真正的AI 攻勢，是堅決把自己變成科技企業。2026 年2 月初，中國的兩大互聯網公司騰訊與阿里巴巴率先打響了新一年的AI 大戰——騰訊的元寶端出10 億元現金激勵，並上線新的AI 社交功能“元寶派”；阿里的千問拿出30 億元的福利，鼓勵用戶在App 裡點外賣。儘管字節跳動在這之前已經靠火山引擎奪得了這一年央視春晚的總冠名，但在紅包與補貼的喧囂中，它並未成為輿論主角。直到農曆春節前一周，視頻生成模型Seedance 2.0 問世，在創作者圈與行業討論中迅速發酵，把它重新推回了討論中心。 2 月14 日，字節跳動陸續上線了豆包大模型2.0 與圖像創作模型Seedream 5.0 Lite。除夕當晚，字節跳動又帶著火山引擎、豆包App，以及Seed 系列大模型來到了春晚舞台上。從舞美視效​​、機器人說話到播出保障，AI 第一次如此全面地滲透進了春晚這場全民狂歡之中。就連互動玩法也被重新設計：觀眾需要調用豆包裡的大模型生成圖片或文字才能獲得紅包福利。紅包裡不只現金，還有3D 列印機、汽車、無人機、機器人等科技產品。這是字節的春節戰役。一位字節跳動人士告訴我們，相較於許多公司把春節當作單純的成長機遇，他們更希望在這個節點，創造關於AI 和科技的集體記憶；也試著告訴外界，做AI 不是只做一兩款應用，而是要服務各個產業。據了解，春晚上送出去的科技產品都是豆包大模型的客戶。官方數據顯示，除夕當天豆包的AI 總互動19 億。除夕前，備受國內外關注的影片產生模型Seedance 2.0 和最新的豆包大模型2.0Pro 也接連第一時間接入豆包（『專家』模式），讓先進的模型能力第一時間落到產品觸達用戶。無論是與新春等生活場景的貼合，或是模型能力上限的提升，可以說，豆包都已具備國民級AI 產品的特徵。在過去幾年的科技業，最有說服力的成長故事從來不是藉補貼堆出來的。 2022 年11 月，ChatGPT 掀起全球AI 熱潮後，靠著模型迭代成為了一個周活躍用戶數達8 億的超級產品；兩年多後走紅的DeepSeek 在沒有任何廣告投放、沒有多模態能力的情況下，穩定維持著超過3000 萬的日活躍用戶數。字節跳動CEO 梁汝波在年初的員工大會上將公司2026 年的關鍵字設定為了「勇攀高峰」。他提到，AI 時代存在著許多重要的機會，字節要追求其中最重要的，去攀登最高的高峰。「真正的AI 攻勢，是堅決把自己變成一家科技企業。」上述字節跳動人士說。 2026 年春節，這家公司正試圖點亮自己的ChatGPT 時刻。從水墨舞台、機器人到算力洪峰，不一樣的春節戰役作為少有的全民級晚會平台，春晚在過去十年裡一直是各互聯網公司重要的新用戶增長渠道，它與一眾互聯網產品之間的合作也幾乎與“紅包” 綁定。從2015 年開始，微信、支付寶、淘寶、京東、抖音、拼多多、快手等知名網路產品均在春晚上發放過紅包福利。但在今年春晚，冠名方火山引擎接到的第一個任務卻是與舞台視覺相關的。一位字節跳動人士告訴我們，今年春晚導演組一直在探索將最新科技和傳統文化藝術結合。因此作為獨家A​​I 雲夥伴，火山引擎及其背後的字節跳動不僅要以贊助方身份參與冠名，還要把AI 與雲端運算能力嵌入到春晚節目的創作與製作流程中。在《駕馭風歌》這個節目裡，歌手身後出現了一幅國寶級的駿馬水墨畫。節目籌備時，導演組並不滿足於將其當作靜態背景，而是希望讓這些馬匹在舞台空間裡動起來，並與歌手互動。為了達成這個效果，他們最初試過一些海外模型，但大多抓不住水墨畫的氣韻；國內模型也同樣受限，水墨風格的高品質訓練語料稀缺，生成出來的畫作風格容易偏離。更棘手的是運動一致性：馬要跑得流暢，卻又必須在每一幀都貼合原畫的筆觸與結構、不變形；馬匹的數量、每匹馬之間的動作都要始終保持一致。轉折來自字節跳動團隊帶來的Seedance 2.0。這個全新的視頻生成模型不僅能更好地遵循物理規律，讓馬的步態更接近真實，關節銜接自然、動作連貫；還能很好地遵循導演的要求，實現像“跑慢一點”“鬃毛飄動輕一點” 這類帶副詞的細微調度指令。它還有另一個重要的突破——模型不再是只能學單張圖的風格，而是能同時學習大量不同風格的多模態素材。一位字節春晚計畫組人士告訴我們，他們把導演組畫的草圖、水墨畫作，還有許多關於馬的影像素材一起餵給模型，透過參考生成的方式，實現畫風與動作能穩定地在同一個標準之下生成。他們也繞過了直接寫提示詞的方式，改用Seedance 2.0 與圖像生成模型Seedream 協作：先由Seedream 生成每一幀關鍵幀，再由Seedance 在關鍵幀約束下生成動態視頻，實現了風格的一致性和運動的連貫性。把理想中的效果表現出來只是第一步。春晚的每個節目迭代頻率都極高，因此字節團隊常遇到的情況是，剛交付一版滿意的畫面，導演組馬上提出：「能不能再往前走一步」。這些節目還需要經歷反覆過審與驗收，這也意味著他們每次都必須拿出更好的版本。 「整體節奏基本上得按'周' 來迭代推進。」上述字節春晚專案組人士說。很快新的難題又出現了。雖然《大鬧天宮》《小蝌蚪找媽媽》等上美影經典作代表了手繪全動畫的巔峰，但在AI 數位化生成的語境下，如何在高幀率、高分辨率的現代廣播標準中，讓水墨這種'無邊界' 的藝術形式在高速大運動下不閃爍、不崩壞，是行業公認的難題。考慮到主流視頻生成模型通常輸出720p 或1080p 分辨率、24 幀/秒的視頻，字節團隊將突破點放在後處理鏈路，借助火山引擎視頻點播的畫質增強能力升級畫質：先通過超分技術，在不改變畫面內容的前提下將較小尺寸的畫面放大至6K 或8K；這套處理也非一刀切的通用演算法，而是能根據每個畫面的畫質與內容特點做專屬優化。最後他們順利解決了這個難題。除了舞美，字節團隊還將AI 能力帶入了春晚的多個環節。舞台上出現的機器人連結了豆包的視覺大模型、文字大模型和語音辨識模型，語音側採用了豆包的合成與復刻模型。模型讓它們不僅有了“大腦”，還裝上了“嘴巴” 和“耳朵”。