Foxconn (Hon Hai Precision Industry)
Last reviewed
Sources
52 citations
Review status
Source-backed
Revision
v4 · 4,853 words
Improve this article
Add missing citations, update stale details, or suggest a clearer explanation.
Suggest edit
CtrlK
Last reviewed
Sources
52 citations
Review status
Source-backed
Revision
v4 · 4,853 words
Add missing citations, update stale details, or suggest a clearer explanation.
Foxconn, officially Hon Hai Precision Industry Co., Ltd. (Chinese: 鴻海精密工業股份有限公司; trading as Hon Hai Technology Group), is a Taiwanese multinational electronics contract manufacturer that is the world's largest builder of NVIDIA AI servers. Founded in 1974 by Terry Gou and headquartered in Tucheng District, New Taipei City, Foxconn shipped roughly 12,000 NVIDIA GB-series rack-scale systems in 2025 for an estimated 43 percent of the global AI rack market, the single largest share of any manufacturer.[1][2][3] It is the world's largest contract manufacturer of electronics and Taiwan's largest private employer, operating more than 230 campuses across roughly 24 countries with a global workforce that swells to around 900,000 during peak seasons. The company trades on the Taiwan Stock Exchange under ticker 2317.[4][5]
Though best known for assembling the iPhone for Apple and consumer products for Sony, Dell, HP, Microsoft, Nintendo, and others, Foxconn became a central player in the artificial intelligence infrastructure economy in the mid-2020s. In the second quarter of 2025 its AI server and cloud-networking business overtook its consumer-electronics (Apple) business as the largest source of group revenue for the first time, with the cloud and networking division reaching about 41 percent of revenue versus roughly 35 percent for consumer electronics.[3][6] The company is building one of the world's largest production lines for NVIDIA's GB200 superchips, deploying NVIDIA-powered humanoid robots on its own assembly lines, and pursuing robotics, advanced semiconductor packaging, electric vehicles, and digital health under chairman Young Liu.[7][8]
Reported revenue for fiscal year 2024 reached approximately NT$6.86 trillion (about US$208 billion), an 11.4 percent increase year over year and, at the time, the highest annual figure ever recorded by the company.[9][10] Full-year 2025 revenue rose further to roughly US$252 billion (up about 18 percent), and first-quarter 2026 revenue reached about US$67 billion, up 29.7 percent year over year, on record AI server demand.[2][3]
Foxconn provides full-stack electronics manufacturing services (EMS), spanning printed circuit board assembly, system-level integration, mechanical components, plastics, connectors, optics, networking equipment, and high-density servers. Since the late 2010s the company has pivoted aggressively toward higher-margin lines, including AI servers, automotive electronics, semiconductors, and industrial robotics, while diversifying production geographically.[4][5]
Foxconn's leadership refers to its strategic framework as 3+3+3: three intelligent platforms (smart manufacturing, smart EV, smart city), three emerging industries (electric vehicles, digital health, robotics), and three core technologies (artificial intelligence, semiconductors, next-generation communications). NVIDIA technologies underpin all three intelligent platforms, with smart manufacturing built on NVIDIA Isaac robotics, smart EV on NVIDIA DRIVE Hyperion 9 and Thor, and smart city on NVIDIA Metropolis video analytics.[11]
