@benitoz Thanks! Great point on compiler-driven. Appreciate you sharing.
English
郭明錤 (Ming-Chi Kuo)
1.3K posts
郭明錤 (Ming-Chi Kuo)
@mingchikuo
香港天風國際證券分析師，分享科技產業趨勢觀察｜TF International Securities (HK) analyst sharing tech trend insights
Taiwan Katılım Mart 2011
348 Takip Edilen221.1K Takipçiler
@mingchikuo great supply chain work as always ! the M9-grade CCL angle is new and important the rest maps to what I published three weeks ago compiler-driven dataflow is the unlock, not SRAM density.
English
Nvidia LPU/LPX racks set for order-of-magnitude growth: application trends, ecosystem integration, and a new PCB cycle
1. My latest supply-chain checks suggest that following Nvidia’s investment in Groq, LPU shipment forecasts have moved significantly higher. Total shipments in 2026–2027 are estimated at 4–5 million units (with 30–40% in 2026 and 60–70% in 2027). Relative to historical annual volumes, this implies order-of-magnitude growth, roughly 10×+.
2. Fast-growing LPU demand reflects two key factors:
(1) Deep integration with Nvidia’s ecosystem (e.g., CUDA), significantly lowering barriers to application development and deployment.
(2) Rapid expansion of ultra-low-latency inference workloads, including AI agents (such as coding agents) as well as emerging real-time, consumer-facing, and physical-AI applications.
3. Nvidia is expected to increase LPU density per rack from 64 to 256 units. This helps preserve ultra-low latency during the decode stage of inference while accommodating expanding KV-cache requirements driven by long-context inference. The new rack architecture is expected to enter mass production in 4Q26–1Q27, with rack shipments projected at 300–500 units in 2026 and 15,000–20,000 units in 2027.
4. Three key areas to monitor regarding LPU integration into the Nvidia ecosystem:
(1) Network architecture: rack-level interconnect via NVLink Fusion and RealScale.
(2) Developer interface: whether Nvidia NIM allows developers to deploy workloads without needing to distinguish between GPUs and LPUs.
(3) Compiler integration: whether TensorRT-LLM supports the compile-first architecture of LPUs.
5. The ramp of LPU/LPX racks also has important implications for the PCB supply chain, with WUS Printed Circuit playing a key role. LPU/LPX racks represent the first large-scale deployment of M9-grade CCL materials. A successful ramp would not only contribute meaningfully to WUS’s 2027 earnings, but also confirm the company’s breakthrough in quartz-glass fabric processing for high-layer-count boards—potentially catalyzing a new growth cycle across the PCB industry.
English
Nvidia LPU/LPX機櫃將迎來數量級成長：應用趨勢、生態整合與PCB新成長週期
1. 我的最新產業調查顯示，Nvidia投資Groq後，LPU出貨規劃大幅上修。2026-2027年LPU共出貨預估約400–500萬顆（2026約30–40%、2027約60–70%），相較過去年度出貨量，將出現10倍以上的數量級成長。
2. LPU需求快速成長主要來自兩項因素：
(1) 與Nvidia生態系 (如CUDA) 高度整合，大幅降低應用開發與部署門檻。
(2) 超低延遲推論需求快速增加，包括AI agents（如coding agents）以及正在興起的real-time、consumer-facing與physical-AI等類型應用。
3. 為維持推論decode階段的超低延遲優勢，並因應長文本推理帶動的KV cache需求快速成長，Nvidia預計將每機櫃LPU數量由目前64顆提升至256顆，以擴大記憶體容量並維持超低延遲效能。新架構機櫃預計於4Q26–1Q27量產，2026與2027年機櫃出貨量分別約300–500與15,000–20,000個。
4. Nvidia生態整合LPU的三個關鍵觀察重點：
(1) 網路架構：NVLink Fusion與RealScale的機櫃互聯。
(2) 開發者介面：Nvidia NIM是否讓開發者在部署時無需區分GPU與LPU。
(3) 編譯整合：TensorRT-LLM是否支援LPU的compile-first架構。
5. LPU/LPX機櫃量產亦對PCB產業有重要意義，關鍵PCB供應商滬電股份扮演核心角色。LPU/LPX機櫃是首度大規模採用CCL M9材料的應用。若順利量產，不僅意味著LPU方案的數量級成長將在2027年為滬電股份帶來顯著貢獻，也代表該公司突破高層板石英布加工技術門檻，有望帶動PCB產業展開新一輪成長週期。
中文
Supply chain check update: Nvidia and WUS Printed Circuit have begun testing next-gen CCL material M10, which could trigger the next upgrade cycle in PCB materials for future AI servers. Key points:
1. The program suggests WUS has an early lead in PCB development for Nvidia’s next-gen rack Kyber and the new Rubin Ultra / Feynman platforms. This could support the company’s future growth momentum.
