報告_GS_台光電2383台燿6274_3Q26_20261028

PDF 原檔：報告_GS_台光電2383台燿6274_3Q26_20261028_original.pdf

原始內容

[讀者資訊已遮蔽]

TAIWAN CCL

Solid expansion on GM/OPM will continue into the next 12+ months, with widening CCL S/D gap; Buy on EMC & TUC

3Q26 overview - solid beat by both EMC/TUC thanks to faster than expected product mix improvement

EMC's 3Q26 core business (OPI) was 6% higher than our estimate (the highest on the Street), and was 15% higher than the BBG consensus, a clean beat thanks to (1) the much more favorable than expected product mix (M9 CCL shipment started from 3Q26, whose ASP is 3-4x higher than the company's average level + increasing M8 adoption rate in the 2H25 new AI server - see here), and (2) some delayed pricing hike happened in 3Q26 but was announced in 2Q26.

For the 3Q GM, it was 0.9ppt higher than GSe and 1.9ppt higher than BBG

consensus, which was up 3.8ppt QoQ (vs. 2Q GM grew 4.5ppt QoQ, with 20-25% pricing hike in average, while we didn't see major pricing hike happen in 3Q26). EMC's 3Q26 OPM was 30.7%, which was 0.1ppt higher than GSe and 1.8ppt higher than BBG consensus, also thanks to the better than expected product mix, despite the higher than expected R&D investment (up 44% QoQ in 3Q) for the substrate CCL and new generation M10/11 CCL.

TUC's 3Q26 core business (OPI) was also a solid beat on our estimate (the most bullish on the Street) at 16% higher than GSe, which was 25% higher than BBG consensus, while this beat was mainly supported by (1) the key ASIC customers' M8.5 CCL demand ramp up (ASP was 2-3x higher than TUC average level), (2) the faster than expected Thailand plant ramp up progress, and (3) the solid pricing uptrend for the non-Low Loss products and power products (30-40% pricing hiked in 3Q per GSe) even these only accounted for 34% of the company's total revenue in 3Q26. TUC's 3Q26 GM was 3.8ppt higher than our bullish estimate and was 5.1ppt higher than the BBG consensus, given the solid product mix improvement and better pricing condition for its mid to low end business. For 3Q26 OPM, it was 31.0% and is higher than EMC (30.7%), which is the first time TUC's OPM surpass EMC, driven mainly by the better operation efficiency. OPM in 3Q26 was

3.7ppt/5.2ppt higher than GSe/BBG consensus, which was in-line with the GM trend.

4Q26 outlook - Further expansion in GM/OPM supported by the 5-20% high-end CCL pricing hike and solid demand on M8/9 CCL

For EMC, considering (1) the 5-20% pricing hike that just announced in early October (our supply chain check suggests M8 pricing will go up by 10% in coming

Goldman Sachs does and seeks to do business with companies covered in its research reports. As a result, investors should be aware that the firm may have a conflict of interest that could affect the objectivity of this report. Investors should consider this report as only a single factor in making their investment decision. For Reg AC certification and other important disclosures, see the Disclosure Appendix, or go to www.gs.com/research/hedge.html. Analysts employed by non-US affiliates are not registered/qualified as research analysts with FINRA in the U.S.

Chao Wang

+886(2)2730-4195 | kuanchao.wang@gs.com Goldman Sachs (Asia) L.L.C., Taipei Branch

Allen Chang

+852-2978-2930 | allen.k.chang@gs.com Goldman Sachs (Asia) L.L.C.

Al Wang

+886(2)2730-4081 | al.wang@gs.com Goldman Sachs (Asia) L.L.C., Taipei Branch e92c7a75ab8b4efbba794e6b187208c8

[讀者資訊已遮蔽]

months of the year), (2) new Kunshan capacity (total 1.2mn sheets/month or 21% of the company's current capacity) ramp up from December, and (3) the mass volume shipment for several new AI projects (Vera Rubin + TPU v8 etc.) with a better demand volume on the most high end M8.5/M9 CCL, we now expect EMC's revenue to go up by 20% QoQ, and expect the GM will further expand to 40.9% (first time surpassing 40%).

For TUC, with an even faster product mix improvement speed (due to the lower base vs. EMC) and the solid pricing hike uptrend for both high end and mid end CCL (15-20% QoQ average pricing hike in 4Q), we expect the company to deliver a revenue growth rate at 18% QoQ, and believe the GM will reach 40.8% (+3.9ppt QoQ) in 4Q26, despite the slowing demand from key AI ASIC customer due to the transition period of the ASIC projects whose capacity will be replaced by other high-end Ethernet switch projects (also M8.5/M9 grade CCL).

