Ming-Chi Kuo

一些關於Super Micro的預測 (地緣政治的角度) https://t.co/zUdmY5mG4m

I get why some people want to see, in the short term, a dramatic boost in iPhone shipments thanks to Apple Intelligence. However, looking back over 20 years of consumer electronics, it's extremely rare to see software or service upgrades alone driving a significant surge in hardware sales. Apple would no doubt love for Apple Intelligence to accelerate new hardware replacement cycles. But if that were Apple's top priority, how would we explain why the years-old M1 iPad and Mac can run Apple Intelligence?

我能夠理解，有些人期望iPhone出貨量能在短期內因為Apple Intelligence而有戲劇性的推升。但從消費電子過去20餘年的歷史來看，僅憑軟體或服務的升級就可以顯著推升硬體出貨量的例子非常、非常的罕見。 我相信Apple應該很樂於見到Apple Intelligence能顯著帶動硬體換機週期，但如果Apple將此視為第一要務，那又如何解釋多年前的M1 iPad與Mac就能運作Apple Intelligence？

Broadcom currently supplies over 300 million Wi-Fi+BT chips (hereafter referred to as Wi-Fi chips) per year to Apple. However, Apple will rapidly reduce its reliance on Broadcom. With new products in 2H25 (e.g., iPhone 17), Apple plans to use its own Wi-Fi chips, which will be made by TSMC's N7 process and support the latest Wi-Fi 7 spec. Apple expects to move nearly all products to in-house Wi-Fi chips within about three years. This move will reduce costs and enhance Apple's ecosystem integration advantages.

目前Broadcom每年出貨3億多個Wi-Fi+BT晶片(以下簡稱Wi-Fi)給Apple，但Apple將開始快速降低對Broadcom的依賴。Apple預計將自2H25新產品 (如iPhone 17) 開始採用自家Wi-Fi晶片，該晶片由台積電N7製程生產且支援最新的Wi-Fi 7，預計約3年內Apple所有產品幾乎改採自家Wi-Fi晶片。Apple此舉除能降低成本，更能提升生態整合優勢。

I believe the likelihood of ARM canceling Qualcomm's license is extremely low. Such a move would be mutually destructive, with no winners. This dispute is likely to end in some form of settlement. https://t.co/9DZTzMwR8K

To avoid any misunderstanding, the 10 million units cut is the sum of three-quarters of cuts (from 4Q24 to 2Q25)

避免誤解，1000萬支的砍單是指共三個季度的砍單總和 (從4Q24到2Q25)

A year ago, I predicted that ASML's EUV orders/shipments in 2024 would be lower than expected. One of the main reasons for ASML's lower-than-expected orders is the limited increase in orders from the memory industry. Although AI server demand is very strong, the production bottleneck for HBM is TSV. Memory makers don't need to buy EUV equipment without an improvement in the overall demand for consumer electronics. This trend is similar to what Largan mentioned in the Q3 earnings call yesterday about declining demand after November.

一年前我預測了ASML 2024年EUV訂單/出貨會低於預期。ASML訂單低於預期的其中一個主因來自記憶體產業訂單增加有限。雖然AI伺服器需求非常強勁，但HBM的生產瓶頸是TSV，故消費電子需求沒有起色下，記憶體業者就沒有必要買EUV設備。而這也與昨天大立光第三季法說提到的11月之後需求下滑的趨勢有共通之處。

市場傳言ASML要對台積電漲價設備價格，但我的理解是台積電反而會對ASML砍價。 Market rumors suggest that ASML plans to raise equipment prices for TSMC. However, my understanding is that TSMC is likely to push for price reductions from ASML.

奇鋐為微軟4Q24 GB200訂單顯著增加3–4倍的領先受益者，4Q24與2025營運動能可望超預期 https://t.co/gVVeRLKlS1

AVC is the early beneficiary from Microsoft’s 3–4 times significant increase in 4Q24 GB200 orders, with better-than-expected 4Q24 and 2025 business momentum https://t.co/kTOG6TP1ra

GB200訂單更新：Microsoft需求高於其他CSP訂單總和，4Q24訂單顯著增加3–4倍；Microsoft供應鏈為投資關注焦點 https://t.co/oV59dpkIOu

GB200 order update: Microsoft’s demand exceeds total orders from other CSPs, 4Q24 orders significantly increase 3–4 times; Microsoft’s supply chain becomes investment focus https://t.co/r0IybwUGlT

Snapdragon 8 Gen 4強勁出貨將顯著推升Qualcomm 4Q24/F1Q25營運動能 https://t.co/f7EHDMJW8s

Tesla's robotaxi mass production timeline aligns with my previous survey result. At the event, Elon Musk stated: "Well, I tend to be a little optimistic with time frames. But in 2026. So yeah. Before 2027. Let me put it that way." Considering Musk's careful choice of words and his specific mention of 2027 as a benchmark for Cybercab production, along with his history of setting optimistic targets, this announcement aligns with my previous survey of Cybercab mass production commencing in 1Q27. 1Q27 is still far off, and the Cybercab production timeline may shift. Key points to watch for $TSLA in the coming months include 1) the next earning…

