ams-OSRAM AG (AMS.SW) - Deep Dive Research Report

Prepared 3 July 2026. Listing venue: SIX Swiss Exchange, ticker AMS. Reporting currency: EUR. Fiscal year ends 31 December; quarterly reporting.

1. What the company does

ams-OSRAM makes the tiny light sources and light sensors that let machines see, and it makes the specialty lamps that light up cars. Strip away the jargon and there are two things in the box. First, semiconductors that either emit light (LEDs, infrared emitters, lasers, VCSELs) or detect it (optical sensors, ambient-light and proximity sensors, the chips that read a signal off light). Second, a legacy but cash-generative business of automotive and specialty lamps - the halogen, xenon and LED bulbs that go into headlights and into aftermarket replacement, plus niche industrial and entertainment lamps.

The reason both live under one roof is a 2020 merger that has defined everything since. The company you see today is the product of a small, clever Austrian sensor-chip firm (ams AG) swallowing a much larger, much older German lighting giant (OSRAM, a name that traces to 1919 and was spun out of Siemens). ams borrowed heavily to do it. The debt from that deal, and the subsequent collapse of a marquee customer project, is the gravitational force that explains the restructuring, the divestments, and the strategic reinvention the company is living through right now.

Here is the value proposition in a concrete example. When you raise a modern smartphone to your face and it wakes and unlocks, an infrared emitter throws a pattern of dots, and a sensor reads the reflection to build a depth map of your face. ams-OSRAM has supplied the emitter and sensor building blocks for exactly this kind of 3D sensing. When a car's adaptive headlight dims just the pixels aimed at oncoming traffic while keeping the road lit, that is a pixelated LED matrix, and ams-OSRAM is one of a very small number of firms that can make automotive-grade versions. When a surgeon's endoscope needs a light source small enough to fit down a tube, or an X-ray detector needs to count individual photons to cut radiation dose, the light-emitting and light-sensing silicon is the kind of thing this company builds.

What makes it hard is not one thing but a stack of them: compound-semiconductor materials science (gallium nitride, gallium arsenide, indium phosphide rather than plain silicon), the ability to shrink an emitter to microscopic scale while keeping it efficient and reliable, decades of automotive qualification credentials that a new entrant cannot buy, and the packaging know-how to put emitter, sensor and driver chip into a module the size of a grain of rice. That combination is genuinely rare. The problem, as the last three years have shown, is that rare capability does not automatically translate into cash when a balance sheet is loaded with acquisition debt and your biggest bet loses its anchor customer.

CEO Aldo Kamper's framing of the current strategy: the company is "uniquely positioned to capture the major inflection points in Digital Photonics across automotive, Augmented Reality smart glasses, biosensing, home and industrial robotics, AI data-center optical interconnects and potentially even visionary applications like laser fusion." (Digital Photonics leader announcement, 2026)

2. Business segments

ams-OSRAM reports in two segments: Semiconductors (roughly two-thirds of group revenue) and Lamps & Systems (roughly one-third). The whole corporate story of 2023 to 2026 is management shrinking, refocusing and re-labelling the Semiconductor segment while milking Lamps & Systems for cash to fund the transition.

2.1 Semiconductors (~67% of revenue)

What it does. This is the heart of the company and the part management wants you to value. It contains two divisions. Opto Semiconductors (OS) makes things that emit light: visible LEDs, infrared LEDs, lasers and VCSELs (vertical-cavity surface-emitting lasers, the emitters behind 3D face-sensing and lidar). CMOS Sensors & ASICs (CSA) makes things that read the physical world and turn it into data: ambient-light sensors, proximity sensors, spectral and colour sensors, position sensors, and application-specific chips with embedded software. The end markets are consumer electronics (smartphones, wearables, AR/VR headsets), automotive (interior and exterior lighting, driver monitoring, sensing), industrial, and medical (endoscopy light sources, photon-counting X-ray sensors).

Core capability. The moat here is compound-semiconductor process knowledge plus miniaturisation. Emitting light efficiently at microscopic scale, in materials that are not silicon, is a craft measured in decades. ams-OSRAM's emitter heritage runs back through OSRAM Opto Semiconductors; its sensor and mixed-signal IC heritage comes from the original ams AG. Very few firms hold both the emitter and the sensor side, and management's entire "Digital Photonics" thesis rests on being able to combine an emitter, a sensor and the processing intelligence in one system.