Foxconn has become the most important single manufacturing partner of the AI infrastructure boom. The company assembles complex liquid-cooled rack-scale AI systems for the largest U.S. cloud providers including Microsoft, Amazon Web Services, Google, and Meta. It shipped about 12,000 NVIDIA GB-series racks in 2025 for an estimated 43 percent of the global AI rack market, a share analysts project could climb to 55-60 percent in 2026 as rack shipments roughly double.[1][2] AI server output reached roughly 1,000 racks per week by late 2025, and Foxconn's U.S. capacity has been ramping toward about 2,000 racks per week, a rate one estimate values at US$200 million to US$300 million in weekly revenue from a single factory.[12][2]
In the second quarter of 2025, Foxconn's cloud and networking segment overtook smart consumer electronics as the largest contributor to group revenue, with quarterly AI server revenue projected to grow more than 170 percent year over year and server-rack sales by about 300 percent in the third quarter of 2025.[3][6][12] Foxconn has guided full-year AI server revenue past NT$1 trillion (about US$33 billion).[6]
Foxconn is the leading manufacturing partner for NVIDIA's GB200 NVL72 rack-scale platform, built around the Blackwell GPU architecture. In October 2024, Foxconn announced construction of what it called the world's largest GB200 production facility in Guadalajara, Mexico, operated through its Foxconn Industrial Internet subsidiary and designed for capacity of roughly 20,000 AI servers annually. The plant began small-scale GB200 production in Q3 2024 and ramped to volume during 2025.[13][14] Foxconn is qualified by NVIDIA as a PBR (Pilot Build Request) Partner, the highest-tier manufacturing classification, for both GB200 NVL72 and the successor GB300 NVL72 platforms.[7]
In 2025, Foxconn began construction of a major AI server manufacturing campus in Houston, Texas, dedicated principally to NVIDIA-based AI server racks for U.S. customers. The Houston plant became the headline location for Foxconn's announcement that it would deploy NVIDIA Isaac GR00T-powered humanoid robots on its production lines beginning in Q1 2026.[15][16]
In November 2025, Foxconn disclosed an expanded direct relationship with OpenAI, shifting from indirect supplier through hyperscaler customers to direct collaborator on AI hardware, including data-center cabling, power systems, and related equipment for U.S. data-center infrastructure, mirroring Foxconn's direct engagements with Google, AWS, and Microsoft.[12] Foxconn also manufactures the AI servers used by Apple for Apple Intelligence (Private Cloud Compute), which run on Apple silicon, and is positioned to remain a primary assembler for Apple's future in-house AI chip targeting 2027 server deployment.[17]
In 2024, Foxconn and NVIDIA announced the Hon Hai Kaohsiung Super Computing Center, a Blackwell-based supercomputer in southern Taiwan using 64 GB200 NVL72 racks containing 4,608 Tensor Core GPUs and projected to deliver more than 90 exaflops of AI performance. Initial operations were expected by mid-2025, with full deployment targeted for 2026.[18]
| Client | Products / role | Notes |
|---|---|---|
| NVIDIA | GB200 NVL72 and GB300 NVL72 rack-scale systems | PBR (Pilot Build Request) Partner status |
| Microsoft Azure | Custom AI training and inference racks | Direct hyperscaler relationship |
| Amazon Web Services | Custom AI accelerator and training racks | Direct hyperscaler relationship |
| Google Cloud | TPU and GPU-based AI server hardware | Direct hyperscaler relationship |
| Meta | Custom AI training and inference servers | ODM partner alongside Quanta and Wiwynn |
| Apple | Apple Intelligence (Private Cloud Compute) servers | Apple silicon-based AI infrastructure |
| OpenAI | Direct collaboration on AI hardware | Announced November 2025 |
