Innolux Corporation: Taiwan's Panel Giant Bets the Fab on Chips and Cockpits

I. Introduction & Roadmap

Somewhere in the Tainan Science Park, a building that spent fifteen years turning sheets of glass into laptop screens is being emptied out. The clean-room gowns are gone. The severance packages have been posted on the noticeboard — three phases, starting in June, with twenty-year veterans entitled to roughly nine months of combined severance and notice pay.¹ The building itself has been sold to a semiconductor packaging company for NT$14.85 billion, and the accounting gain on that sale — about NT$13.3 billion — will do more for its former owner's 2026 earnings per share than almost anything the factory produced in its final years.¹

This is Innolux Corporation, 群創光電. Ticker 3481 on the Taiwan Stock Exchange. Roughly 45,000 employees, NT$226.7 billion of revenue in 2025, and a piece of glass in an enormous number of the car dashboards, monitors, notebooks and medical scanners that pass through ordinary life.²[^3] Most people outside Taiwan have never heard of it. Almost everyone has looked at one of its products today.

Here is the tension that makes the story worth telling. For most of the 2020s, Innolux has been very good at making things almost nobody wants to pay a profit for. Its operating line — the part of the income statement that measures whether the core business earns money — lost NT$7.9 billion in 2024 and NT$4.2 billion in 2025.[^3] The company reported a profit in both years anyway, because the non-operating line, stuffed with asset disposals and investment gains, more than covered the hole.[^3] Now management is doing the logical thing: selling the factories that made it famous, and buying its way into two businesses — semiconductor advanced packaging and automotive smart cockpits — where it hopes the economics are different.

The market has decided this is working. Innolux shares rose roughly 345% over the trailing twelve months into August 2026, re-rating from well below book value to about 1.8 times book.² That is a very large vote of confidence in a transformation that, by management's own account on its most recent calls, has barely begun to show up in revenue.

So the roadmap. We start with a merger born of the 2009 financial crisis that created scale and, according to Taiwanese business press, a culture clash that Foxconn's founder later came to regret. We pass through one of the largest price-fixing prosecutions in US antitrust history — because it explains something durable about how this industry behaves. We look hard at the industry structure that made LCD a value trap, including the moment when Korea's two best panel makers concluded the segment wasn't worth defending. We meet a chairman with a Goldman Sachs résumé and no engineering background who has just consolidated more control over the board than any Innolux leader before him. We follow the money into a century-old Japanese audio brand and out of three Tainan fabs. And then we test the case — why this works from here, and what would break it.

↑ Back to Top

II. Origins: From Captive Supplier to Merger Behemoth (2003–2012)

On November 15, 2009, in a Taipei conference room, two men who had spent their careers on opposite ends of Taiwanese industry shook hands over a NT$199.7 billion share swap.³ One was 郭台銘 Terry Gou, the relentless founder of 鴻海精密 Hon Hai Precision Industry — Foxconn — a man whose reputation was built on squeezing cost out of assembly lines. The other was 許文龍 Hsu Wen-lung, founder of the 奇美集團 Chi Mei Group, a Tainan industrialist famous for playing the violin, collecting art, letting his workers go home at five, and running a company culture that was as close to gentle as Taiwanese heavy industry got.

Hsu called the deal "the most difficult decision in my life."³ Gou called the ambition plainly: "We would like to be in the world's top three."³

Rewind six years. Innolux Display Corporation had been incorporated in 2003 inside the Foxconn orbit, headquartered in Miaoli, for a purpose that was less about glory than about supply chain arithmetic.⁴ Foxconn assembled an enormous share of the world's consumer electronics. Displays were among the most expensive components in a laptop or a monitor, and they were bought from third parties at prices Foxconn did not control. Building a captive panel maker was classic vertical integration: internalize the margin, secure the supply, and gain a seat at the table when panel prices spiked. Innolux listed on the Taiwan Stock Exchange on October 24, 2006.⁴

Then came 2008–2009. The financial crisis crushed panel demand, and the global LCD industry — which had spent the prior decade in an arms race of ever-larger fabs — found itself with far more capacity than the world wanted. Consolidation became the survival strategy. In October 2009, Innolux announced a roughly NT$20 billion share-swap merger with TPO Displays, a specialist in small and mid-sized panels.³ Six weeks later came the far bigger prize: Chi Mei Optoelectronics, Taiwan's number two panel maker and a genuine powerhouse in large TV panels, at a swap ratio of one Innolux share for 2.05 Chi Mei shares — a 22% premium to Chi Mei holders.³

The arithmetic was seductive. The combined entity would command roughly 17% of the global panel market, edging past AU Optronics at 16%, and vault into the world's top three.³ The mergers completed in March 2010, the company renamed itself Chimei Innolux, and in December 2012 simplified to Innolux Corporation.⁴

Then the arithmetic met the org chart.

The merged company operated, in practice, as two companies wearing one badge: the Foxconn-aligned northern plants around Miaoli and Hsinchu, and the Chi Mei southern operation in Tainan. According to a 2016 investigation by 鏡週刊 Mirror Media, the Tainan facilities kept Chi Mei's original benefits — including the famous five o'clock departure — while the northern sites ran on Foxconn time.⁵ The first three years after the merger produced cumulative losses of NT$108.5 billion.⁵ A later recovery clawed back NT$37.6 billion, and by 2016 the company was losing money again.⁵

The governance was the deeper problem. Gou had promised Hsu he would not interfere with Chi Mei's operations — a gesture of respect to a founder he admired. But Chi Mei retained only a 5.74% stake, and Foxconn's own holding was not controlling either, which meant that when Gou wanted to change something as basic as procurement, he could not.⁵ In 2014, a conflict between Gou and then-chairman 段行建 Tuan Hsing-chien over the Kaohsiung plant escalated to the point where five vice presidents sided with Tuan, and Gou backed down.⁵ Mirror Media framed the whole episode as a lasting regret for Gou — and noted that when he later acquired Japan's Sharp, he took six of nine board seats and imposed Foxconn management practices from day one.⁵

That is the founding trade-off Innolux still carries. It bought scale in a business where scale turned out not to be a moat, and it paid for that scale in integration debt that took more than a decade to work through. Keep the lesson in mind: it will reappear, almost verbatim, when we get to a Japanese acquisition in 2025.

Scale, though, was not the only thing the Taiwanese panel industry had been buying in those years. It had also been buying something rather less legal.

↑ Back to Top

III. The Crystal Meetings: LCD's Global Price-Fixing Cartel

They were called, without much irony, the Crystal Meetings.

For roughly a decade beginning in the mid-1990s, senior executives from Taiwanese and Korean LCD makers met in Taipei hotel rooms, restaurants, and karaoke bars to do the one thing that a commodity industry with no pricing power desperately wants to do: agree on prices.⁶ They compared production plans. They set price floors for specific panel sizes sold to specific customers. They did this while their companies publicly described a fiercely competitive market.