2. Sampling began in 1Q26, with preliminary test results expected in 2Q26.
3. M10's target applications include orthogonal backplanes (midplanes) designed to replace the current cartridge architecture, as well as switch blade motherboards for the Rubin Ultra / Feynman platform.
4. Unlike M9, where only Elite Material Co. was qualified, M10 testing currently involves three CCL suppliers. In addition to EMC, one Chinese vendor and one Taiwanese vendor have been added, potentially improving Nvidia’s CCL supply-chain management flexibility.
5. If testing proceeds smoothly, M10 CCL and PCB mass production could begin in 2H27.
6. From a manufacturability and commercialization perspective, the quartz cloth used in M10 may be replaced with Low Dk-2 Glass.
7. WUS is also the primary supplier of 52-layer PCBs for Nvidia’s ultra-low-latency LPU/LPX inference racks, which are expected to enter mass production in 4Q26–1Q27. With strengths in both high-layer PCB manufacturing and advanced materials development, the company appears well positioned to benefit from Nvidia’s upcoming AI server architectures.
English
供應鏈調查更新：Nvidia與滬電股份已開始測試次世代CCL材料M10，將帶動未來AI伺服器的PCB材料新一輪升級週期。幾個重點：
1. 此測試計畫意味著滬電股份在Nvidia次世代機櫃Kyber與新平台 Rubin Ultra / Feynman的PCB之開發上有領先優勢，此優勢有助於該公司未來的營運動能成長。
2. 1Q26已開始送樣與打樣，預計2Q26將取得初步測試結果。
3. M10的主要應用包括用來取代Cartridge架構的正交背板 (midplane / orthogonal backplane)，以及Rubin Ultra / Feynman平台的Switch blade主板。
4. 與M9僅台光電通過認證不同，目前M10測試共三家CCL供應商。除台光電外，另新增一家中國廠與一家台系廠，未來有助於提升Nvidia的CCL供應鏈管理彈性。
5. 若測試順利，M10 CCL與PCB預計將於2H27開始量產。
6. 從量產性與商業化考量，M10目前使用的石英布 (Quartz Cloth)，未來也可能改為Low Dk-2。
7. 滬電股份同時也是Nvidia用於AI推理的超低延遲LPU/LPX機櫃之52層PCB主要供應商 (預計在4Q26–1Q27量產)。在同時具備高層數PCB製造能力與先進材料開發能力之優勢下，該公司可望受益於Nvidia未來AI伺服器架構升級。
中文
Apple’s Q1 2026 earnings call provided a number of insights. Below are the key takeaways from a memory sector perspective:
Earnings call transcript:
seekingalpha.com/article/486418…
1. Apple highlighted two major uncertainties in its hardware supply chain: advanced nodes and memory. The former is defined as a "supply constraint," while the latter is viewed as a "cost pressure."
2. Apple’s comments aligns with my pre-earnings analysis (see quotation below): memory pricing has increased significantly, with supply guaranteed.
3. For AI memory investors, Apple’s comments are relatively less relevant. In contrast, non-AI memory investors should pay closer attention, as over the long term, the non-AI memory market is likely to reach supply-demand balance more efficiently than the AI memory market.
