

Figures for the level (NAICS 33411) are our ingested federal ground truth; company-level and industry-color figures are synthesized from the three Histometrics child primers [4][5][6] and the underlying sources they cite.

1. U.S. Census Bureau, County Business Patterns (CBP) 2023 — NAICS 33411 and children 334111/334112/334118. Establishments (557), employment (31,508), annual payroll ($3.84B), and the per-child splits. https://www.census.gov/programs-surveys/cbp.html
2. U.S. Census Bureau, 2022 Economic Census — Concentration by Largest Firms, NAICS 33411. Firms (384), receipts ($19.84B), CR4 43.9% / CR8 58.6% / CR20 78.0% / CR50 89.2%, HHI 706.8. https://www.census.gov/programs-surveys/economic-census.html
3. U.S. Small Business Administration, Table of Small Business Size Standards (2023). 1,250-employee standard for 334111/334112; 1,000 for 334118. https://www.sba.gov/document/support-table-size-standards
4. Histometrics primer, 334111 Electronic Computer Manufacturing (2026) — and its cited sources (Apple FY2025 results; Dell FY2026, HP FY2025, HPE FY2025, Super Micro FY2025 and IBM FY2025 Form 10-Ks; Gartner 2025 PC shipments, AI-PC and 2026 memory-cost forecasts; Goldman/Introl hyperscaler-capex estimates; MarketsandMarkets generative-AI server market; TrendForce Taiwan-ODM U.S. expansion; CRS/Gibson Dunn export-control and Section 232 analysis; BIS January 2026 licensing update; Census AIES 2023 and FT900 import data; FCC firm-size analysis; EPA electronics-manufacturing rules).
5. Histometrics primer, 334112 Computer Storage Device Manufacturing (2026) — and its cited sources (Seagate, Western Digital, SanDisk and Quantum FY2025 Form 10-Ks and results; Coughlin/Forbes HDD industry updates and capacity shares; TrendForce NAND shares; IDC external enterprise-storage market; NetApp and Pure Storage results; DigiTimes/Morgan Stanley shortage and pricing; BIS Seagate settlement; Micron CHIPS Act award; LTO Program tape-shipment data; JPX Toshiba delisting and Kioxia TSE listing notices).
6. Histometrics primer, 334118 Computer Terminal and Other Computer Peripheral Equipment Manufacturing (2026) — and its cited sources (Zebra FY2025, Logitech FY2026, HP FY2025 and NCR Atleos FY2025 Form 10-Ks; NCR Voyix, Diebold Nixdorf, Corsair results; Xerox/Lexmark completion; Brink's–NCR Atleos merger filings; NCR Voyix hardware-manufacturing transfer; Census M3 shipments and BLS producer price index; IBISWorld domestic-peripheral decline; FCC firm-size analysis; Verifone/Ingenico/PAX ownership; PCI, ADA, Section 508, EPA CRT and Section 301 tariff analysis).
7. Apple Inc., Q4 FY2025 Results (Form 8-K) (2025). Mac segment revenue (~$33.7B). https://www.apple.com/newsroom/2025/10/apple-reports-fourth-quarter-results/