Why it is a segment. It is the growth engine and the technology identity of the group. Management measures its health by design wins (contracts booked that convert to revenue over a product's multi-year life). In FY2025 the company reported an all-time high of over EUR 5 billion in semiconductor design wins (Q4/FY2025 results, 10 Feb 2026). The "core semiconductor portfolio" grew mid-to-high single digits year on year on a like-for-like basis through 2025 even as the reported top line shrank, which is the number management points to as evidence the underlying business is healthy.

Competitive position. In emitters it competes with Nichia, Lumileds, Seoul Semiconductor and Samsung. In optical sensing and mixed-signal it overlaps with STMicroelectronics, onsemi and Broadcom. It wins on automotive-grade emitters and on combined emitter-plus-sensor systems; it is exposed wherever the product commoditises into a pure LED price war, which is why management is deliberately exiting the more commoditised corners.

2.2 Lamps & Systems (~33% of revenue)

What it does. This is the OSRAM lighting inheritance, now pruned to its most defensible core: automotive lamps (halogen, xenon and LED bulbs sold both to carmakers as original equipment and, crucially, into the aftermarket where drivers replace a blown headlight bulb) plus specialty lamps for industrial and entertainment use. Management has been selling off the fringes of this segment (it sold the Entertainment & Industrial Lamps business for EUR 114 million in 2025).

Core capability. Brand, distribution and installed base. OSRAM is a household automotive-lamp name with a global aftermarket channel that a new entrant cannot replicate quickly. The aftermarket in particular is a recurring, high-margin annuity: every car on the road eventually needs a replacement bulb, and the buyer reaches for the brand they recognise.

Why it is a segment, and how it fits. It exists as a separate unit because its economics are the opposite of the semiconductor business: mature, slowly declining as LED headlamps last longer, but cash-rich and predictable. Management is explicit that it is a cash cow to fund the growth bet. The Digital Photonics announcement described traditional automotive lamps, including the aftermarket, as "a reliable cash flow generator for funding growth in its semiconductor business."

Competitive position. Against Koito, Hella/Forvia, Valeo and Philips-branded automotive lighting. It competes on brand and channel, not technology, and the strategic question is only how gracefully it declines as the vehicle fleet electrifies and LED modules replace replaceable bulbs.

3. Products and business detail

The emitter catalogue. Visible and infrared LEDs, high-power lasers, VCSEL arrays, and pixelated LED matrices. Recent named products aimed squarely at the AR/wearables push include the FIREFLY SFH 4030B and SFH 4060B infrared LEDs (the first in class with a fully black package for discreet integration into smart glasses and AR/VR headsets), VEGALED, and the Chip LED SFH 2707 for vital-sign monitoring in wearables (CES 2026 launch). The forward-looking flagship is an ultra-efficient microLED array for "slow and wide" AI optical interconnects demonstrated in March 2026, which uses a highly parallel micro-emitter array to move data between server racks in AI data centres at lower power than conventional laser optics.

The sensor catalogue. Ambient-light, proximity, colour and spectral sensors; position and optical force sensors; and, historically, CMOS image sensors and a broad non-optical analog/mixed-signal sensor line for automotive, industrial and medical use. Two of those lines are being sold (see below), narrowing the catalogue deliberately toward optical.

The lamp catalogue. Halogen, xenon and LED automotive bulbs; specialty and entertainment lamps. This is a physical-product, channel-driven business rather than a design-win business.

Manufacturing. ams-OSRAM runs its own fabs (it is an IDM, an integrated device manufacturer, not fabless), with compound-semiconductor wafer fabrication and specialised packaging. The defining manufacturing episode of the era was the Kulim, Malaysia 8-inch microLED epiwafer fab, announced in 2022 with roughly EUR 800 million of planned spend. When the anchor microLED customer cancelled in early 2024, the fab lost its purpose; ams-OSRAM exited microLED volume production there and took an impairment on the order of EUR 600-900 million, then sought a buyer for the site. This is the single most important operational scar on the company and the reason the whole strategy was rebuilt.

