AI Trends - Gemini 3 became a central focus in the market following its launch [1] - The report discusses the ongoing debate surrounding the AI bubble, focusing on demand and valuation [1] Apple's Strategy - Apple is exploring crisis-driven innovation, including on-device AI capabilities [1] - Apple is potentially developing a foldable iPhone and smart glasses [1]

我在上個月 (11月27日) 接受財經M平方的訪談，廣泛聊聊Apple、消費電子與AI產業趨勢，內容主要分成三個主題：1. Gemini 3 － 發布後迅速成為市場焦點。2. Apple－生存危機驅動創新動機、裝置端AI、折疊iPhone與智慧眼鏡。3. AI泡沫－需求與估值為核心議題。全文連結：https://t.co/ymnhfb2RBS https://t.co/iSrYOfctgq ...

Market Trends & Industry Dynamics - Intel is projected to begin shipping Apple's lowest-end M-series processors as early as 2Q-3Q27, utilizing the 18AP advanced process [1] - Apple's plan hinges on the progress after obtaining PDK 1.0/1.1, expected from Intel in 1Q26 [1] - The lowest-end M-series chips are primarily used in MacBook Air and iPad Pro, with a combined shipment volume of approximately 20 million units in 2025 [1] - Expected shipments of the lowest-end M-series processors are projected to be between 15 million and 20 million units in both 2026 and 2027, potentially influenced by a new lower-priced MacBook equipped with an iPhone processor in 2026 [1] Strategic Implications & Supply Chain - Apple aims to diversify its supply chain by seeking a second supplier, aligning with the "Made in America" initiative, while still heavily relying on TSMC's advanced processes [2] - Securing Apple's advanced process orders holds significant symbolic value for Intel, outweighing the actual revenue and profit contribution [2] - This order signifies that the worst may be over for Intel's IFS business, potentially leading to more orders from Apple and other major clients for future 14A or more advanced processes, improving long-term prospects [2] - The order for the lowest-end M-series processors is not expected to impact TSMC's fundamentals or its leading position in the coming years [2]

Intel expected to begin shipping Apple’s lowest-end M processor as early as 2027There have long been market rumors that Intel could become an advanced-node foundry supplier to Apple, but visibility around this had remained low. My latest industry surveys, however, indicate that visibility on Intel becoming an advanced-node supplier to Apple has recently improved significantly.Apple previously signed an NDA with Intel and obtained the advanced-node 18AP PDK 0.9.1GA. The key simulation and research projects ( ...

Industry Outlook - Intel is projected to begin shipping Apple's lowest-end M processors as early as 2Q-3Q27, utilizing the 18AP advanced process [1] - The actual timeline depends on the development progress after obtaining PDK 1.0/1.1, expected to be released by Intel in 1Q26 [1] - The lowest-end M chips are primarily used in MacBook Air and iPad Pro, with a combined shipment volume of approximately 20 million units in 2025 [1] - Due to potential impact from new low-cost MacBooks with iPhone processors, the projected shipment volume for the lowest-end M processors in 2026 and 2027 is estimated to be between 15 million and 20 million units [1] Strategic Implications - The order for the lowest-end M processors has no impact on TSMC's fundamentals and leading advantage in the coming years [2] - For Apple, this move supports the "Made in America" policy and diversifies its supply chain, reducing reliance on TSMC [2] - For Intel, securing Apple's advanced process order is more significant than the actual revenue and profit contribution, signaling a positive turn for its IFS business [2] - This order suggests that Intel's 14A or more advanced processes may attract more orders from Apple and other major clients, improving its long-term prospects [2]

Days Sales Outstanding (DSO) Analysis - Critics argue Nvidia's DSO increase from an average of 46 days in FY2020-2024 to 53 days in Q3 FY26 indicates financial irregularities [1] - The increase in DSO is reasonable due to accounts receivable concentration, with major customers' share rising from 238% in FY2020-2024 to 65% in Q3 FY26 [2] - Cloud service providers (CSPs) have significant bargaining power, leading to longer payment terms and influencing DSO [3] - Comparing Nvidia's DSO to suppliers with similar CSP customers, such as Arista, Celestica, and Vertiv (whose DSOs typically run above 60-70 days), makes Nvidia's 53-day DSO appear reasonable [3] Inventory Analysis - Critics describe the 32% QoQ increase in Nvidia's Q3 FY26 inventory as a "paradox," citing Q2 FY23 as a counterexample [3] - Q2 FY23 inventory increased by roughly 23% QoQ to USD 3889 billion, contradicting the claim of an 18% decline [4] - The 32% increase in Q3 FY26 inventory aligns with the ramp in upstream capacity, as TSMC's CoWoS average monthly output grew by roughly 25-30% QoQ to around 60 thousand wafers per month (kwpm) in 3Q25 [5] - Work-in-Process (WIP) represented 442% of Nvidia's total inventory in Q3 FY26, surging about 98% QoQ to USD 8735 billion, reflecting the ramp and mass production of the new Blackwell B300 GPU [5]

