Foxconn Group: Control by Nodes, Control by Ownership

First, Bust a Myth: Foxconn Is Not Decentralized — It Is Extremely Centralized

Talking about Foxconn's M&A, many people instinctively reach for the "holding company, autonomous subsidiaries" picture of decentralization. That is wrong — and crucially so. Grasp this point and everything that follows makes sense.

Part 2 of this series, Amphenol, has 130-plus business units, each run like an independent company; Part 3, Constellation, delegates even sub-US$20-million acquisitions to its operating groups, with headquarters acting only as a "coach." That is genuine decentralization.

Foxconn is their opposite. In the Terry Gou era, control was highly personal and militaristic. Gou's famous lines: "Democracy is the least efficient way to get things done," and "Outside the lab there is no high tech, only the discipline of execution." The most famous episode is the firing of veteran lieutenant Hsieh Kuan-hung — according to multiple media reports and a subsequent court ruling, when Hsieh's secretary mis-entered his vacation dates and he was already on a plane (flying to Japan to handle his daughter's school enrollment), Gou ordered him to deplane and return for a meeting; with the cabin door already closed, Hsieh couldn't comply, and Gou announced his dismissal on the spot in the meeting (later ruled an unlawful dismissal). An empire across dozens of countries with over a million employees, whose chairman can fire a business-group general on the spot over one meeting — that does not happen in a decentralized group.

After Young Liu took over in 2019, the "vehicle" of control shifted from the individual to the institution, but its intensity did not loosen. He set up a nine-member operating committee as the decision core (in his own words, "the operating committee leads Foxconn Group"); during the pandemic he personally chaired daily video meetings with hundreds of managers worldwide and allocated resources through a central platform. From 2024 he introduced a rotating-CEO system (the six business-group heads rotate every six months), but major decisions are still settled by the chairman. In other words, Foxconn evolved from "one person's centralization" into "a system's centralization" — central control over the business groups was never given up.

The Visible Surface: Acquisitions Were Never About "Diversification," but "Irreplaceability"

Let's pin down the surface first, then dig deeper. Foxconn's M&A motive is utterly different from the subjects of this series' first four parts.

Berkshire buys "moat assets that need no fixing," Constellation buys "niche software for permanent rent," Danaher buys "industrial companies that get better after transformation" — for them, the target itself is the goal. Not for Foxconn. When Foxconn buys a company, the target is never the goal; "making the core business more irreplaceable" is.

It engineered Innolux's three-way merger in 2010; took ~66% of Sharp for ¥388.8 billion in 2016 (locking in Apple panels, squeezing Samsung); FIT Hon Teng acquired America's Belkin in 2018 (~US$866m, moving from connector parts into consumer brands and channels); it bought Macronix's 6-inch fab in the Hsinchu Science Park in 2021 (entering silicon-carbide wide-bandgap semiconductors); and from 2020 it formed Foxtron with Yulon to enter EVs. Every step locks down, along the value chain, a node that makes the core harder to bypass. This is vertical-extension M&A — its trust is "strategic trust": efficient when the division of labor is clear (technology to the target, scale and capital to the parent), but it bets that "the node it locks down will stay important."

Yet this elegant surface narrative only answers "what Foxconn buys." What truly decides the fate of original shareholders are the next three questions: How does it hold? How does it spin off? And how much does it disclose to you?

Control Mechanic One: The Maze of Cross-Ownership and Holding Platforms

Foxconn does not hold its subsidiaries 100% directly through the parent (2317). Much of its ownership runs through layer upon layer of investment holding platforms — recurring names include Hung Yang Venture Capital, Pao Hsin International Investment, Hua Chun Investment, and Century Technology (Shenzhen Chaoda).

A few publicly compiled examples: Ennoconn is held about 30% via Pao Hsin International; Foxconn Technology (鴻準) and Foxsemicon (京鼎) are held at single-digit percentages via Hung Yang Venture Capital; and in panel maker Innolux, the Foxconn camp's stake is deliberately spread across Gou personally, Foxconn, Foxconn Technology and several VC arms — totaling about 7.7%, with no single entity in absolute control. More subtly, Foxconn Technology is itself a Foxconn holding target yet in turn holds Innolux — subsidiaries holding one another, forming a web outsiders struggle to see through at a glance.

