Gooaye 股癌

Google 的 OCS 省掉光電轉換解決傳輸能耗瓶頸 “Over the past year, OCS in the datacenter has expanded from a singular Google project to a multi-vendor market with real deployments, purchase orders, and a credible path to multi-billion-dollar annual sales,” said Scott Wilkinson, Lead Analyst for Optical Components at Cignal AI. “The revised forecast, the number of active vendors, and the breadth of operator evaluations make it clear that optical circuit switching will be a foundational technology for large-scale AI systems rather than a niche option.” https://cignal.ai/2025/12/optical-circuit-switching-market-to-exceed-2-5b-in-2029/

EP617 | 👾 短期內第二個參與資收的經驗 iHerb黑五優惠開跑！ 美國保健品龍頭 iHerb，5萬種正品最快3天到貨 🔥主委推薦必買清單： Nature's Bounty 鈣鎂鋅→幫助放鬆好入睡 Orgain 有機蛋白粉→健身補給首選 Garden of Life男性益生菌→腸道健康神隊友 iHerb黑色星期五期間限定優惠： 🎁12/10-12/15輸入優惠碼 「GOOAYE24」享 76 折 🎁12/16-12/30輸入優惠碼 「GOOAYE20」享 8 折 馬上去結帳👉https://maac.io/4XuTA 跟iHerb成為LINE好友👉https://maac.io/4XuUn 折扣碼適用地區：台港澳、東南亞、美國、泰國、菲律賓 ※台灣半年內跨境購物 6 次、單筆不超過 NT$2,000 還能享免稅 股癌傳送門：http://linktr.ee/gooaye

朋友公司徵人，有興趣都可投看看

記憶體續噴，然後 RTX 4090 和 5090 的價格倒掛 RTX 4090 (非D版) 有完整的 tensor core 性能和算力，而 RTX 5090D (中國特供) 則為了合規出口，TPP 壓低到美國商務部的紅線下。一個是遊戲卡，另一個賣一張少一張是算力卡的概念 " - 中國國產記憶體 DDR4，32GB 套條在今年 6 月的日常價不過 300 元，目前的價格已經漲到 1169 元 - 中國國產記憶體 DDR5，半年前的價格不到 500 元，現價 2159 元，已超四倍 - 海力士顆粒的 DDR5 套條漲幅小一些，因為本來就賣得比較貴，半年前價格大約 900 元，現在價格 2500 百元" https://zhuanlan.zhihu.com/p/1981043445880602883

https://podcasts.apple.com/us/podcast/%E5%A5%BD%E4%B9%85%E4%B8%8D%E8%A6%8B%E7%9A%84%E7%AC%AC%E4%BA%8C%E5%AD%A3%E7%AC%AC%E4%B8%80%E9%9B%86-%E7%8D%A8%E7%AB%8B%E9%81%8A%E6%88%B2%E7%9A%84%E9%BB%83%E9%87%91%E6%99%82%E4%BB%A3%E8%A6%81%E4%BE%86%E8%87%A8%E4%BA%86/id1535478733?i=1000740134070 nice!

CRM Q3 and special call notes: LLM 帶來了 DIY 實驗浪潮，但企業在嘗試之後意識到，要實現大規模、準確，且結合既有業務流程，包含應用程式和數據的 AI，還是需要解決 "last mile" 問題。salesforce 認為因此，驅動了其 agentforce 和 data Cloud 的市場需求，進而加速了預訂數字和收入增長 對於 CRM 所述的 "last mile" 可理解為企業可以把資料尻去外面串 AI 做 DIY 實驗，但玩一玩會發現：a) 成本過高和耗時, b) 跨部門功能整合串接困難, c) 工作流習慣難更改, d) 產出一致性不好 OAI 發難之初，滿多說法認為 G 社瀏覽器將消亡，隨後發現沒這麼簡單，因為沒考慮到人們的使用習慣不容易改變；又如有一說 "未來無軟體"，需要什麼應用的時候直接 gen 出來就好，這沒考慮成本和一致性標準化的問題 未來絕對怎樣怎樣的評論，拿來說書好用吸睛，但實務運作總是漸進 如果今天是以單兵作戰的角度來說，AI 可以把本來 "不會" 做的事情都弄上 60 分水準。但如果是自己 "會，且專業" 的內容，馬上就能抓出一堆問題。這應該是很多人使用上的體驗，能問通識，但專業項目會失真 為求準確就需要所謂的 "last mile"，去 tune 出有價值和一致性的產出。總結還是成本：花費是成本、做錯被客訴是成本、土炮瞎搞砸招牌也是成本 --- "We bring the last mile. And this is what companies are realizing. They've been experimenting. Many of them -- I was in a conversation yesterday. Somebody said, Miguel, I've been talking to some customers of Salesforce that are building the agentic layer outside Salesforce. I'm like, yes, there are many of them that try to do it, try to get the data because you can do it yourself anything. They get the data out of Salesforce, then they build the logic, then they decide that they want to take some actions, but they don't have the capability to execute the thousands of workflows that the customers have already built on Salesforce. So basically, by doing that -- and by the way, the humans are working here, and they built this agentic layer here, trying to build a cool interactive conversational UI, but a UI that is disconnected from the humans and it disconnected from the execution and is semi-disconnected with the data because the one -- the moment you move the data out of Salesforce into other data lakes, et cetera, it just becomes obsolete. You can do it, but it's expensive. So most of these customers have realized that we need -- they need the last mile. And they turn to us and they said, "let's do it together. Let's start the agentic transformation journey." " We also launched ITSM. I think we're going to have a lot of good discussions in our next earnings around ITSM is booming already. We just launched it. We won dozens of customers. Life Science Cloud was a great example. It just quadrupled or I don't know the specific statistic, but it grew a lot during Q3. And I just want to mention that this is pretty big because we were doing a lot of business in life science and also medical devices, et cetera. And -- but as you know, for many years, we had this great partner that we basically partner and capturing that opportunity. We were capturing the outside commercial side, the opportunity. They were very focused on the commercial side and the clinical side. And then 1.5 years ago, they decided to compete head-to-head with us, and we decided to build solutions to compete with them, and the results have been incredible. I mean we announced yesterday Novartis. We had announced Pfizer of the top 20. Of the top 20, we've already won 5 or 6 and I say 5 or 6 because there's one that is in embargo, but I think there's going to be a press release later today, where we're going to announce another major pharma company going to us. So we are gaining market share from Veeva like there is no tomorrow. We've already -- in addition to the big 20, which there are still many in the air and they are reviewing, they all want the Salesforce platform. They want our agentic capabilities. They want our data cloud to unify data across all the different domains in the business. And they are afraid to move with a small player that is going to build a new platform for them. So many of them want to stay and are staying there's more than 100 life science cloud -- life science customers that have selected Life Science Cloud and are moving off Veeva, and we are just getting started. So that's an example of innovation. "

