Electronic Computer Manufacturing (U.S.)

NAICS 2022 code 334111 — the industry that builds finished electronic computers: desktops, laptops, servers, mainframes, and workstations.

---

1. Overview

Electronic Computer Manufacturing covers the companies that assemble complete, working computers — from a consumer laptop to a room-sized mainframe to the rack-mounted servers that fill an artificial-intelligence (AI) data center. It does not cover the chips inside, the monitors and keyboards around them, or the software that runs on them; those are separate industries (see Section 2).

Why an investor should care: this industry sits at the cash register of the AI build-out. The world's largest cloud operators are spending hundreds of billions of dollars a year on computing hardware, and most of that money flows through a short list of computer makers and their factory partners [10][15]. At the same time the industry has a hard truth built into it — assembling computers is a low-margin, cyclical business. Owners make real money only by controlling cost, brand, or a high-value niche, not by stamping out boxes.

Ways in for investors:

• Public markets: a handful of large, liquid U.S.-listed names (Apple, Dell, HP Inc., Hewlett Packard Enterprise, Super Micro, IBM) plus foreign-listed contract manufacturers.
• Private markets: most of the physical assembly is done by privately held or foreign contract manufacturers (Foxconn, Quanta, Wistron and peers), and a wave of new U.S. assembly plants is being built with private and joint-venture capital [12].

A caution up front: U.S. federal statistics measure only computers physically assembled on American soil, which is a small fraction of what U.S.-headquartered companies actually sell. That gap is the single most important thing to understand about this industry, and Section 3 explains it.

---

2. What it is and how it's structured

In scope (334111): establishments whose main activity is manufacturing and/or assembling complete electronic computers — personal computers, laptops/notebooks, servers, mainframe computers, and workstations. A plant that takes chips, boards, memory, drives, and a chassis and turns them into a finished computer belongs here [1].

What it explicitly excludes — and where those activities are classified instead:

• Peripherals (external monitors, printers, keyboards, mice, scanners): NAICS 334118, Computer Terminal and Other Computer Peripheral Equipment Manufacturing [1].
• The chips and loaded boards inside (semiconductors, sound/video/network cards, add-in boards, solid-state storage devices): NAICS 334412/334413 and related 33441 codes [1].
• Networking and telecom gear (switches, routers): NAICS 334210, Telephone Apparatus Manufacturing [1].
• Blank storage media: NAICS 334613 [1].
• Software, IT services, systems integration, and repair: these are services industries, not manufacturing, and sit outside 334 entirely [1].

Ownership mix. The domestic industry is unusually top-heavy. A very small number of firms account for nearly all output (Section 3), and much of the actual box-building is done by contract manufacturers — also called original design manufacturers (ODMs) — that build hardware to a brand's specification. Dell, for example, states in its FY2026 10-K that it uses contract manufacturers worldwide, purchases components and resells them to those manufacturers, and has substantial outsourced assembly concentrated among a few manufacturers, principally in Asia; its own process consists primarily of assembly, software installation, functional testing and quality control [6]. The familiar U.S. brands mostly design, market, and support computers; the physical assembly is outsourced, historically to Asia and increasingly to new U.S. plants [12]. So "ownership" splits three ways: brand owners (Apple, Dell, HP), server/systems specialists (HPE, Super Micro, IBM), and the factory operators (Foxconn/Hon Hai, Quanta, Wistron, Inventec, Pegatron) that most consumers never see.

---

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

Our ground-truth federal figures for NAICS 334111:

Read that carefully: about 13,000 workers and $12-14 billion in receipts. That is tiny for something as central to the economy as "computers." The reason is not that computers are a small business — it is that the federal manufacturing statistics only count computers physically assembled inside the United States. The U.S. companies that dominate global computing (Apple, Dell, HP) design and sell computers here but have their hardware built abroad; that revenue is recorded under wholesale, retail, or corporate-management categories, not under domestic manufacturing. For comparison, a single U.S. company's computer-hardware line — Apple's Mac — brought in roughly $34 billion in fiscal 2025, more than double the entire domestic industry's measured receipts [5]. So treat the federal figures as a fair measure of on-shore assembly, not of the U.S. computer business, which is an order of magnitude larger once offshore production and imports are counted. U.S. imports of computer hardware and semiconductors topped $450 billion in 2025 [13].

The concentration figures are trustworthy and telling: the four largest firms account for 60.5% of domestic receipts and the top 50 account for 97.8% — a heavily concentrated industry even before you look at the global brand market, which is a tighter oligopoly still. Note that global PC shipment shares (Lenovo, HP, Dell) should not be confused with official U.S. manufacturing-establishment concentration; those shares measure worldwide channel shipments and exclude servers.