他們還提前收集了舞蹈演員的3D 素材，再利用了空間視頻技術，將舞蹈演員製作成3D 數字分身，在舞台上實現真人3D 克隆效果讓觀眾難辨真假。最後又加上了字節自研的4D 高斯演算法、豆包大模型優化光影效果。豆包App 是今年春晚與觀眾互動的核心載體。與往年常見的在網路產品上搖紅包、搶紅包不同，今年互動流程改為，用戶需先在豆包上用大模型生成頭像或新春祝福，再搶紅包。這樣的變化極大地抬高了成本——以往紅包互動形式的主要開銷集中在網絡、IO 和實時連接，現在則在原有基礎上疊加了龐大的算力請求與算力支出。調控算力資源的難題也大幅上升。一位字節的春晚專案組人士保守測算，使用者產生一則新春祝福或一張圖片，一次請求就需要完成10 TOPS（每秒10 兆次操作）的計算量。而以往類似互動請求的計算量僅約1/100000 TOPS，兩者在算力需求上相差了整整100 萬倍。同時，他們在除夕當晚還要應付抖音、今日頭條等產品春節活動帶來的大模型召喚。由於時間緊迫，字節沒辦法靠臨時堆機器頂住算力洪峰。還好火山方舟頂了上來，這是字節統一調度算力的總控台，長期在字節和外部客戶的各類高並發場景中沉澱了資源調度能力。火山方舟的一個特點是將推理、訓練和離線任務放進同一資源池統一統籌，因此在春節算力與流量峰值來臨時，系統能把可延遲的負載錯峰移開，為各類春節活動騰出更多資源。這不是件容易做到的事。它涉及到海量場景、多個機房、多種機型和多類模型，還要持續動態分配資源。一位火山引擎人士把這個過程形容為「裝箱」：一方面，不同流量的延遲要求不同、對異質硬體的依賴不同，約束極多、解空間巨大；另一方面，大模型推理不像傳統CPU 服務那樣資源類型相對統一，同一個推理服務裡也可能混用多種硬體。更麻煩的是，GPU 的遷移不是單點動作，儲存、網路、上游負載平衡等配套資源也要協同遷移，否則算力挪過去也接不住流量。「冠名春晚對字節來說，不再是一場簡單的增長活動，它更像一次真正的技術試煉。」一位字節跳動人士說。AI 時代，靠紅包換成長的模式失靈了字節跳動今年與春晚合作方式的變化，似乎再一次說明：AI 時代的成長邏輯正在改寫。行動互聯網時代，巨頭崛起不是靠時刻引領創新，而是靠在有人驗證了某個需求後，成系統地做出同款產品，以更高效率大量拉來用戶，再根據用戶反饋快速迭代改進體驗。更好的體驗帶來更多的收入，這些收入又被拿來投放，獲得更多用戶，如此循環。撒錢則是最直接的成長手段。「核心是極速成長、極速迭代、極速變現的飛輪。」一位網路公司的產品經理說。對網路產品而言，越快拉新，產品體驗越快提升。 「多數情況下，推薦演算法是發現你和另一個用戶偏好相似，然後把他喜歡的東西展示給你。」一位推薦產品經理說。用戶增加，可供演算法學習、發現關聯的資料隨之增加。這既適用於推薦影片、小說，也適用於推薦商品卡、帶貨直播間。微信支付的崛起是另一個例子，它用紅包把支付變成了社交的一部分，想參與就得綁卡、能轉帳。人越多，越常用，久而久之就成了預設的支付工具。交易型產品同樣受益於此。一個叫車平台會同時補上兩端：一邊補司機提高供給，一邊補乘客拉動需求。司機密度上來後，乘客等待時間縮短；訂單密了，司機空駛更少，平台的履約成本也會隨之下降，用戶體驗也更好。網路產品的成本也不會隨著用戶增加而線性上漲；新增開銷主要在頻寬、儲存和機器資源上，它也會隨著用戶規模的提升而被攤薄。但AI 產品幾乎是網路產品的另一面。2025 年初，字節跳動CEO 梁汝波曾在集團全員會上提到，豆包沒顯出「越多人用越好用」 的網路產品特性。這部分因為chatbot 產品不是社群網路或平台。用戶數量增長帶來的新數據也有限。一個短視頻產品，只要用戶還在上下刷，就會產生一組組數據供推薦算法優化；但chatbot 類產品生成一段回复，只有極糟時，用戶才有動力多點下按鈕反饋。能收集來數據，也不保證它們能讓底層模型更聰明。 「多數用戶的問題高度重合，又沒什麼深度，沒辦法提高模型能力。」一位網路公司的AI 產品經理說。 “例如程式碼方向，公司就會在內部找程式設計師寫案例。”大模型的專業能力早已超過大多數人類，它們不能透過收集普通人的數據來改善自己，就像AlphaGo 不需要和普通人對弈來提昇技能——它後來的變種AlphaGo Zero 甚至連世界冠軍的數據都不用，只是機器訓練機器就能贏過所有人。即便ChatGPT 應用的月活躍用戶量比Claude 高出100 多倍，但這並沒有讓OpenAI 的模型比Claude 好用那怕一倍。AI 還無法像網路產品一樣，靠用戶量攤薄成本。一位深入研究AI 的二級市場人士曾計算，目前的主流AI 產品如果服務1 億日活用戶，每天的模型推理成本就要幾千萬元。這個計算還沒有考慮新的Agent 產品，如果類似Manus 的產品開始流行，單一使用者每天需要的算力可能還會再翻幾倍。最關鍵的是商業化難題。在海外，ChatGPT、Gemini、Claude 砸下了巨額投資以滿足複雜計算，用戶也必須付錢，低一檔17-20 美元/月，高一檔可以到數百美元/月。但願意為軟體服務支付這般費用的中國用戶很少，字節和阿里等公司更多是靠雲端服務賺回在AI 上投入的錢——把模型能力做成雲上的API/託管服務與行業解決方案，企業按調用量、並發、存儲與算力時長付費。根據官方數據，2025 年12 月，火山引擎上的豆包大模型日均Token 處理量超過50 兆，半年成長超200%。成長的動力不僅來自字節旗下豆包、即夢等AI 應用快速發展，還有一群外部客戶在深入使用大模型：累計使用上兆Token 的超過100 家，比全球雲端運算巨頭AWS 還多了一倍。當下，AI 產品體驗提升幾乎全部來自底層模型能力提升。 AI 程式碼編輯器Cursor 能出圈，前提是它接入了Claude 系列模型程式碼能力大幅提升。 OpenAI 的Deep Research 體驗驚艷，也是靠著底層模型學會了長鏈思考、逐步解題。