4. Had Apple defined memory as a “supply constraint” rather than a “cost pressure” during last week’s call, it likely would have benefited non-AI memory stocks—an outcome some short-term/bullish investors were looking for. Coincidentally, non-AI memory stocks across various markets saw a correction following the earnings call. Yet, few recent discussions have interpreted this pullback through the lens of Apple’s guidance/implications for demand.
5. Given Apple’s strong bargaining power and the fact that iPhone DRAM and NAND flash consumption accounts for around 20–25% of the global smartphone memory market, Apple’s perspective may not be an ideal proxy for broader non-AI/smartphone memory demand. In comparison, outlooks from Qualcomm and MediaTek provide a more useful reference for evaluating medium- to long-term non-AI/smartphone memory demand trends.
6. In my view, valuations of non-AI memory stocks may not yet fully reflect the strong earnings momentum expected this year. Looking further ahead to 2027–2028, AI memory demand should continue to offer better visibility.
郭明錤 (Ming-Chi Kuo)@mingchikuo
A few quick thoughts on Apple/iPhone memory price hikes: 1. The 1Q26 LPDDR price hikes mentioned in the news are pretty close to what I’ve heard. NAND flash increases, however, are a bit lower. 2. iPhone memory pricing is now negotiated quarterly instead of every six months, so expect another hike in 2Q26. Right now, the 2Q26 QoQ increase looks similar to 1Q26. 3. For most non-AI brands, even if you’re willing to pay up, there’s no guarantee you’ll get the supply. The fact that Apple can lock in a deal like this shows just how much leverage they have. 4. Higher memory costs will hit iPhone gross margins. But Apple’s playbook is clear: use the market chaos to their advantage—secure the chips, absorb the costs, and grab more market share. They’ll make it back later on the services side. 5. The cost pressure from memory could be a hot topic for investors and analysts at Apple’s earnings call this week. What Apple says could actually shake up other industries’ stocks more than Apple’s own or its suppliers'. 6. Apple’s current plan for 2H26 new iPhone 18 models is to avoid raising prices as much as possible—at least keep the starting price flat, which is helpful for marketing. 7. Apple has realized that it’s not just memory and T-glass—other components could also run short as the AI server boom continues to squeeze the rest of the supply chain. trendforce.com/news/2026/01/2…
English
Apple上週法說 (Q1 2026) 資訊量豐富，以下從記憶體角度整理重點：
法說會議記錄：
seekingalpha.com/article/486418…
1. Apple硬體供應鏈的兩大不確定性為先進製程與記憶體；前者屬「供應限制」 （supply constraint），後者為「成本壓力」 （cost pressure）。
2. Apple的說法與我在法說前的分析一致 (見下面引用)：記憶體報價顯著上漲，但能得到供應保障。
3. AI記憶體投資人可相對忽略Apple的觀點，非AI記憶體投資人則需關注，因長期來看，非AI記憶體市場達到供需平衡的效率應該高於AI記憶體市場。
4. 若Apple在上週法說中定義記憶體是「供應限制」，而非「成本壓力」，應有利於非AI記憶體股，這是部分短線/樂觀投資人想要看到的結果。很巧的是，Apple法說後不同市場的非AI記憶體股價同時出現修正，而在近期非AI記憶體股價修正相關討論中，較少看到從Apple法說/需求面來解讀的觀點。
5. 考量到Apple具備絕強的議價力，且iPhone的DRAM與NAND Flash用量均約佔手機記憶體市場的20–25%，故較不適合單憑Apple的觀點解讀非AI／手機記憶體需求。相較下，Qualcomm與聯發科對手機市場的展望，更具中長期非AI／手機記憶體需求趨勢之參考價值。