Hon Hai Precision Industry was founded in 1974 by Terry Gou, then aged 23, with about US$7,500 borrowed from his mother. Gou rented a shed in Tucheng and used the capital to buy plastic injection-molding machines, initially producing knobs for Western television sets. By the early 1980s the firm had moved into connectors for personal computers, supplying U.S. and Japanese customers. The company adopted Foxconn as its trade name in 1985.[5][19] Gou opened the company's first mainland China factory in Shenzhen in 1988. The Shenzhen Longhua complex, sometimes called "Foxconn City," eventually grew to host hundreds of thousands of workers and became one of the largest single industrial sites in the world.[5]
Foxconn began assembling Apple iMacs in the early 2000s and was selected to assemble the original iPhone in 2007. Apple has since become Foxconn's single largest customer, contributing roughly 40 percent of group revenue. The company's Zhengzhou campus in Henan Province, sometimes called "iPhone City," remains the world's largest single iPhone assembly site.[4][20] Beyond Apple, Foxconn assembles or has assembled products for Microsoft, Sony, Nintendo, Amazon, Google, Dell, HP, Cisco, BlackBerry, Nokia, Xiaomi, and many others. By the early 2010s, industry estimates credited Foxconn with assembling roughly 40 percent of the world's consumer electronics by volume.[4]
In June 2019, Terry Gou stepped down as chairman to pursue (unsuccessfully) the presidency of Taiwan in the Kuomintang primaries. He was succeeded by Young Liu (劉揚偉), formerly head of Foxconn's S Business Group (semiconductors). Liu has overseen the company's pivot toward AI servers, automotive electronics, and semiconductor packaging, and has accelerated geographic diversification toward India, Mexico, the United States, and Vietnam. In 2023, Foxconn introduced a rotating CEO structure under Liu's chairmanship.[21][22]
In 2017, Foxconn announced a US$10 billion LCD manufacturing project in Mount Pleasant, Wisconsin, promising up to 13,000 jobs. The "Wisconn Valley" project attracted roughly US$3 billion in promised state and local incentives and was championed by President Donald Trump at a 2018 groundbreaking. The plant never reached anywhere near its announced scale. Foxconn renegotiated in 2021, slashing investment commitments to US$672 million and jobs to roughly 1,454 by 2025. Portions of the original 2,400-acre site have since been sold to Microsoft for a separate AI data center project.[23]
| Customer | Products assembled by Foxconn | Notes |
|---|---|---|
| Apple | iPhone, iPad, Mac, Apple AI server hardware | Largest customer, ~40% of revenue |
| Microsoft | Xbox consoles | Long-running console partner |
| Sony | PlayStation 4, PlayStation 5 | Multiple PlayStation generations |
| Nintendo | Switch, Switch 2 consoles | Every home Nintendo console since GameCube |
| Dell | PCs, servers, storage | Top-five customer historically |
| HP | PCs, printers, servers | Top-five customer historically |
| Amazon | Kindle, Echo, AWS servers | Long-running EMS relationship |
| Pixel smartphones, server hardware | Pixel and AI infrastructure | |
| Xiaomi | Smartphones | Major Android customer |
| Cisco | Networking and switching hardware | Networking ODM |
Foxconn Industrial Internet Co., Ltd. (FII), a subsidiary listed on the Shanghai Stock Exchange (601138.SH) since 2018, consolidates much of Foxconn's data center, AI server, networking, and industrial internet business. By 2024, FII reported double-digit revenue growth driven by AI server demand, and analysts at Omdia projected that Foxconn (largely through FII) would surpass Dell as the world's largest server vendor by revenue. By 2025, FII had also overtaken Huawei in cloud and networking revenue.[24][25]
Foxconn has been one of the world's most aggressive deployers of industrial robotics inside its own factories, motivated by labor cost increases in China, recurring criticism over working conditions, and the need to operate at extremely high yield and consistency for premium electronics customers.