在Nvidia公布Q3 FY26財報後，有觀點質疑該財報中的應收帳款週轉天數（Days Sales Outstanding；DSO）與庫存（Inventory）異常且疑似詐欺。但我認為，該財報結果反而符合財報常識與產業現況。1. 應收帳款週轉天數：(1) 質疑者認為Nvidia的DSO由FY2020－2024的平均46天增加到Q3 FY26的53天，此惡化是金融異常與詐欺。(2) 然而，上述並沒有考慮到應收帳款的客戶集中度變化，從FY2020－2024的平均23.8%，顯著提升到Q3 FY26的65%。(3) DSO隨應收帳款的客戶集中度顯著提升而增加很合理，這反映了大客戶的議價力，特別Nvidia目前面對的是付款條件向來較久的各大CSP。(4) 質疑者的另一錯誤是將Nvidia與其他主要客戶非CSP的廠商之DSO來比較。事實上，其他主要客戶為各大CSP的廠商，如Arista、Celestica與Vertiv等，其DSO普遍在60－70天以上，這是另一個驗證Nvidia DSO為53天的合理證明。2. 庫存：(1) 質疑者將Nvidia的庫存金額在Q3 FY26增加32% QoQ解釋為庫存悖論，並舉出Q2 F ...

Days Sales Outstanding (DSO) Analysis - Critics argue Nvidia's DSO increase from an average of 46 days in FY2020-2024 to 53 days in Q3 FY26 indicates financial irregularities [1] - The increase in DSO is reasonable due to a shift in accounts receivable concentration, with major customers accounting for 65% in Q3 FY26, up from an average of 238% in FY2020-2024 [2] - Cloud service providers (CSPs) have significant bargaining power and historically operate with longer payment terms, justifying the rise in DSO [3] - Nvidia's 53-day DSO is reasonable when compared to suppliers like Arista, Celestica, and Vertiv, whose DSOs typically exceed 60-70 days [3] Inventory Analysis - Critics describe the 32% QoQ increase in Nvidia's Q3 FY26 inventory as a "paradox" [3] - Q2 FY23 inventory increased by roughly 23% QoQ to USD 3889 billion, contradicting claims of a decline [4] - The 32% increase in Q3 FY26 inventory aligns with the ramp in upstream capacity, with TSMC's CoWoS average monthly output growing by roughly 25-30% QoQ to around 60kwpm [5] - Work-in-Process (WIP) represented 442% of Nvidia's total inventory in Q3 FY26, surging about 98% QoQ to USD 8735 billion, reflecting the ramp and mass production of the new Blackwell B300 GPU [5]

在Nvidia公布Q3 FY26財報後，有觀點質疑該財報中的應收帳款週轉天數（Days Sales Outstanding；DSO）與庫存（Inventory）異常且疑似詐欺。但我認為，該財報結果反而符合財報常識與產業現況。1. 應收帳款週轉天數：(1) 質疑者認為Nvidia的DSO由FY2020－2024的平均46天增加到Q3 FY26的53天，此惡化是金融異常與詐欺。(2) 然而，上述並沒有考慮到應收帳款的客戶集中度變化，從FY2020－2024的平均23.8%，顯著提升到Q3 FY26的65%。(3) DSO隨應收帳款的客戶集中度顯著提升而增加很合理，這反映了大客戶的議價力，特別Nvidia目前面對的是付款條件向來較久的各大CSP。(4) 質疑者的另一錯誤是將Nvidia與其他主要客戶非CSP的廠商之DSO來比較。事實上，其他主要客戶為各大CSP的廠商，如Arista、Celestica與Vertiv等，其DSO普遍在60－70天以上，這是另一個驗證Nvidia DSO為53天的合理證明。2. 庫存：(1) 質疑者將Nvidia的庫存金額在Q3 FY26增加32% QoQ解釋為庫存悖論，並舉出Q2 F ...

VR200 NVL144 Rail Survey UpdateAlthough competition is everywhere—and rails are classified as a higher-replaceability Group C component in Nvidia’s AI server parts categories—my latest supply chain checks indicate that the VR200 NVL144 currently in development, testing, and assembly is using rails solely supplied by King Slide. The part-number prefix is the same as that of the GB300 NVL72. ...