Geographies. Co-headquartered in Premstätten, Austria (the ams heritage) and Munich, Germany (the OSRAM heritage), with global manufacturing, design and sales. It sells into every major electronics and automotive supply chain worldwide. Roughly 20,000 employees.

The 2026 portfolio surgery. Two divestments reshaped the product footprint:

• Sold the non-optical analog/mixed-signal automotive, industrial & medical sensor business to Infineon for EUR 570 million cash (announced 2026, expected close Q2 2026), removing roughly EUR 220 million of annual revenue and about EUR 60 million of adjusted EBITDA, and cutting pro-forma leverage to about 2.5x.
• Sold the CMOS Image Sensor business to indie Semiconductor for EUR 40 million cash (May 2026).

The strategic logic in one line: shed everything that is not optical emitting-and-sensing, use the proceeds to pay down acquisition debt, and reconcentrate on the "Digital Photonics" core.

4. Customers

Who buys. Three broad camps. Consumer electronics OEMs (smartphone, wearable and AR/VR makers) buy emitters and sensors for face-sensing, ambient-light control, health sensing and eye-tracking. Automotive OEMs and Tier-1 suppliers (and, separately, the mass aftermarket) buy LEDs, sensors and lamps. Industrial and medical OEMs buy specialty light sources and sensors for endoscopy, X-ray detection, machine vision and robotics.

How the buying works. In semiconductors the sales cycle is long and front-loaded. A customer's engineering team qualifies a component into a product design; once designed in, the part ships for the multi-year life of that product. This is why management reports design wins as the leading indicator - a EUR 5 billion design-win year books revenue that lands across future years. The decision-maker is an engineering and procurement committee weighing performance, reliability, automotive qualification, roadmap and price. In the lamp aftermarket the "customer" is ultimately a driver reaching for a familiar brand in a parts shop, mediated by distributors.

Why they choose ams-OSRAM. Automotive-grade reliability credentials that take years to earn; the ability to supply a combined emitter-plus-sensor system rather than a single part; and, in emitters, genuine compound-semiconductor performance. In lamps, brand and channel.

Switching costs. High in automotive semiconductors, where a qualified part is locked in for a vehicle platform's life and requalifying is costly and slow. Lower in commodity consumer LEDs, where a design can be re-sourced between generations - which is precisely the vulnerability the microLED cancellation exposed.

Concentration. The company's history with a single dominant consumer customer (widely understood to be Apple, described in past coverage as its largest client) is the cautionary tale of this business. That customer's cancellation of the microLED watch project in 2024 blew a EUR 700 million hole and killed the Malaysia fab. Management's current strategy is in part an explicit answer to that concentration: diversify across AR glasses, biosensing, robotics, automotive and AI data centres so that no single project cancellation can repeat the damage.

Contract structure. A mix of design-win-driven multi-year component supply (predictable once won, lumpy to win), development agreements for emerging categories (the 2026 AI photonics and AR agreements are development deals, not yet volume production), and recurring aftermarket lamp sales.

5. Competitive landscape

ams-OSRAM does not face one competitor; it faces a different set in each product line, and its competitive position ranges from strong (automotive emitters, combined optical systems) to exposed (commodity LEDs, general mixed-signal).

In emitters (LEDs, lasers, VCSELs): Nichia (the Japanese firm that invented the blue and white LED, the quality benchmark), Lumileds (US, high-performance LEDs, emerged from a 2022 bankruptcy restructuring), Seoul Semiconductor (Korea, which overtook Nichia in backlight LED share), and Samsung's LED arm. ams-OSRAM wins in automotive-grade and specialty emitters where reliability and qualification dominate; it loses where the product is a commodity backlight LED sold on price.

In optical and mixed-signal sensing: STMicroelectronics (strong in time-of-flight and 3D sensing), onsemi (image sensing and power), and Broadcom (optical components). ams-OSRAM has deliberately handed onsemi's neighbour Infineon its non-optical sensor line and sold its CMOS image sensor line, conceding those fights to concentrate on optical.

In the emerging AI optical-interconnect arena: the incumbents of datacom optics - Coherent, Lumentum, Broadcom, and specialist silicon-photonics players - already ship the laser optics that ams-OSRAM's "slow and wide" micro-emitter approach aims to displace for rack-to-rack links. This is a bet on a different architecture, not an established position.