如果大模型能持續進步，許多精心調教後的產品能力成為龐大模型的一部分，使用者直接說幾句話就能實現想要的效果，那麼大模型本身就是終極產品。 ChatGPT 與DeepSeek 的成功都在說明，靠投流、補貼去搬運用戶的老辦法越來越吃力，真正能拉開差距的，還是模型與產品體驗的硬實力。唯一的解法：更堅定地成為一家科技公司2017 年的一次CEO 面對面上，有字節跳動的員工問演算法技術負責人楊震原：公司在人工智慧上與BAT 的差距在那裡？楊震原回答說，「今日頭條本來就是一家人工智慧公司。」他進一步解釋，做資訊分發就是需要人工智慧，今日頭條想透過機器和人結合的方式提升創作、分發、討論等每個環節的效率。當時，AI 在字節跳動的產品中已經有一些應用，例如推薦、內容創作、抖音的AR 效果、 內容審核、廣告系統、標題自動產生封面、時光相簿和東方IC 對圖片的篩選功能等等。2022 年底，ChatGPT 橫空出世帶來了新的AI 熱潮。字節跳動是這一輪投入最大的中國科技巨頭之一。它迅速訂購了大量GPU、組成全新的AI 部門，從infra、數據、模型、產品到人才全方位追趕矽谷公司。兩年後，字節的模型數量、迭代速度和表現明顯提升。技術的躍升很少一蹴而就，背後需要堅定的長期投入。一位字節跳動人士回憶，Seedance 1.5 推出時，外界對它的評價並不高：人物在運動中容易崩壞，細節也不夠穩定。後來Seedance 2.0 的效果明顯提升，有人猜測這是團隊調整了Benchmark，把更影響用戶體驗的指標排在了更前面。「其實大家都忽略了基礎工作的重要性。」上述人士說。以運動崩壞為例，團隊需要反覆驗證不同方案，找到更有效的路徑，這類工作很耗時，卻是必須補齊的基礎能力。 “就像一個很聰明的小朋友，未來也許能解大學題，但如果基礎知識沒學完，現在硬讓他做，結果肯定不會好。”另一個例子是豆包的語音合成模型（Seed TTS）。 2024 年，團隊在做Seed TTS 1.0 時的目標是讓模型能自然流暢地復刻說話者特徵，尤其在跨語種時也能保持相似度與韻律。當1.0 達到可用水準後，新的問題又暴露出來了：模型的語氣表達太平淡。但這個難題在當時無法快速解決，於是他們在研發2.0 版本時，把重點放在了將理解與思考能力融入端到端語音鏈路。直到2025 年中Seed TTS 2.0 發布，模型才具備了更強的情感表達能力。字節也是少有願意在基礎研究上投入資源的中國科技公司。一位接近字節高層的人士曾告訴我們，字節目前的infra （工程化能力）已經比國內任何一家公司都要強。但和全球比，最大的問題是缺少OpenAI 裡那種能提出方向、能做前沿探索的人，例如GPT 4o、Sora。「中國過去沒有真正的企業研究院是因為民營企業獲利能力有限，現在終於可以試一試了。」上述人士說。2025 年1 月下旬，字節正式設立了代號為「Seed Edge」 的研究項目，核心目標是做比預訓練和大模型迭代更長期、更基礎的AGI 前沿研究。該計畫設定了更寬鬆的考核機制：字節本來每半年考核一次績效，而Seed Edge 則在專案取得突破進展後，再做最終評估。字節在考察AI 專案時，ROI 依然重要，但周期更長了。一位字節AI 產品團隊人士告訴我們，字節現在會以AI 產品一定周期後的單用戶價值作為考核係數，以計算未來收益，不同產品的考核周期不同，長的甚至可以到5 年；但字節也沒對這種考核方式做強制要求。商業與科技史上無數次驗證過了，想要拿下使用者、市場，或是更大的機會，靠的永遠是技術的躍遷與產品的硬實力。二十多年前，Google 因擔心微軟殺入搜尋市場，在內部製訂了「芬蘭計畫」。這套計畫的核心不是守城，而是逼自己持續做出更有創意、把體驗打到極致的產品——即使在Windows 這樣的主場上，也要做出比微軟自備工具更好用的瀏覽器。最終，在沒有產品迭代優勢、沒有網路效應的背景之下，Google 硬是靠著出色的工程技術能力做出了Chrome 並贏得了戰爭的勝利。在與微軟的競爭中，Google 也鍛造出一套強大的商業化系統AdSense，把源源不斷的現金流與生態網路牢牢綁定在自己身上，並在此後多年逐步沉澱為難以撼動的障礙。2024 年，OpenAI 聯合創始人、前首席科學家伊爾亞·蘇茨克維(Ilya Sutskever) 曾提到，「2010 年代是擴大規模的時代，現在我們再次回到了需要奇蹟和新發現的時代。」對字節跳動來說，它用13 年長成中國最大的互聯網公司上限和新發現的時代。」對字節跳動來說，它用13 年長成中國最大的互聯網公司； (晚點LatePost)

今年春節，AI產業「無人入眠」。阿里巴巴、字節跳動、騰訊、月之暗面等中國AI企業發動了一場史無前例的春節攻勢。截至2月17日（大年初一）凌晨，戰局初定。阿里以巴巴「晶片+雲端+模型」的垂直整合優勢，打出了一場B端與C端的雙線大捷。在B端，阿里除夕開源全新一代大模型千問Qwen 3.5-Plus，不僅性能對標GoogleGemini 3 Pro，更祭出了令市場咋舌的1/18「地板價」。而在C端，戰況更為激烈。最新數據顯示，春節期間超過1.3億人湧入千問App體驗AI購物，千問日活（DAU）在短短三個月內飆升至7300萬量級，迅速追平了字節跳動“豆包”積累三年的身位。這場春節大戰的背後，是AI價值重心的轉移——從以Chat為代表的“陪聊”，轉向以Agent為代表的“辦事”。阿里正試圖運用其獨特的「通雲哥」全棧能力，解決困擾產業的兩大難題：推理成本的摩爾定律與商業變現的「最後一公里」。B端「降本奇蹟」：性能媲美Google旗艦，成本僅1/182月16日除夕，Qwen 3.5-Plus正式開源。在大模型「百模大戰」進入深水區後，阿里不再單純卷參數，而是卷「單位算力經濟性」。阿里此次發表的Qwen 3.5-Plus，採用創新的稀疏混合專家（MoE）架構，總參數3970億，但啟動參數僅170億。