報告_UBS_台灣論壇Day1科技硬體_20260915

PDF 原檔：報告_UBS_台灣論壇Day1科技硬體_20260915_original.pdf

圖片清單（已驗證 2026-09-16）

抽出圖片經 size 篩選後無 40KB 以上候選（多為 logo／評等標記），本頁不提供可嵌入的真資料圖。

原始內容

APAC Technology

Taiwan Summit 2026: Day 1 Tech Hardware takeaways

Key Takeaways from Day 1 of the UBS Taiwan Summit 2026

The UBS Taiwan Summit 2026 kicked off with over 340 investors and 65 companies across the Taiwan / tech supply chain spanning AI infrastructure, servers, power, networking and components. In our day 1 report, we include the detailed bullets from the meetings from Chenbro, Compal, Delta, Inventec, Lenovo, Lotes, and Zhen Ding. Companies continue to see AI demand strong heading into 2027, with constraints continuing to come from supply rather than end demand. While investors remain focused on GPU server deployments, companies increasingly highlighted inference, ASICs, power infrastructure and rack-level integration as adjacent growth vectors. Memory shortages remain a key industry bottleneck, but are also creating consolidation and share-gain opportunities for larger players. Overall, companies remains constructive on AI and general servers amid weakness in notebooks and consumer electronics.

AI Infrastructure Growth stays healthy

Feedback from Chenbro, Compal, Inventec and Lenovo was positive on the AI infrastructure outlook, with demand extending well beyond current GPU training clusters. Companies highlighted growing adoption of inference platforms, accelerating Vera Rubin builds into Q4, increasing AI server penetration and expanding deployments from both hyperscale and Neocloud customers. Inventec noted server revenue growth of 40-50% with customer capacity reservations extending into 2028, while Lenovo's AI server opportunity pipeline expanded sharply and Compal sees AI servers becoming a meaningful growth driver, quickly expanding from low single digit percent of 2025 to target 30-40% of sales by 2027. Several companies also cited custom silicon and ASIC opportunities as growing faster than the broader server market.

Inference, ASICs and Rack-Level Integration key themes

The next phase of AI infrastructure growth appears increasingly tied to inference deployments, custom silicon and higher-value rack-level content. Inventec highlighted growing customer interest in lower-cost inference platforms versus NVL72 clusters, while Chenbro continues to move up the value chain from chassis products toward rack engineering and integration. Companies repeatedly emphasized that customer decisions are increasingly shifting from the server level to the rack level, creating opportunities in liquid cooling, power distribution, manifolds, busbars and systems integration. ASIC programs were highlighted across multiple meetings, with growing adoption among hyperscalers and increasing engagement on next-gen platforms.

Power infrastructure strong demand constrained by supply

Delta views AI infrastructure growth is increasingly constrained by power delivery and component availability rather than by customer demand. Power ICs and semiconductors remain bottlenecks, while rack power requirements continue to rise significantly with future AI platforms. Management expects accelerating adoption of liquid cooling, DCDC conversion and eventually high-voltage DC architectures as power density increases. We believe the company is passing a near-term low point in August with sales up in Sept, up double digits QoQ in Q4 and above seasonal in Q127 into ramps of new platforms, with GMs slightly in Q3 and a bit further in Q4 as it passes on higher component costs.

Memory remained one of the most frequently discussed challenges across Day 1 meetings. Compal described DRAM and NAND as the industry's largest constraint, while

Equities

Global

Technology

Randy Abrams

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

Diana Chang

Analyst diana.chang@ubs.com

+886-2-8722 7335

Nicolas Gaudois

Analyst nicolas.gaudois@ubs.com +65-6495 5148

Sunny Lin

Analyst sunny.lin@ubs.com +886-2-8722 7346

Jerry Su

Analyst jerry.su@ubs.com +886-28-722 7306

Irene Chang

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

Annie Chen

Associate Analyst annie.chen@ubs.com +886-2-8722 7281

Inventec and Lenovo similarly pointed to server memory shortages as the primary factor limiting shipments. Companies generally expect tight conditions to continue well into 2027, particularly for AI and enterprise server applications. At the same time, larger manufacturers believe procurement scale is becoming an important competitive advantage, creating opportunities for share gains as smaller competitors struggle to secure sufficient component supply.

