報告_UBS_光寶科2301_首次覆蓋_20260914

PDF 原檔：報告_UBS_光寶科2301_首次覆蓋_20260914_original.pdf

圖片清單（已驗證 2026-09-16）

原始內容

Initiation of Coverage

Lite-On Technology

A stronger contender in AI power

Lite-On better positioned to capture AI growth; Initiate at Buy, PT NT$400

Lite-On is a Taiwan-based power electronics supplier spanning cloud power, optoelectronics and auto, increasingly focused on AI server power. We expect Lite-On to capture more of the expanding AI power electronics opportunity as execution improves, its customer base broadens and margins rise. Cloud power (48% of our 2026 sales forecast) should benefit from higher content per watt alongside growth in GW of AI hardware deployed. We forecast the AI power TAM to grow at a 69% CAGR to US $91.2bn in 2030 (link), supported by transitions to 400/800V DC architectures. These drivers underpin our above-consensus sales and earnings forecasts. We initiate coverage with a Buy rating and NT$400 target, implying 20x 2028E EPS versus peers at 22x.

Late entrant gaining momentum; ASICs open up its power-rack opportunity

Lite-On has improved execution, securing NVIDIA approved-vendor qualification and supplying three of the top five CSPs, with a leading position in power supplies for one of the ASIC servers. Our checks point to a first dedicated 48V ASIC power rack ramp from late 2026, followed by a substantially higher-content 800V/MW-class architecture with the next generation from late 2027. We expect penetration into the remaining two hyperscalers within 12-18 months. Cloud & AIoT revenue growth accelerated from -2% in 2024 to +48% in 2025 and +60% in 2026 YTD as design wins scaled. Stronger focus, R&D investment and balance sheet support should help Lite-On sustain a top-three position in data-centre white-space power and capture AI-driven growth.

Portfolio repositioning sets up the next phase of margin expansion

Lite-On has restructured its portfolio toward higher-growth, higher-margin markets. Divestitures lifted group gross margin from the mid-teens to approximately 20% in 2021. We expect AI to drive the next phase of expansion, with Delta's AI ramp illustrating the potential benefits from a richer product mix and operating leverage. LiteOn's margin expansion should accelerate in 2026-27 as AI exposure increases, with gross margin potentially reaching 27-29%. Cloud & AIoT OpM of 20%, vs. 9% for Opto and 7.5% for IT/CE, highlight the opportunity for further mix improvement. Greater operating efficiency in the server business should also support operating margins, while R&D spending remains at 5-6% of sales.

Valuation: AI growth supports a higher trading multiple

Our NT$400 price target applies 20x to our 2028E EPS of NT$20, equivalent to 25x our 2027E EPS of NT$16 versus peers at 29x. Lite-On has historically traded at 8-14x P/E, but we believe its transition into a key AI power supplier warrants a higher multiple. Our EPS grows at a 43% CAGR to NT$20 from 2025-2028E (vs. 57% for Delta) from the combination of AI revenue growth and margin expansion, supporting the case for further re-rating. Growth is supported by higher capex at NT$18bn in 2026 (vs. NT $4.6bn past 3 yrs avg.) to more than double its capacity by the end of 2026.

Source: Company accounts, LSEG Eikon, UBS estimates. Metrics marked as (UBS) have had analyst adjustments applied. Valuations: based on an average share price that year, (E): based on a share price of NT$ 277.00 on 11-Sep-2026 21:27:42 CST

Equities

Taiwan

Diversified Technology Services

Trading data and key metrics

Randy Abrams

Analyst randy.abrams@ubs.com +886-2-8722 7338

Irene Chang

Associate Analyst irene.chang@ubs.com +886-2-8722 7280

Lite-On Technology - 2301. TW Price

Pivotal Questions

100

2024

UBS VIEW

EVIDENCE

WHAT´S PRICED IN?

Upside/Downside Spectrum

Company Description

2025

TWD 277.0

11 Sep

540.0

Upside to Downside

Upside:

37.6 to 1

+95%

Base:

Thesis Map UBS Research THESIS MAP a guide to our thinking and what´s where in this report

270.0

Downside: -3%

Q: Can Lite-On better participate in the industry's strong AI power electronics growth?

After lagging early in the AI cycle, Lite-On has improved execution, securing NVIDIA qualification and supplying 3 of the top 5 CSPs, including leading on Trainium servers and targeting shipments to the other top 2 hyperscalers in the next 12-18 months. Cloud & AIoT growth accelerated from -2% YoY in 2024 to +48%/+60% in 2025/'26 YTD, supported by design wins, R&D, and capacity expansion. Lite-On can remain a top-three white-space supplier as AI TAM grows. However, Delta Electronics (2308.TW) should retain ~50% share given incumbency, R&D, technical support, and ecosystem ties.

Q: Is Lite-On able to further improve its margins?

Lite-On gross margin has improved from 17% in 2020 to 23% in 2025, with OpM rising from 6% to 10%. We view this as largely self-help, trimming the legacy low-margin business, with the payoff from AI product only beginning to show. We expect GM can further improve to above 25% while OpM stays at 15-18% as AI exposure rises and the company becomes more vertically integrated, moving from a component supplier toward a system integrator.

Q: Would the mature business drag on Lite-On's operations?

We see the mature business as less important to growth rather than becoming a material drag on group earnings as Opto-electronic BU and traditional power supply BU lowers from 55% of 2025 sales to 43%/35% of 2026-27 sales. We also view this business growing single digits YoY as it offsets mature growth in PC/game console with growth into LEO (3% of sales but doubling YoY), industrial IoT for its optocouplers and LED into more Auto/EV applications and tuck-in acquisitions. The business is also a cash cow so on mild growth can maintain high-single-digit OpM.

We expect Lite-On to benefit from AI power electronics TAM expansion and margin improvement as execution, market penetration, and customer diversification improve. Cloud power, representing 48% of 2026 sales, should drive strong growth through content/W expansion and healthy growth in GW hardware deployments. These factors should support earnings growth above Street expectations. We initiate with a Buy rating and NT$400 target, based on 20x 2028E EPS.

Our bottom-up analysis of AI capex and supply chain inputs across ODMs, foundry, back-end, memory, and fabless/design services, plus rack upgrades for higher TDP , supports a strong power industry TAM outlook. At OCP APAC 2026, Lite-On showcased a full 1MW TDP rack solution after Q3 shipments of a lower-spec power rack solution to an ASIC vender, underscoring its shift from component supplier to system integrator. We expect the AI power market to grow at a 69% CAGR in 2025-30, with Lite-On among the top-three suppliers and gaining share faster than the industry.

The stock has rallied +66% YTD with the market re-rating it as part of the AI supply chain on its solid results +21.4% QoQ in Q226 as evidence. The stock now trades at 22x next-12-month consensus PE, above its historical range of 8-14x. However, we believe further upside exists supported by our global TAM growth and Lite-On's traction improved to captured the industry growth.

Source: UBS

Founded in 1975, Lite-On Technology is Taiwan's first listed electronics company and a global leader in power management and optoelectronic solutions. Initially a pioneer in LED technologies, it is now a key supplier of AI infrastructure, cloud computing, automotive electronics, 5G, and AIoT solutions.