---

4. The investable universe

There are relatively few pure-play ways to buy this industry, and no single stock is only "a U.S. computer-assembly company." The large public names are diversified technology firms for which finished computers are one segment. The dedicated box-builders are mostly listed in Taiwan.

Publicly traded (U.S.-listed) — brand and systems makers

Publicly traded (foreign-listed) — global brand and the box-builders

A note on Nvidia (NVDA). Nvidia is classified as a semiconductor company, not a computer maker, and is not in NAICS 334111. But it increasingly sells complete rack-scale "computers" (its GB200/GB300 systems) and captures an estimated ~90% of AI-accelerator spending [10], which shapes the economics of everyone in this industry. Investors treat it as an adjacent, dominant force rather than a pure computer-manufacturing play.

Private / other owners. Most physical U.S. assembly is done by privately held operations or the U.S. subsidiaries of the foreign ODMs above. The current wave of new domestic AI-server plants — in Texas, Wisconsin, and California — is being built by these contract manufacturers, often as joint ventures with brand customers, and is largely outside public-equity reach except through the parent listings [12]. Private firms in the space are generally specialty, rugged, boutique or white-box manufacturers rather than peers of Dell, HP or Lenovo.

---

5. How the money works

This is a manufacturing business, so the economics run on capacity utilization, input costs, product mix, and cyclicality — not on rate base, occupancy, or fee rates. The key levers:

• Thin hardware margins, won on scale and cost. Assembling a PC or a commodity server earns low-single-digit to mid-single-digit operating margins. Dell's fiscal 2026 Client Solutions Group generated $51.0 billion of revenue at a 5.6% segment operating margin; HP's fiscal 2025 Personal Systems reported $38.5 billion at 5.3% [6][7]. Owners make money by buying components cheaply at huge volume, running factories at high utilization, turning inventory fast, and holding little of it. Dell states it normally holds minimal component and product inventories, which supports working-capital efficiency but increases vulnerability to shortages [6]. A pile of unsold laptops loses value by the week, so tight working-capital and supply-chain management is the whole game.
• Average unit price (AUP) and mix. The most important recent lever is mix shift toward high-value AI servers. When a server carries eight expensive graphics processing units (GPUs), its price can be 10-20x a plain server, so revenue dollars explode — Dell's fiscal 2026 AI-optimized-server revenue reached $24.7 billion, up 166% [6]. But watch the trap: because the pricey GPU is bought from a chip vendor and passed through, gross-margin percentage often falls even as gross-margin dollars rise. Dell's consolidated gross margin fell to 20.0% in FY2026, principally because mix shifted toward AI-optimized servers [6]. Super Micro's gross margin fell to 11.1% from 13.8%, citing competitive pricing and product mix [9]. AI-server revenue is big but structurally lower-margin than legacy gear.
• Infrastructure carries better margins than PCs. Dell's Infrastructure Solutions Group posted an 11.7% operating margin in FY2026 [6], roughly double its PC segment — but that business includes higher-margin storage and networking, not just servers.
• Backlog and bookings. For AI servers, orders are placed months ahead and reported as backlog. Backlog and book-to-bill ratios are the leading indicators investors watch, because they signal revenue that is already sold. Dell warns that customer readiness, component transitions and large orders make AI demand and shipment timing non-linear [6].
• The high-margin exception: mainframes. IBM's Z systems are the outlier — a proprietary, mission-critical computer sold on multi-year cycles with rich software and services attached. Infrastructure segment margins run above 20%, far above commodity hardware, and a new model (the AI-oriented z17) can lift revenue for years [11].
• Brand and integration. Apple earns computer-industry-leading margins on the Mac not by manufacturing cleverly but by controlling the chip, the operating system, and the brand — vertical integration that a box-builder cannot copy.
• Recurring revenue attach. The brands increasingly bolt services, financing, warranties, and software subscriptions onto the hardware sale to smooth out the cyclicality of the box itself.

Bottom line: the box is a low-margin, cyclical commodity; the profit lives in mix, scale, attached services, and the rare proprietary platform.