The US Department of Justice eventually brought one of the largest cartel prosecutions in American antitrust history. AU Optronics and two of its executives were criminally convicted after an eight-week trial in San Francisco, and in September 2012 the company was sentenced to a $500 million criminal fine — at the time matching the largest antitrust fine ever imposed on a corporation in the United States.⁶ Former AUO president Hsuan Bin Chen and former executive vice president Hui Hsiung each received three years in prison and $200,000 fines.⁶ LG Display had already pleaded guilty in 2008 and paid $400 million.⁶

Innolux's own entanglement came through the entity it had just absorbed. Chi Mei Optoelectronics had been a participant, and the newly merged Chimei Innolux inherited the liability. In December 2011, California Attorney General Kamala Harris announced a $553 million nationwide settlement resolving state and class claims over LCD price fixing covering conduct from 1999 to 2006; Chimei Innolux Corporation and Chi Mei Optoelectronics USA were among the settling defendants, alongside Samsung Electronics, Sharp, Hitachi and HannStar.⁷ Eight state attorneys general coordinated the settlement with a national class action, and claims opened to consumers in February 2012.⁷ A separate direct-purchaser class settlement of roughly $405 million ran in parallel.

Why spend six minutes of a business story on a cartel that ended two decades ago? Because it is the cleanest available evidence about the nature of this industry, and it is evidence that no management presentation will ever give you.

Cartels form where the economics are unbearable. Firms with genuine pricing power — a patented drug, a dominant operating system, a beloved brand — do not need to meet in karaoke bars. LCD makers did, because the underlying business had a structural defect: enormous fixed costs, multi-year capacity lead times, and a product that buyers could switch between suppliers almost costlessly. When demand softened, every player's incentive was to keep the fab running at full tilt and cut price, because marginal cash contribution beat idle depreciation. The collective result was value destruction. "Industry discipline" in LCD has historically meant collusion, not rational capacity planning.

That matters today for a very practical reason. When you hear a panel executive in 2026 talk about the industry finally behaving rationally — coordinated maintenance shutdowns, utilization discipline, supply cuts to support prices — you should ask what mechanism enforces that discipline, and whether it survives the first quarter in which one player decides to grab share. In an industry whose historical answer to that question was a hotel room in Taipei, and whose current largest players answer to a different set of incentives entirely, the honest answer is: not much.

Which brings us to the shape of the industry Innolux actually competes in today.

↑ Back to Top

IV. Industry Structure: The Global LCD Supercycle

Picture the capital cycle in slow motion. A panel maker decides in year one to build a new-generation fab. Construction and equipment take two to three years and consume several billion dollars. The fab opens in year four — into whatever demand environment happens to exist then, which nobody could forecast in year one. Because the capital is sunk and depreciation runs whether the line moves or not, the fab runs. Prices fall. Everyone's margins compress. Two or three years later, capacity finally gets rationalized, prices spike, and somebody decides to build a new fab.

That is the LCD supercycle, and it has run more or less continuously since the 1990s. It is a textbook capital-intensive commodity: multi-year lead times, high operating leverage, and — in the base case for TV and IT panels — essentially no durable pricing power. The interesting question is never whether the cycle turns. It is who is left standing when it does.

The Korean verdict

The most instructive datapoint in the modern history of this industry is not a Chinese expansion. It is a Korean exit.

Samsung Display, arguably the most technically capable display manufacturer ever built, fully exited LCD production in June 2022 and sold its remaining LCD patent portfolio to TCL CSOT. LG Display, its equally formidable domestic rival, took longer but reached the same conclusion: in September 2024 it agreed to sell its 80% stake in the Guangzhou large-LCD plant to TCL CSOT for 10.8 billion yuan, about $1.5 billion, with closing in March 2025.[^9] That transaction ended LG Display's presence in large LCD entirely; it had already stopped LCD TV panel production at home by the end of 2022, and now retains only premium LCD for IT and automotive, with proceeds redirected into OLED.[^9]

Read that carefully. Two companies with world-class process technology, deep capital, and every incentive to defend a business they invented looked at the returns available in large LCD and walked away. That is not a cyclical judgment. It is a structural one.

The Chinese flood

What replaced them is a supply base with a different cost of capital. 京东方 BOE, TCL CSOT and HKC — "China's big three" — now anchor a mainland Chinese industry that controls more than 70% of global LCD production capacity, a position built on a decade of state industrial policy, provincial subsidies, and acquisitions like the Guangzhou purchase.[^10] By some measures China accounts for roughly three-quarters of global display production capacity by area.[^10]

This is the single most important fact in the Innolux investment case, and it is worth stating in plain language. When your largest competitors are willing to build capacity that does not earn its cost of capital, because the policy objective is domestic supply-chain control and employment rather than return on invested capital, then your own scale is not a defense. You cannot out-invest a balance sheet that does not require a return. The only responses available are to exit, to specialize, or to lose money slowly.

The Taiwan duo

Taiwan's two survivors, Innolux and AU Optronics, have chosen versions of specialize-and-exit. AUO is the larger by revenue — NT$280.25 billion in 2024 against Innolux's NT$216.5 billion — though Innolux carries a comparable workforce and asset base.⁸[^3] The two have periodically been the subject of merger speculation going back more than a decade, and it has never happened.

What is striking in 2026 is that they are de-scaling in parallel rather than consolidating. Innolux has been selling Tainan fabs into the semiconductor packaging boom and closing its notebook-panel line; AUO sold its Taoyuan Huaya plant to Quanta and its Kaohsiung C5E colour-filter facility to ASMedia, with Huaya slated for closure in the first half of 2027.⁹ TrendForce's read of the combined effect: Taiwan's share of global LCD monitor panel supply falls below 10%, notebook panel share to roughly 20%, with TV panels holding near 22% because that segment still pays.⁹ Innolux is pointing its exit proceeds at semiconductor packaging; AUO at co-packaged optics and Micro LED. Same diagnosis, different prescriptions, no merger.

Five forces, honestly applied

Run the standard framework and the picture is bleak in the core business. Rivalry is brutal and structurally subsidized. Buyer power is high: TV brands and PC OEMs are concentrated, sophisticated, and face near-zero switching costs on commodity panels, which is exactly why panel ASPs move with supply rather than with anything the supplier does. Supplier power is modest — glass, driver ICs and polarizers have their own concentration, but they are not the binding constraint. Barriers to entry, in the specific sense that matters, are low: they are not technical, they are financial, and Chinese provincial governments have repeatedly demonstrated a willingness to clear them. And the substitute threat is real — OLED continues to take the premium end of TV, phone and increasingly IT displays, shrinking the addressable high-value segment of the LCD market from above.

Where does a 7 Powers lens find anything real? Not in commodity TV and IT panels. There is no cornered resource, no counter-positioning, no branding, and the scale economies that would normally be Innolux's defense are precisely what a state-backed rival can replicate.