Our Thesis in Pictures

Our Thesis In Pictures

Sources for exhibits above: Company data, UBS Research

Rising power density and AI build-out compound the AI server power TAM from US$6.6bn in 2025 to US$91.1bn in 2030E, a 69% CAGR, offering substantial upside for Lite-On

We believe Lite-On has improved its traction so it can better participate in the AI power industry as a late entrant; its business has been accelerating since 2025.

Optimizing business portfolio with higher AI exposure to capture the further growth upside

Lite-On margins have improved on divestment and the server mix increase is driving the next expansion

Executive Summary

We initiate coverage of Lite-On Technology with a Buy rating and NT$400 price target, based on 20x our 2028E EPS of NT$20.0, or approximately 25x our 2027E EPS of NT $16.0. The stock has already re-rated materially as investors recognize its improving participation in AI power, but we believe the change in the earnings base is structural. Lite-On has spent several years exiting lower-return businesses, improving its margin structure and redirecting investment toward Cloud & AIoT, where AI server power is now becoming a meaningful growth and earnings driver. Our target multiple remains below the roughly 22x level at which the broader power-supply peer group trades, reflecting Lite-On's later entry and still-developing customer breadth. We forecast revenue growth of 31%/38%/25% in 2026/27/28 and EPS growth of 63%/47%/25%, supported by a rising cloud mix and operating leverage.

Lite-On is moving from proving participation to scaling in AI power

The core change in our thesis is that Lite-On no longer needs to prove that it can enter AI server power; it increasingly needs to demonstrate how far it can scale. AI server PSU shipments began ramping in late 2024/2025, customer qualifications have broadened, and the company is moving beyond individual power modules into larger power shelves, BBU/energy-management systems and eventually dedicated high-voltage power racks. Importantly, our checks suggest the power-rack opportunity is already becoming tangible on custom ASICs: we expect Lite-On to begin ramping a dedicated 48V power rack for its top ASIC client from late 2026, before stepping up to an ~800V/MW-class architecture with the following generation from late 2027. Lite-On remains well behind Delta in absolute AI power scale and breadth, but our forecasts do not require it to close that gap. The base case assumes only modest GPU-server share and stronger positions in selected custom-ASIC programs, leaving additional CSP wins as potential upside rather than a prerequisite.

AI power growth is a watts story, not an 800V adoption call

Our companion AI Power report (link) estimates industry AI server power-electronics TAM rising from approximately US$6.6bn in 2025 to US$91.1bn in 2030, a 69% CAGR. AI server rack shipments grow only around 12% CAGR in our model; the much larger drivers are a roughly 39% CAGR in average rack power and 55% CAGR in annual IT GW deployed, supplemented by higher power-system content per watt. This distinction matters for Lite-On. We expect strong 2026-27 AI power growth even if NVIDIA's initial transition to dedicated 800V power racks is slower than previously expected. Higher accelerator shipments, the transition at its top ASIC customer to a richer 48V followed by 800V power rack, AMD's move toward rack-scale systems and increasing customASIC rack density should drive substantial growth before widespread native 800V adoption, which should start with its ASIC customer from late 2027. The 400/800V transition becomes a more meaningful incremental content opportunity from 2028 onward rather than the foundation of our near-term estimates.

ASIC exposure is a differentiator, while concentration is a risk

Lite-On is largely exposed to custom-ASIC power relative to the industry. We estimate ASIC programs contribute roughly two-thirds of its modeled AI power exposure over much of our forecast period, compared with custom ASICs representing roughly 4050% of industry AI power TAM. This creates both opportunity and concentration risk. ASIC systems are moving toward more accelerators per rack, higher BBU attachment and increasingly rack-level power architectures, all favorable to Lite-On's content opportunity. However, project timing, customer qualifications and share shifts can have a greater effect on Lite-On than on a more diversified supplier such as Delta.

Mix and leverage should support a second leg of margin improvement

Lite-On's first margin expansion came from restructuring and exiting lower-return businesses. We believe the next phase can be driven increasingly by product mix and operating leverage. Cloud & AIoT already represented approximately 55% of 2Q26 sales and should continue to increase as AI server power ramps. We forecast revenue rising to c.NT$219bn in 2026, NT$300bn in 2027 and NT$360bn+ in 2028, while gross margin remains in the mid-to-high 20s and operating margin expands as higher-value AI power products scale. Our EPS forecasts rise to approximately NT$10.9/NT$16.0/NT$20.0 in 2026-28E. The key debate, in our view, is therefore moving from whether Lite-On can participate in AI power to whether it can sustain customer wins, execute a large capacity build and convert rapid sales growth into the margin profile implied by our forecasts.

Initiate with a Buy rating and NT$400 target based on 20x 2028E EPS, vs. peer group at 22x.

The ramp of AI server shipments has begun. Lite-On does not need to prove its market presence; any new client wins could represent incremental upside.

AI server power-electronics TAM rising from approximately US$6.6bn in 2025 to US$91.1bn in 2030, a 69% CAGR.

Lite-On has higher ASIC than GPU server power exposure, vs. the industry roughly 40/60 mix.

The company's initial margin expansion was driven by the exit from lower-margin businesses and is now being supported by AI server opportunities.

Strengthening the core while pivoting to Cloud

Founded in 1975, Lite-On Technology is Taiwan's first listed electronics company and a global leader in power management and optoelectronic solutions. The company has spent more than a decade reshaping a broad electronics portfolio toward businesses where it has stronger technology content, customer stickiness and return potential. Following earlier consolidation of group entities, it exited or divested display, cameramodule, SSD and imaging activities and has increasingly concentrated resources around power management, connectivity and optoelectronics. The result is a fundamentally different company from the legacy PC/peripheral power supplier that investors historically associated with Lite-On.

Figure 1: Lite-On has streamlined its business through divestment and consolidation

Source: Company data

Cloud & AIoT is becoming the earnings center of gravity

Lite-On operates through three main business groups: Cloud & AIoT focused on the data center power and energy storate business, Opto-electronics consisting of visible and invisible LEDs, IT and consumer electronics which includes PC power supply, game console and new applications such as LEO power supply.

Figure 2: Key Product line: Cloud/AIoT driving expansion, with mild growth from IT/Consumer and Opto

Source: Company data

We expect Cloud & AIoT mix to increase as AI-server revenue ramps, while the IT/ consumer and optoelectronic businesses grow at more moderate rates. This is important not only for revenue growth but also for group profitability: the higher-value Cloud mix should progressively dilute the impact of lower-margin legacy consumer products.

Cloud & AIoT (55% of Q226 revenue)

Cloud & AIoT represented approximately 55% of 2Q26 revenue. This segment splits into two complementary businesses: Cloud, which targets data-center infrastructure buildout, and AIoT, which targets edge-AI connectivity demand.