這種「以小勝大」的架構革新，直接帶來了推理效率的指數級提升。數據顯示，其最大推理吞吐量提升至19倍，顯存佔用降低60%。更具殺傷力的是推理效率。 Qwen 3.5-Plus的API價格低至每百萬Token 0.8元。在同等性能​​水準下，價格僅為GoogleGemini 3 Pro的1/18。阿里此舉意在透過極致的性價比，加速企業級AI的滲透率，同時對那些依賴API轉售或缺乏底層算力優化的中間層模型廠商形成「降維打擊」。自2023年以來，阿里已開源超400個模型，全球下載量突破10億次，其策略非常清晰：用開源透過規模效應攤薄成本，用低價建構生態護城河。國民級AI助理誕生：1.3億人「喚醒」Agent2月17日最新數據顯示，春節期間用戶向千問發出了50億次「千問幫我」的指令，1.3億人首次體驗了AI購物。這項數據標誌著「千問幫我」已成為AI時代的用戶新習慣，千問一躍成為國民級AI助理。成長數據的顆粒度透露出驚人的滲透力：票務爆發：隨著春節檔電影上線，千問電影票訂單季增372倍，交通票務成長超7倍；下沉市場：來自三、四線城市的AI訂單量暴漲782倍，所有AI訂單中近一半來自縣城；跨越鴻溝：近400萬60歲以上用戶體驗了AI購物，證明了「一句話點單」正在抹平數位落差。QuestMobile數據顯示，2月7日千問DAU已達7352萬，逼近豆包的7871萬。豆包走了三年的路，千問只用3個月就跑完了。這背後是阿里生態的「集團軍作戰」。目前，淘寶、支付寶、飛豬、高德、大麥等核心資產都已接入千問，未來也將上線AI搭計程車、儲值等功能。相較於OpenAI和谷歌，阿里擁有「從意圖到執行」的完整履約網絡——這正是矽谷巨頭們目前最眼紅的能力。「通雲哥」鐵三角：掌握核心算力定義權阿里為何能同時支撐B端的「價格戰」和C端的「億級並發」？底氣源自於其被稱為「通雲哥」的硬派科技策略。「通雲哥」即通義實驗室（模型）、阿里雲（基建）與平頭哥（晶片）的黃金三角。市場消息顯示，阿里已決定支持平頭哥未來獨立上市。在全球AI軍備競賽中，垂直整合是降低成本的唯一解。谷歌透過「TPU+Cloud+Gemini」驗證了這個邏輯，而阿里則更進一步。透過自研晶片「真武」與千問模型的深度耦合，阿里實現了一台軟硬一體的「AI超級電腦」。這種底層架構的統一，使得阿里能夠最大化壓榨算力效率。在FP8、FP32精度應用策略下，訓練成本大幅降低，訓練速度提升10%。對市場而言，阿里的敘事邏輯正在改變。過去，市場擔憂阿里在AI時代缺乏類似ChatGPT的現象級入口。但這春節的戰果表明，阿里正在透過「最強開源模型+最全實體生態」的雙輪驅動，走出一條不同於矽谷的道路：不只是做會聊天的Chat，更要做能辦事的Agent。當AI從“嚐鮮”走向“實用”，擁有算力定價權和商業履約能力的阿里，或許才剛剛展露其在AI時代的真實力量。 (國際金融報)

千問3.5"以小勝大"背後，是全新架構的效率代際躍遷"源神"啟動！阿里巴巴正式開源Qwen3.5-Plus，性能指標直接對標Gemini 3 pro與GPT 5.2等頂級閉源模型。千問團隊憑藉這一代際跨越，坐上全球最強開源模型交椅。核心突破在架構。397B總參數，推理時僅啟動17B，這種精巧設計讓其全面超越上代兆參數的Qwen3 Max，部署視訊記憶體佔用降低60%，推理效率大幅提升，最大推理吞吐量可提升至19倍，速度也更快了。過去束縛大模型商業化的算力瓶頸被打破。支撐點是阿里長期堅持的基礎技術創新。研發團隊深度應用混合專家架構，拋棄簡單參數堆砌。預訓練階段採用統一架構同時處理多模態資料，建構出同等性能等級中參數量最小的模型底座。但千問3.5隻是開始。這次發佈揭開了阿里多模態戰略佈局的一角，千問研發團隊正在探索文字、視覺、語音與操作反饋的全模態融合，這條大一統架構路徑，指向通用人工智慧（AGI）。科技估值體系、企業AI應用滲透率、開源生態話語權，都將在此刻重新定義。01. 千問3.5，更小模型，更強性能發佈即開源——"極致性價比"參數規模曾被視為提升智能的唯一路徑。千問3.5用資料終結了這一迷信。更精簡的模型實現了更強性能，同時帶來顛覆性定價。Qwen3.5 Plus的API價格降至0.8元/百萬Token，同等性能下僅為Gemini 3 pro的十八分之一。頂級AI能力跳出了高算力、高成本的閉環。個人開發者、初創團隊、中小企業都能以極低門檻獲取世界級智能基礎設施，AI向下滲透的管道被打通。性價比背後是模型基礎能力的全面升級。千問3.5具備真正的跨模態理解能力，標誌著千問系列從語言模型進化為原生多模態大模型。行業內多數"多模態"方案仍停留在工程拼裝：先訓練語言模型，再外掛視覺或音訊模組，各模組通過適配層勉強對齊。有些產品只是在統一入口後用路由分發任務給不同單模態模型。這類方案能應對基礎任務，但無法實現真正的多模態融合，甚至會出現視覺能力增強、語言能力下降的降智現象。Qwen3.5走了另一條路。從預訓練第一天起，模型就在文字與視覺混合資料中聯合學習。視覺與語言在統一參數空間內深度交織。模型看到圖像能理解深層語義，讀到文字能在神經網路中建構對應畫面。中間翻譯環節被抹除，資訊折損降到最低，千問具備了人類等級的跨模態理解力。這種原生融合帶來了寬廣的能力邊界。細粒度視覺定位能做到像素級精度，長時序理解長達2小時視訊內容的演變過程與內在因果關聯，跨模態生成，能把手繪草圖輕鬆轉化為可運行的前端程式碼。模型進化真正意義上重構了工作流。千問3.5可以作為視覺智能體，自主操控手機與電腦系統，跨多個應用完成複雜多步驟指令。從模式識別到跨模態邏輯推理，Qwen3.5推動大模型從文字工具進化為理解現實世界的基座，為更自然的多模態生成與複雜推理打下技術基礎。02. 千問3.5"以小勝大"背後全新架構與效率的代際躍遷過去兩年，大模型行業陷入算力競賽。參數從千億堆到兆，性能緩慢爬升，成本指數級膨脹。