Flexible OLED and the Race to Micro-LED: Samsung Display and CMOS Image Sensors

Every screen is a factory problem before it is a design problem. A phone owner never sees the power-management IC or the modem. They see the display, and everything else is judged by how the display behaves. Underneath, that display is a capital-intensive, yield-sensitive manufacturing job: thin-film transistor arrays deposited layer by layer over enormous substrates, patterned at sub-micron precision, and sealed against oxygen and moisture with a tolerance measured in single failed pixels out of tens of millions.

The same physics, run backwards, gives you the image sensor. This note covers flexible OLED using Samsung Display as the worked example, where micro-LED stands in the source material, the CMOS image sensor as the mirror-image business, and what it all means for India. The facts come from the displays and optoelectronics chapter of my ESDM book, and I keep its gaps as gaps.

Three display technologies, three manufacturing logics

TFT-LCD is the older and still higher-volume technology, and it remains the backbone of televisions and monitors. Fabs are classed by glass generation. A Gen 8.5 fab processes mother glass of roughly 2,200 mm by 2,500 mm. A Gen 10.5 fab, built to maximise large TV panels per sheet, processes roughly 2,940 mm by 3,370 mm. Pixels do not emit light. A separate backlight shines through liquid crystal cells. The backlight has become a differentiator: MiniLED backlighting, with 1,000 or more independently controlled local-dimming zones, lets an LCD approximate the contrast and HDR performance of a self-emissive panel while keeping LCD's mature, cheaper manufacturing base.

AMOLED inverts the design. Each subpixel is its own light source, made of organic emissive material deposited on a thin-film-transistor backplane, with no backlight. A pixel that is off emits nothing, which gives true blacks, and the structure allows real mechanical flexibility. The backplane matters as much as the emitter. Low-temperature polycrystalline silicon (LTPS) and the newer low-temperature polycrystalline oxide (LTPO) backplanes are deposited on polyimide instead of rigid glass, and LTPO allows a refresh rate that varies from 1 Hz to 120 Hz depending on content. That is the mechanism behind an always-on display that barely draws power showing a static clock.

Micro-LED is a self-emissive technology built from inorganic light-emitting elements. The book names it as a category but carries no manufacturing cost, yield or shipment data for it. It exists in industry discussion and pilot production, and the source corpus does not document it as a mass-market volume category the way TFT-LCD and AMOLED are. So the "race to micro-LED" in my title is a race the book can describe only from the AMOLED side. I will not invent numbers for the challenger.

The stencil that decides AMOLED cost

The organic layer is patterned with fine metal mask (FMM) thermal evaporation, and this step is the biggest reason AMOLED yield and cost behave so differently from a conventional semiconductor process. The mask is a physical stencil, typically an invar alloy sheet with an opening for every subpixel, held close to the substrate while red, green and blue organic material is evaporated through it in three separate passes, under ultra-high vacuum. Any thermal expansion mismatch between mask and substrate, any sag across a large panel, or any misalignment between the three colour passes turns directly into a visible pixel defect. There is no downstream correction step of the kind photolithography offers.

That is why FMM patterning has historically capped practical AMOLED substrate sizes well below the Gen 8.5 and Gen 10.5 glass that LCD fabs already run. Extending self-emissive manufacturing to larger sheets, the step that would let AMOLED compete in televisions, remains an ongoing engineering problem rather than a solved one.

Capital and concentration

A modern Gen 8.6 AMOLED fab needs roughly $3 billion to $4.5 billion of capital expenditure, the same order of magnitude as a leading-edge semiconductor fab. That shapes where panels are made. More than 90% of global panel fabrication capacity sits in China (BOE, CSOT, Visionox), South Korea (Samsung Display, LG Display) and Taiwan (AUO, Innolux). The global flat-panel market is roughly $120 billion to $140 billion. AMOLED now accounts for more than 50% of premium smartphone panels shipped, per DSCC tracking, even though TFT-LCD keeps the larger area in televisions and monitors.