EP616 | 🥚 買你心中卻有別人 主委一直被問到爛的組合回來了🔥 → https://reurl.cc/9R9keV 說真的，盤勢可以震，我體力不能震 邏輯很簡單，盤勢有停損，體力也要有補給 補對東西，狀態直接噴回多頭，什麼都對了 📉身體像走勢 該補不補就一路往下破 📈但補對東西 狀態能站回季線，維持在多頭區間 長線績效靠的是穩定，不是一次爆發 這兩年我都堅持補： 米鴻這組📦把我體力打回上升軌道 吸收快、濃度高，就像打進一根漂亮長紅 止跌反彈很明顯🔥 【MIHONG®司令B群黑紅瑪卡💊】 #膠囊版 #日常持續 ☑️ 專利祕魯95%高純度MacaPro黑紅瑪卡 ☑️ 600%濃縮，吸收效率像突破壓力線 ☑️ 草本調節機能，維持活力不亂震 ☑️ 足量L-精胺酸＝體力的基本面 ☑️ 兩顆含2400mg濃縮精華 ☑️ 七大無添加＋第三方檢驗 【MIHONG®極濃司令黑紅瑪卡】 #粉包版 #快速反應 #硬朗升級 🏆 瑪卡拉到6770mg＝強勢補量長紅 🏆 人蔘皂甘40mg＝補能量的主力買盤 🏆 專利南非醉茄＋L-精氨酸＝撐住走勢 🏆 草本精華，幫助滋補強身 🏆 新粉包＝想補就補，不拖拉好攜帶 兩款男女可食 我們要的不是短線暴衝 而是把「體力K線」守在均線之上 以「延續」體力取代「預支」 投資要看趨勢，身體也要維持趨勢 補給做對，狀態自然不破底👌 不用碼、最高現折$1980→ https://reurl.cc/9R9keV 股癌傳送門：https://linktr.ee/gooaye

- 有篇SSRN(社會科學研究網)的working paper最近很有名 - 標題Giving Up，討論現在年輕家庭放棄購房的文化與影響 - 使用美國的微觀數據發現，當租房者對買房絕望時，會產生的行為模式 - 1: 消費增加 - 資產較低的群體，信用卡支出顯著高於同等資產的屋主 - 減少了儲蓄的動力，就會把錢花在娛樂跟小奢侈品上，作為安慰 - 2: 工作努力下降 - 放棄購房的人在工作中盡最大努力意願較低 - 反應一種不同的work-life balance，也更常quiet quitting - 3: 參與高風險投資 - 既然買不起房，損失本金也變得不那麼可怕 - 更多人轉去投資加密貨幣，試圖博取低機率暴富的機會 - 這些會在生命週期中累積，放棄者跟懷抱希望者的財富差距擴大 - 其中，有沒有『希望』是根據數學模型，判斷退休前能買到房的機率 - 亂花錢而買不起房，與買不起房而開始亂花錢，是個交錯都有的狀態 https://www.facebook.com/ytsai.realtor/posts/pfbid0xNtfdMu8Z2Y13iwTHMvfEdrTAaVtqMag4UnNPPAuVCTyaokQE7wK5jMLhi9x8qREl?locale=zh_TW

amkor 最近滿兇的，osat 在先進製程的擴充程度短期會大於 fab，後者委外後段的訂單越來越多 比較需要注意的是良率的改善是否符合期待，若有，則市場目光有機會從 foundry 設備股移轉到 osat 上，雖然也有滿多重疊，但有新的機會 --- amkr @ubs forum - 先進封裝領域大多數 OSAT 集中在台灣或少數在中國，而 Amkor 在韓國（持續擴張）和未來的美國有可選性，供應鏈多元化是客戶對產生興趣的關鍵 - 2.5D 平台已量產，HDFO 平台開始 ramp 了，Bridge Type 則在開發階段 - Compute Segment 佔總銷售額的 20%，CAGR 12%，主要由先進封裝領域驅動 - 2026 年將繼續投資於增量產能 - 與 TSMC 的合作關係非常牢固，TSMC 在美國擴大封裝設施將為 Amkor 創造更多機會，特別是在 substrate - 越南的記憶體和 SiP 產品 ramping 順利 - 運算部門在 Q4 略有下滑，但對明年的需求感到積極，對 Q1 的表現感到樂觀