6. 我認為非AI記憶體的股價，可能還沒充分反映今年的強勁獲利動能。不過，展望2027-2028，AI記憶體的需求應該還是有較好的能見度。
郭明錤 (Ming-Chi Kuo)@mingchikuo
簡單聊一下Apple/iPhone的記憶體漲價： 1. 這篇新聞提到的1Q26 LPDDR漲價幅度跟我認知的接近，NAND Flash的漲價幅度略低。 2. iPhone的記憶體報價現在是按季談，而非半年，所以iPhone的記憶體報價在2Q26還會再漲一次。目前看2Q26的QoQ漲價幅度約跟1Q26接近。 3. 絕大部分非AI產業的品牌客戶，就算願意付錢也不見得能得到記憶體供應保證，所以Apple能談成這樣已經算很強了。 4. 記憶體報價上漲會影響iPhone的毛利率。但Apple的策略是，趁記憶體市場混亂時，透過確定能拿到貨與吸收成本的優勢，提升市佔率，後續再用服務業務賺回來。 5. 記憶體報價上漲導致成本上升的議題，應該是投資人與分析師對本週Apple法說的關注重點之一。Apple對這件事的說法，可能對其他產業的股價造成的波動，會遠高於對Apple自身與其供應鏈股價的影響。 6. Apple目前對2H26新款iPhone 18的定價策略是「盡可能不漲價」，至少起始價不動，有利行銷宣傳。 7. Apple已經意識到，在記憶體跟T-glass後，可能還會有其他零組件也會受到AI伺服器產業的影響而供應短缺。 trendforce.com/news/2026/01/2…
中文
A few quick thoughts on Apple/iPhone memory price hikes:
1. The 1Q26 LPDDR price hikes mentioned in the news are pretty close to what I’ve heard. NAND flash increases, however, are a bit lower.
2. iPhone memory pricing is now negotiated quarterly instead of every six months, so expect another hike in 2Q26. Right now, the 2Q26 QoQ increase looks similar to 1Q26.
3. For most non-AI brands, even if you’re willing to pay up, there’s no guarantee you’ll get the supply. The fact that Apple can lock in a deal like this shows just how much leverage they have.
4. Higher memory costs will hit iPhone gross margins. But Apple’s playbook is clear: use the market chaos to their advantage—secure the chips, absorb the costs, and grab more market share. They’ll make it back later on the services side.
5. The cost pressure from memory could be a hot topic for investors and analysts at Apple’s earnings call this week. What Apple says could actually shake up other industries’ stocks more than Apple’s own or its suppliers'.
6. Apple’s current plan for 2H26 new iPhone 18 models is to avoid raising prices as much as possible—at least keep the starting price flat, which is helpful for marketing.
7. Apple has realized that it’s not just memory and T-glass—other components could also run short as the AI server boom continues to squeeze the rest of the supply chain.
trendforce.com/news/2026/01/2…
English
簡單聊一下Apple/iPhone的記憶體漲價：
1. 這篇新聞提到的1Q26 LPDDR漲價幅度跟我認知的接近，NAND Flash的漲價幅度略低。
2. iPhone的記憶體報價現在是按季談，而非半年，所以iPhone的記憶體報價在2Q26還會再漲一次。目前看2Q26的QoQ漲價幅度約跟1Q26接近。
3. 絕大部分非AI產業的品牌客戶，就算願意付錢也不見得能得到記憶體供應保證，所以Apple能談成這樣已經算很強了。
4. 記憶體報價上漲會影響iPhone的毛利率。但Apple的策略是，趁記憶體市場混亂時，透過確定能拿到貨與吸收成本的優勢，提升市佔率，後續再用服務業務賺回來。
5. 記憶體報價上漲導致成本上升的議題，應該是投資人與分析師對本週Apple法說的關注重點之一。Apple對這件事的說法，可能對其他產業的股價造成的波動，會遠高於對Apple自身與其供應鏈股價的影響。
6. Apple目前對2H26新款iPhone 18的定價策略是「盡可能不漲價」，至少起始價不動，有利行銷宣傳。
7. Apple已經意識到，在記憶體跟T-glass後，可能還會有其他零組件也會受到AI伺服器產業的影響而供應短缺。
trendforce.com/news/2026/01/2…
中文
My latest supply-chain surveys indicate that Carl Zeiss SMT will significantly expand its EUV and higher-ASP immersion DUV optical system capacity by 20-25% YoY and 40-50% YoY, respectively, in 2027 to meet robust demand from ASML. Coupled with stronger-than-expected shipment outlooks for 2026, ASML's revenue is projected to reach €38–40bn in 2026 and €45–47bn in 2027, outperforming market consensus of ~€34–36bn and ~€41–43bn.