India's policy response is direct. Under the India Semiconductor Mission the government offers a 50% central capital subsidy for companies setting up AMOLED or TFT-LCD display fabs, and Gen 6 AMOLED facilities have been proposed for evaluation in Telangana and Gujarat. That is a different ambition from the SMT assembly and box-build work India has already scaled. A display fab is front-end, semiconductor-adjacent manufacturing with wafer-fab-like demands on cleanrooms, yield discipline and ultra-pure water. Samsung's existing operations in Noida are assembly and finishing, not panel fabrication. Domestic Indian panel output is marked evidence required in the book, pending MeitY display fab approvals. The Gen 6 projects are proposals under evaluation, not operating capacity.

Samsung Display: flexible OLED as a worked example

No company shows the AMOLED stack more completely than Samsung Display, the South Korean subsidiary that has for over a decade been the leading producer of flexible OLED panels. Its position is described by two numbers, both accurate, pointing in different directions.

Across the broader medium and small OLED category, covering phones, watches, laptops and automotive displays, Samsung Display holds roughly 55%, verified via Omdia, with a wider range of 50% to 65% in the same source's framing. In the smartphone-only OLED segment its share is closer to 28%, per Counterpoint Research data from 2025, as BOE, Visionox and CSOT compete harder there. Read together, it is still the leader in OLED broadly, while the highest-volume application has become genuinely contested.

Five stages from glass carrier to bendable module

The backplane at stage 2 is the most consequential engineering choice. LTPO combines conventional LTPS transistors with IGZO (indium gallium zinc oxide) oxide transistors on one backplane. The hybrid exists to solve a power problem. An IGZO transistor leaks far less current than an LTPS transistor when holding a static image, which lets an LTPO panel drop to 1 Hz without visible flicker. An LTPS-only panel is locked at a fixed 60 Hz or 120 Hz whatever is on screen. Samsung Display's own specifications put the saving at 15% to 20% versus an equivalent LTPS panel. In a phone or watch, where the display is typically the largest power draw, that matters.

One more component is manufactured and integrated as part of the same product: the display driver IC. It takes the digital pixel data from the phone's application processor and turns it into the analog gate and data voltages that set each subpixel's brightness. On a flexible OLED panel the driver is typically bonded straight onto the flexible substrate using chip-on-film or chip-on-plastic packaging instead of sitting on a rigid board, a choice forced by the same flexibility requirement that drives the polyimide and LTPO decisions. Display manufacturing and semiconductor packaging have converged in the flexible-panel era.

Ultra-thin glass and the panel generations

Ultra-thin glass (UTG) is what makes Samsung's foldables durable instead of merely bendable. At roughly 30 micrometres, verified against both Schott and Samsung Display's specifications, it replaces the plastic cover window of earlier flexible panels with a real glass layer thin enough to fold, giving the Galaxy Z Fold and Z Flip a scratch-resistant surface closer to a normal phone. LTPO plus UTG is rated for more than 200,000 fold cycles.

The last row is the closest the book comes to the "post-OLED" frontier, and it is worth noticing what it is: OLED on a silicon wafer, not micro-LED. The customer list is also telling. Samsung Display supplies flagship panels to Apple for the iPhone, to Samsung Electronics' own Galaxy Ultra and Z Fold lines, and to premium automotive makers building digital cockpit displays. An outside competitor and an internal sibling sit side by side as customers. Samsung Display behaves as a semi-independent supplier within the group, winning outside business on manufacturing merit.

The mirror business: CMOS image sensors

A display turns an electrical signal into light. The reverse job, turning light into a signal a processor can use, belongs to the CMOS image sensor (CIS). Modern sensors use back-illuminated (BSI) and 3D-stacked architectures, where the pixel array, the supporting logic and, in the most advanced parts, a dedicated DRAM layer are made as separate silicon tiers and bonded together. Stacking lets pixel counts and readout speed keep climbing as pixels shrink, because logic and memory no longer compete with the photodiode array for the same silicon area.