Stock Implications - raising Delta to a Key Buy Call

With Delta poised for re-accelerating growth into Q426 on new platform launches and reaffirming confidence in architectural upgrades adding power content to supplement over AI growth consistent with our APAC Focus (link), we add the stock as a Key Buy Call in APAC. Elsewhere in the hardware supply chain, we stay positive on hardware/ component suppliers that command a leading or rising share position in growing segments at a supportive valuation (King Slide, Unimicron, Kinsus, and Chenbro) alongside our top ODM picks Hon Hai and Wiwynn.

Taiwan Summit day 1 hardware takeaways

Chenbro (8210.TW)

Strong AI demand outlook reaffirmed as Chenbro moves up the value chain. UBS recently initiated Chenbro at BUY (link) as the company moves beyond L3-L6 chassis and components into higher-value rack-level engineering and integration. Denser, more customized AI systems are pushing design decisions from the server to the rack level, and management sees no meaningful demand slowdown over the next three to five years. We believe revenue growth of more than 40% YoY in 2026 and approximately 30% in 2027 is achievable, with recent softness primarily reflecting supply chain constraints and inventory digestion rather than any deterioration in underlying AI demand.
Near-term ramp affected by inventory digestion, while 1H26 profitability remained strong. August sales were NT$2.46bn, - 0.2% MoM /+40.5% YoY. Its recent shipment growth was constrained by ODM inventory digestion due to the component mismatch caused by shortages in memory, CPUs and other key components. Management anticipates existing inventories to be largely cleared by end-September, supporting stronger orders into October and broader newproduct pull-ins in 4Q26. While rack business is margin dilutive, H126 margin remained strong thanks to higher NRE contribution. Management reiterates its mid to long term gross margin target of 28-29%.
Capacity expansion supports rack ramp, with execution and engineering talent as key differentiators. Chenbro is expanding in Vietnam and Malaysia, with Vietnam expected to support the CDU rack ramp around 2Q27. Annual capex of NT$2.5-3.0bn in both 2026 and 2027 is expected to be self funded through free cash flow. As rack-level integration becomes increasingly complex, production capacity, customization capabilities, manufacturing quality, and scarce engineering talent should emerge as key competitive differentiators.

Compal (2324.TW)

AI server becoming a meaningful growth driver with a path to 30-40% mix in 2027. Compal's server revenue reached 7-8% of sales in Q226 (vs. 5% in Q126) and remains on track to achieve management's 10% FY26 target as early as Q326. AI servers represented ~70% of server revenue in Q226 led by neo-cloud customers (Humain, GMI, Verda and Data Section) and is expected to contribute a roughly similar ~10% of sales in next year, while the main growth driver would come from its US enterprise customer, where it expects to gain share as a second source supplier. The company also highlighted traction with hyperscalers, starting with L6 motherboards in 2H26 with longer-term growth is supported by potential expansion into L10/L11 systems and rack-level solutions, as well as ASIC opportunities still under discussion.
PC demand weakened more than expected in Q3 amidst continued ASP hikes . Management now expects notebook shipments to decline more than 10% QoQ in Q326 vs. earlier expectations for flattish demand, citing increasing weakness across both commercial and consumer end markets due to pull-ins in 1H already building up inventory. Revenue performance should remain more resilient as notebook ASPs have increased from roughly US$600-650 historically to US $850-1,000 currently as memory prices surge. The higher price at this stage could continue to pressure units into 2027. Growing AI PC penetration is a bright spot, representing roughly 50% of shipments in 1H26 and >20% meeting Microsoft's >40 TOPS Copilot+ definition.
Memory remains the industry's largest constraint and inflation driver . Management views memory as a far more significant issue than CPUs shortage, with DRAM and NAND costs rising sharply and now accounting for roughly 4050% of notebook BOM costs versus only ~10% historically. The shortage is forcing OEMs to reduce specs on entry-level systems, with purchasing increasing adoption of lower-memory and lower-storage configurations to preserve affordability. While leading OEM customers such as Dell and HP have largely secured memory supply, smaller competitors remain vulnerable, creating share opportunities for larger vendors. Management believes memory tightness could extend into 2027 and remain a key determinant of industry profitability and end-demand elasticity.
Aggressive capacity expansion and elevated CapEx support AI server

ambitions. Compal is significantly expanding its manufacturing footprint to support AI infrastructure growth, including the new Daxi facility in Taiwan, which adds L11 and rack-level assembly capabilities and effectively doubles AI server capacity. The company is also ramping its Texas facility, which will provide L10/L11 manufacturing capability and support large US enterprise customers, while Vietnam expansion remains focused on notebook and electronics manufacturing. FY26 CapEx guidance stands at NT$18bn, more than double historical levels, with approximately NT$6-7bn allocated directly toward AI server investments. Management indicated CapEx is unlikely to decline next year as investment priorities remain heavily skewed toward AI servers and US localization.