The one place the framework does find something is automotive and industrial displays. A dashboard cluster is not a monitor panel. It has to survive fifteen years of temperature cycling, meet functional-safety qualification, be readable in direct sun, and it is designed into a vehicle platform three to five years before that vehicle reaches a showroom. Once a panel is qualified into a program, replacing it means re-qualifying — expensive, slow, and risky for the automaker. That produces genuine switching costs and a form of process power built on accumulated qualification know-how. It is not a wide moat. Automotive panels are still panels, and Chinese suppliers are qualifying into programs too. But it is the only part of Innolux's business where the customer has a reason to care who makes the glass.

That distinction — commodity versus everything else — is exactly how Innolux now organizes its own financial disclosure. Which is where the numbers get interesting.

↑ Back to Top

V. Inside the Business Today: Segments and Where the Money Actually Is

There is a single slide in Innolux's March 2026 investor deck that tells you more about this company than the entire rest of the presentation. It splits revenue into three buckets — commodity, non-commodity, non-display — and puts a gross margin next to each.

For full-year 2025: commodity products (TV, IT, tablets, mobile phones) were 55% of revenue at an average gross margin below 5%. Non-commodity commercial products were 16% of revenue at 16–20% gross margin. Non-display — X-ray detectors, automotive, and the fledgling packaging business — was 29% of revenue at 11–15%.[^3]

Sit with that for a moment. More than half of Innolux's revenue earns a gross margin that, after operating expenses running around 10% of sales, cannot possibly cover its own overhead.[^3] The company's entire margin structure depends on the 45% of revenue that isn't commodity display. And that commodity margin has been deteriorating: in the first half of 2025 the deck showed commodity gross margin at 6–10%; by the full year it had slipped below 5%.[^13][^3]

The mix shift, at least, is real and fast. Commodity revenue share fell from 70% in the second quarter of 2024 to 50% by the fourth quarter of 2025.[^13][^3] Non-display rose from 22% to 35% over the same window.[^13][^3] By the first quarter of 2026, non-display products were 44% of sales against 56% for display.¹⁰ Two years is a short time to move roughly twenty points of revenue mix in a heavy-manufacturing business, and it deserves credit as execution.

The reason it can move that fast, though, is worth naming: the numerator is growing, but the denominator is also being deliberately shrunk. Closing notebook lines and selling fabs mechanically raises the non-display percentage without any new customer buying anything. Both things are happening at once. Mix improvement driven by growth and mix improvement driven by amputation look identical in a pie chart and are not remotely the same thing for a shareholder.

Automotive: strong, but not a duopoly

Automotive display is the segment Innolux talks about most, and it is a genuine strength — the company is consistently named among the leading global suppliers alongside LG Display, Samsung Display, BOE and AUO.¹¹

It is also a segment where claims should be handled carefully. Independent market research puts the top five automotive display players at a combined 38–45% of global revenue in a market worth roughly $13.6 billion in 2025 — and that top-five group includes vertically integrated module specialists and sits alongside automotive tier-one suppliers like Continental, Bosch, Denso and Visteon who own the customer relationship.¹¹ A market where the top five share less than half the revenue is moderately concentrated, not a duopoly. Company and press claims of outright category leadership are claims requiring evidence, not established facts. What is well supported is that Innolux is a significant player in a growing, higher-margin category — and that the category is going premium, with LTPS TFT and OLED forecast to exceed half of automotive display revenue in 2025, which is a technology mix where Innolux is not the natural leader.¹¹

FOPLP: the fourth leg, still very small

The newest business is the hardest to explain and the easiest to over-hype, so it is worth doing properly.

When a chip comes off a wafer, it has to be packaged — connected to the outside world and protected. Traditionally that happens on the round 12-inch silicon wafer itself. Fan-out panel-level packaging does the same job on a large rectangular glass panel instead. The advantage is geometry: circles waste area at the edges, rectangles don't. Innolux's own figures put a 620x750mm panel at 6.6 times the usable area of a 12-inch wafer, with meaningfully lower cost per unit — up to roughly 25% cheaper for smaller chips.[^3] And the equipment that patterns fine lines onto big sheets of glass is, conveniently, the equipment Innolux already owns from making LCD panels.

That is the strategic logic, and it is elegant: a dying asset base repurposed for a booming end market. Innolux is pursuing three process families — chip-first (already in mass production, aimed at RF chips, power management ICs and automotive radar), RDL-first (finer patterns for AI and HPC chiplets), and TGV, or through-glass via, which drills holes through a glass substrate for large AI and HPC processors.[^3][^13] The company also secured a validation win: reports in late 2025 indicated Innolux won an order to package radio-frequency chips for SpaceX's low-Earth-orbit satellites, with the related capacity running full.¹²

Now the discipline. On the March 2026 investor call, management stated plainly that cumulative panel-level packaging revenue amounted to only hundreds of millions of NT dollars against a company with a NT$200-billion-plus revenue base — immaterial, in their own words — and that the 2026 priority was securing customer certification for the advanced RDL-first and TGV processes over the next one to two years rather than maximizing volume.¹³

That is an unusually non-promotional thing for a management team to say while its stock is quadrupling on exactly this story. It is a genuine credibility marker, and it should be weighed as one. It is also, read the other way, a clear statement that the thing the market is paying for does not yet exist at scale.

The plot, told through profits

Strip the segments away and Innolux's recent financial history reads as a single violent arc. The 2021 pandemic boom — work from home, school from home, everyone buying a monitor — produced record profitability. The unwind produced a net loss of roughly NT$28 billion in 2022 and NT$18.6 billion in 2023.¹⁴ Then a fragile recovery: revenue of NT$216.5 billion in 2024 and NT$226.7 billion in 2025, up 4.7%.[^3]

But the operating line never recovered. Innolux posted an operating loss of NT$7.9 billion in 2024 and NT$4.2 billion in 2025.[^3] Both years reported a bottom-line profit only because non-operating income — NT$16.1 billion in 2024, NT$4.5 billion in 2025 — filled the gap.[^3] Full-year 2025 net profit attributable to owners was NT$250 million on NT$226.7 billion of revenue. Earnings per share: NT$0.03.[^3]

The recovery narrative and the earnings narrative are two different stories that happen to be running at the same time. Understanding which one you are being told is the central analytical task with this company.

↑ Back to Top

VI. The Reckoning: 2022–2023 Losses and the Strategic Pivot Decision

On August 18, 2022, Innolux's management sat down for an investor call with a set of numbers that were about to get worse. The second quarter had produced a net loss of NT$4.7 billion. The third quarter would nearly triple it, to NT$12.7 billion.¹⁴ By April 2024, the company had reported its eighth consecutive quarterly loss.⁴

What management announced on that August call was a specific, and revealing, pair of decisions. Capacity utilization would be cut to roughly 70%. Capital expenditure would stay flat at NT$26 billion.¹⁵

Consider what that combination says. Cutting utilization is the correct short-term response to an inventory glut: stop making things nobody is buying. Holding capex flat, in the same breath, is a bet — a bet that the downturn was a demand-timing problem rather than a structural one, and that the assets being funded would be needed when it passed. Chairman 洪進揚 Jim Hung's framing on the call was consistent with that: notebook pricing and volume growth had slowed, 2021's pandemic-inflated volumes had created an impossible comparison base, and the replacement cycle from those pandemic purchases would arrive from late 2023 into 2024.¹⁵ Slow inventory clearance was named as the core headwind.¹⁵

The replacement cycle did eventually arrive, more or less on the predicted schedule. What did not arrive was pricing. Panel prices stayed under pressure because Chinese capacity kept growing into a market that did not need it, and Innolux's operating line stayed negative through 2024 and 2025.[^3] Management got the demand call roughly right and the profitability call wrong, because profitability in this industry is not primarily a demand variable.