1. Cloud-related power supply:

• ˜ Key Products
• ˜ Power supply: The company has a product portfolio spanning component (power supply units, DC-DC converters), module (power shelves, supercapacitors), and system (power racks).
• ˜ Energy management system : Lite-On provides an integrated battery backup unit (BBU) solution.
• ˜ Components : Cooling components, rack chassis.
• ˜ Key Application and Use Cases: The company is a key supplier of datacenter "white space" power systems, with a customer base spanning CSPs, NeoClouds, and enterprises. The company is migrating toward higher-end server power supply; within total cloud sales, >25% was coming from AI server product while the company is expecting the total AI mix to increase to 30% by the end of '26.

2. AIoT infrastructure

• ˜ Key Products
• ˜ Networking modules and solutions : The acquisition of U-MEDIA in 2026 expands Lite-On's total connectivity solution with FWA (Fixed Wireless Access), building on its existing 5G O-RAN and AI-RAN solutions.
• ˜ Smart/IoT devices : High-end video surveillance, access control, and alarm systems.
• ˜ Key Application and Use Cases : As AI compute and mobile network traffic scale, high-speed, low-latency connectivity for edge AI becomes essential. LiteOn offers enterprise, campus, and consumer solutions spanning hardware and software to capture this growth.

Opto-electronics (15% of Q226 revenue)

Opto-electronics is Lite-On's oldest business segment, the company started out as an LED supplier. Over time, it has actively reshaped this business group into three subsegments.

1. Visible light solution

• ˜ Key Products: automotive LED, SMD LED, flash LED, UV LED, LED display.
• ˜ Key Application and Use Cases: the company is targeting higher-margin use cases by leveraging existing client relationships, such as data center UV LED and automotive LED displays.

2. Invisible light solution

• ˜ Key Products: Photocoupler, optical sensor, InP components.
• ˜ Key Application and Use Cases: in addition to its legacy photocoupler and optical sensor business, Lite-On announced a strategic investment in DenseLight in 2026. The two companies aim to jointly develop optical engines and modules for 1.6T/3.2T optical communications and CPO architectures; revenue contribution isn't expected before 2H2027, so frame this as a forward-looking optionality thesis, not a near-term revenue driver.

3. Auto electronics

• ˜ Key Products: EV charger, ADAS solution, advanced Auto display LED.
• ˜ EV charger: the company offers both AC and DC EV charging solutions, with experience handling a wide power range (30kW-180kW modules) and multivoltage input compatibility (240V-480V).
• ˜ ADAS solution: comprehensive camera modules including in-cabin monitoring (driver/passenger status) and road-environment sensing (front/rear surround-view modules).
• ˜ Key Application and Use Cases: the company supplies charging solutions and ADAS modules to automakers in the US, EU, and China, enabling their smartmobility platforms.

Information Technology & Consumer Electronics (30% of Q226 revenue)

This segment is the company's cash cow, combining the legacy, cash-generating consumer-electronics power supply business with the high-growth LEO (Low Earth Orbit) power module business.

1. Consumer electronics

• ˜ Key Products: PC Power modules with a >40% market share, along with game console power modules and other consumer electronics products.
• ˜ Key Application and Use Cases: Lite-On partners with all major PC brands and game console suppliers.

2. LEO power module

• ˜ Key Products: ground-station/user-terminal power modules for LEO satellite networks, ruggedized outdoor design (IP67-level protection, surge resistance, extended hold-up time, Test-to-Fail validated against extreme conditions).
• ˜ Key Application and Use Cases: Lite-On is the primary ground-power supplier to the two leading LEO satellite constellation operators with >50% market share. Management guided 2026 LEO-related shipment volume to at least double YoY, with demand trending up quarter over quarter.

M&A to expand solution breadth in its business lines

The company's long-term strategy is centered on three key growth pillars: 1) Cloud and AI infrastructure, 2) Automotive 3) clean energy solutions. The company has been actively rounding out its product portfolio and expanding into high-growth markets through M&A.

Its M&A strategy focus on the key technology trends driving growth in the next decade: 1) AI/data-center power supply and energy management, 2) physical/edge-AI connectivity opportunities, and 3) industrial automation. Recent investments are intended to broaden the capabilities surrounding Lite-On's core power franchise. The DCX partnership adds liquid-cooling expertise that can complement power delivery in high-density systems, while DenseLight adds InP optical-component capabilities to LiteOn's existing packaging and integration strengths. We view these as opportunities to expand system participation and customer engagement, rather than assuming that investment announcements immediately translate into material revenue or earnings.

Figure 3: Selective M&A expands Lite-On's solution breadth around power, racks, cooling and connectivity

Source: Company data

Lite-On is better participating in a fastgrowing AI power market

Rising power density and AI buildout expand power TAM

The AI power opportunity is being driven by a fundamental increase in compute and power density at every level of the system. Accelerator package power rises as vendors integrate more compute dies and memory around each package; accelerator counts and scale-up bandwidth increase at the rack level; and the resulting rack moves from tens of kilowatts toward hundreds of kilowatts and ultimately megawatt-class designs.

Our companion AI Power analysis demonstrates why physical rack counts increasingly become the wrong denominator. From 2025 to 2030 we forecast AI server rack units increasing at only around a 12% CAGR, but average rack/system power at roughly 39%, producing a 55% CAGR in annual IT GW deployed. Power-system content per watt adds another approximately 9% CAGR.

Together these variables drive our AI server power TAM from approximately US$6.6bn in 2025 to US$91.1bn in 2030.

Figure 4: AI Power TAM growing at a 69% CAGR to US$91.1bn in 2030E

Source: UBS estimates, company data

Power conversion is becoming larger, shared and increasingly systemlevel

The linear relationship between compute density and power load has hit what we call the "Amperage Wall" as AI accelerators transition from the "server era" - discrete units consuming 500-1,000W - to the "AI Factory era," defined by rack-scale systems consuming 120-140kW. We see three structural challenges that make a fundamental redesign of power architecture increasingly unavoidable:

1. Current scales faster than power loss can be managed. As the current increases, resistive losses and heat generation rise exponentially.
2. Rising compute density leaves ever less physical space for power. The pursuit of higher compute density results in a massive thermal design power (TDP) load per rack, with the IT rack already packed with GPUs and cooling component, leaving little room for additional power-delivery system which encouraging the migration of power electronics into a dedicated sidecar power rack.
3. Synchronized GPU cluster create severe transient load spike. AI training workloads are highly synchronized. Thousands of GPUs/AI ASICs can shift simultaneously between utilization states, creating instantaneous spikes in power consumption within milliseconds.

We believe the industry is undergoing a structural changes with the 4 phases of data center architecture evolution (Figure 5 I A p o w e r d l i v s y f m P S U t a c n z h k ), AI power delivery is moving through a multistage transition rather than changing from today's 54V racks to native 800V in one step.

Lite-On therefore has several avenues to participate before widespread native 800V adoption. Its addressable content includes PSUs, power shelves, controllers, BBU/energy storage, protection, rack hardware and DC/DC products. Over time, dedicated power racks add further system-integration opportunity.

Ploss = I 2 R. This shows the exponential relationship - if you double the current (flow rate), the loss is quadrupled (friction/heat)

PHASE 1

Rack-level AC-DC

AC conversion remains inside IT rack.