Growth strategy prioritizes scale, enterprise penetration and long-term profit expansion . Management's strategic focus has shifted from maximizing gross margin percentage toward maximizing absolute gross profit dollars. While moving from L6 motherboards into L10/L11 systems and racks could dilute margin rates due to GPU pass-through content, server economics remain superior to PCs and create significantly larger profit pools. Compal intends to deepen relationships with enterprise customers, hyperscalers and ASIC opportunities while avoiding direct competition with key customers in the largest neo-cloud deployments. Beyond servers, management continues to see potential from AI PCs, connectivity, smart devices and other AI-adjacent opportunities, but views AI infrastructure as the company's most important long-term growth platform.

Delta (2308.TW)

Growth to bottom from July, accelerate into Q4. Delta affirmed a fundamentally stronger growth profile as AI infrastructure now represents more than half of the business, reducing the relevance of traditional consumerelectronics seasonality. Despite originally expecting 3Q up mid-to-high teens, supply constraints on Power ICs have tempered that outlook to 10% to low-teens in-line with latest UBS expectations after the down August sales. Sales are expected up slightly in September and then Q426/Q127 are expected to reaccelerate to above seasonal into new product ramps and more supply, potentially implying to double digit QoQ growth in Q426 and above seasonal Q127 (vs. normal high single/low teens QoQ decline.
Supply, Not Demand, Remains The Key Constraint. The message was consistent throughout the session that growth is being capped by component availability rather than customer demand. Power ICs and power semiconductors remain the primary bottlenecks, with vendors still unable to fully meet Delta's requirements despite increased capacity commitments and long-term contracts. Management expects shortages to persist for at least the next several quarters, limiting the industry's ability to satisfy AI infrastructure demand, while 4Q shipments should accelerate meaningfully even if supply remains insufficient to close the demand gap. The company indicated demand continues to accelerate faster than supply, with the supply-demand gap likely widening before narrowing.
HVDC Adoption Appears Increasingly Inevitable. While HVDC contributes only around 1% of revenue today, production and shipments have commenced, with demand already exceeding available supply. As rack power increases from 140kW on GB300 to nearly 250kW on VR200 performance configurations and potentially higher on Rubin Ultra, management believes space and efficiency constraints will increasingly force power systems out of the rack and into side-car or centralized architectures. Delta views HVDC as the logical solution to address power-loss, thermal and density challenges, and noted that higher-voltage architectures represent a multiple-times higher value opportunity on a per-watt basis than existing solutions. The company also views 3 hyperscalers considering 400V with two adopting it for NVIDIA racks initially before later converging on 800V. Importantly, management sees little content difference between 400V and 800V standards, but expects future architectures to increasingly favor highervoltage solutions as power densities continue rising.
Content Expansion Extends Well Beyond PSU Growth. Delta sees several layers of content growth as AI platforms evolve. Rack power budgets continue to rise, PSU counts remain elevated due to redundancy requirements, and nextgeneration systems create incremental opportunities in HVDC, side-car power systems, liquid cooling, DC-DC conversion and solid-state transformer technologies. Management highlighted liquid cooling at 12% of 1H sales, with full-year growth now expected to exceed 70-80% although slightly lower as a %
of sales in 2H from a high base. DC-DC is also on track to reach roughly 8% of revenue versus less than 3% last year and growing further in 2027.
Working through near-term Cost And Quality Headwinds. Management acknowledged margin pressure from power component inflation, with 2Q and 3Q impacted by higher input costs before customer price pass-throughs begin to take effect. It still expects operating margins to improve sequentially through a better mix and scale benefits, with management maintaining confidence in a secular upward margin trajectory. On the recently disclosed aluminum capacitor issue, Delta emphasized that the root cause was traced to a supplier fabrication problem, the affected products represented only a small volume, replacement actions were implemented immediately using a second source supplier, and the company expects no financial impact or change in customer demand patterns.
Aggressive Investment Signals Confidence in Multi-Year Opportunity. Delta continues to invest aggressively behind the AI power opportunity while simultaneously expanding into adjacent growth areas. Capex was increased by 40% from NT$50bn to NT$70bn, driven by incremental investment in power supplies, cooling, and solid oxide fuel cell (SOFC) initiatives. Management is also negotiating a new overseas campus, expanding U.S. assembly capacity in Plano, Texas, and continues to scale liquid cooling and DC-DC power businesses, both of which are outperforming original expectations.
Monetizing a very small % of Delta Thailand. Following the presentation, Delta's Singapore subsidiary holding 43% of 61% stake in Delta Thailand announced US$500mn of 2027 bonds at -2-4% yield to maturity at 15-20% above the placement price and US$1bn of 2031 bonds at -0.5 to -1.5% rate at 3040% above the placement price on its Delta Thailand stake, both representing 1.1% of Delta Thailand's shares if fully exchanged. Together, Delta is receiving US $1.5bn cash now for a 1.1% stake if Delta Thailand appreciates 30-40% from here.