Barriers to entry are high in compound-semiconductor emitters (materials, fabs, decades of process knowledge, automotive qualification) and in the OSRAM lamp brand and channel, but low in commodity LEDs where Asian volume producers compete purely on cost. The structural shift underway is twofold: LED lighting commoditising at the low end (pushing ams-OSRAM upmarket), and a possible new high-value frontier opening in AR wearables and AI-datacenter photonics (which the company is racing to reach before its balance sheet forces its hand).

Market caps are approximate order-of-magnitude references as of mid-2026 and move constantly; they are a peer-size guide only.

6. Industry

Demand drivers. ams-OSRAM sits at the intersection of several distinct cycles. Consumer optical content per device rises with each smartphone and wearable generation that adds sensing (face unlock, health monitoring, eye tracking). Automotive semiconductor content per car rises structurally as headlights go pixelated, interiors gain ambient lighting and driver-monitoring, and sensing multiplies - even as unit car volumes stay cyclical. Industrial and medical demand tracks capex and healthcare spend. And a genuinely new driver is emerging: AI data-centre buildout, where the volume of optical interconnect needed to link accelerators is exploding.

Size and growth. The broad semiconductor market was cited at roughly USD 637 billion in 2026, growing toward USD 1.15 trillion by 2033 (~8.8% CAGR) in third-party market data. The opto-semiconductor and optical-sensing slices ams-OSRAM plays in are smaller and more fragmented, with growth concentrated in automotive optical content and, prospectively, AR and datacenter photonics rather than in commodity LED lighting, which is mature and price-deflationary.

Position in the supply chain. ams-OSRAM is an integrated device manufacturer near the front of the chain: it designs and fabricates the emitting and sensing silicon, packages it into components and modules, and sells to OEMs and Tier-1s. That IDM model means it carries fab capex and utilisation risk directly, which is why the Malaysia overbuild hurt so much.

Regulation and qualification. The binding "regulation" here is automotive qualification (AEC-Q standards) and medical certification, which act as barriers protecting incumbents. There is no single government policy gating the business, though it is exposed to semiconductor trade and export dynamics between the US, Europe and Asia.

Cyclicality. High. This is a cyclical industry layered on top of two cyclical end markets (consumer electronics and autos). Inventory corrections, smartphone demand swings and auto production cuts all flow straight through. The last down-cycle, compounded by the microLED cancellation, is what forced the restructuring. The offset management is building is a mature, counter-cyclical aftermarket-lamp cash flow underneath the volatile semiconductor top line.

7. Growth triggers

Drawn strictly from the six most recent concalls and their accompanying releases.

• AI photonics / optical interconnects for AI data centres. Entered the AI market via a development agreement with a leading AI photonics partner, built on the "slow and wide" micro-emitter array demonstrated in March 2026 (Q1 2026 concall, 7 May 2026). Management frames the opportunity as very large but long-dated.

  Management described the revenue potential as a "high triple-digit million" opportunity, with meaningful inflows not expected before 2030 (Q1 2026 concall, 7 May 2026).

• AR / AI smart glasses content. Broader component portfolio for AI-enabled smart glasses, with new black-package FIREFLY infrared emitters and eye-tracking illumination; management targets EUR 50-100 of content per device (CES 2026, and Q1 2026 concall, 7 May 2026). Repeated as a strategic pillar across the Q4 2025 and Q1 2026 calls.

• Micro-emitter array AI optical interconnect advancing to product development (announced March 2026; reiterated Q1 2026 concall, 7 May 2026).

• Design-win backlog conversion. All-time-high semiconductor design wins exceeding EUR 5 billion in FY2025 feed future revenue as programmes ramp (Q4/FY2025 concall, 10 Feb 2026). Design-win records were flagged as improving across the Q2, Q3 and Q4 2025 calls.

• "Simplify" transformation and savings program, ~EUR 200 million. Launched to drive further cost-out and organisational simplification, targeted to deliver by 2028 (Q4/FY2025 concall, 10 Feb 2026).