GRID

415/480 VAC

UPS

PDU

AC to AC

IT RACK

230/400 VAC

48/50/54 VDC

AC/DC

PSU& Power

IBC/PDB

shelves

Rack Busbar

12/6 VDC

Sub-1V VDC

POL/VRM

GPU / CPU

Al load

Figure 5: AI power delivery evolves from server PSUs toward centralized shelves and dedicated power racks

PHASE 2

feed multiple IT racks

PHASE 3

Facility 800 VDC

PHASE 4

Source: UBS, Based on industry feedback at OCP Asia, Semicon Taiwan, Computex and GTC San Jose.

Lite-On is scaling quickly after entering AI server power later than Delta

The AI trend accelerated after ChatGPT was introduced in 2023, with Delta moving first and rapidly capturing approximately 60% of the AI power supply market, supported by its early presence in NVIDIA GPU racks. We estimate the company is now one of the few suppliers with meaningful qualifications across the high-power PSU, rack-shelf and BBU opportunity, with additional power-rack products in qualification.

Figure 6: Delta entered AI server power earlier; Lite-On began scaling rapidly from late 2024

Source: Company data, UBSe

Figure 7: AI server power becomes an increasingly meaningful share of Lite-On sales

Source: Company data, UBSe

By comparison, Lite-On entered this cycle as a diversified conglomerate, with legacy exposure in optoelectronics and consumer power, while its AI server power business only

POWER RACK

IT RACK

started to scale from end of 2024 as it started shipping PSUs into GPU servers followed by ASICs in 2205. Server power supply + 49% YoY in 2025 after two consecutive year of decline and the company now supplies 3 of the top 5 CSP's GPU demand. AI-related revenue exceeded 25% of total sales in 2Q26 and, with management targeting above 30% by the end of as new ASIC platforms ramp. We believe roughly 70% of cloud revenue comes from hyperscaler customers, with the rest from enterprise and neocloud deployments.

We model AI-related power revenue growing considerably faster than the broader company through 2026-28. As that happens, the contribution of legacy consumer power declines rapidly as a percentage of the total business even if those revenues remain broadly stable.

Figure 8: Delta retains greater scale, but Lite-On is gaining participation from a much lower base

Source: Company data, UBSe

ASIC projects anchor Lite-On's AI growth

Custom ASICs are a particularly important part of the Lite-On thesis. We estimate ASIC platforms represent ~65% of Lite-On's modeled AI power exposure over our forecast period, versus ~45% for the industry, indicating material over-indexing to ASICs. Exposure is concentrated in AWS, which we believe accounts for 35% of Lite-On cloud revenue, including traditional server and AI server power supply across GPU and ASIC servers. This is supported by a high BBU attach rate, with deployments approaching 100% BBU penetration alongside PSUs.

Figure 10: Industry ASIC mix stable at ~45%

Source: UBSe

Figure 11: AWS anchors the current ramp; additional ASIC customers to drive further growth

Source: UBSe

Figure 9: Lite-On now ramping its server exposure through 2025-26

Source: Company data, UBSe

In-rack

800VDC

We expect H226 growth to remain robust despite the elevated H126 base, as a new ASIC project begins scaling in 3Q26. The ASIC 48V power rack is Lite-On's first complete rack-level solution and carries a higher margin profile than standalone PSUs. We view this as a meaningful step up the value chain, shifting Lite-On from component supplier to system-level solution provider by integrating power components, rack infrastructure and proprietary software.

8.5kW

The attraction of ASIC exposure is not merely unit growth. AWS, Google, Meta and other hyperscalers are all moving toward denser racks, more integrated scale-up systems and higher system power. These changes increase the need for larger power shelves, distribution, storage and eventually external power racks. The risk is concentration. Delays in a single major ASIC program or a share shift at an important customer can materially affect Lite-On's annual growth. This is why we treat additional CSP wins as upside rather than embedding aggressive diversification in our base case.

Near-term growth comes from 48/54V power, but transitioning toward HVDC 2026

Lite-On is positioned relatively early in the industry's move toward higher-voltage DC architectures, but we would separate this longer-term opportunity from the near-term earnings thesis. The company's roadmap progresses from 5.5kW-class PSU modules and 33kW rack shelves toward 8.5-18.3kW modules, approximately 110kW-class shelves and eventually 400/800V dedicated power racks. Its demonstrated architecture combines power conversion, rack distribution, protection, BBU/energy storage and downstream DC/DC.

Figure 12: Lite-On's roadmap moves from 33kW shelves toward 110kW conversion and MW-class power racks

Source: Company data

Lite-On's 800VDC power rack is in qualification, with sample shipments expected in 4Q26 and some revenue for the Rubin generation from 2027 and then ramping more in 2028 with its top ASIC customer. Lite-On demonstrated its rack solution, including:

1. PSU module: PSU capacity was 5.5kW for Blackwell and is expected to upgrade to 18.3kW for Rubin, while content per watt remains broadly stable at US$0.120.13. As installed capacity rises with higher IT rack TDP , overall power content per PSU could increase by ~80%.
2. Power shelves: The power shelf hosts the PSU module and PMU, taking facility AC input and outputting +/-400V or 800V DC. System integrators such as Lite-On typically supply the chassis, backplane, and cooling system, increasing captured content value.
3. Battery Backup System: BBUs are designed to bridge sudden power failures, providing up to 45 seconds of runtime during AC loss before generators start.

Power Rack

800VDC

660kW

Power Rack

2.5kW

DCDC Brick

50 VDC Input

12 VDC Output

Supercapacitors offer an alternative, addressing millisecond-scale load transients via ramp-rate control and smoothing GPU power fluctuations.

1. DC/DC convertor: While the power rack delivers HVDC to the IT rack, additional Phase 2 IT rack shelves step down high voltage to the lower voltage required by compute nodes. Current versions convert 800V to 54V, with potential to move further to 6V before compute nodes can be HV native fed.

Figure 13: Lite-On demonstrated it HVDC power rack solution

Source: UBS

Energy storage adds content, but architecture determines where that value sits

For Lite-On, BBU and CBU is both an opportunity and an architectural swing factor. Today, high BBU attachment on some hyperscale platforms materially increases content per rack. Our industry work estimates the combined BBU/CBU market grows substantially faster than the broader server-power market as AI load density rises.Longer term, pooled LFP UPS/BESS architectures could shift some battery value away from individual IT or power racks, but power-electronics demand should remain, given requirements for battery management, bidirectional conversion, controls, protection, and DC distribution. Suppliers with broader system-level capabilities may capture more value than vendors limited to standalone rack-level battery modules.

Power Racks: ASIC provides the first tangible ramp, ahead of NVIDIA

Our Power TAM work suggests NVIDIA's initial shift to dedicated power racks may be slower than expected. Vera Rubin/Oberon systems at ~225-230kW can still be supported by several hundred kilowatts of installed 54V rack-level shelf capacity, reducing near-term urgency to externalize conversion and giving operators more time to qualify high-voltage infrastructure and safety standards.

This is not seen as particularly negative for the overall Lite-On thesis. The bulk of 202627 growth in our model comes from increasing AI accelerator deployment, larger rack power, higher-capacity PSUs and shelves, BBUs and richer ASIC systems-not from widespread native 800V adoption.