Foxbot is the brand name for Foxconn's in-house industrial robotic arms, first deployed in factory production around 2011. In a widely cited 2011 statement, Terry Gou said the company would install one million robots within three years across its mainland China factories. The literal one-million figure proved aspirational. By the mid-2010s, Foxconn had deployed roughly 40,000 fully operational Foxbot industrial robots, alongside hundreds of thousands of pieces of other automated equipment, and has since produced Foxbots in-house at around 10,000 units per year.[26]
Foxconn's automation strategy is typically described in three phases: automate individual workstations performing tasks that are dangerous or unpopular with human workers; automate complete production lines; then automate entire factories with skeleton crews for logistics, testing, and inspection.[26] By 2016, the company reported ten fully automated production lines and that Foxbots had displaced as many as 60,000 workers at its Kunshan plant alone. By the mid-2020s, Foxconn operated multiple "lights-out" factories.[27][26]
Foxconn earned its first World Economic Forum (WEF) Manufacturing Lighthouse designation in 2019 and has since accumulated multiple Lighthouse and Sustainability Lighthouse certifications across China and Vietnam. The Lighthouse program, run jointly by the WEF and McKinsey, recognizes factories exemplifying Industry 4.0 transformation through big data, AI, and IIoT.[28][29] In 2024, Foxconn's Ingrasys subsidiary received the WEF designation as the world's first AI server Lighthouse factory, recognizing its AI-powered automated optical inspection, predictive maintenance, and digital twin tooling. By 2025, Foxconn was operating around eight WEF-recognized Lighthouse factories alongside roughly 80 internal corporate-Lighthouse pilot sites.[30][28]
Foxconn's lights-out factories use private 5G networks, AI-driven optical inspection, machine learning-based predictive maintenance, and self-maintenance systems for production tools. The Shenzhen FII lights-off factory was among the earliest WEF-recognized lights-out electronics sites, and the model has been extended to plants in Chengdu, Zhengzhou, and Vietnam.[31][28]
Foxconn uses NVIDIA Omniverse to build digital twins of its factories, simulating new production lines virtually before physical commissioning. The company has publicly demonstrated Omniverse-based simulation of robot path planning, Isaac Sim training environments for autonomous mobile robots, FoundationPose for object pose estimation, and Metropolis vision AI for quality inspection.[32]
In November 2025, at Hon Hai Tech Day 2025, Foxconn and Alphabet's robotics subsidiary Intrinsic announced a multi-phase joint venture to build a new generation of AI-driven, flexible robotic systems for electronics assembly. The venture targets electronics assembly, inspection, machine tending, and logistics, using Intrinsic Flowstate as the development environment and the Intrinsic Vision Model (IVM) for perception. Initial deployments are intended for U.S. factories, including the Houston AI server site.[33][34]
Foxconn has separately partnered with eBots on embodied AI robots for adaptive precision manufacturing, with one such robot passing factory acceptance tests in early 2025, and with FARobot, whose "Swarm Autonomy" solution enables fleets of autonomous mobile robots to coordinate inside Foxconn's Lighthouse network plants.[35][36]
| Initiative | Type | Year | Notes |
|---|---|---|---|
| Foxbot industrial robotic arms | Articulated industrial robot | 2011 onward | In-house design; tens of thousands deployed |
| WEF Lighthouse factory designation (Shenzhen) | Industry 4.0 lights-off plant | 2019 | First Foxconn Lighthouse certification |
| FARobot Swarm Autonomy | Multi-AMR fleet coordination | 2024 | Deployed across Lighthouse network |
| Ingrasys AI server Lighthouse factory | AI server assembly | 2024 | First AI server Lighthouse worldwide |
| Vietnam lights-out plants | Lights-out final assembly | 2023 | Two plants with WEF Sustainability Lighthouse |