• "Re-establish the Base" (RtB) savings running ahead of plan. Roughly EUR 135 million run-rate realised by Q1 2025, building to about EUR 220 million by year-end 2025, described as a year ahead of plan (Q1 2025 concall, 30 Apr 2025; Q4/FY2025 concall, 10 Feb 2026).

• Deleveraging via divestments. Infineon sale (EUR 570 million) and indie CMOS image sensor sale (EUR 40 million) cut pro-forma leverage toward 2.5x and fund the pivot (Q1 2026 concall, 7 May 2026).

• Path to positive group free cash flow in 2027. Management explicitly guided a path to positive FCF in 2027 (Q1 2026 concall, 7 May 2026), building on FY2025 FCF of EUR 144 million.

• 2030 financial targets. Semiconductors mid-to-high single-digit revenue CAGR with adjusted EBITDA margin at or above 25%; group FCF above EUR 200 million and leverage below 2 (Digital Photonics announcement, 2026).

8. Key risks

• Balance-sheet and refinancing overhang. The whole restructuring exists because the 2020 OSRAM acquisition left ams-OSRAM heavily indebted, and the 2024 impairments deepened the hole. Deleveraging depends on divestment proceeds landing (Infineon's EUR 570 million requires antitrust clearance and a Q2 2026 close) and on the core business generating cash. If a divestment slips or the cycle turns before leverage is under control, financing cost and covenant pressure return. Mechanism: high fixed interest against a cyclical, shrinking-then-reinventing top line.

• Customer-concentration repeat. The microLED cancellation showed how a single consumer customer's decision can vaporise a EUR 700 million investment. The AR-glasses and AI-photonics bets are again exposed to a small number of very large potential customers whose product roadmaps can change overnight. Mechanism: a development agreement is not a volume contract; a partner can walk before it ever ramps.

• The AI-photonics bet is unproven and long-dated. Management itself says meaningful revenue is not expected before 2030, and the incumbents (Coherent, Lumentum, Broadcom) are entrenched in datacenter optics. Mechanism: money and management attention go into a category that may not pay for half a decade, if the "slow and wide" architecture wins at all.

• Shrinking to grow. The company is deliberately selling revenue (Infineon's ~EUR 220 million, indie's CMOS line) and reporting a top line that declines even as the "core" grows. If the divested EBITDA and stranded costs bite harder than the savings programs deliver, reported profitability sags during the transition. Management flagged exactly this: FY2026 adjusted EBITDA is expected to be dragged by divestment one-offs, stranded costs, higher precious-metal prices and a weaker dollar (Q4/FY2025 concall, 10 Feb 2026).

• Cyclicality and FX. Revenue rides consumer-electronics and auto cycles simultaneously, and a large slice of sales is dollar-linked while the company reports in euros; management repeatedly cited USD weakness as a headwind (Q2 2025 and Q4 2025 calls). This is specific because the divestments make the remaining base smaller and therefore proportionally more sensitive to each swing.

• Lamp secular decline. The cash cow funding the pivot is a mature business that shrinks as LED headlamps outlast replaceable bulbs. Mechanism: if the aftermarket annuity erodes faster than expected, the funding source for the growth bet thins just when it is needed most.

9. Walk the talk

The six calls used, most recent last:

1. Q4/FY2024 - reported 11 Feb 2025
2. Q1 2025 - reported 30 Apr 2025
3. Q2 2025 - reported 31 Jul 2025
4. Q3 2025 - reported 18 Nov 2025
5. Q4/FY2025 - reported 10 Feb 2026
6. Q1 2026 - reported 7 May 2026

The most recent is 7 May 2026, within 90 days of today.

The story across these six calls is one of a management team that inherited a crisis, set deliberately conservative operational targets, and has so far hit or beaten them - while the strategic reinvention on top is still unproven.

At the Q4/FY2024 call (Feb 2025), management guided to return to positive free cash flow, promising FY2025 FCF "exceeding EUR 100 million" on improved earnings, lower capex and stable working capital, and flagged that the "Re-establish the Base" savings were already ahead of plan. That is a specific, datable promise.