Our feedback suggests Lite-On should begin ramping dedicated power racks at its top ASIC customer from late 2026, We expect the initial architecture to remain 48V, with

~255kW of PSU capacity and ~170kW of BBU capacity supporting a 72-ASIC system. Despite remaining at 48V, moving power conversion and backup power into a dedicated rack materially increases Lite-On's content, which we estimate rises from ~US$54k for the comparable rack-level architecture to ~US$117k per power rack, with content/W increasing from ~US$0.75 to ~US$0.98.

The larger step is expected with Amazon's 4th Gen system, which we believe will launch in late 2027. Based on multiple industry supply-chain conversations, we understand Amazon is targeting an ~800V architecture and materially higher rack power. Our expectation is driven in part by feedback suggesting a transition from the current 2-die chipset configuration to a higher-power 4-die design. Lite-On has been developing an approximately 1.2MW power rack capable of supporting two IT racks, which could lift system content to ~US$306k at roughly US$0.93/W.

Importantly, this means slower near-term adoption of dedicated power racks on NVIDIA does not invalidate the opportunity. Trainium provides a tangible path to commercialization first, while NVIDIA, other ASIC platforms and ultimately 800V architectures broaden the addressable market thereafter.

Figure 14: Lite-On content per W and total content ramping with ASIC 48V/800V power racks

Source: UBS

Beyond the Trainium ramp, two main platforms should drive power rack adoption:

1. OCP's Mount Diablo platform was the first widely adopted open-standard implementation, initially targeting approximately 72kW of power-rack capacity. Based on our industry feedback, the power rack is in the qualification stage right now with the sample rack shipping in Q4 and mass production in H127.
2. NVIDIA's own reference architecture is materially more aggressive, centered around a 660kW power-rack design incorporating six 110kW rectifier shelves and 480kW of battery backup capacity. While OCP introduced the architecture earlier to establish an open ecosystem and accelerate adoption, NVIDIA's higher-power implementation is designed for the next generation Rubin Ultra, given that the current power shelves design can still support the Vera Rubin generation.

We think industry concerns over NVIDIA's 2027 power rack implementation delays should have a limited impact on Lite-On, given its stronger relative position in ASIC. We expect robust growth, supported by strong GW of compute deployments and healthy ~60% power upgrades on the VR200 platform. Our sensitivity analysis indicates NVIDIA power-rack penetration has only a modest effect on 2027 power TAM, with our baseline assuming limited 10% penetration. A slower transition mainly shifts content mix and timing rather than reducing underlying watts.

Figure 15: 2027 NVIDIA sensitivity to power rack adoption in 2027

Source: UBS

Figure 16: 1% impact on the power TAM for each 10ppt swing in power rack adoption in 2027

Source: UBS

By 2028 and beyond, the architecture becomes more important. Higher-density NVIDIA generations, denser ASIC racks and eventual movement toward 400/800V distribution increase the amount of external conversion, protection and distribution content. For Lite-On, qualification in these systems therefore represents a meaningful opportunity.

Figure 17: Our base case TAM assumption suggest a 10% power rack penetration will bring TAM to US$38.2bn in 2027

Source: UBS

Existing client engagement supports our base case; diversification creates upside

We believe Lite-On is rooting its presence within the AI power industry with 3 of the top 5 CSP engagement and also Neocloud and Enterprise order. We expect growth from GPU/ASIC to support Lite-On cloud revenue to grow +74% YoY in 2026 and remain robust at +67% in 2027, supported by next-generation ASIC platforms with higher power requirements, increasing adoption of NVIDIA's high-voltage rack architecture, and additional CSP customer wins.

Our base case assumes Lite-On captures 3-4% of total GPU server power supply market and 17-20% of AWS ASIC server power supply, taking AI power revenue to NT$27bn/ NT60bn in 2026/27E, +177%/123% YoY. We also factor in additional upside from other ASIC vendors, as we believe Lite-On could establish engagements with Google in 2027 and Meta in 2028, lifting company sales to another level.

Figure 18: Lite-On should see ASIC ramp faster than GPU power supply

Source: UBSe. * The market share includes all power content (PSU, power shelves, power management system, power distribution and energy storage and back up system)

Our upside case, which considers Lite-On winning further ASIC market share from new opportunities tied to new CSP clients' GPU server demand, brings 2026-27E AI server revenue to NT$28.5bn/NT$73.4bn,implying upside of 6%/22% to sales and +NT $0.15NT$1.03 to our 2026-2027E EPS.

Figure 19: Lite-On upside scenario: 2% higher AI share in ASIC and GPU projects drives 40-50% higher AI power sales

Source: UBSe

Our downside scenario assume Lite-On secure NVIDIA qualification later than expected and takes a smaller share during the initial HVDC power ramp, alongside weaker tracton on other ASIC projects. This brings 2026/27 sales to NT$22/NT$48.2bn in 2026/2027, implying downside of -19%/-20% to sales and -NT$0.41/NT$-1.01 to our 2026-2027E EPS.

Figure 20: Lite-On downside scenario: late qualification and 1-2% share loss is 20-30% downside to AI power sales

Source: UBSe

Financial outlook: Cloud power drives a larger, more profitable earnings base

Lite-On is entering the next phase of its business transformation: portfolio restructuring has improved the underlying margin base, while the expansion of cloud power now provides the catalyst for revenue and earnings growth. Earlier disposals reduced exposure to lower-return businesses; the next stage depends on scaling customer programs, increasing power content supplied per deployment and converting the growing cloud mix into operating leverage. Our forecasts are principally driven by cloud power, with more moderate growth in optoelectronics and IT/consumer applications.

We forecast revenue rising from NT$166.1bn in 2025 to NT$218.1bn in 2026, NT $299.8bn in 2027 and NT$373.6bn in 2028-growth of 31%, 38% and 25%, respectively. Gross margin expands from 22.9% in 2025 to 25.2%/26.4%/26.6% in 2026-28, while operating margin increases from 10.1% to 13.5%/15.3%/15.5%. This supports EPS of NT$10.85/NT$16.00/NT$20.00, representing approximately 44% compound annual growth over 2025-28 versus 31% revenue growth. We view the earnings opportunity therefore as not just a larger addressable market, but a greater proportion of revenue coming from higher-value power products.

Importantly, the margin forecast does not extrapolate the exceptional 27.2% gross margin achieved in 2Q26. We model a more measured improvement as cloud products scale, allowing for continued engineering investment, capacity expansion and the costs of introducing new systems. The key execution test is whether Lite-On can translate qualifications and customer demand into repeatable volume production while maintaining pricing discipline and manufacturing efficiency.

Figure 21: Lite-On sales accelerating from 2025 at expanding margins

Source: Company data. UBSe

Figure 22: Lite-On mix shifting to higher growth cloud/ AIoT

Source: Company data. UBSe

Lite-On revenue model builds up the key drivers by product

Our revenue forecast is built by product and business line: Within cloud, we model power supplies and power shelves, BBUs, cooling, rack-related hardware and other products, alongside the AIoT/connectivity business. Power supplies and shelves remain the largest revenue contributor, while storage and the expansion into broader system content provide additional growth. Cooling and rack products broaden the opportunity without needing to become the principal earnings drivers in the near term. We forecast the cloud subtotal rising from approximately NT$60.5bn in 2025 to NT$105.5bn in 2026, NT$176.0bn in 2027 and NT$242.1bn in 2028. Including AIoT, segment revenue increases from NT$75.1bn to NT$123.5bn/NT$195.8bn/NT$262.7bn, lifting its share of total revenue from approximately 45% in 2025 to 57%/65%/70%.