The market is led by Sony Semiconductor Solutions with over 45% of global CIS, built on the BSI-stacked architecture Sony pioneered commercially, alongside Samsung's ISOCELL, Omnivision and SK hynix. The most visible use is the phone camera, but the same architecture underpins in-cabin automotive cameras, security systems and industrial machine vision at different resolution and cost points. The fabrication shares equipment, cleanrooms and process skill with the advanced-node and packaging capability those companies already run for other products, which is why CIS is a strategic adjacent business for logic and memory players.

Samsung is on both sides of the electron-photon interface. Its System LSI division makes ISOCELL sensors, including new 200-megapixel and 50-megapixel parts launched around the fourth quarter of 2025, while the same parent supplies roughly 55% of medium and small OLED. The image sensor business is smaller and more contested than the display business, and the System LSI unit faced seasonal softness in the same period per company commentary. Both are precision semiconductor manufacturing built around the same interface, running in opposite directions.

A quieter category sits below both: discrete indicator LEDs, infrared emitter and detector pairs for proximity and remote sensing, and photodiodes for ambient light and optical bio-sensing. They turn light into current and current into light exactly as a pixel does, one component at a time. A wearable's green and infrared LED and photodiode pair for heart-rate and blood-oxygen sensing is the example. The book has no market size, volume or supplier concentration data for this category, so I treat it as a real presence and a documented gap. The same is true of optical transceivers, which the book names in its taxonomy without market detail.

What pulls the stack into volume: wearables and India's assembly base

Displays and optoelectronics reach volume through the products that need them, and wearables and hearables have pushed miniaturisation hardest. Smartwatches and true wireless earbuds are among the highest-volume consumer categories, with aggressive annual refresh cycles, tiny form factors, low-power designs and unforgiving cost targets. The book cites IDC for more than 500 million wearables shipped a year.

The packaging that results is a high-density system-in-package under 10 mm by 10 mm, integrating a Bluetooth Low Energy SoC, a power-management IC, MEMS sensors and the passives in one moulded package small enough to vanish into an earbud stem or watch case. Power comes from miniature lithium-ion coin cells of 30 mAh to 300 mAh, driven by ultra-low-quiescent-current PMICs. Because these products sit against skin and meet sweat, rain and submersion, they have to hold IP67 or IP68, achieved with liquid silicone rubber overmoulding and conformal nano-coatings on the board. The adjacent smart-home category pulls a related set: multi-protocol microcontrollers for Matter, Zigbee, Wi-Fi 6E/7 and Thread, and voice SoCs with MEMS microphone arrays and Class-D amplifiers.

India has built a real position here. Boat, Noise and Fire-Boltt have worked with EMS providers Dixon Technologies, Optiemus Electronics and Syrma SGS to localise assembly under the Phased Manufacturing Programme. More than 75% of domestic wearable sales are now met by domestic assembly, per ICEA and IDC India. The book is careful about what this does and does not show. It supports the assembly value chain of brands, EMS partners and the PMP framework. It does not give the average component count of a TWS earbud, which is marked evidence required pending a teardown analysis. So the 75% describes where assembly happens, not how much of the bill of materials is Indian.

What to take away

• Display fabs are semiconductor-scale bets: $3 billion to $4.5 billion for a Gen 8.6 AMOLED fab, with over 90% of capacity in China, Korea and Taiwan.
• Samsung Display holds about 55% of medium and small OLED but about 28% of smartphone OLED. Both figures are right, and the gap is the Chinese makers closing in.
• LTPO (15% to 20% power saving) and ultra-thin glass (200,000+ folds) are the two differentiators behind flagship and foldable panels.
• Micro-LED is a named category with no cost, yield or shipment data in the source. What the book does show at the frontier is OLED-on-silicon for headsets, not micro-LED volume.
• Sony holds over 45% of image sensors, and the sensor and the display are the same electron-photon physics run in opposite directions.
• India's wearable win is assembly, not panels or sensors. A Gen 6 display fab remains a proposal under evaluation.

The lesson for anyone sizing a display opportunity is to ask which side of the fab door the activity sits on. Assembly is scaled and real in India. Panel fabrication has a 50% subsidy on offer and no confirmed capacity yet.