Through Q1 2025 (Apr), Q2 2025 (Jul) and Q3 2025 (Nov), they repeatedly confirmed the ">EUR 100 million FCF" outlook rather than quietly walking it back. Q1 delivered 16.4% adjusted EBITDA above the guidance midpoint with about EUR 135 million of run-rate savings already booked; Q2 came in at the midpoint (18.8% EBITDA on EUR 775 million) despite currency headwinds and executed the first deleveraging step by selling Entertainment & Industrial Lamps for EUR 114 million; Q3 beat the midpoint again (19.5% EBITDA, EUR 43 million FCF). The pattern is consistent: guide to a range, land at or above the midpoint, reaffirm the annual number.

At the Q4/FY2025 call (Feb 2026) the promise was cashed in. FY2025 FCF came in at EUR 144 million, comfortably above the ">EUR 100 million" pledged a year earlier, and RtB run-rate savings reached about EUR 220 million, described as a year ahead of plan.

Guided a year earlier: FY2025 free cash flow "exceeding EUR 100 million" (Q4/FY2024 concall, Feb 2025). Delivered: EUR 144 million (Q4/FY2025 concall, 10 Feb 2026).

That is a clean kept promise on the operational, cost-and-cash side, backed by a full year of consistent quarterly beats. On the same call they launched the EUR 200 million "Simplify" program and reported a record EUR 5 billion design-win year, and they were candid that FY2026 revenue would soften on divestments and FX rather than dressing it up.

The Q1 2026 call (May 2026) kept the pattern (EUR 796 million revenue in the upper half of guidance, 16.5% EBITDA) and added the new, higher-risk strategic promises: the AI-photonics development deal and a stated path to positive group FCF in 2027. These are the commitments still to be tested.

The honest read: on the things they can control - cost-out, cash generation, deleveraging, guidance discipline - this management has done what it said, quarter after quarter, and beat its headline FCF pledge. The RtB savings ran ahead of plan and the FCF target was exceeded. Where credibility is not yet earned is on the strategic reinvention: the microLED disaster was inherited chaos they cleaned up, but the AI-photonics and AR-glasses bets are development agreements with 2030 revenue horizons, and none of that has converted to volume yet. This is a team that has been consistently conservative and accurate on operations; the jury is out on whether the Digital Photonics vision pays off. Judge them a delivered-what-they-said management on the base business, and an unproven one on the growth story.

10. Shareholder friendliness index

Dividends. ams-OSRAM has not been a dividend payer over the last three years. Given the heavy losses (a net loss of roughly EUR 1.6 billion in FY2023 driven by a EUR 1.3 billion goodwill impairment, followed by the 2024 microLED write-offs), no ordinary dividend was paid for FY2023, FY2024 or FY2025. Capital preservation and debt reduction, not distribution, have been the entire priority. (Any "dividend" figure appearing on US ADR aggregator pages reflects ADR mechanics or historic pre-2023 payments, not a current ordinary distribution.)

Buybacks and dilution. There have been no share buybacks; the last three years ran hard in the opposite direction. In late 2023 the company executed a heavily discounted rights issue of about 724 million new shares at CHF 1.07, raising roughly CHF 800 million (~EUR 802 million) to shore up the balance sheet, sharply increasing the share count. A 10:1 reverse share split followed in September 2024, which cut the nominal count for cosmetic/price reasons but did not return capital. MoatMap records zero buybacks in the trailing ~90 days (since 4 Apr 2026), consistent with the multi-year picture: over the last three years the company issued equity rather than retiring it. The only "capital return" adjacent activity has been divestment proceeds (Entertainment & Industrial Lamps EUR 114 million, Infineon EUR 570 million, indie EUR 40 million) routed entirely to debt reduction, not to shareholders.

Verdict: Hoards capital (of necessity) - no dividend and no buyback in three years, and shareholders were diluted by a large 2023 rescue rights issue, because every euro is going to repair a balance sheet strained by the OSRAM acquisition and the microLED write-off.

11. Insider activities

Venue and source. ams-OSRAM lists on SIX; the authoritative source is the SIX Exchange Regulation registers (management transactions under Art. 56 Listing Rules, and significant-shareholder disclosures under Art. 120 FinfraG). The MoatMap disclosure database is the spine for the last 12 months below; the most recent entries (through 18 Jun 2026) are within the cross-check window and are consistent with SIX significant-shareholder filings.