Key drivers behind the upside vs. consensus are as follows:
1. 2026 Growth Drivers:
➢ EUV and DUV shipments are estimated at 67 and 355 units, surpassing consensus of 53–55 and 310–320 units.
➢ Driven by strong 2nm demand, TSMC has upwardly revised its 2026 EUV orders twice: from an initial 22 units to 25 units last October, and currently to 28 units.
➢ To capture robust demand from Chinese memory makers, ASML plans to launch a new immersion DUV model NXT:1965i in 4Q26. As a down-spec version of the 1980i series, it complies with U.S. export controls while addressing Chinese clients' needs, serving as a key growth driver for 4Q26 and 2027.
2. 2027 Growth Drivers:
➢ 2027 capacity for both EUV and immersion DUV is currently fully booked thanks to the strong demand; further shipment upside will hinge on ASML’s supply-side improvements.
➢ Based on Zeiss SMT’s expansion, ASML’s 2027 shipments are projected to reach 80–85 EUV systems and 380–400 DUV systems.
➢ Boosted by the new 1965i launch, China’s procurement of high-ASP immersion DUV is expected to grow ~40% YoY in 2027.
3. Additional Upside Potential to Monitor:
➢ Potential ASP hikes for EUV and DUV systems amid prolonged supply tightness.
➢ Ongoing upward revisions for the new 1965i model from Chinese memory suppliers.
➢ Potential Intel Upside: The above estimates exclude incremental EUV demand from Intel. Although Intel has not officially placed orders (explaining its previously lower-than-expected Capex guidance), it has entered discussions with ASML for additional bookings. Intel's incremental EUV demand is projected at 20–30 systems over 2026–2027, of which 3–5 are High-NA EUV.
English
我最新的供應鏈調查顯示，Carl Zeiss SMT (蔡司半導體) 為滿足ASML的強勁需求，將顯著擴增2027年EUV與較高ASP的浸潤式DUV光學系統產能分別約20-25% YoY與40-50% YoY，加上2026年EUV與DUV出貨優於預期，故預估ASML 2026與2027年營收將分別達到約380-400億與450-470億歐元，優於市場共識的340-360億與410-430億歐元。
以下是優於市場共識的關鍵原因：
1. 2026年關鍵動能：
➢ 預估EUV與DUV出貨分別達67台與355台，優於共識的53-55台與310-320台。
➢ 台積電因2nm強勁需求，接連兩次上修2026年EUV訂單，從最初的22台，於去年10月上修到25台，再到目前的28台。
➢ ASML為滿足中國記憶體廠商的強勁需求，預計在4Q26推出新款浸潤式DUV NXT:1965i。此新款DUV由1980i系列降規而來，同時符合美國出口規範並更能滿足中國記憶體業者需求，將貢獻4Q26與2027年動能。
2. 2027年關鍵動能：
➢ 受益於強勁需求，2027年的EUV與浸潤式DUV目前已銷售一空，能否再提升出貨量取決於ASML的供應改善狀況。
➢ 根據Zeiss SMT的擴產規劃，ASML在2027年的EUV與DUV出貨將分別達到80-85台與380-400台。
➢ 受益於新款DUV 1965i，中國對高單價浸潤式DUV採購量量在2027年成長約40% YoY。
3. 需關注潛在上調可能：
➢ ASML因供應緊張而調升EUV與DUV價格。
➢ 來自中國記憶體廠商強勁的需求，新款浸潤式DUV 1965i訂單仍在上修中。
➢ 以上分析未計入來自Intel的新EUV需求。Intel雖未就新需求正式下單 (故先前法說的資本支出低於預期)，但已跟ASML開始協調加單事宜。目前預估Intel對EUV額外需求在2026與2027年共20-30台 (其中3-5台是High-NA EUV)。
中文
Wistron, a leading AI server assembler, reported its FY25 financial results. Gross margin (GM) for FY25 and 4Q25 came in at 6.1% and 5.6%, respectively, both below street consensus of 6.5% and 6.8%.