The accountability question

It is worth being direct about a pattern here rather than either accusing or excusing.

Across the 2022 and 2023 investor communications, the explanations for the downturn are consistently external: pandemic base effects, inflation, inventory cycles, geopolitical supply-chain restructuring, and later tariffs and a strong Taiwan dollar.¹⁵[^13] These were all real. They were also all things happening to Innolux rather than things Innolux did.

What is not readily found in the public record is a clear instance of leadership explicitly owning the internal decisions — capacity timing, capex pacing, the choice to hold NT$26 billion of spending flat into a deteriorating cycle — that made the downturn worse than it needed to be. That absence is not proof of evasion; Taiwanese investor communications are generally less confessional than American ones, and management's subsequent behavior (deep capex cuts, aggressive asset disposal) suggests the lesson was learned even if it was not narrated. But an investor assessing management credibility should note the pattern rather than assume it away: this is a team whose public accounting of a very bad two years attributes causation almost entirely outside the building.

The 666 Blueprint

Against that, there is a genuine consistency worth crediting. Since 2018, Hung has organized his strategic communication around what he calls the 666規劃藍圖, the 666 Blueprint: three consecutive six-year phases. 2018–2024 was "stabilize and seek profitability." 2025–2030 is "breakthrough transformation, expand reach" — the automotive and packaging pivot. 2031 onward is framed as "collective strength, panel sustainability."¹⁶

Whatever one thinks of the branding, the discipline of it is unusual. Management has been describing 2025–2030 as the transformation window since 2018, which means the Pioneer acquisition, the fab sales and the packaging push are not opportunistic reinventions dressed up after the fact — they are broadly what was advertised, on roughly the schedule advertised. In an industry where strategy tends to be whatever the last quarter's panel price made necessary, narrative consistency over eight years is a real, if partial, credibility asset.

The question is whether the person holding that narrative has the capability, and the checks, to execute the hard part.

↑ Back to Top

VII. Jim Hung and the New Innolux: Leadership, Ownership, Governance

There is a photograph from a few years back of Innolux's chairman standing in a Taipei MRT station wearing Pikachu ears, holding hands with his chief operating officer, promoting a new display product.¹⁷ It is not the image one associates with the head of a NT$400 billion industrial company. That is rather the point.

洪進揚 Jim Hung did not come up through a fab. His background is finance — Credit Suisse, Goldman Sachs, BNP Paribas — a career spent explaining companies to institutional investors rather than running clean rooms. He joined Taiwan Cement in 2013, entered the Hon Hai orbit in 2017, and was reportedly recruited personally by Terry Gou, who valued precisely the thing Innolux most lacked: credibility with the capital markets.¹⁷ He became chairman in June 2018 and added the CEO title that October, succeeding 王志超 Wang Chih-chao, a panel industry veteran from the Chi Mei lineage. Hung became, at the time, the youngest chairman in the Taiwanese panel industry.¹⁷

His early record was strong. Innolux had been through seven consecutive quarters of losses when he arrived; by the pandemic boom year the company posted record net profits.¹⁷ The GVM profile that captured the Pikachu moment also captured his self-assessment: asked about his reputation as a 金童, a golden boy, he rejected the label.¹⁷ The style is modern, media-fluent, deliberately informal — a conscious signal that, as he put it, Innolux is different now.

The obvious question follows. The pivot Innolux is executing is not a marketing problem. It is a technically demanding move into semiconductor packaging, where yields, certification cycles and process control are the entire game, competing against ASE, SPIL, Powertech and ultimately TSMC. A chairman whose comparative advantage is capital markets fluency is exactly the right person to sell fabs at good prices and finance an acquisition. Whether he is the right person to supervise a TGV process qualification is a different question, and it depends heavily on who else is in the room.

The May 2025 board election

In May 2025, that question got a concrete answer.

At the board election that month, Hung secured his third term as chairman — and the veterans went out. Wang Chih-chao, the former chairman, and 丁景隆 Ting Ching-lung, a former executive vice president, both panel industry lifers, unexpectedly lost their board seats.¹⁸ CEO 楊柱祥 Yang Chu-hsiang was moved out of his role and reassigned as chairman of a subsidiary.¹⁸ The technology advisory office that Ting had led was dissolved that same month.¹⁸ Wang retained an advisory title and the chairmanship of the company's Micro LED subsidiary; that was the extent of the veteran presence remaining.¹⁸

工商時報 Commercial Times reported the sequence bluntly as Hung achieving full command, with competing internal power centers eliminated.¹⁸

There is a legitimate reading in which this is good. Innolux's history — the post-merger paralysis, the 2014 episode where five vice presidents blocked the controlling shareholder — is a case study in what happens when a company has too many veto points. A transformation this severe, involving plant closures and thousands of severance packages, is not executed by committee.

There is another reading, and an investor should hold both. The board just removed the deepest display-technology expertise from its own governance layer, and dissolved the internal technology advisory function, at the precise moment the company is betting its future on a technically demanding new domain. Concentration of authority raises execution speed and removes the friction that catches mistakes. When the person holding that authority is not a technologist, the loss of technical counterweight is not a footnote.

Who actually owns this company

The ownership structure is genuinely unusual for an Asian industrial group of this size, and it changes how you should read everything above.

Innolux is not founder-controlled or family-controlled. Free float is roughly four-fifths of the register. Institutional ownership sits near 16%, with BlackRock around 4.2% and Vanguard around 2.5% — index money, not activist money. Insider and management ownership is under 1%. Hon Hai's historic stake, around 6.8% as of 2021, and Chi Mei's residual holding — roughly 2.6% and declining as Chi Mei sold down and booked losses on the position — remain influential but are nowhere near controlling.

The practical consequence: there is no entrenched family to blame and no controlling shareholder to appeal to. Governance here rests almost entirely on a board, and that board just concentrated power in one executive. Meanwhile, management's own equity stake is small enough that alignment runs through compensation design rather than through personal wealth in the stock — a structural fact worth knowing when a chairman is deciding whether to sell factories and buy Japanese companies.

The dividend that deserves a hard look

Which brings us to a capital allocation decision that should be stated plainly rather than glossed.

Dividends were cut to near-nothing through the 2022–2023 losses. They have since been restored. For fiscal 2025 — a year in which Innolux earned NT$0.03 per share — the board declared a cash dividend of NT$1.00 per share, funded NT$0.50 from retained earnings and NT$0.50 from capital reserve.¹³

That is a payout more than thirty times the year's earnings per share, with half of it drawn from the balance sheet rather than from anything the business generated. Management framed it as maintaining the prior year's level.¹³ That framing is accurate and it is also the tell: the decision being optimized was dividend continuity, not earnings-linked distribution.