---

6. What drives demand

• AI data-center build-out (the dominant driver today). The largest cloud operators plan on the order of $600-725 billion of capital spending in 2026, up sharply from ~$410 billion in 2025, with roughly half going to servers and chips [10]. Goldman Sachs estimates cumulative hyperscaler capex of about $1.15 trillion across 2025-2027 [10]. One market estimate puts the generative-AI server market alone at ~$104 billion in 2025 [15]. This is the demand wave lifting Dell, HPE, Super Micro, and the ODMs.
• PC replacement cycles. The consumer/enterprise PC market runs in multi-year refresh waves. 2025 shipments grew ~9% to more than 270 million units [14]. Microsoft ended standard Windows 10 support on October 14, 2025 [19], giving enterprises a concrete security and compatibility reason to refresh older systems. Gartner estimated "AI PCs" reached ~31% of shipments (~78 million units) in 2025, though it subsequently observed that many buyers were future-proofing fleets and that local-AI features had not yet produced clear productivity gains over cloud-based AI [14][20]. The 2025 rebound was partly demand pulled forward, not necessarily a new long-term unit-growth regime.
• Enterprise and cloud IT budgets. Corporate server and storage refresh, plus general cloud expansion, set the baseline for non-AI hardware.
• Consumer spending and product cycles. For Apple and the PC brands, household budgets, new form factors, and upgrade timing move volumes.
• Government and defense procurement. Federal and defense buyers are a meaningful, policy-sensitive slice of high-assurance computing demand.

---

7. Regulation

Computer manufacturing is lightly regulated as a product (mainly safety, energy-efficiency, and electromagnetic standards), but it is heavily shaped by trade and national-security policy — arguably more than any other consumer-hardware industry right now:

• Export controls on advanced computing. The U.S. Bureau of Industry and Security (BIS) restricts exports of the most advanced AI chips and computers, using license requirements and an "Entity List" of blocked buyers, chiefly aimed at China [16]. Policy has whipsawed: 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, allowing some advanced-chip sales to approved Chinese customers subject to capacity, compliance and testing conditions [21]. This directly affects what AI systems U.S. makers can sell where.
• Section 232 tariffs on semiconductors. A 25% tariff on certain advanced semiconductor articles was imposed under Section 232 (national-security) authority, with notable carve-outs — including chips used in U.S. data centers, R&D, and domestic manufacturing [16]. Because finished AI computers are mostly chip cost, tariff design flows straight through to system prices.
• The CHIPS and Science Act subsidizes domestic semiconductor fabrication rather than final computer assembly, but it reshapes the upstream supply chain that this industry depends on.
• Government procurement rules (Buy American provisions, the Trade Agreements Act) govern which computers federal agencies may buy and create an incentive for U.S.-based assembly.
• Environmental compliance. Electronics manufacturing facilities meeting relevant source-category and threshold definitions may be subject to EPA greenhouse-gas reporting, and end-of-life computers create metals and e-waste-management exposure [22].

The net effect is that a computer maker's addressable market, cost structure, and factory location decisions are now partly set in Washington, not just in the market.

---

8. Competitive dynamics and consolidation

• A concentrated, oligopolistic market. Domestically the top four firms hold 60.5% of receipts [3]; globally, PCs are led by Lenovo (~27%), HP (~21%), and Dell (~15%) [14], and servers by Dell, HPE, Super Micro, and the ODMs. Barriers are scale, supply-chain relationships, and channel reach rather than technology secrets.
• The ODM-direct threat. The biggest structural shift is hyperscalers buying "white-box" servers directly from contract manufacturers (Foxconn, Quanta, Wistron), cutting out the branded middleman. This pressures the traditional server vendors' margins and pushes them to add software and services value. Nvidia has also moved to centralize AI-server assembly among selected partners, reshaping who builds what [12].
• Reshoring of final assembly. Tariff risk and customer preference are pulling AI-server assembly onshore: Foxconn, Wistron, Quanta, Pegatron and Inventec have announced or begun U.S. plants (Houston and Dallas, Texas; Wisconsin; California) [12]. This is one of the few places domestic NAICS 334111 employment could actually grow.
• Consolidation history. The sector has already restructured: Hewlett-Packard split into HP Inc. (PCs/printers) and Hewlett Packard Enterprise (servers) in 2015, and Dell acquired EMC in 2016 to build its infrastructure arm. Expect further vertical moves (into software, services, and cooling/power) rather than simple horizontal mergers.