企業部署需要昂貴GPU叢集，日常推理燒錢，中小企業付不起帳單，端側裝置跑不動模型，技術越強大，落地越困難。Qwen3.5顛覆了這條路徑。競爭重點從"誰的參數更多"轉向"誰的模型更聰明"，用更小規模通過技術創新獲取超越規模的智能。這不是偶然，是阿里大模型團隊多年技術積累的集中兌現。四項核心突破共同驅動了這場代際躍遷。第一，混合注意力機制。傳統模型處理長文字時，每個Token都要與所有歷史上下文做全量計算，算力消耗呈平方級增長。Qwen3.5讓模型學會有詳有略地閱讀——動態分配注意力資源，略讀無關資訊，精讀關鍵節點。效率與精度雙重提升。第二，極致稀疏的混合專家架構。傳統稠密模型每次推理都要啟動全部參數，規模越大開銷越高。Qwen3.5將397B參數庫中與任務最相關的專家子網路精準啟動，每次推理僅用17B參數——用不到5%的算力調動全域知識儲備。第三，原生多Token預測。模型不再一個字一個字輸出。預訓練階段就學會了對後續多個Token進行前瞻性聯合預測。在長文生成、程式碼補全、多輪對話等場景中，推理速度接近翻倍。第四，系統級訓練穩定性最佳化。通義團隊獲NeurIPS 2025最佳論文的注意力門控機制是關鍵。在Transformer注意力輸出端嵌入基於Sigmoid的智能開關，防止稀疏資訊被噪聲淹沒，抑制無效擾動放大。這削弱了注意力下沉負面效應，消除了對特定位置編碼的過度依賴，長上下文泛化能力質變。配合歸一化策略與專家路由初始化，新架構在大規模訓練中表現穩健。四項技術疊加帶來化學反應：訓練成本降90%，相比Qwen3 Max視訊記憶體佔用降60%，推理吞吐量最高提升19倍。原生多模態融合使訓練提速10%、啟動記憶體減少50%。大規模強化學習框架讓智能體訓練效率提升3到5倍。當其他廠商還在跑分榜上爭論小數點，Qwen3.5已經降維打擊，競爭維度變成了誰的模型更好用、更實用、讓更多的企業毫無壓力地用得起。03. 模型+基礎設施，"powered by阿里雲"成為AI時代印記在電話會中，皮查伊正式確認了外界猜測已久的傳聞——Google與蘋果達成深度合作。這不僅穩固了Google在iOS生態中的地位，更可能成為其AI模型分發的關鍵管道。卓越模型如果脫離生態與基礎設施，只能是實驗室展品。阿里採取雙管齊下戰略：研發端高頻推出SOTA級模型，生態端堅持開源，讓頂尖模型免費可得。千問產品矩陣覆蓋從輕量端側到千億旗艦，全線遵循Apache 2.0協議完全開放。這將核心技術直接推向全球數千萬開發者，千問迅速成為最受歡迎的開源大模型。開源資料顯示優勢明顯。千問官方開源模型超400個，全球開發者衍生的微調與垂直定製模型突破20萬個，總下載量超10億次，在熱度、活躍度與生態廣度上全面超越Meta的Llama生態。李飛飛團隊、愛彼迎等頂尖力量已將千問深度嵌入其AI引擎。當開源模型在邏輯推理與跨模態理解上追平甚至超越頂級閉源模型時，免費且最強成為開發者無法拒絕的選項。競爭主軸從單點跑分轉向生態系統的整體替代。企業端市場，阿里雲與平頭哥構築了技術護城河。平頭哥真武晶片針對MoE混合專家架構做了底層指令級最佳化，這種軟硬體深度耦合實現了模型演算法、晶片與雲網路的協同效能爆發。真武晶片對路由分發機製做硬體級加速，滿足千問大模型對大規模稀疏計算的需求。軟硬協同釋放了晶片算力潛能，資料中心叢集效率大幅提升，企業微調訓練與高並行推理時間被壓縮，這是API價格降至0.8元/百萬Token的底氣。軟硬協同與開源生態反哺推動阿里雲進入新增長期。Omdia《中國AI雲市場，1H25》報告顯示，2025年上半年中國AI雲市場規模達223億元，阿里雲以35.8%市場佔比排名第一，超過第二至第四名總和。Omdia預測2025-2030年中國AI雲市場復合年增長率為26.8%，其中MaaS層增速最快，複合增長率超72%，2030年將達177億元規模，2025年阿里雲在中國雲市場份額從33%升至36%，優勢持續擴大。財務資料印證這一趨勢。近期財報顯示，阿里雲公共雲收入同比增長34%，AI相關產品收入連續9個季度保持三位數增長。資本市場對中國科技資產的舊有低估偏見正在被徹底拋棄，全球資金對中國科技股的定價邏輯正經歷著重塑。縱觀科技產業史，真正定義時代的往往不是最貴的閉源產品，而是被最多人使用的基礎設施。Linux定義了伺服器時代，Android定義了移動網際網路時代。開源、免費、無處不在的特質，將使千問成為AI產業默認的底層基石。"powered by 阿里雲"正深深烙印在每個智能應用的程式碼深處，成為這個時代的技術印記。04. 結語這次阿里"源神"啟動的是AI產業的範式切換。千問3.5用技術創新重構行業規則，當最強性能遇見最低成本，AI真正從少數人走向所有人的生產力工具。這場由阿里雲主導的技術革命，正在重寫全球AI的遊戲規則。 (硬AI)

據“市說新語”微信公眾號消息，2月13日，市場監管總局約談阿里巴巴、抖音、百度、騰訊、京東、美團、淘寶閃購等平台企業，要求有關平台企業嚴格遵守《中華人民共和國反不正當競爭法》《中華人民共和國價格法》《中華人民共和國消費者權益保護法》《中華人民共和國電子商務法》等法律法規規定，主動落實主體責任，進一步規範平台促銷推廣行為。提醒相關平台企業要杜絕各種形式的“內卷式”競爭，共同維護公平競爭市場環境，促進平台經濟創新和健康發展。 (證券時報)