Reasonable people can defend this. The company held NT$52.8 billion of cash at year-end against a net-debt-to-equity ratio of essentially zero, so affordability was not in question.[^3] Fab sales were about to generate large cash proceeds. A Taiwanese retail and institutional shareholder base places real weight on dividend continuity, and breaking a payout can cost a stock its holder base.

But an investor should name what it is: income smoothing. Paying out of capital surplus in a near-breakeven year is a choice to make the shareholder experience look steadier than the business is. It works fine as long as the transformation lands. It becomes an unforced error if the operating line stays negative and the capital surplus keeps funding distributions that operations cannot. Separately, market data show Innolux's share count fell roughly 7% over the past year — a second, quieter form of capital return that has drawn far less attention than the dividend.²

Now to the transaction that will define whether any of this was the right call.

↑ Back to Top

VIII. Betting the House: CarUX, Pioneer, and the Smart-Cockpit Play

Pioneer Corporation is 88 years old. For most of the second half of the twentieth century it was one of the great names in Japanese audio — home stereo separates, the laserdisc, car speakers that a generation of enthusiasts installed themselves. It also, by the 2010s, was a company that had lost its consumer businesses, gone through financial distress, and ended up owned by the Swedish private equity firm EQT Partners, running primarily on automotive audio, navigation, mapping and human-machine interface software, with deep relationships across Japanese automakers including Toyota.¹⁹

On June 27, 2025, Innolux's automotive subsidiary CarUX Holding announced it was buying 100% of Pioneer from EQT for approximately NT$37.7 billion — about $1.3 billion — in cash, funded through a bridge loan and a syndicated facility.¹⁹ Pioneer had generated ¥240 billion, roughly $1.7 billion, of revenue in fiscal 2024.¹⁹ The deal closed on December 1, 2025.²⁰ In the resulting structure, Innolux holds 85.91% of CarUX Holding, which in turn owns CarUX and Pioneer outright, with Pioneer's organizational structure, operations, employment and brand explicitly maintained.[^3]

The strategic logic

The logic is easy to state and genuinely coherent. Selling a display panel to an automaker makes you a component vendor: you compete on price and specification, and the tier-one supplier who integrates your glass into a cockpit module captures the systems margin. Owning the audio, the HMI software and the integration capability moves you up the stack to tier-one status yourself, where the customer is buying an in-cabin experience rather than a part number.

The complementarity is unusually clean on paper. CarUX's customer base skews American and European OEMs; Pioneer's skews Japanese — so the combination diversifies customers rather than overlapping them.[^3] CarUX brings display and integration; Pioneer brings audio, multimedia and HMI software.[^3] There are vertical integration savings available, since CarUX can supply display units directly into Pioneer products, plus procurement scale on overlapping materials.[^13] And Pioneer's manufacturing footprint outside Greater China has acquired new strategic value in a world where global automakers are actively seeking non-China supply options.[^13]¹³ Combined revenue is targeted at approximately NT$100 billion annually — roughly doubling CarUX's standalone scale.²⁰

The price and the skepticism

This is Innolux's largest acquisition since the founding merger of 2009–2010, and it is being paid for in cash by a company whose operating line loses money. That is a meaningful commitment of the balance sheet, and it shows: short-term debt jumped from NT$13.7 billion at the end of September 2025 to NT$33.5 billion at year-end, and long-term debt from NT$4.9 billion to NT$18.3 billion — the acquisition financing landing on the books.[^3] Net debt to equity moved from -10% to roughly zero.[^3] Innolux remains conservatively capitalized, but it has spent its net cash position.

The market's reaction to the announcement was notably restrained — the stock moved only modestly, with contemporaneous analyst commentary describing limited upside from the deal. The skepticism was not about strategic logic. It was about integration: a Taiwanese hardware manufacturer absorbing a Japanese consumer-legacy brand with its own management culture, its own union relationships, and a decade of private-equity restructuring behind it.

Hung's own language suggests he knows exactly where the risk sits. Asked what convinced Pioneer's leadership, he answered with a sports metaphor about how star talent alone does not win — "we're a team."¹⁶ The integration plan he described is granular in a way that reads as lessons learned: a three-phase, 100-day milestone structure covering cost optimization, cross-selling to global automakers, and combined audio-visual product development targeted at CES 2027.¹⁶

Read that against Section II and the subtext is unmistakable. The last time this company acquired its way to scale, the acquired culture kept its own hours for years and the merged entity lost NT$108.5 billion in three years.⁵ Innolux is now attempting a cross-border integration across a larger cultural distance, with a chairman who has explicitly promised not to run over the acquired company. Whether the 100-day milestone discipline is genuinely different from Gou's 2009 promise of non-interference, or the same mistake with better project management, is the single most important open question in this deal — and it will not be answerable from a slide deck. Watch CarUX+Pioneer revenue and margin, not the press releases.

Financing the deal, of course, required cash. Which is where the factories come in.

↑ Back to Top

IX. Turning Fabs Into Cash: The TSMC Deal and the "Three Plant Sales"

In August 2024, Innolux sold a building.

Not just any building: Nanke Fab 4, a 5.5-generation LCD plant in Tainan that the company had shut down in 2023 after concluding it could no longer compete on cost.²¹ The buyer was TSMC, which paid NT$17.14 billion — about $531 million — for the buildings and manufacturing facilities.²¹ TSMC did not announce detailed plans at the time, but the context was unambiguous: chairman C.C. Wei had been telling investors the company aimed to more than double CoWoS advanced packaging capacity in each of the next two years to meet AI chip demand, and the facility became TSMC's AP8 advanced packaging site.²¹²²

Step back and appreciate the trade. An idle commodity LCD fab — a stranded asset in a structurally impaired industry — was converted into cash at a price set by the hottest capacity shortage in global semiconductors. Innolux did not need to be good at advanced packaging to capture that value. It needed to own real estate, clean rooms and utilities in the Tainan Science Park at the exact moment the AI buildout ran out of places to put packaging capacity. That is opportunistic capital recycling of a high order, and management deserves straightforward credit for it.

The 2026 encore

In 2026, Innolux ran the play three more times — what the Taiwanese press dubbed 賣廠三連發, the three consecutive plant sales.

In March, a module plant in Tainan's Xinshi district went to ChipMOS for roughly NT$880 million, generating a disposal gain of about NT$659 million; the facility covered some 34,000 square metres and had previously been leased to Corning.²³¹ Next, Nanke Fab 2, an older fourth-generation line, was sold for roughly NT$6.3 billion with an estimated gain near NT$5.8 billion.¹ Then on April 15, the big one: Nanke Fab 5, the 5.5-generation plant, sold to 日月光 ASE for NT$14.85 billion, with disposal gains estimated at approximately NT$13.3 billion.¹

Combined proceeds across the three: about NT$22.06 billion. Combined disposal gains: roughly NT$19.76 billion — an estimated NT$2.4 per share contribution to 2026 earnings.¹ The buyer set is itself the story: ChipMOS and ASE are OSAT companies, and ASE has been expanding aggressively, projecting its advanced packaging revenue would double from $1.6 billion to $3.2 billion with a further $1.5 billion of investment.²² The board authorized Hung directly to execute disposals at prices no lower than 90% of appraised value, specifically to move faster.²³

Behind the transactions sat the human cost: a three-phase severance program beginning in June and targeted for completion in August 2026.¹

The quality-of-earnings problem

Here is where an investor has to be uncomfortably precise.