Key update - EMC now plan to expand capacity to 15.25mn sheets/month from 5.85mn sheets/month capacity today, with solid demand EMC's BOD has approved additional capacity expansion plans extending into 2029/30E. EMC now targets year-end capacity of 8.70/10.45/11.95/15.25mn sheets/month in 2027/28/29/30E (implying 27% 2026-30E CAGR), which is the first time EMC announced its new capacity expansion plan after 1H28. EMC mentioned that capacity expansion is supported by LTAs with customers who secure volumes (EMC does not proactively initiate LTA negotiations). EMC indicated that despite its proactive capacity expansion planned for 2027-30E, it would still be insufficient to meet with customer demand secured under signed LTAs (demand under LTA extends at least 3Y ahead).

EMC's near-term capacity expansion remains intact, as the capacity expansion in Kunshan (with additional 1.2mn sheets/month, 21% of current capacity) could come online as early as Dec 2026 and continue through May 2027 (with new capacity coming online every month), the expansion was originally scheduled in 2 phases, with 600k sheets/month each coming online in 1Q27 and 3Q27. The accelerated ramp-up will offset the rescheduling of Guanyin capacity expansion, where the 300k sheet/month expansions originally scheduled for 4Q26 have been deferred to 2H27, mainly due to shortages of construction capacity in Taiwan as well as long equipment lead time + Taiwan government's strict policy on labor safety related requirement (however, we only see delay in Taiwan, while all other plans are basically in-line or earlier than the company's original expectation).

Investment view - Solid margin beat + expanding margin, suggesting a stock performance outlook; buy in the dip for both EMC & TUC

We notice that the BBG consensus for 2027 EMC/TUC GM is only at 38%/34%, and we see upside risk to consensus EPS estimates for both EMC/TUC as the GM in 4Q could both surpass 40-42%, without any possibility to see the GM to decrease in coming quarters as the high-end CCL supply demand gap to further expand in coming years (see here). As a result, we believe consensus 2027 GM could lift from 34-38% to 40-42%+ for both companies in coming days, suggesting 10-25% EPS upward revision opportunity, and recommend taking advantage to buy in this valuation dip (EMC/TUC only trades at 15x/12x of our bullish 2027 EPS estimate, which was far below the past 3 year range of e92c7a75ab8b4efbba794e6b187208c8

[讀者資訊已遮蔽]

20-30x).

Moreover, EMC has mentioned all of its new capacity in coming years (capacity will expand by 2.6x in 2030 vs. 2026; see more details above) is still not large enough to support customers' LTAs volume demand, suggesting an even more favorable high-end CCL pricing outlook in coming years (all the way to 2030), which is in-line with what we discussed in our TPCA takeaway report. We expect EMC/TUC will rise the like for like base pricing for M7+ CCL by 10-40%, and we now expect the similar pricing hike ratio in 2027/28, with more room to adjust if the supply shortage issue continues.

For the new CCL generation, we exect EMC to start shipment on M10 grade CCL from 2H27, which is only for single customers' AI GPU projects, but still implies a CCL migration trend in-line with our bullish view. For M9, we only see HGX R200 UBB and VR200 LPU adopt this solution in 2026, but see multiple new projects in 2027 will adopt this high ASP project (including several 1.6T ethernet switch projects, and Trainium 4/ TPU v9 could also see at least 50% adoption rate of the M9 CCL; moreover, the VR300 rack should have 50%+ of its CCL content related to M9 grade CCL). We believe this new technology migration trend going into 2027 will further drive both companies' ASP and profit in coming quarters, suggesting an even more favorable outlook for EMC/TUC's EPS growth in coming years.

Overall, we maintain our bullish view on high-end CCL players' business outlook and margin expansion trend, and expect investors to start focusing on this solid margin expansion trend after this clean beat on GM/OPM in 3Q26; maintain Buy on EMC/TUC, with a new TP of NT$11,700/NT$3,460 (from NT$10,200/NT$3,010).

Source: Company data, Goldman Sachs Global Investment Research, Bloomberg

e92c7a75ab8b4efbba794e6b187208c8

[讀者資訊已遮蔽]

Exhibit 2: TUC earnings comparison table

Source: Company data, Goldman Sachs Global Investment Research, Bloomberg

Exhibit 3: We expect both EMC and TUC GM to start the next round of expansion driven by solid AI demand and tightening S/D outlook

Exhibit 4: Both EMC and TUC will expand more than 100%+ capacity in coming 2 years, while both companies' future capacity has all been booked by high-end customers Average monthly CCL capacity (k sheets/month)

Source: Company data, Goldman Sachs Global Investment Research

Earnings revisions

EMC

We revise up our 2026/27/28E net income estimates by 13/22/22% to factor in 3Q26 actual results, a more favorable pricing outlook for high end CCL and the faster than expected high-end CCL demand growth, suggesting a better GM outlook (revising up 2026/27/28E revenue by 8/22/24%, and revise up 2026/27/28E GM by 1.5/1.6/0.7ppt).

Also, we introduce our 2029E earnings estimates with this report.