1. These results validate my earlier view on the structural impact of AI server design trends on assembler profitability. GM was also pressured by the product transition (GB200 NVL72 → GB300 NVL72).
2. 2026 assembly GM positives to watch: 1) ~200% YoY shipment growth for the GB NVL72 series in 1H26, where scale should be supportive; and 2) Mass production of new ASIC AI servers with higher assembly GM starting in 2H26.
3. On the downside for 2026 assembly GM: Beyond the aforementioned Nvidia AI server design trends, rising component prices will also dilute reported GM.
4. Wistron and its largest AI server customer, Dell, rallied together last September on buy-side expectations of Wistron’s market share gains and GM expansion (from ~5.5–6% to 6%+). While Wistron’s shares have remained elevated in recent months following an earlier rerating, it is worth noting that Dell’s share price has retraced to its September starting point amid profitability/GM concerns, following sell-on-news after a sharp backlog increase.
郭明錤 (Ming-Chi Kuo)@mingchikuo
AI server assembler Wiwynn’s recent 4Q25 gross margin miss (7.2% vs. consensus estimates of 8–8.3%) has triggered a share price correction and reignited investor concerns. While rising component costs can mechanically dilute reported gross margins, it is more critical to examine the underlying profitability trends in server assembly through the lens of structural changes in AI server design. Nvidia continues to increase the level of design integration in AI servers to boost token output per unit of space and power, addressing the constraints of limited data-center space and scarce electricity supply. Highly integrated designs also help improve manufacturing efficiency, which in turn benefits supply-chain management and reduces maintenance costs. VR200 NVL72 is a representative example of this high-integration approach. Based on my industry surveys, its compute tray adopts a cable-less design, resulting in an ~40% reduction in the number of component items vs. GB300 NVL72. However, higher integration also comes with a greater share of Nvidia-specified components and a shrinking scope for customization, both of which are unfavorable to assembler profitability. My industry survey indicates that components classified as “customization not allowed” in the VR200 NVL72 compute tray have risen to 20–22%, well above the 5–7% level seen in GB300 NVL72. Moreover, VR200 NVL72 has directly eliminated "ODM-designed" component items entirely. In fact, this trend toward higher integration began with GB300 NVL72. My industry checks suggest that white-box GB200 NVL72 assembly gross margins were still around 18–20% in 1Q25, but by 1Q26, white-box GB300 NVL72 margins are expected to decline sharply to 6–8%. Importantly, this trend is likely to persist. Looking ahead, Kyber rack designs are expected to feature an even higher degree of integration and face greater margin pressure than the current Oberon generation. Assemblers will not remain passive in the face of these challenges. Beyond straightforward improvements in manufacturing efficiency, there are currently three practical approaches to addressing margin pressure in the server business, although their effectiveness remains to be seen: 1. optimizing product mix by increasing exposure to higher-margin ASIC AI servers; 2. securing NPI opportunities to enhance bargaining power with customers during the early stages of mass production; and 3. pursuing vertical integration strategies—ranging from qualifying in-house or group-affiliated server components through Nvidia validation, to expanding the business scope toward providing full data-center build-out solutions.