NT$2.4 per share of disposal gains, in a company that earned NT$0.03 per share in 2025 from everything it does. Those gains are real cash and real value creation — selling a depreciating asset at a good price to a buyer who values it more is exactly what good capital allocation looks like. They are also, by definition, non-recurring. There are a finite number of fabs.

So when 2026 full-year results arrive showing Innolux comfortably profitable, that statement will be true and it will mean almost nothing about whether the core business works. "Innolux is profitable again" and "Innolux's core operations are healthy again" are two different propositions, and in 2026 only the first is likely to be demonstrable. The disciplined way to read the next several years of results is to strip non-operating income out entirely and look only at the operating line — the same line that lost NT$7.9 billion and NT$4.2 billion in the two prior years.[^3]

The flip side: genuine capex discipline

The encouraging counterpart is that the shrinkage is not only opportunistic. Capital expenditure has been cut hard and deliberately: NT$16.1 billion in 2024, NT$11.7 billion actually spent in 2025, and guidance of roughly NT$13 billion for 2026, explicitly framed as asset-light optimization.[^3]¹³ Management also quantified the structural benefit of the closures rather than just the accounting gain — roughly NT$1.4 billion of annual maintenance cost savings, about 0.4% of revenue.¹³

And the shrinkage continues on a published schedule. Fab 3, the notebook-panel line, is slated for closure between 2027 and 2028, with production of notebook, monitor and medical panels consolidating into the Zhunan T2 site and Nanke Fab 3.⁹²⁴ Innolux is not being forced out of commodity display. It is walking out, on a timetable, while the exit still has resale value.

That is the strategically important distinction. Panel makers who wait until the assets are worthless discover that nobody wants a fab. Innolux is selling into demand.

↑ Back to Top

X. Optionality Note: The India Bet

One more asset deserves a brief mention, sized correctly.

On February 14, 2023, Innolux disclosed an agreement to transfer TFT-LCD panel and module technology to India's Vedanta, supporting what would be India's first integrated flat-panel display factory — covering TFT, colour filter and cell front-plane processes plus module assembly.²⁵ The plan contemplated a fab modelled on Innolux's own Fab 8B, likely running Gen 8.6 substrates at an assumed 60,000 substrates per month, with Vedanta engineers already training on site in Taiwan.²⁵ Vedanta would bear the capital cost, estimated in the billions of dollars. Innolux's role was to license technology and supply engineering know-how without operating the plant, in exchange for a one-time transfer fee and ongoing per-display royalties.²⁵

The structure is the attraction. This is optionality bought almost entirely with intellectual property rather than capital — a hedge on India's display-manufacturing incentive push and a source of geographic diversification at minimal balance-sheet risk. If India builds a domestic display industry behind protective incentives, Innolux earns royalties from it rather than competing with it.

It is also, three years on, still pre-production, and Vedanta's broader semiconductor and display ambitions have moved slowly. Treat this as a call option with a long expiry and an uncertain strike: real, worth knowing about, and not a pillar of the investment case in 2026.

The pillars are in Tainan and Tokyo. And in 2026, the market decided it liked what it saw.

↑ Back to Top

XI. Current State: The 2025–2026 Results and the Rally

Fiscal 2025 was, on the surface, a recovery year. Revenue rose 4.7% to NT$226.7 billion. Gross profit improved 31.6% to NT$18.7 billion, and gross margin expanded from 6.5% to 8.2%. The operating loss narrowed by nearly half.[^3]

Beneath the surface it was brutal. Net income attributable to owners fell about 96% year over year to NT$250 million, from NT$6.5 billion in 2024, because the non-operating income that had carried 2024 — NT$16.1 billion of it — collapsed to NT$4.5 billion in 2025.[^3] EPS went from NT$0.76 to NT$0.03.[^3] The fourth quarter was worse still: a NT$1.17 billion operating loss and a small net loss attributable to owners.[^3]

In other words, 2024's headline profit was a non-operating artifact, and when the artifact shrank, so did the profit. The operating business improved modestly and remained loss-making throughout.

The first quarter of 2026

Then something genuinely changed. First-quarter 2026 revenue came in at NT$66.6 billion, up 19.2% year over year, with operating profit of NT$1.5 billion — swinging from the prior quarter's NT$1.17 billion operating loss — and net profit of NT$1.79 billion.¹⁰ Non-display products reached 44% of sales.¹⁰

The drivers matter as much as the number. Management attributed the revenue strength to customers building inventory ahead of major global sporting events including the 2026 FIFA World Cup, and to a scramble driven by rising memory chip prices, which raised the cost of finished devices and pulled panel purchasing forward.¹⁰ The FOPLP and SpaceX news added to sentiment.¹⁰ Innolux's own market commentary described panel prices rebounding on World Cup stocking and maintenance-driven supply cuts, with 2026 panel area demand expected to grow roughly 6% year over year led by ultra-large sizes.[^3]

Read that honestly. An operating profit is an operating profit and it is the first one in a long while. But World Cup stocking is a pull-forward, not a new demand level; maintenance-driven supply cuts are a temporary condition; and memory-cost-driven pre-buying can reverse just as fast. The Q1 inflection is real and it is also substantially cyclical. Whether it is durable is precisely what the next set of results will answer.

The market did not wait. Innolux shares hit the daily 10% limit-up on May 12, 2026, closing at NT$35.50, outperforming a TAIEX that rose 0.72%; AUO rose 7.4% in sympathy.¹⁰ By mid-June the stock traded above NT$64, and the 52-week high reached NT$72.60.

The re-rating, and its fragility

As of August 10, 2026, Innolux traded at NT$52.30, a market capitalization of about NT$418 billion, and a price-to-book ratio of 1.82 — against 7.99 billion shares outstanding and book value per share of roughly NT$28.²[^3] The trailing 52-week price change was approximately +345%.²

The re-rating is the whole story in one number. This stock spent years trading at roughly half of book value — the market's assessment that Innolux's assets were worth less than their carrying value, which for idle LCD fabs was a defensible view. It now trades at nearly twice book. The market has revised its opinion not of the panel business but of what the assets can be converted into, and of the earnings power of a company that is 44% non-display and heading higher.

That revision has raised the bar considerably. At half of book, almost nothing needed to go right. At 1.8 times book, the transformation has to actually deliver.

And the shareholder base knows it. On July 7, 2026, Innolux fell NT$5 to NT$64.20 — a 7.23% single-session drop — as foreign institutional investors reversed from buyers to sellers and dumped NT$7.74 billion, 113,600 lots, making it the worst casualty of a day when the TAIEX fell 2.31% and foreign investors sold NT$54.7 billion market-wide.²⁶ AUO was second on the same sell list.²⁶ From the 52-week high, the stock has since given back roughly a quarter of its value.