| eBots embodied AI robot | Adaptive precision manufacturing | 2025 | Passed Factory Acceptance Test |
| Intrinsic JV | Flexible robotic assembly | 2025 announcement | Multi-phase US-focused JV |
| NVIDIA Isaac GR00T humanoids (Houston) | Humanoid robot deployment | 2026 (planned) | Q1 2026 production deployment |
Foxconn's most prominent foray into humanoid robots is its co-development program with NVIDIA, centered on NVIDIA's Isaac GR00T general-purpose humanoid foundation model and the Jetson Thor robotics compute platform. The program targets Foxconn's own production lines as the first deployment environment.[15][16]
Foxconn first publicly previewed humanoid robot ambitions at Hon Hai Tech Day 2024 (HHTD24) in October 2024, positioning humanoid robots as a flagship pillar of its smart manufacturing platform. At Hon Hai Tech Day 2025 (HHTD25) in November 2025, Foxconn unveiled a bipedal humanoid prototype with a dexterous robot hand, jointly developed with NVIDIA. The robot was shown statically with a focus on hand dexterity, including pick-and-place, screw fastening, and small-part handling. Foxconn also indicated plans for a second, lower-cost variant using a wheeled mobile base.[37][38]
In October 2025, at NVIDIA's GTC keynote in Washington, D.C., Foxconn confirmed humanoid robots would be deployed on its production lines at the Houston AI server plant. The robots will run NVIDIA's Isaac GR00T N foundation model on Jetson Thor compute, with on-board perception via NVIDIA FoundationPose, and were trained in NVIDIA Isaac Sim and Omniverse. Initial deployment was set for the first quarter of 2026, marking the first known use of humanoid robots in Foxconn's production lines in the company's 50-year history.[15][16] Foxconn has signaled plans to expand humanoid deployment to additional U.S. AI server sites in Texas, Wisconsin, and California, alongside flexible robotic arms developed with Intrinsic, as part of chairman Young Liu's long-term "AI factories" vision.[15][33]
| Event / partner | Date | Highlights |
|---|---|---|
| Hon Hai Tech Day 2024 (HHTD24) | October 2024 | First public humanoid roadmap; smart manufacturing platform anchored by NVIDIA Isaac |
| NVIDIA GTC Washington keynote | October 2025 | Foxconn-NVIDIA humanoid co-development announced; Houston plant identified as first deployment site |
| Hon Hai Tech Day 2025 (HHTD25) | November 2025 | Bipedal humanoid prototype unveiled with dexterous hand; second wheeled variant disclosed |
| Houston AI server plant rollout | Q1 2026 (planned) | First known humanoid robots on Foxconn production lines |
| NVIDIA GTC (post-HHTD25) | Following HHTD25 | Foxconn humanoids and Vera Rubin NVL72 modular data center showcased |
Under Young Liu, Foxconn has expanded into semiconductors, including IC design, foundry partnerships, and advanced packaging through the S Business Group. Foxconn has established a strategic alliance with Japanese partners on panel-level fan-out (FOPLP) packaging and announced a EUR 250 million investment in 2025 to build Europe's first fan-out wafer-level packaging (FOWLP) facility through a joint venture in France with Thales and Radiall, intended to serve aerospace, defense, telecommunications, and AI workloads. Foxconn has also indicated interest in advanced packaging demand for AI accelerators and HBM-related applications.[39][40]
Foxconn launched the MIH (Mobility in Harmony) open-source electric vehicle platform in October 2020 as the foundation for a broad EV manufacturing services strategy. The MIH Consortium has attracted approximately 2,300 partners worldwide.[41] In November 2021, Foxconn agreed to acquire the Lordstown, Ohio assembly plant from Lordstown Motors. In May 2022, the two companies closed an asset purchase agreement and entered a joint venture, MIH EV Design LLC (55% Foxconn, 45% Lordstown), to co-develop EV programs. Lordstown Motors subsequently went bankrupt, but the Ohio plant has continued to operate under Foxconn ownership as a hub for North American EV programs.[42][43] The MIH platform competes in part with the contract-EV strategies pursued by Tesla suppliers and other automotive contract manufacturers.