English
AI伺服器組裝大廠緯創公布2025年財務業績，關鍵之一的毛利率在2025年與4Q25分別為6.1%與5.6%，均低於市場共識的約6.5%與6.8%。
1. 緯創的財務業績除驗證我先前提到AI伺服器設計趨勢對組裝廠獲利之結構性影響，另一原因是產品轉換 (GB200 NVL72 ⭢ GB300 NVL72) 對組裝廠的毛利率也有壓力。
2. 2026年組裝廠的毛利率利多可觀察：1) GB NVL72系列在1H26出貨顯著成長約200% YoY，規模擴大應有幫助、與2) 組裝毛利率較高的新款ASIC AI伺服器在2H26開始量產。
3. 2026年組裝廠的毛利率挑戰方面，除上述提到Nvidia AI伺服器設計趨勢外，還有零組件漲價會稀釋帳面毛利率。
4. 緯創與其最大AI伺服器客戶Dell在去年9月股價一同起漲，反應法人調升緯創的市占率與毛利率 (2025年的5.5-6% ⭢ 6%+)。緯創股價雖受益於重新評價 (rerating) 故近期仍在高檔，但需關注Dell股價在顯著成長的積壓訂單 （Backlog） 利多出盡後，因市場對其獲利能力/毛利率擔憂，股價已回到去年9月起漲點。
郭明錤 (Ming-Chi Kuo)@mingchikuo
近期AI伺服器組裝廠商緯穎公布低於預期的4Q25毛利率 (7.2% vs. 市場預期的8-8.3%)，導致股價下跌，此事又引起投資人關注。雖零組件漲價也會稀釋帳面上的毛利率，但更重要的是從AI伺服器設計的本質上去檢視組裝的獲利能力趨勢。 Nvidia持續提升AI伺服器的設計整合程度，以提升每單位空間與電力的Token產出，來因應資料中心有限空間與稀缺電力挑戰。此外，高度整合的設計也有助於改善生產效率，進而有利供應鏈管理與降低維修成本。 VR200 NVL72就是設計高度整合的例子，我的產業調查顯示，其Compute tray導入無線化設計（Cable-less），零組件項目顯著減少約40% (vs. GB300 NVL72)。 但隨整合程度提高，伴隨而來的是Nvidia指定用料比重提升，與客製化空間也遭到壓縮，這些都不利組裝廠商的獲利能力。我的產業調查顯示，VR200 NVL72 compute tray「不允許客製化規格」的零組件項目佔比提升至20-22%，遠高於GB300 NVL72的5-7%，且VR200 NVL72更直接取消「ODM設計」的零組件項目。 事實上，提高整合程度的設計趨勢自GB300 NVL72就已開始。我的產業調查指出，白牌GB200 NVL72在1Q25的組裝毛利率尚維持在18-20%，但到1Q26，白牌GB300 NVL72的毛利率預估將顯著下滑至6-8%。 值得注意的是，上述趨勢只會繼續延續，未來Kyber機櫃的設計整合程度與毛利率壓力將比現在的Oberon更高。 組裝廠商們當然不會做坐以待斃。面對上述伺服器事業的毛利率挑戰，除單純改善生產效率，實務上目前有3種做法應對，不過實際幫助仍需觀察。3種做法是： 1. 改善產品組合：提升毛利率較高的ASIC AI伺服器出貨比重、 2. 爭取NPI以提高量產初期對客戶的議價力、與 3. 垂直整合策略：小至爭取自家/集團內的伺服器零組件通過Nvidia驗證，大至將事業範疇拓展至提供資料中心建置方案。
中文
Recent market chatter suggests HP may raise its 2026 notebook shipment outlook. My latest supply-chain surveys indicate this is not the case. HP has only verbally asked suppliers to extend lead times from about one month to six months, while full-year shipment volume remains largely unchanged. Since there’s no upfront payment, suppliers are unlikely to truly build inventory at that level—though some incremental stocking may occur, creating the appearance of order growth.
Two key takeaways:
1. A notable increase in HP notebook shipments is unlikely with no meaningful improvement in memory supply or end-market demand. This looks more like a move to get ahead of potential component price hikes by building inventory early.