This distinction matters for the investment case: legacy businesses do not need to shrink for the portfolio to improve. Faster cloud growth improves mix, while higher-value products and operating efficiency increase the earnings contribution from that mix.

Figure 23: Cloud & AIoT drives approximately 90% of incremental revenue in our 2025-28 forecast

Source: UBSe, Company data

Margins improved on divestment; Server mix drives the next upgrade

Lite-On's margin expansion began in 2018, following the divestiture of its casing and camera module businesses, and subsequently its image BU and SSD businesses in later years. These actions lifted the company's gross margin from the mid-teens to approximately 20% in 2021. We believe the next stage is different. Rather than relying primarily on further divestments, margin expansion should increasingly come from richer product mix, scale and operating leverage. Higher-power AI products are not simply larger versions of commodity PC PSUs. Moving from roughly 5.5kW modules toward 18kW+ modules requires higher power density, three-phase input, more demanding thermal management, dynamic response, redundancy, telemetry and system coordination. Power shelves, BBUs and rack-level systems add further control, protection and integration content.

Figure 24: AI mix and operating leverage support further margin expansion

Source: Company data, UBSe

This phase of margin is already best illustrated by Delta. Delta's GM/OPM had remained stable in the mid-twenties/teens; however, as AI shipments began to scale, GM

expanded to above 30% and is trending toward the 40% range, while OPM accelerated from approximately 10% to 20% in 2024-25 and is likely to remain in the low-twenties going forward. We expect a similar expansion trajectory for Lite-On which began in 2025, with further acceleration in 2026/27 as exposure to higher AI gross margins increases. We believe Lite-On could scale its gross margin toward 27-29% and expect the company's operating margin to expand as operating efficiency improves across its server businesses - Cloud/AIoT at 20% (vs. Opto at 9%, and IT/CE at 7.5%). Meanwhile, R&D expense should continue to increase in line with sales at 5-6%, reflecting the company's sustained investment in new products.

Q2 results inflected with stronger participation in the DC growth trend

Lite-On Q226 sales +21% QoQ, +30% YoY to NT$52.7bn with gross margin expanded 6.5ppts QoQ to 27.2% driven by the AI power system. The company current revenue mix was 55% of Cloud&AIoT, 15% of Opto-Electronics and 30% of IT/CE. Cloud & AIoT led the growth, up 73% YoY in Q2, and is also the company's most profitable segment at an 11% operating margin, contributing 50% of total company operating profit, while Opto and IT/CE operated at 10% and 9% operating margins, respectively. Q2'26 EPS came in at NT$3.14, up 90% QoQ and 126% YoY - underscoring the impact of the company's continued progress on business restructuring and its evolution into a solutions provider within the data-center power industry.

H226 growth momentum should continue with 2027 ramp.

We expect H2'26 growth to be driven by a strong ASIC server ramp, bringing Q3 sales to +10.5% QoQ and full-year sales to +31.3% YoY, to NT$218bn. The next-generation 8.5kW PSU should ship with BBUs at a 1:1 ratio, driven by Trainium 3, while the 800VDC rack supporting the GPU platform should complete qualification in Q4, with smallvolume shipments in late 2026 - bringing Cloud & AIoT growth to 43% HoH / 79% YoY in H2'26. We expect the LEO business to grow from a low base, doubling YoY in H2'26 to reach approximately 3% of total company revenue. Importantly, the company has not seen material shortages or component pricing pressure affecting its core business, having entered long-term agreements (LTAs) with its GaN and MOSFET suppliers, backed by the long order visibility provided by its CSP client relationships.

Figure 25: Lite-On '26/'27/'28 sales +31%/38%/25%, EPS at NT$10.8/16.0/20.0

Source: Company data, UBSe

After years of restructuring and shifting toward higher-margin businesses, Lite-On's Q2 GM improved 5.5ppts to a record 27.2%, though this partly reflected power pushed out from Q1. We see the company sustaining GM above 25% as the data-center power product mix continues to rise. The risk around component price hikes looks manageable as the company has locked in sufficient component supply at reasonable pricing through the end of 2026. For upcoming projects it is able to renegotiate pricing with clients based on higher input costs to support margins. We believe the company has now passed the transition period, with every product line profitable and starting to benefit from economies of scale alongside improving efficiency. We forecast the company to deliver EPS of NT$10.85 /NT$16.00NT$20.00 in 2026-28.

Lite-On balance sheet supports its capacity expansion

On its Q2 earnings call, Lite-On increased its 2026 capex guidance to NT$18.0bn from the prior NT$13bn, significantly above its 2021-25 average of approximately NT$4.3bn, or 2.7% of revenue. e manufacturing flexibility outside China with additional capacity in US and Taiwan. We view the willingness to invest materially ahead of demand as evidence of customer visibility, while recognizing that under-utilization would become a risk if AI deployments slow materially.

Figure 26: financial summary - revenue, capex, free cash flow, and dividend trends

Source: Company data, UBSe

The company has historically maintained substantial net cash and a high dividend payout, providing room to fund the current expansion without compromising financial flexibility. Lite-On's net cash position, which moved within a NT$24bn-39bn range through the 2010s, equivalent to net cash per share of roughly NT$10-17, is expected to decline during the 2026-28E capex ramp, with net cash per share falling toward approximately NT$6-7.

The company's historical payout ratio of around 70% is likely to be maintained, with dividend per share projected to increase to NT$8.11 in 2026E from NT$4.52 in 2025 and remain at NT$7.72 in 2027E, despite deeply negative free cash flow that year. We view this as a credible indication of management's confidence in the medium-term returns from this investment cycle, supported by a historically strong balance sheet.

Figure 27: Capex steps up sharply in 2026 to support AI power capacity

Source: Company data, UBSe

Figure 28: Lite-On drawing down net cash to accelerate capacity additions

Source: Company data, UBSe

What´s Priced In?

WHAT´S PRICED IN?

Source: Refinitiv Workspace

Valuation: The re-rating is partly done, but room for earnings to catch up

We initiate coverage of Lite-On with a Buy rating and a PT of NT$400, applying a 20x multiple to our 2028 EPS estimate of NT$20, or 25x our 2027 EPS forecast of NT$16, compared with peers trading at 29x. Historically, the stock has traded in an 14-20x P/E range. However, we believe Lite-On is evolving into a key AI supplier and is better positioned to benefit from the rapid expansion of the AI power TAM. This underpins our 2026/27/28 EPS forecasts of NT$10.85/NT$16.00/NT$20.00 and an EPS CAGR of 67.1% over 2025-29E.