Source: Company data, Goldman Sachs Global Investment Research

e92c7a75ab8b4efbba794e6b187208c8

[讀者資訊已遮蔽]

Exhibit 5: EMC earnings revision table

Source: Goldman Sachs Global Investment Research

TUC

We revise up our 2026/27/28E net income estimates by 12/20/27% to factor 3Q26 actual results, faster than expected high-end CCL demand growth and better high end CCL pricing outlook, which suggest a better than expect GM expansion tend going forward (revising up 2026/27/28E revenue by 3/12/17%, and revise up 2026/27/28E GM by 3.1/4.5/6.1ppt).

Also, we introduce our 2029E earnings estimates with this report.

Exhibit 6: TUC earnings revision table

Source: Goldman Sachs Global Investment Research

Valuation EMC

Maintain Buy on EMC, with a new TP of NT$11,700 (from NT$10,200), based on the same 27x 2H27-1H28E EPS (in line with the peak P/E multiple for AI component suppliers in past 3 years), and factoring in our latest earnings revisions.

e92c7a75ab8b4efbba794e6b187208c8

[讀者資訊已遮蔽]

Exhibit 7: EMC P&L table

Source: Company data, Goldman Sachs Global Investment Research

TUC

We maintain Buy on TUC, with new TP of NT$3,460 (from NT$3,010) which is based on an unchanged 22x on 2H27-1H28E P/E, which is +2x STDV higher than the past 3-year industry average P/E multiple, and factoring in our earnings revisions.

Exhibit 8: TUC P&L table

Source: Company data, Goldman Sachs Global Investment Research

Investment Thesis & PT methodology and risks

Investment Thesis - Elite Material

EMC, a key supplier of high-end HDI material (70%+ market share) and SLP material (90%+ market share), has been focusing on the high-speed switch/server CCL market for e92c7a75ab8b4efbba794e6b187208c8

[讀者資訊已遮蔽]

3-5+ years. We are positive on its market share trajectory in the server industry and expect EMC to increase its share from <10% in the Purley generation to 15-20%/20+% in the Whitley/Eagle Stream platforms. EMC is also proactively expanding its share in the switch market (40%/30%+ in 2024/23 vs. <5% before 2019), with an expanding AI customer base (only one 400G customer in 3Q20, to being one of the largest suppliers in 2023), which should continue to benefit the company's top/bottom line growth over time. We believe EMC's leading position in the AI server CCL market should drive strong revenue growth and GM/OPM expansion in the long term, given that it is the major supplier for all Nvidia/Google/AWS AI server projects, and we expect it to maintain its leadership position. Considering its leading position in the high-end CCL industry, solid production skills and earnings growth outlook, we are Buy rated on the stock. With the shares trading at a 2026E P/E below AI server component suppliers and supported by a solid growth outlook, we view valuation as attractive.

Key downside risks include: (1) RCC to replace all high-end smartphone HDI design; (2) rising trade tensions, which could lead to weaker smartphone and server shipments; and (3) rising competition from mainland China peers.

Price Target Risks and Methodology - Elite Material

Valuation methodology: Our 12m TP of NT$11,700 is based on 27x 2H27-1H28E P/E (in line with the peak P/E multiple for AI component suppliers in the past 3 years).

Key downside risks: (1) RCC to replace all high-end smartphone HDI design; (2) rising trade tensions, which could lead to weaker smartphone and server shipments; and (3) rising competition from mainland China peers.

Investment Thesis - Taiwan Union Technology Corp.

TUC is a key high-end global CCL supplier that focuses on M7+ grade high-speed CCL (key applications include high-end switches (100G+) and AI server materials), with 20%+ market share in the past two years. The company continues to expect its unit market share in the server industry to increase from 15% in Whitley and Purley processors to higher in the Eagle Stream generation, and plans to launch two types of products for low-end and high-end customers (T2A and T2C) to gain market share. Also, strong growth in 400G/800G switch shipments and new 800G switch shipments, scheduled for launch in 2025-27, should continue to be a key demand driver for TUC, in our view. Additionally, the company is working on high-end AI projects with CSP and enterprise players, which we believe should drive revenue momentum in the long term. We view the stock as undervalued vs. Taiwan CCL peers on a P/E basis and rate it Buy.

Key downside risks include: (1) slower-than-expected share gains in the low-loss CCL segment; (2) rising trade tensions, which could lead to weaker server and switch shipments globally; and (3) rising competition from mainland China peers.

Price Target Risks and Methodology - Taiwan Union Technology Corp.

Valuation methodology: Our 12m TP of NT$3,460 is based on a 22x 2H27-1H28E P/E (+2x STDV higher than the past 3-year industry average P/E multiple).

e92c7a75ab8b4efbba794e6b187208c8

[讀者資訊已遮蔽]

Key downside risks: (1) slower-than-expected share gains in the low-loss CCL segment; (2) rising trade tensions, which could lead to weaker server and switch shipments globally; and (3) rising competition from mainland China peers.

e92c7a75ab8b4efbba794e6b187208c8

[讀者資訊已遮蔽]

圖片清單（已驗證 2026-10-29）