Inventec (2356.TW)

Server Demand Remains Robust Despite Memory Constraints. Management continues to see strong server demand, with total server unit shipments expected to grow more than 20% YoY (up from high single digits at the start of the year) and revenue up 40-50% due to higher AI content and richer configurations. Two top US hyperscalers (Search/E-Commerce) remain the largest growth drivers, while enterprise servers now represent less than 30% of mix. The primary constraint remains memory, with DRAM, SSDs and some CPUs limiting shipments in 3Q, although management expects allocation conditions to improve modestly in 4Q and believes underlying demand remains stronger than current shipment levels.
Inference Driving Next AI Server Expansion Wave. Customer interest is increasingly shifting toward inference-optimized platforms, including CPX and Vera-based solutions, which offer significantly lower costs versus NVL72 systems. Management highlighted that only a limited number of customers are deploying NVL72 given its high cost, while CPX, HGX and Rubin-based platforms are gaining broader customer traction. Inventec expects initial Rubin shipments in late 2026 with more meaningful volume beginning in 1Q27-2Q27, supporting another leg of AI infrastructure growth.
ASIC Programs Expanding Rapidly. Beyond GPU servers, management emphasized strong momentum in custom silicon and ASIC programs across the top two hyperscalers on ASIC for CPU and GPU (L6). Management noted that ASIC shipment volumes already exceed AI server volumes on a unit basis, although GPU servers generate higher revenue due to significantly higher system value. Looking into 2027, engagement with both of these customers is increasing, particularly around liquid-cooled infrastructure and next-generation rack deployments.
2nd AI GPU platform ramping in 2027. Inventec expects the new AI GPU program to become increasingly meaningful beginning in 2027. The company is supplying motherboard and subsystem content for platforms serving customers including Microsoft, Oracle and Meta, with motherboard shipments beginning in Q426 before meaningful rack deployments ramp in 2Q27. Management highlighted close cooperation on system design and manufacturing capacity for both CPU and GPU AI platform growth over the next several years.
AI Capex And Global Expansion Reflect Long-Term Confidence. Management is aggressively expanding capacity to support expected AI demand, with roughly US$1 billion of capex planned across 2026 and 2027. Customers have already reserved capacity extending into 2028, driving investments across Taiwan, Thailand and Houston. Inventec estimates Taiwan ODMs collectively command roughly 90% share of AI server manufacturing, while increasing requirements for power availability, liquid cooling infrastructure and local North American production create higher barriers to entry and support continued industry consolidation.
Notebook Outlook Softer, But Revenue Mix Improving. In contrast to servers, notebook demand remains more cautious, with customers becoming increasingly conservative on 2027 forecasts. Inventec expects notebook shipments to decline high-single digits in 3Q due to earlier pull-ins and sees flattish-to-down trends in 4Q. However, component inflation and a more commercial-focused customer mix are driving higher ASPs, allowing notebook revenue to grow at a double-digit rate despite lower unit volumes. The company also expects a better corporate mix and increasing AI PC penetration to support profitability.