Control Mechanic Two: "One Chicken, Many Dishes" — Spin Off, but Never Let Go

Foxconn's most-watched capital-market maneuver is spinning subsidiaries off to list in different markets — colloquially, "one chicken, many dishes."

The textbook case is Foxconn Industrial Internet (FII). In 2018, Gou packaged the group's communications-network equipment, cloud-service equipment and industrial robotics and floated them on the Shanghai A-share market — clearing review in just 36 days, the fastest IPO on record at the time, raising about RMB 27.1 billion. It jumped ~44% on day one to a market cap near RMB 390 billion, briefly surpassing parent Foxconn to become the highest-valued tech stock on the Shanghai exchange.

See the pattern? Spin off, yet every stake stays far above the control threshold. That is the essence of "one chicken, many dishes" — letting subsidiaries raise capital independently and earn local valuations, while keeping an iron grip on control through absolute majority ownership. For the parent, it is textbook capital engineering.

But for original shareholders, there is another side. FII broke below its issue price not long after listing, stayed below its offering price for a long stretch, and saw its P/E fall to about 11x at one point — far below the ~39x of comparable Luxshare Precision. The rosy promise of "spin-offs unlocking value" did not keep paying off. The spin-off moved assets into another market, but whether the premium truly flowed back to the parent's original shareholders is a question mark (more on this in Section 8).

Seven-Plus Listed Companies: Division of Labor, or Overlap?

Foxconn is not one company but a "federation of sub-groups." Besides the parent (2317.TW), the group has multiple independently listed companies across Taiwan, Hong Kong and mainland China. In theory each guards a slice of the value chain; in practice, overlap and internal competition are not rare.

Beyond these, the group's map also includes panel maker Innolux, PCB maker Zhen Ding (and its A-share spin-off Avary Holding), touch-panel maker GIS, plus Portwell, Foxsemicon, Cosmos, MTI and other listed companies. Within the display chain (Innolux → Sharp → GIS) there is both vertical division of labor and internal competition.

How Much Is Skill, How Much Is Capital-Market Engineering? The Related-Party-Transaction Question

This is the question most worth asking — and most in need of care: of Foxconn Group's growth and profit, how much comes from genuine skill against outside customers, and how much from intra-group buying and selling in service of the capital markets?

Look at FII's public figures. Its net margin has long been only about 4%–5%, direct material costs run around 90% of revenue, R&D spend has long been under 2% of revenue, and its top five customers (unnamed) together account for roughly 60%–70% of revenue. Chinese financial media drew blunt conclusions: a piece syndicated on Tencent News called it "a trillion-cap 'assembly plant'" and "one of the most overvalued names on the A-share market"; Wallstreetcn questioned that it "isn't short of cash — listing was more likely to reward financial investors."

Another fact worth noting: in October 2023, Chinese tax and natural-resource authorities launched tax audits and land-use inquiries into Foxconn's enterprises across several provinces, and on the news FII fell limit-down that day. This is a regulatory fact, but the authorities published no conclusive characterization — we note it, but do not conflate it with the related-party-transaction suspicion above.

The Financial Engineering of M&A: Sharp's Price Cut, and Terry Gou's "Personal Stake"

Vertical-extension acquisitions often feature striking financial engineering. The Sharp deal is a double case study.

Layer one: the hard skill of cutting the price. Foxconn's originally agreed terms were about ¥488.8 billion, ¥118 per share; but right before signing, Sharp disclosed a list of contingent liabilities of about ¥350 billion. Foxconn immediately hit pause and reopened due diligence, leaving Sharp waiting about a month, and ultimately cut the price to ¥88 per share (around 75% of the original) — taking ~66% for ¥388.8 billion (¥288.8bn in common shares + ~¥100bn in preferred shares). That is elegant financial discipline.

Layer two is an interest structure worth noting neutrally. According to multiple media reports, Terry Gou had become a major shareholder of Sharp's Sakai plant (SDP) in his personal capacity as early as 2012; the 66% taken in the 2016 Sharp acquisition comprised Foxconn, Foxconn's Cayman and Singapore subsidiaries, plus Gou's personal 8.45%. That is, the founder personally and the listed company held the same asset at the same time.