---

9. Risks

• Cyclicality and an AI-capex "digestion" risk. The current boom rests on a handful of hyperscalers spending unprecedented sums, increasingly funded by debt [10]. If AI returns disappoint or budgets pause, orders can fall fast — this is a boom-bust industry, and the AI wave is the largest bet in it.
• Commoditization and thin margins. Most products are substitutable; a maker that loses its cost or scale edge has little to fall back on.
• Customer concentration. For server makers, a few cloud giants drive demand; losing one program is material, and those same customers can bypass the brand via ODM-direct.
• Supply-chain and component bottlenecks. The industry depends on Taiwan-centered chip and assembly capacity and on scarce inputs — high-bandwidth memory (HBM) and advanced packaging (CoWoS) were reported sold out through 2026 [10]. GPU and memory cost inflation squeezes margins. Gartner projects combined DRAM and solid-state-drive prices to rise ~130% by the end of 2026, which it forecasts will reduce worldwide PC shipments by ~10% during 2026 [23].
• Trade and export-control whiplash. Sudden tariff or licensing changes can strand inventory, close markets, or reroute production overnight (Section 7).
• Technology obsolescence. Product generations turn over quickly; unsold or superseded hardware loses value rapidly.
• Concentrated sourcing. Dell notes that many components come from foreign vendors, assembly is concentrated among a few Asian contractors, and some critical products remain single- or limited-sourced [6].

---

10. How to invest, and the outlook

Public-market routes.

• Broad, diversified exposure: the mega-cap names (Apple, IBM) give computer-hardware exposure wrapped inside much larger services/ecosystem businesses — lower volatility, but the computer line barely moves the whole.
• Higher-beta AI-hardware plays: Dell, HPE, and especially Super Micro are the most direct listed bets on the AI-server cycle — more torque, but lower margins and more cyclicality [6][8][9].
• PC-cycle plays: HP Inc. and Lenovo track the consumer/enterprise refresh cycle [7][14].
• The picks-and-shovels adjacency: many investors gain "computer" exposure through the chip layer (Nvidia and peers), which captures the fattest margins in the value chain but is a different industry [10].
• Foreign-listed manufacturers: the Taiwan ODMs (Foxconn, Quanta, Wistron) are the purest exposure to physical assembly volumes, accessible via foreign listings or ADRs.

Private-market routes. The clearest private angle is the U.S. reshoring build-out — new AI-server assembly plants, plus the data-center real estate, power, and cooling around them — typically reached through the ODM parents, private infrastructure funds, or joint ventures rather than standalone equities [12]. Diligence should establish who owns component inventory and non-cancellable purchase commitments; customer and supplier concentration; dependence on rebates; warranty reserves; domestic versus imported production; and whether AI-server growth remains profitable after accelerator pass-through, expedite costs and write-downs.

Near-term outlook (forward-looking). The base case is continued strong dollar demand: hyperscaler capex is guided sharply higher into 2026 [10], the PC refresh still has room (though memory-cost headwinds may suppress 2026 shipments [23]), and reshoring should add domestic assembly. The watch-items are margin and durability — GPU-heavy revenue is lower-margin than it looks, the boom depends on a few debt-funded buyers, and policy can turn quickly. In short, the volumes look set to grow, but this remains a cyclical, thin-margin, capital-disciplined industry where the winners are the ones who manage cost, mix, and working capital best — not simply the ones who ship the most boxes.