This is a flow-driven, sentiment-driven, foreign-institution-dominated holder base sitting on an enormous gain in a name whose fundamentals have not yet caught up to the price. That combination does not predict direction. It does predict volatility.

↑ Back to Top

XII. Bull vs. Bear: Why Innolux Wins From Here, Why It May Not

Time to war-game it properly, taking both sides seriously.

The bull case

The diversification is quantified, disclosed and moving fast. This is not a vague "we're moving up the value chain" narrative; it is a segment disclosure showing commodity revenue falling from 70% to 50% of the mix in six quarters, with a documented margin gap — sub-5% gross margin in commodity versus 11–15% in non-display and 16–20% in non-commodity commercial products.[^13][^3] If the mix keeps shifting at anything like that rate, the arithmetic alone repairs the P&L without a single price increase anywhere.

The packaging entry has customer validation without promotional overreach. A satellite RF chip order is a demanding qualification, and management explicitly refused to characterize the resulting revenue as material.¹²¹³ The combination of a real customer and honest sizing is exactly the pattern you want to see early in a genuine new business, and the opposite of the pattern that precedes disappointment.

The cost position in packaging is not imaginary. Glass panels genuinely offer more usable area than round wafers, and Innolux already owns the fabs, the tools and the large-substrate handling expertise.[^3] This is an authentic adjacency, not diversification for its own sake.

The automotive position is real and now systems-level. Post-Pioneer, CarUX is a tier-one smart-cockpit supplier with complementary customer bases across three regions and a NT$100 billion combined revenue target.²⁰[^3] Automotive is the one display segment with genuine qualification-based switching costs.

Capital allocation has been genuinely opportunistic. Selling four fabs into the AI packaging boom at NT$22 billion-plus of proceeds and roughly NT$19.8 billion of gains, while cutting capex, is textbook recycling of declining assets.¹²¹

And there is no entrenchment risk in the ownership structure. No founding family, no controlling block, an overwhelmingly floating register — the structure that most commonly blocks value-creating change simply is not present.

The bear case

The core business is still majority-commodity LCD, and it is losing money at the operating line. Two consecutive years of operating losses, in what were supposed to be recovery years, against competitors who control more than 70% of global capacity and do not require a return on it.[^3][^10] Nothing in the bull case changes the fact that today's Innolux still depends on a segment where the structural verdict has already been delivered by Samsung and LG Display.

OLED keeps eating from the top. Every premium application that migrates from LCD to OLED shrinks the addressable high-value LCD market — and in automotive specifically, the shift toward LTPS and OLED is where Innolux's incumbency is weakest.¹¹

Reported profitability is being manufactured by disposals. NT$2.4 per share of 2026 EPS from fab sales, in a company whose operations earned NT$0.03 per share in 2025.¹[^3] The transformation and the earnings are not the same thing, and there is a finite supply of fabs.

The Pioneer integration is unproven, capital-intensive, and drew a tepid market reaction. It consumed the company's net cash position, and this management team's institutional history with post-merger integration is the worst chapter in its own past.[^3]⁵

Governance concentrated at the worst possible moment. One non-technical chairman with full command, the display veterans off the board, the technology advisory office dissolved — during a pivot into semiconductor process technology.¹⁸

And the valuation has already priced the outcome. A 345% move to 1.8 times book prices a successful transformation that has not been demonstrated at scale.² The sell-side consensus rating is a hold with an average price target well below the current price, which tells you the analyst community has not followed the stock up.²

One more item belongs on the risk radar, and it is quiet: Innolux's Altman Z-Score sits at 1.42 — a heavy-asset, thin-margin, cyclical manufacturer's profile.² With NT$52.8 billion of cash and roughly zero net debt after the Pioneer financing, near-term solvency is not the issue.[^3] But this is a business whose balance sheet strength depends on continuing to convert assets to cash, and the conversion pipeline is finite.

The activist stress test

No activist campaign exists at Innolux today. But run the screen a skeptical fund would run, and the ingredients are all visible: insider ownership under 1%, a governance concentration event that removed independent technical counterweight, a dividend paid half out of capital surplus in a near-breakeven year, reported profits dominated by one-time gains, and a valuation that has run well ahead of operating fundamentals.

The first question such an investor would ask is not hostile — it is simply the right question, and it is the one the August 2026 results should be read against: how much of this year's profit is recurring operating income, and how much is fab-sale gains? The second would be: if the answer is mostly fab sales, what does the operating business look like in 2028 when there are no more fabs to sell?

The three KPIs that matter

Everything above compresses into three things worth tracking, and only three.

One: non-display and non-commodity share of revenue, and the gross margin of each. This is the transformation, measured. Innolux discloses it every quarter in the same format. If the share keeps climbing while those margins hold, the thesis is working. If the share climbs only because commodity revenue is being amputated, and the non-display margin drifts down as competition arrives, it is not.

Two: capacity utilization and blended ASP in the remaining commodity base. This is the drag, measured. The commodity business does not need to be great; it needs to stop consuming the profits of everything else. Watch whether commodity gross margin recovers from below 5% or stays there.

Three: CarUX plus Pioneer revenue and margin against the NT$100 billion target. This is the acquisition, measured, and it is the cleanest possible test of whether the integration lesson from 2010 was actually learned.

Notice what is not on the list: quarterly EPS. In a company reporting large disposal gains, headline earnings are the least informative number available.

↑ Back to Top

XIII. Playbook: Business and Investing Lessons

Four things generalize from this story.

Scale in commodity manufacturing is not a moat when a state actor will subsidize a rival past profitability. This is the lesson of the entire LCD industry, and Korea delivered the verdict most clearly: two of the most capable display manufacturers ever built examined the returns available in large LCD and exited entirely rather than defend it.[^9] Note what that implies about the standard defense of scale businesses. Scale protects you against competitors who need to earn their cost of capital. Against competitors who don't, scale simply means you own more of the wrong asset. The relevant question for any capital-intensive business is never "am I big?" but "is my largest competitor's capital priced the same as mine?"

Post-merger integration debt compounds, and it compounds quietly. The 2009–2010 merger delivered its promised scale and its unpromised culture clash within eighteen months, but the cost kept accruing for a decade — in paralysis, in blocked decisions, in the observation that the southern plants still ran on the old schedule years later.⁵ The reason this matters analytically is that integration failure rarely appears as a single write-down. It appears as a company that is slightly slower than it should be, for years. When the same company attempts a larger, more culturally distant integration fifteen years later, the historical base rate is the most useful prior available.

Capital recycling can be excellent capital allocation, and it is not operating improvement. Selling declining assets into someone else's boom is genuinely value-creating — Innolux monetized four stranded fabs at prices set by an AI capacity shortage that had nothing to do with panels.²¹¹ Applaud it. Then separate it. The discipline is mechanical: find the operating line, ignore everything below it, and ask whether the business would be profitable if it sold nothing. When management is doing something clever with the balance sheet, that is precisely when the income statement most needs to be read from the top down rather than the bottom up.