Tesla robotaxi量產時間與先前調查結果一致。 Elon Musk在發佈會說到：Well, I tend to be a little optimistic with time frames. But in 2026. So yeah. Before 2027. Let me put it that way. 從Musk謹慎的措辭和使用 2027 年作為Cyber​​cab生產時程的分水嶺，並考慮到他過去較常提出樂觀目標的記錄，我認為發布結果與先前對1Q27季開始Cyber​​cab量產的調查一致。 距離1Q27還久，Cybercab的量產時間還有改變的可能。$TSLA的未來數月內觀察重點包括：1) 下一個財報會議、2) 美國總統大選結果、與 3) Tesla能否在2025年在加州與德州提供無人監管的FSD。

As Nvidia's GB200 nears mass production, more components are gaining new suppliers, which helps reduce costs and improve quality control. However, it also signals rising competition within the supply chain. Two recent examples of key power components are: 1. Integrated base controllers: Previously supplied only by Delta Electronics, MPS has become the second supplier. 2. Voltage regulators: Initially supplied exclusively by MPS, Infineon and TI have now been added as new suppliers.

隨著Nvidia GB200接近量產之際，越來越多的零組件增加新供應商，有利降低與成本與品質控管，但也意味著供應鏈競爭壓力浮現。舉兩個近期關鍵電源零組件的例子 1. 整合基座控制器：原本是台達電獨供，現在新增MPS。 2. 電壓調節器：原本是MPS獨供，現在新增Infineon與TI。

iPhone 16 shipment updates 1. Assembly orders almost remain unchanged. Suppliers had been asked to continue producing two Pro models during China's National Day holiday, indicating that demand for the Pro models has generally met expectations so far. 2. Some component orders for mid-November onwards have been slightly reduced (mainly for the 16 base model and 16 Plus), but by less than 3-5%, which is essentially negligible. 3. Demand for the 16 base model and 16 Plus remains lackluster compared to last year. 4. The 16 Pro's shipments are similar to last year's 15 Pro, but it's worth noting that the current shipping times for the 16 Pro ar…

iPhone 16出貨更新： 1. 組裝訂單幾乎沒有變動，供應商被要求在中國十一長假期間不休假繼續生產兩款Pro系列，意味著兩款Pro機型的需求至今大致符合預期。 2. 部分零組件有略砍11月中旬後的訂單 (主要是16標準版與16 Plus)，但幅度小於3-5%，基本上可忽略。 3. 16標準版與16 Plus的需求相較去年依舊沒有太大起色。 4. 16 Pro的出貨量與去年15 Pro相似，但需注意目前16 Pro的出貨時間較15 Pro短。 5. 下一個觀察重點是Apple Intelligence在10月下旬上線後，對美國市場需求/出貨的影響。 6. 維持iPhone 16在4Q24 8,800萬-8,900萬支的生產預估，略低於去年的9,000萬-9,100萬支。

Nvidia halting development of GB200 NVL36*2 (dual-rack 72 GPUs version); long-term AI and Nvidia trends remaining positive, but frequent product plan changes likely softening short-term trading sentiment https://t.co/smmZyIyRzJ

Nvidia停止GB200 NVL36*2 (72GPU雙櫃版 ) 開發；AI與Nvidia長期趨勢仍正向，但產品規劃頻繁改變恐讓短期交易氣氛降溫 https://t.co/9l1UBSAQrl

About Meta's AR glasses Orion: 1. Meta's internal product roadmap indicates that Orion could enter mass production as early as 2027. However, as an EMS/NPI supplier has not yet been selected and many technical issues remain resolved, the production timeline will likely be later. 2. Before Orion's mass production, Meta may launch a lower-spec version (e.g., using LCoS technology instead of Micro LED). Meta's strategy is to continuously test and gather market feedback to rapidly improve products rather than wait for perfection before release. 3. Orion is positioned to replace smartphones and tablets. Its key selling point is the integration of …

關於Meta的AR眼鏡Orion： 1. Meta內部的產品路線顯示，Orion最快在2027年量產。但因目前尚未決定EMS/NPI廠商，且考慮到諸多技術問題，故實際量產時間應會更晚。 2. 在Orion量產前，Meta可能會推出較低規格的版本 (如採用LCoS以取代Micro LED)。Meta的策略就是會不斷測試與收集市場反應，以此快速改善產品，而不會等到產品完美後才發布。 3. Orion的定位就是要取代手機跟平板，關鍵賣點為整合AI與空間運算。

我的產業調查也顯示，Qualcomm目前正與Intel交涉併購事宜。但此併購若發生，對Qualcomm可能是一場災難 https://t.co/hwaxH3s9Wx

My industry survey also indicates that Qualcomm is currently in takeover talks with Intel. However, if this acquisition were to go through, it could prove disastrous for Qualcomm https://t.co/fJ7EozISve

華為三折手機Mate XT的調查與分析；出貨預估調升但需求熱度能否延續為關鍵 https://t.co/5wn3SAghlG

Survey and analysis of Huawei’s tri-fold phone Mate XT; shipment estimates raised, but whether demand can continue is key https://t.co/eMEJ9fUu6O