大中华半导体行业：中国台湾专场调研纪要

Research Analysts intelliGotakeawaysfromTaiwanFocus Semiconductor Edge Al player—intelliGo Technology, a company Aaron Jeng, CFA - NITBaaron.jeng@nomura.com+886(2) 21769982 We discuss business dynamics with intelliGo (7749 TT, Not rated) on 6 Jan. Beloware the major takeaways: Donnie Teng - NIHKdonrie.terng@nomura.com+852 2252 1439 Company briefing: IntelliGo is a non-typical fabless company that specializes in Alacoustic sclutions. Its core competencies include neural network algorithmdevelopment, signal processing, and IC designing. IntelliGo mainly focuses on EdgeAl, with its solutions being adcpted in PC/NB, garning consoles, gaming headsets,TwS headsets, speakerphones, wearables, and smart loT devices, The companywas first listed on the TWSE in June 2D25 vivian.yang@norura.ccm+886(2) 21769970 Eric Chen, CFA - NITBsic.chsm@ncmurs.ccm)+886(2) 21769965 Read-across for the fabless sector: intelliGo is cne of very few edge Alsemiconductor fabless/IP companies listed in Asia with key funders frorn MediaTek(2454 TT, Neutral). "Voice’ is the company's key focus while Al is ils approach behindprcducl cornpelitiveness. As such, it has generated 70-a0% gross margin, cornbiningcontributions from both chips and IPs. Voice interactions are more irnportant now thanever between human and machines with Al's popularities, The interactions wculd notonly happen belween hurnan phones, nolebooks, TVs, garne consoles, but also carsand robotics in the future. It would be worth watching if the growing Al popularitywould be spreading inlo edge Al, On cur Al cycle view, please reler to our Dec 2025Af Serti ard Server Anchor report. Our AI Sermi lop picks are TSMC (2330 TT, Buy)ASPEED (5274 TT, Buy), ASE (3711 TT, Buy), with trading buy for KYEC (2449 TT,Neutral), GUC (3443 TT, Neutral), and MediaTek (2454 TT, Neutral) Al Capability: InlelliGo provides niative Al rmodel develcprmenl capability (rather thanrely cn open-scurced solutions) with IPs for model training. Once the algorithrn isready, , the company's engineers will help with porting to cross platfoms/variousedge devices. Also, il provides tailor-made solutions lo fine-tune models/parameterswith custormized integration techniques (Fig. 7 ) Business model: intelliGo mainly provides IPs licensing (100% GM), but it could alsooffer tailored AI Edge ASIC solutions (lower GM, ~70%) as required by customers (Fig. 2 ), Sales by product: Game console now accounts for 35-40% of its tolal sales, as it isproviding ASIC solutions to a major customer, and the company believes ramping upprcduclion cculd pressure GM but would bring incrernental profit dollar. PC/Smarlphone/Cyberon (unlisted) ccrmbined account for 23-25% of total sales withall IP solutions. Others (mainly TwS, monilors, elc) account for the rest with mixedbusiness models (Fig. 3). Altnough its end devices are mainly consumer electronics, the company does notexxect any meaningful impact from memory shorlages/price hikes in 2026E,based on its existing cornmercial arrangernents. However, it believes the impactmight be relevant if the shortages persist in 2027E Financials: intelliGo's 1Q-3Q25 sales grew more than 100% y-y. Managementexpects to continue the growth trajeclory with visibily in 2028E. It also expects asimilar revenue pattern for 2025E and 2026E (sequential growth in 1Q-3Q, followedby a mild normalization in 4Q). 2026 oullook: Other than contribulions frorn ils existing custorners (visibility aslong as 5 years+ frormn its largest client, thanks to product cycle), managementshould nearly doutble from 2025 to 2025E, In addilion, the (ormpany anticipatesalso expects higher penetration in smartphones and notebook as the business Frd.rlon Gorrplsls: 2325-0-17 c6 78 UTC that one Al smart-glssses customer and one humanoid robotics customer willenter mass production in 2026E. Management expects strong revenue momentum in 2026E. While the pace ofgrowth could moderate as its revenue base expands, it believes new human-to-machine opportunities will begin bearing fruit, supporting the next robust growthcycle. Long-term strategy: intelliGo is proactively developing human-to-machine products.end devices including glasses, as well as automotive, and humanoids. This is anextension of its existing human-to-human solutions. Revenue contribution might notbe considerable in the initial stage, according to the company, , depending on whenthe inflection point is. For existing business, the company is working on increasingpenetration rate and market share. Interaction processes include: 1) the machine receives voice (and voice needs toneed to decide whether inquiries have to be solved on edge or cloud beforecloud part of the third stage (Fig. 4 ).Al glasses: intelliGo adopts technology such as Sound Event Detection (SED) tosatisfy emerging demand from Al glasses in order to meet specific circumstances(e.g. the detection will then trigger other functions on the glasses).Automotive: intelliGo will follow through Medi