Governance concentration during a technical pivot is a signal, not a footnote. The market rewards decisiveness, and there are real cases for it here — this company's history is partly a history of too many veto points. But the specific configuration matters: authority concentrated in a non-technical executive, simultaneous with the removal of the board's technical expertise and the dissolution of the technology advisory function, during a bet on semiconductor process technology.¹⁸ The general principle is that capability and incentive alignment matter more than a clean equity story when execution risk is high. A widely-held register with no entrenched family is a governance strength on paper; it is worth less when the board that represents it has just reduced its own capacity to challenge the chairman on the substance of the strategy.

Innolux is currently running a live experiment on all four lessons at once. The results start arriving shortly.

↑ Back to Top

XIV. Epilogue & What to Watch

On August 18, 2026 — eight days from now — Innolux will hold its second-half institutional investor conference online, with Jim Hung as designated spokesperson, reporting second-quarter results and the outlook for the rest of the year.²⁷ June 2026 revenue had already come in at NT$21.7 billion, up 17.3% year over year and reportedly a five-year high on automotive and advanced packaging strength.²⁸

That call is the first real test of whether the first-quarter inflection was durable or a World Cup-shaped bump. Three things are worth listening for specifically.

Whether the operating line stays positive without help. Q1 delivered NT$1.5 billion of operating profit.¹⁰ A second consecutive positive quarter, achieved while panel prices normalize after the sporting-event stocking, would be the first genuine evidence that the mix shift is doing structural work rather than riding a cycle.

Whether the disposal program completes on the reported terms, and what replaces it. Fab 5's proceeds and the Fab 3 closure timeline through 2027–2028 are the remaining known steps in the shrink-and-recycle plan.¹⁹ After that, the balance sheet stops generating one-time earnings, and reported profit has to come from operations.

Whether CarUX and Pioneer produce revenue scale-up without margin dilution. The 100-day integration milestones described in December 2025 should be complete; the CES 2027 combined product target is the visible deliverable.¹⁶ Any language that softens the NT$100 billion combined revenue target, or that starts explaining integration timelines rather than reporting integration results, deserves attention.

The larger frame is this. Innolux's story stopped being about LCD panels some time ago. What it is now is a wager — that a finance-trained chairman with newly consolidated control can convert a structurally impaired commodity manufacturer into a semiconductor packaging and automotive systems company, fast enough to justify a stock that has already risen 345% in anticipation of his success.²

The assets are being sold on schedule. The mix is shifting on schedule. The narrative has been consistent since 2018. What has not yet happened, in any year of the transformation so far, is a full year in which the core operating business earned money.[^3] Everything else is preparation for that moment. Whether it arrives is the only question that ultimately matters here — and unlike the disposal gains, it cannot be arranged.

References

1. 群創賣廠三連發 五廠賣給日月光 處分利益近200億元 — 工商時報 Commercial Times, 2026-04-16 ↩↩↩↩↩↩↩↩↩↩↩

2. Innolux Corporation (TPE: 3481) Statistics & Valuation — Stock Analysis, accessed 2026-08-10 ↩↩↩↩↩↩↩↩↩

3. Innolux announces Chi Mei acquisition via a share swap — Taipei Times, 2009-11-15 ↩↩↩↩↩↩

4. InnoLux Corporation — Wikipedia, accessed 2026-08-10 ↩↩↩↩

5. 【整頓群創】6年前的強強合併 現在成了郭台銘心中的痛 — 鏡週刊 Mirror Media, 2016-11-28 ↩↩↩↩↩↩↩↩↩

6. LCD Price Fixing Conspiracy — Federal Bureau of Investigation ↩↩↩↩

7. Attorney General Kamala D. Harris Announces Half-Billion-Dollar Nationwide LCD Settlement — California Attorney General, 2011-12-27 ↩↩

8. AUO posts profit on factory sale, forecasts stable growth in 2025 display sector — Digitimes, 2025-02-13 ↩

9. 面板雙虎轉型！群創、友達賣廠退出規模戰 牽動全球面板供需版圖 — 壹蘋新聞網 Nextapple, 2026-08-04 ↩↩↩↩

10. Innolux shares soar on improving bottom line in Q1 — Focus Taiwan, 2026-05-12 ↩↩↩↩↩↩↩

11. Omdia: LTPS TFT and OLED display revenues to exceed 50% of automotive display market in 2025 — Omdia, 2025-07 ↩↩↩↩

12. Innolux Leads FOPLP Race with Industry's Largest Substrate, Mass Production Reportedly Set in 1H25 — TrendForce, 2025-03-04 ↩↩

13. 群創法說會重點內容備忘錄：未來展望趨勢 20260327 — 富果 Fugle, 2026-03-27 ↩↩↩↩↩↩↩

14. Innolux sees loss widen in 3Q22 — Digitimes, 2022-10-28 ↩↩

15. 面板庫存去化緩慢！群創下調產能至七成，資本支出維持260億不變 — TechNews 科技新報, 2022-08-18 ↩↩↩↩

16. CarUX 收購 Pioneer、FOPLP 逐季放量，群創洪進揚親解轉型戰略 — TechNews 科技新報, 2025-12-23 ↩↩↩↩

17. 群創董座洪進揚：別叫我金童！ — 遠見雜誌 Global Views Monthly ↩↩↩↩↩

18. 群創內部角力不斷 洪進揚全面掌權 — 工商時報 Commercial Times, 2025-05-16 ↩↩↩↩↩↩↩

19. Innolux subsidiary CarUX to acquire Japan's Pioneer — Taipei Times, 2025-06-27 ↩↩↩

20. CarUX finalizes Pioneer purchase, creating new force in smart cockpit tech — Digitimes, 2025-12-10 ↩↩↩

21. TSMC to acquire Innolux Tainan plant for NT$17bn — Taipei Times, 2024-08-16 ↩↩↩↩↩

22. Panel Maker Innolux Fab 5 Reportedly Eyed by ASE, OSAT Firms — Micron Out of Picture — TrendForce, 2026-02-12 ↩↩

23. 群創模組廠8.8億賣南茂 市場：南科二廠、五廠要出售了 — 經濟日報 Economic Daily News ↩↩

24. 《DJ獨家》群創資產活化，Fab3傳明年H2關閉 — MoneyDJ via Yahoo Stock Taiwan ↩

25. Innolux signs LCD technology transfer agreement with India's Vedanta — Omdia, 2023-02 ↩↩↩

26. 面板雙虎遭狙殺！群創遭外資狂掃11萬張被迫吃草 友達轉型題材同步失色也遭倒 — Yahoo奇摩股市, 2026-07-07 ↩↩

27. 【公告】群創光電2026年下半年法人說明會 — Yahoo奇摩股市 / MoneyDJ, 2026 ↩

28. Innolux 1H26 revenue hits 5-year high on auto, advanced packaging — Digitimes, 2026-07-14 ↩