---

Sources

1. U.S. Census Bureau, 2022 NAICS Definition — 334111 Electronic Computer Manufacturing (2022). https://www.census.gov/naics/?details=334111&input=334111&year=2022
2. U.S. Census Bureau, County Business Patterns, NAICS 334111 (2023). Establishments, employment, and payroll. https://www.census.gov/programs-surveys/cbp.html
3. U.S. Census Bureau, 2022 Economic Census — Concentration by Largest Firms, NAICS 334111 (2022). Firms, receipts, and CR4/CR8/CR20/CR50 shares; HHI suppressed. https://www.census.gov/programs-surveys/economic-census.html
4. U.S. Small Business Administration, Table of Small Business Size Standards (2023). 1,250-employee standard for NAICS 334111. https://www.sba.gov/document/support-table-size-standards
5. Apple Inc., Q4 FY2025 Results (Form 8-K) (2025). Mac segment revenue. https://www.apple.com/newsroom/2025/10/apple-reports-fourth-quarter-results/
6. Dell Technologies Inc., Form 10-K, FY2026 (2026). Client Solutions Group, Infrastructure Solutions Group revenue and margins, AI-server revenue, outsourcing model. https://www.sec.gov/Archives/edgar/data/1571996/000157199626000008/dell-20260130.htm
7. HP Inc., Form 10-K, FY2025 (2025). Personal Systems revenue and operating margin. https://www.sec.gov/Archives/edgar/data/47217/000004721725000071/hpq-20251031.htm
8. Hewlett Packard Enterprise, Form 10-K, FY2025 (2025). Total and server-segment revenue. https://www.sec.gov/Archives/edgar/data/1645590/000164559025000130/hpe-20251031.htm
9. Super Micro Computer, Inc., Form 10-K, FY2025 (2025). Revenue, gross margin. https://www.sec.gov/Archives/edgar/data/1375365/000137536525000027/smci-20250630.htm
10. Introl / Goldman Sachs estimates, Hyperscaler CapEx 2026 and AI Infrastructure Outlook (2026). Hyperscaler capex, GPU spend share, HBM/CoWoS constraints. https://introl.com/blog/hyperscaler-capex-600b-2026-ai-infrastructure-debt-january-2026
11. International Business Machines Corp., Form 10-K, FY2025 (2026). Infrastructure segment revenue and IBM Z performance. https://www.sec.gov/Archives/edgar/data/51143/000005114326000010/ibm-20251231_d2.htm
12. TrendForce / The Diplomat / DIGITIMES, Taiwan ODMs and U.S. AI-Server Assembly Expansion (2025). Foxconn, Quanta, Wistron, Pegatron, Inventec U.S. plants and revenue growth. https://www.trendforce.com/news/2025/02/19/news-taiwans-ai-server-manufacturers-target-texas-for-u-s-expansion-amid-tariff-concerns/
13. Tom's Hardware / U.S. Census Bureau FT900, U.S. Trade Deficit and AI Hardware Imports (2026). >$450B computer-hardware and semiconductor imports in 2025. https://www.tomshardware.com/tech-industry/manufacturing/us-trade-deficit-hits-a-record-usd1-2-trillion-as-ai-hardware-imports-surge-under-the-trump-administration
14. Gartner, Worldwide PC Shipments Full-Year 2025 (2026). ~270M units, +9.1%; Lenovo/HP/Dell share. https://www.gartner.com/en/newsroom/press-releases/2026-1-20-gartner-says-worldwide-pc-shipments-increased-9-point-3-percent-in-fourth-quarter-of-2025-and-9-point-1-percent-for-the-full-year
15. MarketsandMarkets, Generative AI Server Market (2025). ~$103.9B in 2025, projected to $448.6B by 2030 (34% CAGR). https://www.marketsandmarkets.com/ResearchInsight/generative-ai-server-companies.asp
16. Congressional Research Service / Gibson Dunn, U.S. Export Controls and Section 232 Tariffs on Advanced Semiconductors and Computing (2025–2026). BIS licensing, Entity List, 25% Section 232 tariff and carve-outs. https://www.congress.gov/crs-product/R48642
17. Federal Communications Commission, FCC Rulemaking Citing Census Economic Census Firm-Size Data (2026). 148 firms, 128 below SBA threshold. https://docs.fcc.gov/public/attachments/FCC-26-38A1.pdf
18. U.S. Census Bureau, Annual Integrated Economic Survey, NAICS 334111 (2023). Sales/shipments/revenue $14.07B. https://data.census.gov/table/AIESBASICTIMESERIES.AIES31BASIC01?codeset=naics~334111&g=010XX00US
19. Microsoft, Windows 10 Support Lifecycle Notice (2025). End of support October 14, 2025. https://support.microsoft.com/en-us/windows/deployment/updates/lifecycle/windows-10-support-has-ended-on-october-14-2025
20. Gartner, AI PCs Will Represent 31 Percent of Worldwide PC Market by End of 2025 (2025). AI-PC forecast and adoption caveats. https://www.gartner.com/en/newsroom/press-releases/2025-08-28-gartner-says-artificial-intelligence-pcs-will-represent-31-percent-of-worldwide-pc-market-by-the-end-of-2025
21. Bureau of Industry and Security, Department of Commerce Revises License Review Policy for Semiconductors Exported to China (2026). January 2026 policy update for H200/MI325X-class chips. https://www.bis.gov/press-release/department-commerce-revises-license-review-policy-semiconductors-exported-china
22. U.S. Environmental Protection Agency, Subpart I — Electronics Manufacturing. GHG reporting and e-waste. https://www.epa.gov/ghgreporting/subpart-i-electronics-manufacturing
23. Gartner, Surging Memory Costs Will Reduce Global PC and Smartphone Shipments in 2026 (2026). DRAM/SSD price forecast +130%, PC shipments forecast -10.4%. https://www.gartner.com/en/newsroom/press-releases/2026-02-26-gartner-says-surging-memory-costs-will-reduce-global-pc-and-smartphone-shipments-in-2026