跟對人比做對事更重要。蔡崇信，馬雲背後的男人。我們都知道他很早就加入阿里巴巴了，是“十八羅漢”之一，蔡崇信當年放棄百萬年薪，選擇拿500元月薪和馬雲創業的故事至今讓人津津樂道。最近，蔡崇信又給我們揭露了更多幕後故事。蔡崇信自曝：見馬雲1小時就入夥，曾被“開除”過近日，蔡崇信在史丹佛大學商學院的一場訪談中又聊到了那些年和馬雲的往事。圖源：蔡崇信訪談視訊截圖公開資料顯示，蔡崇信，祖籍浙江，台灣出生，加拿大籍華人，擁有耶魯大學經濟學及東亞研究學士學位、耶魯法學院法學博士學位和美國紐約州執業律師資格。在遇到馬雲之前，蔡崇信已然是一位職場精英，年薪高達70萬美元。他自曝了加入阿里的原因，並透露自己僅與馬雲交談一小時，便決定投身這家初創公司。蔡崇信表示，1999年接觸阿里時，公司尚未註冊、零收入，而馬雲僅擁有一個域名與初創網站。但他被馬雲的領導氣質與溝通能力打動，認為教師出身的馬雲擅長識人、育人與凝聚團隊。“創業要找你願意共事、能追隨、可學習的人，對我而言，馬雲既是夥伴也是導師。”不得不說，馬雲的魅力再一次得到了驗證。蔡崇信和馬雲早期照片然後，蔡崇信還透露了一段鮮為人知的創業初期往事，他回憶，當時是當了三個月的首席營運官，然後馬雲“解僱”了他。馬雲說，“你根本不懂如何營運。”他便轉為首席財務官。青雲發現，過去，馬雲似乎也提到過這段“開除”蔡崇信的故事。網傳視訊中，他是這麼說的，蔡崇信只當了28天COO（首席營運官）就被我開除了。“我為什麼判斷他當COO，他是律師出身，這個檔案指令下得非常清楚，我覺得他可能當COO，後來當了28天被我開掉了。什麼搞的，律師，檔案寫的話人家會生氣你知道嗎？寫得很嚴肅，下達命令沒有情感，沒有EQ（情商），IQ真高，EQ不行，這也是個大問題。”馬雲聊調整蔡崇信崗位這段視訊後面，馬雲繼續聊到了自己為什麼認為“CFO永遠不能當CEO”，他表示，“當CEO他條理清晰，但是我們做CEO的人，其實有的時候是有個混沌之間的擺動，好像太極的混沌之間，陰和陽，黑和白之間是有些度的。一個優秀的財政官越優秀，越從數字上看問題，除非他不是個好CFO。所以往往不是一個非常好的CFO，他有可能成為一個傑出的CEO，一個傑出的CFO很有可能在當CEO的過程中有自己的漏洞，這是有道理的。”對於蔡崇信擔任首席營運官的時間，兩人的說辭有些出入，不過可以肯定的是，蔡崇信確實經歷過換崗。蔡崇信換崗一事或許也在告訴我們，創業團隊的核心，並不是毫無分歧，而是能正視差異、各司其職、彼此成就。只有成員‌各司其職‌，基於自身的專業與優勢承擔起明確的責任，團隊的運轉才能高效而有序。阿里官網顯示，蔡崇信於1999年加入阿里巴巴集團，為公司創始人之一，自公司成立以來一直為董事會成員。他曾擔任首席財務官至2013年，並擔任執行副主席至2023年9月，目前擔任阿里巴巴集團主席。圖源：阿里集團官網蔡崇信擔任公司CFO的那段日子，多次幫助阿里完成融資，在公司關鍵時刻力挽狂瀾。這也是他“阿里財神爺”之稱的由來。1999年，馬雲初創阿里，曾嘗試過37次融資，但均以失敗告終。那時公司甚至連500元的最低工資都難以按時發放。蔡崇信加入後，憑藉個人人脈說服高盛投資500萬美元，直接讓瀕臨破產的公司起死回生。馬雲在很多次場合中都對蔡崇信表達了感激，他提到：“我最感激的人就是他。”志同道合的人往往都是惺惺相惜的，馬雲和蔡崇信或許就是這樣一對搭檔。蔡崇信跨界體育投資，悶聲賺大錢蔡崇信在阿里幹了20多年，當年他的決定成就了自己的命運。在2025胡潤百富榜中，蔡崇信以830億元的財富位列第57名，無疑是位人生贏家。圖源：胡潤百富官網說到蔡崇信的投資版圖，他十分熱愛籃球，並捨得為此投資，成功將體育愛好轉化為了大生意。福布斯發佈的《2025年全球最富有的運動隊老闆》顯示，截至2025年3月7日，蔡崇信憑藉121億美元的身家，位居排行榜第18名，身家同比大漲42%。2018年，蔡崇信以11.5億美元購入NBA布魯克林籃網隊49%的股份，次年又豪擲13.5億美元拿下剩餘51%股份，一舉成為球隊全資擁有者，創下了NBA歷史上的最高球隊交易記錄。2019年，蔡崇信和他的妻子吳明華以1000萬美元收購了紐約自由人隊。去年6月，據媒體消息，紐約自由人隊以4.5億美元（超32億元人民幣）估值出售股權，創下了女子職業體育俱樂部的最高估值紀錄。按照估值計算，從出手收購至今，這家球隊暴漲了30倍。蔡崇信夫婦也狠狠賺了一筆。值得一提的是，蔡崇信在體育界的活躍，可能還為阿里與NBA達成合作這事有所推動。去年10月10日和12日，NBA中國賽在中國澳門舉行，由布魯克林籃網隊對陣菲尼克斯太陽隊。在比賽前夕，蔡崇信、馬雲、姚明等人被曝出席了籃網隊的一場內部晚宴。馬雲、蔡崇信出席籃網隊內晚宴在宴會結束後不久，蔡崇信與NBA副主席兼首席營運官馬克·塔圖姆共同宣佈，NBA中國與阿里雲達成了一項多年合作，將通過人工智慧與雲技術，重塑中國球迷體驗和參與NBA的方式。蔡崇信表示，“體育與科技的交融，正處在一個激動人心的時代。人工智慧與雲是絕佳的連接器，不僅連接著球員與球迷、品牌與消費者，也將全球視野與本土文化緊密相連。通過與NBA中國的合作並引入阿里的最新技術，我們將為NBA球迷帶來前所未有的超凡體驗，點燃他們的激情。”值得一提的是，馬雲和蔡崇信一樣，也對籃球有著濃厚的興趣。2024年年初，馬雲就被拍到身穿球衣現身法國NBA常規賽，與蔡崇信、吳明華夫婦一同觀賽。馬雲現身觀看法國NBA常規賽上文提到的NBA中國賽，馬雲也去看了，當天他打扮休閒，頭戴棒球帽、穿格子衫，左側坐著成龍、貝克漢姆。當然，蔡崇信也同樣在場。去年10月，馬雲觀看NBA中國賽阿里放下包袱，成功套現300億再把目光看向阿里，今年，阿里繼續自己的主線任務——收縮戰線，集中資源佈局核心業務。近日，三江購物發佈公告披露，公司持股5%以上股東杭州阿里巴巴澤泰資訊技術有限公司已完成此前披露的減持計畫，本次累計減持公司股份1643.04萬股，佔公司總股本的3%，減持套現總金額約2.52 億元。本次減持完成後，阿里澤泰對三江購物的持股比例由原本的30%降至27%。天眼查App顯示，阿里澤泰的控股股東是淘寶（中國）軟體有限公司，後者直接持股比例為98.3051%。據不完全統計，截至2025年底，阿里通過一系列密集的減持與資產剝離，已完成逾300億元的資金回籠。阿里資產處置的核心目的是聚焦電商和AI兩大核心業務，就AI領域，有了“真金白銀”的投入，阿里千問的成長速度超乎想像。