NVIDIA is Foxconn's most strategically important partner across the AI stack: AI server manufacturing as a GB200 NVL72 and GB300 NVL72 PBR Partner with the world's largest GB200 production line in Guadalajara, Mexico; industrial AI through NVIDIA Omniverse, Isaac Sim, Isaac AMR, Metropolis, and FoundationPose across Foxconn factories; humanoid robotics co-development running Isaac GR00T on Jetson Thor; joint work on an 800 VDC power architecture for next-generation AI factories; internal use of NVIDIA RTX PRO Servers across Foxconn operations; and the GB200 NVL72-based Hon Hai Kaohsiung Super Computing Center.[7][13][32][15][44][45][18]
The November 2025 multi-phase joint venture with Alphabet's Intrinsic targets flexible, AI-driven robotics for electronics assembly, inspection, machine tending, and logistics, using Intrinsic Flowstate and the Intrinsic Vision Model.[33][34] In 2025, Foxconn also announced a strategic partnership with SAP to accelerate AI-powered manufacturing and supply chain operations using SAP Business AI.[46] Other smart manufacturing collaborators include Siemens, GE Vernova, Breakthrough Energy, Thales Alenia Space, and Microsoft, several of which have appeared as Hon Hai Tech Day keynote partners.[11]
Foxconn has been one of the most aggressive multinational manufacturers in pursuing diversification away from concentration in mainland China. The company's 2024 and 2025 disposal-of-capital announcements consolidated roughly US$840 million of new commitments across the United States, Mexico, India, and Europe, with subsequent Taiwan Ministry of Economic Affairs approvals covering additional US$2.2 billion-plus projects in the United States and India.[47][48]
In India, Foxconn invested heavily in 2024 and 2025 in iPhone manufacturing in Tamil Nadu and Karnataka, smartphone component fabs in Uttar Pradesh, and a US$1.49 billion smartphone components facility through Yuzhan Technology (India), reinforcing its role in Apple's supply chain rebalancing.[49][48] In Mexico, FII AMC MEXICO scaled significantly in 2024 and 2025 to host the world's largest NVIDIA GB200 production line in Guadalajara, secured by a US$241 million equity acquisition.[13][47] In the United States, Foxconn revitalized its Ohio (Lordstown) site for cloud and networking, established Project ETA (DE) LLC with US$735 million for data center module production, and committed to AI server expansion in Texas (Houston), Wisconsin, and California.[47][15] Foxconn has also expanded in northern Vietnam with multiple Lighthouse and Sustainability Lighthouse-certified factories, and maintains long-running operations in the Czech Republic, Brazil, and Slovakia.[4][28]
Foxconn's revenue has compounded steadily since the early 2000s, driven by the iPhone era and, more recently, the AI server boom. Selected milestones:
| Year | Approx. revenue | Notable context |
|---|---|---|
| 2010 | NT$2.99 trillion | Early iPhone-era ramp |
| 2018 | NT$5.29 trillion | Pre-pandemic peak |
| 2023 | NT$6.16 trillion | Post-pandemic adjustment |
| 2024 | ~NT$6.86 trillion (~US$208 billion) | Record year; +11.4% YoY; AI server inflection |
| 2025 (FY) | ~US$252 billion (consolidated, +18% YoY) | Cloud and networking >40% of revenue |
| Q1 2026 | ~US$67 billion (+29.7% YoY) | Record quarter on AI server demand |
Full-year 2024 net profit reached NT$152.7 billion, equivalent to earnings per share of NT$11.01, the highest in 17 years. The company announced a record cash dividend of NT$5.80 per share, the largest since its 1991 listing on the Taiwan Stock Exchange.[9][50] Through the AI server boom, Foxconn's cloud and networking products division became its largest segment, accounting for more than 40 percent of revenue by 2025, while quarterly operating profit rose sharply on improving AI server scale and margins.[3][2]
Foxconn's global workforce typically ranges between roughly 750,000 and 900,000 employees during peak production seasons. The company operates more than 230 campuses across approximately 24 countries, with the heaviest concentration in mainland China (Shenzhen, Zhengzhou, Chengdu, Chongqing, and Kunshan). Foxconn ranked 32nd on the Fortune Global 500 in 2024.[51][52][5]
Foxconn has faced extensive scrutiny over working conditions in its China factories, notably after a series of worker suicides at the Shenzhen Longhua complex in 2010, which prompted the installation of safety nets, wage increases, and counseling programs. Subsequent reporting has documented continued concerns over overtime, dormitory conditions, and discipline at peak production periods, including during pandemic-era lockdowns at Zhengzhou in late 2022.[4]
The Wisconsin Mount Pleasant project, originally promoted as a US$10 billion LCD facility, has been widely characterized as a failed industrial promise that produced only a fraction of the announced investment and employment.[23]