Figure 30: Lite-On has re-rated toward Delta as investors recognize the AI power transition

Source: Refinitiv

Figure 31: Lite-On stock price vs. earnings

Source: Refinitiv

The stock has risen sharply as the market has begun to recognize Lite-On's role in AI power, and we believe the re-rating is supported by the earning expansion. At approximately 24x next-12-month consensus earnings at the time of writing, Lite-On trades above its own historical range of 14-20x but remains below the roughly 29x level of the power-supply peer group. Relative to Delta, Lite-On currently trades at a 28% P/E discount, vs an average of 60% since 2002 (Figure 30 i L t e - O h n s a r d o w l v D c g z A I p ). We think a modest discount remains appropriate as Delta retains substantially greater AI power scale, customer diversification and grid-to-chip breadth. The valuation case therefore rests on Lite-On delivering the revenue and margin expansion in our model. If it does, earnings growth could support further upside even with only modest multiple expansion, on our estimates.

LIte-On is trading at NT$277, 22x NTM PE, above the upper end of its post-GPT PE range of 14-20x.

Current 22x NTM PE remains at a discount to the industry leader (Delta at 25x) and industry peers at 29x.

351

30

CFROI (%) (Economic Return) |

20

25

20

15

15|

10

10

2011

Sep

50

40

30|

20

10

• 10

2005

Sep

Valuation

Price TWD

315.00

upside

Warranted Price

604.66

+92%

Figure 32: Lite-On still trades below the power-supply peer average on forward earnings

2019

HOLT Price to Book

Sep

Dividend Yield

4.3x

Sep

2.57%

2012

Sep

2013

20P4

Sep

2015|

Sep

2016

Sep

2018

:020

2021

Sep

2022|

Sep

2023

Sep

2024

Sep

2025|

2026

• EPS Average in US$, Share price as of 9/11

Source: Company data, UBSe, Rifinitiv Workspace

HOLT cross-check

Our HOLT analysis also indicates potential upside if Lite-On converts its current investment cycle into structurally higher returns and free cash flow. We view this as a cross-check on the fundamental earnings valuation rather than the principal basis for the price target.

Figure 33: HOLT upside depends on sustained AI return

Source: HOLT. Note: HOLT is a corporate performance and valuation service of UBS.

The key question is whether Lite-On should be valued on its historical mixed-electronics returns or the higher growth and returns of Cloud & AIoT. On a HOLT-based PE, Lite-On trades at 29x, above its 15.7x long-term median, reflecting its shift toward AI.

Figure 34: Holt PE of 29.3x is above its long-term median of 15.7x

Source: HOLT. Note: HOLT is a corporate performance and valuation service of UBS.

Operational

Economic PE

›ep

Lite-On Technology - 2301. TW Price

Upside/Downside Spectrum

Upside/Downside Spectrum

Data: UBSe

UPSIDE (NT$540): Additional CSP programs across GPU and ASIC systems raise AI power share, while improved utilization and richer system content support stronger operating leverage. AI power sales exceed our base case, lifting EPS. In this scenario, total sales grows +50.6%/25.3% YoY to NT$339.9bn /NT$425.9bn, with the rising AI mix and operating leverage lifting GM/OpM from 25.2/13.5% in 2025 to 26.6/16.7% in 2028. EPS rises to NT$11.45/NT$19.43/NT$24.53 in 2026-28. Applying a 22x 2028 EPS gives our upside case valuation of NT$540 or 29x target multiple to our 2027 EPS, implying 95% upside.

BASE (NT$400): Lite-On maintains roughly 3-4% GPU-server power share and stronger positions in selected ASIC programs, grows Cloud & AIoT rapidly, and expands group margins as capacity utilization improves. In the base scenario, total sales grows +37.5%/24.6% YoY to NT$299.8bn /NT$373.6bn, with the rising AI mix lifting gross margin from 22.9% in 2025 to above 26% in 2027-28. Our modeled EPS is NT$10.85/NT $16.00/NT$20.00 in 2026-28. Applying a 20x 2028E EPS gives our PT of NT$400 or 25x target multiple to out 2027 EPS, implying 44% upside.

DOWNSIDE (NT$270): Customer qualification delays, smaller Lite-On shares in selected new programs, and a slower initial HVDC transition materially weigh on revenue and margin growth, despite positive AI power demand. In this scenario, sales growth slows to +29.5%/27.1% YoY to NT$255.1bn /NT$324.3bn, and a slower AI mix alongside reduced utilisation compresses GM/OpM to 26.4/14.1% in 2028. EPS falls to NT$9.13/NT $12.32/NT$15.90 in 2026-28. Applying a 17x 2028 EPS gives our downside case valuation of NT$270 or 22x target multiple to our 2027 EPS, implying 2% downside.

Figure 35: UBS upside, downside, base case scenario

Source: Company data, UBSe

WD 277.1

11 Sep

540.0

• 400.0

270.0

Upside to Downside

37.6 to 1

Upside:

+95%

Base:

+44%

Downside: -3%

Key risks to our investment thesis

• ˜ The principal risk is a slower AI infrastructure deployment cycle. Our industry work shows deployed IT GW is the single largest determinant of the power TAM, so a meaningful reduction in hyperscaler accelerator deployment would affect all power suppliers.
• ˜ Customer concentration is a second risk , particularly Lite-On's large custom-ASIC exposure and reliance on a relatively small number of hyperscalers. Program delays or share changes can materially affect annual revenue.
• ˜ Execution on capacity expansion also matters . Lite-On is investing substantially ahead of forecast demand; if utilization ramps more slowly, depreciation and fixed operating costs could limit expected margin improvement.
• ˜ Competitive pricing remains a risk as more suppliers qualify for high-power products . We believe the move toward larger, more engineered modules and systems should moderate pure commodity pricing pressure, but it will not eliminate it.
• ˜ Architecture is another source of uncertainty. A slower transition to external power racks is not particularly damaging to our near-term thesis because 48/54V systems still require substantial power content. Longer term, however, the precise mix between rack-level BBUs, pooled UPS/BESS, ±400V, single-bus 800V and facility-level conversion will determine where value accrues within the supply chain.
• ˜ Rising competition and slower spec upgrades would impact earnings. Faster progress from Chinese power suppliers, FX movements and weaker legacy IT/consumer demand could create additional earnings variability.

Lite-On Technology (2301.TW)

Source: Company accounts, UBS estimates. (UBS) metrics use reported figures which have been adjusted by UBS analysts. 1 Cash EPS (UBS, diluted) is calculated using UBS net income adding back depreciation and amortization.

Lite-On Technology (2301.TW)

Source: Company accounts, UBS estimates. (UBS) metrics use reported figures which have been adjusted by UBS analysts.

Forecast returns

Company Description

Founded in 1975, Lite-On Technology is Taiwan's first listed electronics company and a global leader in power management and optoelectronic solutions. Originally a pioneer in LED technologies, the company has evolved into a key supplier of AI infrastructure, cloud computing, automotive electronics, 5G and AIoT solutions.

Valuation Method and Risk Statement

We value Lite-On Technology using PE methodology. Key risks to our investment thesis:

The principal risk is a slower AI infrastructure deployment cycle. Our industry work shows deployed IT GW is the single-largest determinant of the power TAM, so a meaningful reduction in hyperscaler accelerator deployment would affect all power suppliers.