The Transparency Deficit Toward Original Shareholders: It Demands Your Trust, Yet May Not Tell You the Whole Truth

Pull the threads together and the series' sharpest contrast emerges: Foxconn talks strategic trust, division of labor, and empowerment to the companies it acquires; but toward its own original shareholders, its disclosure and transparency are repeatedly called insufficient.

At the 2018 shareholder meeting, foreign institutional figure Chou Shang-yi challenged Foxconn's opaque financial structure to its face, asking it to disclose the operations and transformation KPIs of each business and division — the foreign side even flew from London to Taipei to make its case. Gou's reply was candid, and telling: "We don't disclose — that isn't dishonesty; it's that this data is confidential, it's big data, and one calculation would reveal the connections." At the same meeting, financial media compiled "the 5 key questions Terry Gou didn't answer," including transformation KPIs, succession plans, and the heavy institutional dumping of FII shares after its listing — most of them sidestepped.

The institutional signals align: in the latest Taiwan corporate-governance evaluation, TSMC and others held a top-5% spot for a 12th straight time, while Foxconn did not make the top-5% list. As noted, all of Foxconn's directors and supervisors hold only about 0.02%, compliant only via an exemption clause — fully legal in form, but "formal compliance" and "substantive transparency" have never been the same thing.

The FII Spin-off: Who Got the Juiciest Cut?

The most concrete transparency dispute centers on the FII spin-off. Per CM Media, what went into FII were 2017's orders worth over US$10 billion from 13 major customers, of which Apple orders made up about 73% — meaning Foxconn carved out a big slice of "the diamond in the crown," its Apple orders, and packed them into this subsidiary headed for a China listing. That report cited an (unnamed) investment expert who likened FII's underwriting at a low P/E on the A-share market to "handing money to Chinese retail investors"; a CommonWealth Magazine headline asked even more directly, "Do Foxconn's minority shareholders get any benefit?"

To be fair: Foxconn keeps a 84%+ stake, FII's profit and loss still flow back to the parent's books via equity-method accounting, and dividends return per the holding ratio (FII and Avary Holding have paid hundreds of millions of RMB in cumulative dividends since listing, of which Foxconn as major shareholder receives a sizable share). So one cannot flatly conclude "minority shareholders got nothing." But the issue is this: when the most core customer orders are moved into a subsidiary listed in another market under another valuation, the "Foxconn" held by the parent's original shareholders is in essence diluted into a thinner broth — and the gains and losses of that process were never fully and transparently explained to them.

Four Trust Models Compared: Foxconn Group — Transparent to Whom?

The row to underline is the last one. The trust of the first three is two-way: they make promises to the acquired, and they disclose honestly to their own shareholders. Foxconn's trust is more one-way — it asks the market to trust its execution and integration ability, but what it is willing to disclose to original shareholders is often less than the trust it demands. This is not to say it breaks the law (its disclosure is mostly lawful and compliant), but that between "formal compliance" and "substantive transparency," Foxconn has chosen a position that keeps outside shareholders perpetually guessing.

The Taiwan Lesson: Strategic Trust Should Not Only Face Outward — It Should Also Face Inward

Foxconn is Taiwan's most daring and largest M&A player, beyond dispute. It extended "supply-chain thinking" into a complete grammar of acquisition, executed at world-class level. But this article's purpose is not to praise — it is to honestly lay out the empire's other side, because this is precisely the shared challenge of many of Taiwan's family/founder-led groups.

Why Does Sharp Deserve a Piece of Its Own?

By now you've understood the "control mechanics" of Foxconn's acquisition empire. But what happens when this grammar meets a Japanese company with a century of pride, its own union and craftsman culture? How is strategic trust built — and how does it unravel inch by inch after the panel node devalues, until even Tai Jeng-wu, the man who once rescued Sharp, ends up in court against Terry Gou?

That is a complete story of "how strategic trust is built, and how it collapses" — worthy of a whole piece. See you in Part 6.