2. If other brands follow HP and start building inventory as well, the short-term appearance of stronger notebook component orders will become more pronounced.
English
近期市場傳言HP要調升2026年筆電出貨預估。我最新的產業調查顯示並非如此，HP僅是口頭上要求供應鏈提前備貨時間從約1個月增加到6個月，但全年出貨量無太大變化。對供應鏈來說，因為HP沒有付錢，所以不大可能真的這樣備貨，但多少會因為HP的要求多備一些，故有加單的假象。
兩個重點：
1. 因記憶體供應與市場需求無顯著改善，故不預期HP筆電出貨量會明顯成長。HP此舉較像是因應未來數個零組件價格可能上漲故提前備貨。
2. 如果其他品牌也跟進HP提前備貨，那筆電零組件短期內加單的假象會更顯著。
中文
@8teAPi No, the land is still TSMC’s. It’s putting money down upfront to secure capacity on that land.
English
😂 Jensen buys land for TSMC becoming their landlord way in advance of them building capacity for him (also reducing his future expenses and their future margin)
郭明錤 (Ming-Chi Kuo)@mingchikuo
TSMC announced 2026 capex of US$52–56bn at today’s earnings call, with strong Nvidia demand as the key driver. 1. Ahead of the call, sell-side analysts and media generally expected 2026 capex of US$45–52bn, while the buy side was already at US$53–56bn. 2. TSMC is typically conservative on capex guidance, so final 2026 spending is very likely to beat buy-side pre-call expectations. 3. Most customers negotiate for guaranteed capacity. Nvidia CEO Jensen Huang negotiates for land. In November last year, he visited Tainan and told TSMC he was willing to pay to secure the reserved P10 and P11 land next to Fab 18. At the time, Fab 18 plans were only clearly defined for P1–P9, and Jensen Huang’s ultimate goal was to secure the P12 land as well. 4. Before Huang raised the idea of securing P10 and P11 land, buy-side consensus for TSMC’s 2026 capex was US$45–50bn. 5. Even with tight advanced-node and packaging capacity, TSMC stays highly disciplined on demand approval, often approving less than customers’ most optimistic requests. Want the most aggressive capacity? Put more cash down, book the land, and pay to grow capacity—just as Huang did. That said, with resources increasingly constrained across the board, even if other customers try to follow Huang’s playbook, securing the most aggressive capacity is becoming harder and harder.
English
@benitoz No, the land is still TSMC’s. It’s putting money down upfront to secure capacity on that land.
English
@mingchikuo Wow. I didn’t know that. Jensen literally bought empty land and all future fab capacity. It’s like field of dreams. Only he can make such bets.
English
TSMC announced 2026 capex of US$52–56bn at today’s earnings call, with strong Nvidia demand as the key driver.
1. Ahead of the call, sell-side analysts and media generally expected 2026 capex of US$45–52bn, while the buy side was already at US$53–56bn.
2. TSMC is typically conservative on capex guidance, so final 2026 spending is very likely to beat buy-side pre-call expectations.
3. Most customers negotiate for guaranteed capacity. Nvidia CEO Jensen Huang negotiates for land. In November last year, he visited Tainan and told TSMC he was willing to pay to secure the reserved P10 and P11 land next to Fab 18. At the time, Fab 18 plans were only clearly defined for P1–P9, and Jensen Huang’s ultimate goal was to secure the P12 land as well.
4. Before Huang raised the idea of securing P10 and P11 land, buy-side consensus for TSMC’s 2026 capex was US$45–50bn.
5. Even with tight advanced-node and packaging capacity, TSMC stays highly disciplined on demand approval, often approving less than customers’ most optimistic requests. Want the most aggressive capacity? Put more cash down, book the land, and pay to grow capacity—just as Huang did. That said, with resources increasingly constrained across the board, even if other customers try to follow Huang’s playbook, securing the most aggressive capacity is becoming harder and harder.
English
@Alex_Intel_ No, the land is still TSMC’s. It’s putting money down upfront to secure capacity on that land.
English