Customer concentration is a second risk, particularly Lite-On's large custom-ASIC exposure and reliance on a relatively small number of hyperscalers. Program delays or share changes can materially affect annual revenue.

Execution on capacity expansion also matters. Lite-On is investing substantially ahead of forecast demand; if utilization ramps more slowly, depreciation and fixed operating costs could limit expected margin improvement.

Competitive pricing remains a risk as more suppliers qualify for high-power products. We believe the move toward larger, more engineered modules and systems should moderate pure commodity pricing pressure, but it will not eliminate it.

Architecture is another source of uncertainty. A slower transition to external power racks is not particularly damaging to our near-term thesis because 48/54V systems still require substantial power content. Longer term, however, the precise mix between rack-level BBUs, pooled UPS/BESS, ±400V, single-bus 800V and facility-level conversion will determine where value accrues within the supply chain.

Rising competition and slower spec upgrades would impact earnings. Faster progress from Chinese power suppliers, FX movements and weaker legacy IT/consumer demand could create additional earnings variability.

HOLT methodology

The HOLT methodology does not assign ratings or a target price to a security. It is an analytical tool that involves use of a set of proprietary quantitative algorithms and warranted value calculations, collectively called the HOLT valuation model, that are consistently applied to all the companies included in its database. The HOLT valuation model is a discounted cash flow model. Third-party data (including consensus earnings estimates) are systematically translated into a number of default variables and incorporated into the algorithms available in the HOLT valuation model. The source financial statement, pricing, and earnings data provided by outside data vendors are subject to quality control and may also be adjusted to more closely measure the underlying economics of firm performance. These adjustments provide consistency when analyzing a single company across time, or analyzing multiple companies across industries or national borders. The default scenario that is produced by the HOLT valuation model establishes a warranted price that represents the expected mean value for a security based upon empirically derived fade algorithms that forecast a firm's future return on capital and growth rates over an extended period of time. As the third-party data are updated, the warranted price updates automatically. A company's future achieved return on capital or growth rate may differ from HOLT default forecast. Additional information about the HOLT methodology is available upon request.

CFROI, CFROE, HOLT, HOLT Lens, HOLTfolio and 'Powered by HOLT' are trademarks or registered trademarks of UBS in the United States and other countries. HOLT is a corporate performance and valuation service of UBS.

HOLT Business Resilience Dashboard (HBRD) Disclaimer

The HOLT Business Resilience Dashboard (HBRD) is constructed by applying systematic processes on publicly available records and third-party data to generate quantitative measures. The HBRD consistently assigns scores using a transparent combination of metrics and weights across financial, environmental and governance inputs. Individual companies can be scored on a global-size and/or region-size relative basis. The HBRD is also translated into five textual assessments (e.g. 'Highest' or 'Lowest') based on their quintile percentile rank. For further details on the HBRD methodology HOLT Business Resilience Primer please contact UBS Securities LLC, 11 Madison Avenue, New York, NY 10010, USA, Attention: Investment Research. The HBRD does not constitute or imply any endorsement or recommendation of any individual company. UBS has not sought to independently verify information obtained from public and third-party sources and makes no representations or warranties as to accuracy, completeness or reliability of such information due to inconsistent reporting of data by companies. Additionally, as global laws, guidelines and regulations in relation to the tracking and provision of such data are evolving, all such disclosures are made on a non-reliance basis and are subject to change. Any figures, values or other data published in this report are neither designed nor administered in line with the standards of the IOSCO Principles for Financial Benchmarks or the EU Benchmarks Regulation (EU) 2016/1011 or any other benchmark related regulatory guidelines and as such you are prohibited from using those figures, values or data in a way which would amount to use of a financial benchmark under those regimes (or any equivalent rules or regulations in any global jurisdiction). HOLT transferred to UBS on October 1, 2024 - renamed as UBS HOLT. The UBS HOLT Lens website and materials will be transitioning to UBS branding over time, with expected full redesign to UBS in time.

Quantitative Research Review

UBS Global Research publishes a quantitative assessment of its analysts' responses to certain questions about the likelihood of an occurrence of a number of short term factors in a product known as the 'Quantitative Research Review'. The views for this month can be found below. Views contained in this assessment on a particular stock reflect only the views on those short term factors which are a different timeframe to the 12-month timeframe reflected in any equity rating set out in this note. For previous responses please make reference to (i) previous UBS Global Research reports; and (ii) where no applicable research report was published that month, the Quantitative Research Review which can be found at https://neo.ubs.com/quantitative, or contact your UBS sales representative for access to the report or the Quantitative Research Team on ubs-quant-answers@ubs.com. A consolidated report which contains all responses is also available and again you should contact your UBS sales representative for details and pricing or the Quantitative Research Team on the email above.

Lite-On Technology

Required Disclosures

This document has been prepared by UBS Securities Pte. Ltd., Taipei Branch, an affiliate of UBS AG. UBS AG, its subsidiaries, branches and affiliates, including former Credit Suisse AG and its subsidiaries, branches and affiliates are referred to herein as "UBS".

For information on the ways in which UBS manages conflicts and maintains independence of its UBS Global Research product; historical performance information; certain additional disclosures concerning UBS Global Research recommendations; and terms and conditions for certain third party data used in research report, please visit https://www.ubs.com/disclosures. Unless otherwise indicated, information and data in this report are based on company disclosures including but not limited to annual, interim, quarterly reports and other company announcements. The figures contained in performance charts refer to the past; past performance is not a reliable indicator of future results. Additional information will be made available upon request. UBS Securities Co. Limited is licensed to conduct securities investment consultancy businesses by the China Securities Regulatory Commission. UBS acts or may act as principal in the debt securities (or in related derivatives) that may be the subject of this report. This recommendation was finalized on: 11 September 2026 02:34 PM GMT. UBS has designated certain UBS Global Research department members as Derivatives Research Analysts where those department members publish research principally on the analysis of the price or market for a derivative, and provide information reasonably sufficient upon which to base a decision to enter into a derivatives transaction. Where Derivatives Research Analysts co-author research reports with Equity Research Analysts or Economists, the Derivatives Research Analyst is responsible for the derivatives investment views, forecasts, and/or recommendations. Quantitative Research Review: UBS Global Research publishes a quantitative assessment of its analysts' responses to certain questions about the likelihood of an occurrence of a number of short term factors in a product known as the 'Quantitative Research Review'. Views contained in this assessment on a particular stock reflect only the views on those short term factors which are a different timeframe to the 12-month timeframe reflected in any equity rating set out in this note. For the latest responses, please see the Quantitative Research Review Addendum at the back of this report, where applicable. For previous responses please make reference to (i) previous UBS Global Research reports; and (ii) where no applicable research report was published that month, the Quantitative Research Review which can be found at https:// neo.ubs.com/quantitative, or contact your UBS sales representative for access to the report or the Quantitative Research Team on ubsquant-answers@ubs.com. A consolidated report which contains all responses is also available and again you should contact your UBS sales representative for details and pricing or the Quantitative Research team on the email above.