Lenovo (0992.HK)

ISG (Margin Expansion expected to be sustainable. Management expressed confidence that its Infrastructure Solutions Group (ISG) can sustain high-singledigit operating margins and ultimately surpass 10%, supported by both structural improvements and mix enhancement. The record 9.1% operating margin in the June quarter (vs. 3.6% in F4Q) was driven not only by favorable component pricing but, more importantly, by ESMB restructuring initiatives, including cutting more than half of legacy SKUs, optimizing sales incentives, improving pricing discipline, and focusing resources on higher-return opportunities. Management noted enterprise server margins are already approaching the high teens, while mix improvements from enterprise and neo-cloud customers should continue supporting profitability.
AI Server Revenue Set To Accelerate Sharply In 2H. Management highlighted a dramatic increase in AI server pipeline opportunities to US$54 billion from US$21 billion the prior quarter, reflecting accelerating customer demand. While AI servers remain a relatively small percentage of ISG revenue today and carry lower margins than traditional enterprise infrastructure, management indicated volumes should more than double HoH in 2H as NVL and Vera Rubin platforms ramp and Lenovo begins shipping first-wave OEM deployments. The company is seeing particularly strong momentum from neo-cloud providers (such as IREN, NSCale, Sharon), with customers that were previously deploying several hundred million dollars of infrastructure now scaling toward multi-billion-dollar deployments, providing confidence in a multi-quarter AI growth trajectory.
Enterprise Business Emerging As a Profit Engine. Beyond AI, management emphasized that the traditional enterprise server business is becoming an increasingly important earnings contributor. Enterprise operating margins have reached nearly high-teens levels, benefiting from an ongoing refresh cycle as newer servers can replace the compute capacity of six to seven older-generation systems. Demand is broad-based across roughly 6,000 enterprise customers, and management believes secular modernization trends should support growth for multiple years. The company also highlighted growing traction in storage, which represents roughly one-third of the non-AI server market and carries higher profitability than compute servers, with the recently acquired Infinidat expected to strengthen Lenovo's presence in higher-end storage over the next 1.5-2 years.
Memory Shortages Creating Share Gain Opportunity. Lenovo views memory constraints as likely to persist across the server market for an extended period, particularly for DRAM used in AI and enterprise servers. Management stated server DRAM availability remains severely constrained and is unlikely to ease materially over the next two years, allowing Lenovo to leverage its scale and procurement capabilities to gain market share. The company highlighted that it recently achieved the number one position in x86 server unit share, aided by hyperscaler volumes and strong component sourcing. By contrast, memory conditions in PCs and smartphones appear to be gradually improving due to its ability to leverage growing China supply outside the US market, with management expecting supply conditions to ease later this year and into next year, potentially supporting demand recovery while reducing pricing pressure on consumer products.
Business Mix Shifting Toward Higher-Value Infrastructure. Lenovo's results increasingly reflect a transition from a PC-centric company to a diversified infrastructure platform. IDG continues to perform well, holding a 25% global PC market share and generating 55% AI PC mix, while ISG nearly doubled revenue YoY and SSG delivered a 24% operating margin. Within ISG, the mix is evolving beyond hyperscalers toward faster-growing neo-cloud customers, which management characterized as more profitable due to the additional manufacturing, integration, testing, and service content Lenovo provides. Combined with rapid growth in AI infrastructure, storage, enterprise services, and Infrastructure-as-a-Service offerings, management sees significant scope for both revenue and profit dollars to expand even if headline margins fluctuate as AI server mix increases.

Lotes (3533.TW)

Q3 remains on track for QoQ growth despite ODM component shortages. August revenue came in at NT3.2 bn, down 5% MoM and tracking slightly behind at 65%/60% of UBSe/street's Q3 estimates, as ODMs repeatedly adjusted pull-in schedules in the past few months as they are struggling to secure various components. Management expects September revenue to track better with component supply improvement, and remains confident that Q3 should still grow QoQ, despite NB/desktop demand has further softened. New products including SoCAMM and QD, should still serve as the key incremental growth drivers for H226 as the faster ramp of SoCAMM is now contributing to ~2% of Aug sales (vs. previous >1% expectations), while QD for liquid cooling is now around 1% of total sales and potentially reaching 3% of sales by year-end.
NPO confidence improving, but contribution remains a 2027 story. Management expressed greater confidence in its NPO socket program with the qualification now in the final stages and mass production targeted for Q327. Key customer decisions, including final OSAT/supplier allocations, remain pending but are expected to gain clarity by end of 2026. Separately, its subsidiary Lintes is involved in an optical-solution and fiber-related NPO program, partnering with optical ecosystem players while outsourcing the socket for optical engine to Lotes. Although NPO sockets carry significantly fewer pins than CPU sockets (multihundred vs. 5,000-6,000), management noted the higher data-transfer requirements support higher value-per-pin economics, suggesting an attractive profitability profile once volumes ramp in 2027.
GMs to improve as ASP for new products will reset. Management acknowledged that gross-margin pressure from elevated raw material costs is likely to persist for the next few quarters. It has suspended the quarterly price concession for legacy products, while ASP for newer products will be reset given the higher material cost. These should help to stabilize its GM and further improve once the new products start to ramp. While QD currently generates a lower-thancorporate level gross margin, management expects profitability to improve as volume scales and it further moves into mass production, albeit likely remaining below the company's 40%+ corporate GM level. As new products become a larger part of the revenue mix and raw material inflation gradually eases, we believe margin pressure should progressively moderate from current levels.