1月15日，在千問App產品發佈會上，阿里巴巴集團副總裁、阿里千問C端事業群總裁吳嘉便表示，上線兩個月，千問C端月度活躍使用者突破1億。今年春節檔，千問也是卯足了勁往前衝。最近這段時間，大廠之間上演了一場“AI紅包大戰”。阿里方面直接入30億元推出“春節請客計畫”，上線僅9小時AI訂單量便突破1000萬單，由於活動過於火爆，使用者一度把伺服器干“崩潰”了。另外，千問還拿下了四大衛視春晚獨家冠名。甚至馬雲也悄然現身千問項目組，疑似為團隊加油打氣。馬雲現身千問項目組前不久，馬雲現身杭州阿里巴巴總部，看照片透露的地點，馬雲現身的地點是阿里千問項目組。對於阿里內部員工來說，馬雲就像“定海神針”，他的現身想必能給員工帶來信心和動力。總的來說，網際網路競爭已經進入下半場，圍繞AI，大廠之間已經開啟了新的較量。至於誰能笑到最後，讓我們拭目以待。 (科技頭版)

近日，在史丹佛大學商學院的一場訪談中，阿里巴巴董事會蔡崇信首次公開談及一段創業往事：他早年加入阿里後曾擔任三個月 COO，隨後被馬雲直言 “不懂營運”，並調整崗位。蔡崇信透露，1999年接觸阿里時，公司尚未註冊、零收入，而馬雲僅擁有一個域名與初創網站。他被馬雲的領導氣質與溝通能力打動，認為教師出身的馬雲擅長識人、育人與凝聚團隊。在他看來，“創業要找你願意共事、能追隨、可學習的人，對我而言，馬雲既是夥伴也是導師。”蔡崇信稱，自己加入阿里前僅與馬雲交談了一小時，便決定投身這家初創公司。他還披露，當時自己是當了三個月的首席營運官，然後馬雲“解僱”了他，馬雲稱，“你根本不懂如何營運。”於是，他便轉為首席財務官。訪談期間，蔡崇信還談到了自己為何2013年離開阿里日常管理崗位，他表示，這是他和馬雲共同商定的策略，當時阿里業務已涵蓋B2B、淘寶、支付寶、雲端運算等多個類股，他們意識到需要給陸兆禧、張勇等年輕一代機會。官方網站顯示，蔡崇信於1999年加入阿里巴巴集團，為公司創始人之一，自公司成立以來一直為董事會成員。蔡崇信曾擔任首席財務官至2013年，並擔任執行副主席至2023年9月，目前擔任阿里巴巴集團主席。蔡崇信是阿里巴巴合夥人制度的創始合夥人之一。他同時是阿里巴巴集團關聯公司螞蟻集團的董事。以下為訪談原文(英文原文在後)：主持人傑裡米：Joseph Tsai，歡迎你。Joseph Tsai（蔡崇信）：謝謝你，傑裡米。很高興來到這裡。主持人傑裡米：很榮幸你能來。當我準備這次採訪並試圖理清思路時，阿里巴巴已經變得多麼龐大和全球化。真正讓我印象深刻的是，在你的職業生涯中，你是如何跨越文化的，嗯，東方和西方，法律和商業，體育和科技。所以我想從你的第一次跨文化專業經歷開始。你在台灣出生和長大，13歲時搬到美國，就讀於紐澤西州的勞倫斯維爾學校。那段經歷對你來說是什麼感覺？Joseph Tsai（蔡崇信）：嗯，當時很害怕。嗯，我去之前的第一次嗯，嗯，你知道我從未上過七年級。我在台灣上小學。然後我媽媽說，喬，我要送你去國外上學。但是你必須學會說英語。而且我當時根本不會說英語。所以呃，我去上了英語補習班一年，跳過了七年級。所以我八年級時去了勞倫斯維爾。呃，在那段補習班的日子裡，我基本上就是背單詞，但我連一句完整的英語都說不出來。這有點像，想像一下你並不真的懂這門語言。你不知道如何溝通，然後你就置身於一個全新的文化環境中，就像剛著陸。我的意思是就像剛下船一樣，對吧？然後降落在這裡，一個非常獨特的環境。你知道，寄宿學校，那是一所全男生寄宿學校，所以這有其自身的困境而你只能適應。我想我當時是想成為他們中的一員，我想融入其中。我看起來不一樣，我穿得不一樣，說不同的語言。嗯，所以對我來說，被其他人接受是非常重要的。嗯，然後很快，我就開始投身體育運動了。嗯，不幸的是我被棒球隊開除了。想像一個來自台灣的孩子，棒球小聯盟，你知道我沒有進球隊，沒有進游泳隊，嗯，然後我說該死，我要嘗試全新的事物。嗯，我踢足球，美式足球，然後進入長曲棍球，嗯，因為我不會打棒球，這是一項春季運動。長曲棍球，棒球是春季運動。我進不了棒球隊，所以我去打長曲棍球了。嗯，有趣的是昨天我跟傑裡米，我覺得他當時很擔心。你知道的，所以他想做個準備會議。所以我們昨天花了一小時準備這次談話。主持人傑裡米：是的，我倆都熱愛體育，我是湖人隊鐵桿球迷。很明顯，喬是個體育迷、網迷和自由迷。Joseph Tsai（蔡崇信）：是的，我不會怪你的。但不管怎樣，湖人隊現在做得更好了。恭喜你，主持人傑裡米：非常感謝。非常感謝。Joseph Tsai（蔡崇信）：然後我們發現我們都是橄欖球運動員，主持人傑裡米：身材矮小的線衛。喬，你找到感覺了。你從早期的體育挑戰中恢復過來了。你進入了長曲棍球。你曾就讀於耶魯大學的本科和法學院，對嗎？你在一家稅務律師事務所工作，最終進入了私募股權行業。但喬的投資生涯才4年，他就遇到了一個名叫馬雲的人，他正在建立一家公司。與其他17位聯合創始人一起，它是一個線上交易平台，早期零收入就只有幾千名使用者。那麼，是什麼讓你確信那是做出改變的合適時機？當時不僅僅是零收入，Jack（馬雲）甚至連公司都沒有。它甚至還沒有註冊成立。所以Jack，Jack當時只有一個域名和一個新網站。我還記得我第一次去看他的時候，我爬上了二樓的公寓。那裡有一堆鞋子。你知道的，在外面。你要脫下鞋子，你走進去，然後你知道，我們聊了大約一個小時。他就有那種魅力，那種領導氣質。Jack以前是個老師，我認為老師能成為很好的領導者，因為你必須善於溝通，你還需要具備識別優秀人才的能力，發掘並培養優秀學生，就像老師所做的那樣。所有這些品質都逐漸顯現出來，而這正是我覺得非常吸引人的地方。我沒能理解這個商業計畫，因為當時網際網路對每個人來說都是新的，他說的是利用網際網路讓中國所有製造東西的公司都參與進來，把東西交易到海外，讓外國買家和買家來網站購買中國的東西。