Important nature-of-the-data caveat. Every transaction MoatMap captured over the last 12 months is a significant-shareholder threshold disclosure (asset managers and banks crossing the 3%/5% FinfraG notification lines), not an open-market purchase by a director or officer. These are index funds, active funds and bank positions moving across disclosure thresholds - they carry almost none of the conviction signal that a CEO or board-member open-market buy would. There is no visible director or officer open-market buying or selling in the window, which for a Swiss-listed company of this size is itself the notable fact.

(All rows: SIX Art. 120 FinfraG significant-shareholder disclosures, dates as shown.)

Reading the "buys." The heavy cluster of BlackRock disclosures (12 crossings across March to June 2026) plus Deutsche Bank moving to 4.21% and Fidelity/FIL building toward ~4% reflect institutions repositioning around the stock as the Q1 2026 AI-photonics news drove a sharp share-price move (the stock reportedly jumped ~29% on the Q1 print). Rising institutional ownership is mildly constructive as a demand signal, but it is not insider conviction. None of these are managers buying their own company's shares with personal money. The FinfraG mechanism means a single fund can appear multiple times simply by oscillating across a threshold, which inflates the apparent activity count.

Sells. No material insider sells recorded in the window; the disclosures tagged "other" are position notifications, not directional exits.

Net assessment. Activity is broad-based among institutions and concentrated in one name (BlackRock) purely by disclosure mechanics. There is no director or officer open-market buying or selling to read a signal from, which means the strongest signal in this section is simply absent. Net read: neutral. Institutional accumulation around a re-rating is worth noting, but it should not be mistaken for insider conviction, and the lack of any management open-market purchase during a strategic inflection is a mild "watch this space" rather than an endorsement.

12. Scenarios

Bull case. The divestments close cleanly, Infineon's EUR 570 million lands, and leverage drops through the target toward the 2030 goal of below 2. Freed from balance-sheet fear, management proves the Digital Photonics thesis: the AR smart-glasses wave arrives, ams-OSRAM's black-package infrared emitters and eye-tracking illumination win designs at multiple headset makers at EUR 50-100 of content per device, and the customer base is genuinely diversified so no single cancellation can repeat 2024. The "slow and wide" micro-emitter interconnect graduates from development agreement to a real design win inside AI data centres, opening the high-triple-digit-million opportunity management has dangled. The record EUR 5 billion design-win backlog converts to revenue, the core semiconductor business compounds at mid-to-high single digits toward a 25%+ EBITDA margin, group free cash flow turns firmly positive in 2027 and climbs past EUR 200 million. The lamp cash cow declines gracefully, funding the transition exactly as planned. A company that spent three years in intensive care re-rates as a focused optical-photonics growth story.

Base case. Management keeps doing what it has done: hitting the midpoint of guidance, delivering the cost-out, cutting debt, and reporting a top line that softens on divestments and FX while the "core" grows quietly underneath. FY2026 profitability is dragged by the divestment one-offs, stranded costs and precious-metal and dollar headwinds management already flagged, so the reported numbers look uninspiring even as the underlying business tightens. Positive free cash flow arrives around 2027 roughly as promised. The AR-glasses content ramps modestly and the AI-photonics deal stays a development project with real revenue still over the 2030 horizon - a credible option, not yet a cash engine. The business ends up smaller, cleaner, less indebted and more focused than three years ago, a competently run mid-cap optical semiconductor and automotive-lamp company whose growth-story upside is still a call option rather than a delivered result.

Bear case. The cycle turns before the balance sheet is fixed. A consumer-electronics or auto downturn hits the now-smaller, more concentrated revenue base harder than expected, and dollar weakness compounds it. A divestment slips or clears at a worse price, leaving leverage stuck and refinancing costs biting. The AR-glasses opportunity underdelivers - headset volumes disappoint or a large customer changes its roadmap, echoing the microLED cancellation - and the AI-photonics bet, still years from revenue, absorbs cash and attention while entrenched datacenter-optics incumbents keep the sockets ams-OSRAM hoped to win. The savings programs deliver on paper but stranded costs and a shrinking lamp annuity erode the funding underneath the pivot. Instead of a focused growth story, the market sees a serially restructuring company that keeps selling revenue to survive, and the 2030 targets quietly join the microLED fab in the graveyard of ambitious plans.