Zhen Ding (4958.TW; Non-covered)

AI & Networking ramping up. Zhen Ding is accelerating its structural pivot away from consumer electronics (historically ~95% mobile/wearables) toward AI servers and active optical modules, which management expects to account for nearly 50% of relevant capacity next year. ZDT anticipates optical module and server business will strengthen from Q226 and more than doubled in 2026 (optical module/server/IC substrate sales mix up from 12% in 2025 to 20%+/45-50% in 2026/2030F). Module shipments are guided to ramp from >100M units this year to doubling next year. In AI servers, board complexity is expanding rapidly up to 28 layers, backed by proprietary 10-year mSAP experience and direct tier-1 engagement with Nvidia, Google, and Microsoft.
Supply Chain De-Risking & Raw Material Management. Zhen Ding has diversified its critical upstream vendor base for ultra-low-loss glass cloth (T-Glass) across Taiwan Glass, Nan Ya Plastics, and Grace, mitigating single-source constraints. With high-speed PCB moving through M6 to M8/M9 laminates and low-loss resin systems (such as PKP/PTFE), technical execution is shifting toward

multi-pass plating (2-3 plating cycles to achieve line-space filling and high layer integrity), where tight supplier partnerships and process know-how serve as key entry barriers.

Disciplined CapEx & Overcapacity Mitigation via Long-Term Agreements. In response to investor concerns regarding industry CapEx discipline and potential supply gluts, management noted that high-end AI PCB and substrate manufacturing has a typical 2-to-3-year gestation cycle, limiting immediate downside risk. CapEx commitments are closely tied to customer LTAs and capacity reservations-particularly in optical modules and custom server programs. Crucially, pricing structures remain flexible: LTAs do not lock in rigid long-term discounts, allowing Zhen Ding to capture premium delivery bonuses amidst tight industry supply, supported by record-high R&D intensity (1H R&D hit a historical peak of 8.7% of sales).
Strong 2H/FY27 Rebound underway. 1H performance reflected typical offseason seasonality alongside heavier upfront opex (headcount additions, footprint setup, and peak R&D investments), leading to an estimated full-year margin dilution of 12-13% YoY. However, operating leverage and capacity utilization are poised to inflect sharply in 2H and into next year as AI volumes ramp.
Capacity Expansion & Global Footprint. To meet robust multi-year demand, the company is on track to complete six new fabs across three manufacturing campuses by the end of 2026. The new Thailand site (K01) and associated smart fab expansions are scheduled to begin phased qualification and official volume ramps starting 1Q27 (following campus groundbreakings in mid-2026). It is building out its high-end PCB capacity, investing NT$80bn in 2026 with more strategic investments in 2027/2028 to focus on iHDI/mSAP , HLC, IC substrates, and advanced FPCs, with production to come online in phases through 2030.

Valuation Method and Risk Statement

Investing in the technology sector involves a high degree of risk. Rapid technological changes, increasing competition and exposure to macroeconomic cycles are among the many risks faced by investors in technology stocks. Moreover, it is extremely difficult to project the financial results of tech companies since their operating models are highly volatile and unpredictable. Finally, valuing technology stocks can prove challenging, as neither traditional nor non-traditional valuation measures have provided much insight into how these stocks trade. Valuation methodologies in the sector include target price-book-multiples based on the long term average ROE and the cost of equity.

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: 14 September 2026 06:05 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.