嗯，這是一個批發市場。所以阿里巴巴的第一個網站實際上是英文寫的。因為它是為世界其他地區打造的。嗯，談話一個小時後，我說，Jack，你知道我需要去洗手間休息一下。於是我走進洗手間，水槽上有十把牙刷。所以我意識到人們只是睡在那裡，住在那裡，睡在睡袋裡。嗯，有點像創業文化。而我決定，嗯，你知道，我會這樣對年輕人說，當你想換工作，或是當你找到新工作想做點不同的事情時，找到你想合作的人。那些你想追隨的人，以及你想學習的導師，對我來說，傑克既是朋友，也是商業夥伴，也是導師。他教會了我很多東西。我覺得我可以從他身上學到很多。而且當時，你知道在網際網路時代，如果公司名字帶.com後綴，所有人都以為你會在6個月內成為百萬富翁。但對我來說，那個人是傑克。那種人格魅力，那種領導特質正是源自他曾經的教師經歷。主持人傑裡米：是啊，你和傑克的合作顯然挖到了金礦。在與傑克的合作中。史丹佛大學的很多同學都在考慮創辦公司和尋找聯合創始人。那麼，正確的公式是什麼？他們應該在聯合創始人身上尋找什麼？Joseph Tsai（蔡崇信）：首先，我認為他們應該，找一個能相處的人。比如你是否願意和這個人出去喝杯啤酒，在下班後。我認為這是第一次考驗。我認為接下來重要的是互補性，也就是他們擅長而你不擅長的方面。以及你擅長而他們不擅長的方面。保持這種互補心態。無論用什麼詞來形容。能夠互相彌補確實非常重要，這樣創始人就不會互相妨礙。阿里巴巴最初有18位創始人，我認為我們都很好地相互補充，我們帶來了不同的技能組合。我認為當你考慮創辦公司時，就是，你知道，你會想出去找18個人一起創業，但我認為只有兩三個創始人有時會成為問題，因為隨著業務規模的擴大，你的文化會被稀釋，而且周圍沒有創始人與員工建立接觸點，而有18個創始人，我們有這個優勢。你知道，我們能夠，每位創始人可以接觸更多員工，所以，我覺得這很重要。還有就是那些你想朝夕相處的人並共度美好時光，因為創業初期你不會經常回家或者說工作與生活的平衡不會是你該關注的重點。主持人傑裡米：是的，我想我應該問一下，你在公寓裡也有放牙刷嗎？Joseph Tsai（蔡崇信）：是的，嗯，我後來帶了自己的牙刷。主持人傑裡米：所以我想談談這一點，關於有這麼多位聯合創始人，總共有18位，這非常非常獨特。嗯，並於1999年加入阿里巴巴擔任首席財務官時，你才是真正進入公司。Joseph Tsai（蔡崇信）：我當時不是CIO。而是當了三個月的首席營運官，然後傑克解僱了我，因為他說，“喬，你根本不懂如何營運。我當時就想，"好吧，沒關係。那我就當首席財務官吧。主持人傑裡米：好吧，又一次，你從棒球反彈到長曲棍球，從首席營運官再到首席財務官。現在，這群聯合創始人彼此之間已經有些聯絡了，而你算是新加入的人。你是受過美國教育的律師兼投資人，你來自台灣，他們都是中國大陸人，你是怎麼找到方法跨越那種文化？Joseph Tsai（蔡崇信）：我認為首要的是，我通常更傾向於做一個傾聽者。我懷著謙卑的心態投入這項工作，我是去學習的。我對網際網路一無所知。我從未在一家以中國大陸員工為主的中國公司工作過，因為我之前的所有工作經驗要麼是在美國，要麼是在香港。所以我覺得自己沒有立場來指手畫腳，所以我傾聽了很多，然後試著去再次嘗試融入勞倫斯維爾的生活，試著成為人群中的一員，成為社區的一份子。我認為這對我幫助很大。我認為你必須全力以赴，你要帶著謙遜的態度去工作，因為總會有人，你會發現你的同事，你的合作夥伴比你更聰明，他們能教會你很多。你需要以這種心態對待工作。主持人傑裡米：是的，進入並成為一個傾聽者，並且保持謙遜，這顯然對你很有幫助。隨著時間的推移，傑克真的覺得這個超乎尋常的人物是中國科技行業的面孔。你全程都在那裡，但主要是在幕後。為什麼？Joseph Tsai（蔡崇信）：嗯，因為他更擅長站在舞台的最前面。他更擅長演講。他是一個更出色的精神領袖。嗯，對吧？我認為這非常重要。我原認為自己這些年來有所進步。在與人溝通方面，但這對他是天生的，只要站上舞台，讓所有員工共處一室，然後直接宣佈，看，這就是我們的方向。這就是我們這麼做的原因。而且他非常有說服力，所以更擅長這個。主持人傑裡米：這是另一個很好的例子，可以找到一個具有互補技能的人，共同致力於同一項任務。現在，據我所知，你和傑克是飛到沙丘路這邊只有幾分鐘的路程，對吧？向投資者推銷並籌集種子輪。你還記得那次行程嗎？Joseph Tsai（蔡崇信）：處理程序不算順利，我想我們搞砸了。我們開了大約15次會。我們連一個投資者都沒有找到。我想這是因為在史丹佛商學院，他們教你帶著PPT之類的東西去展示。對吧？你得向人們展示些什麼比如演示或類似的，比如演示什麼的。我們什麼都沒有，我們空手而來。人們會說，Jack，你的商業計畫是什麼？Jack說：我沒有商業計畫。他當時確實說過，我沒有商業。但當時中國的趨勢是，事後看來，你本可以看到的。我的意思是事後看來你本可以預見。這是網際網路技術的交匯點，也是中國加入世界貿易組織的時期，這意味著中國與世界其他地區的貿易往來將呈爆炸式增長。如今你來到中國，中國是一個製造業強國。一切都是中國製造的，而且質量越來越好，越來越高，技術越來越高。嗯，但種子在那時已經播下，那是一個即將加入世貿組織的時代。想像一下讓所有這些製造商和貿易公司把他們的產品都搬到網上，讓世界其他地方看到，這就是當時的一個商業構想。但傑克不僅僅談論業務。他說："我的使命是讓天下沒有難做的生意。”這就是我們當時的使命，這就是我們告訴投資者的話。至今這仍是我們的使命。你可以訪問我們的網站，阿里巴巴官網寫道“讓天下沒有難做的生意”。主持人傑裡米：太不可思議了。所以這是一個沒有商業計畫的任務優先領導者。你們從那趟旅行回來採納那些投資人的反饋了嗎？Joseph Tsai（蔡崇信）：並沒有。因為投資者看起來很困惑，他們沒有給我們很多反饋。他們有點沉默寡言。我想他們只是想直接拒絕我們。但那次經歷很重要，因為我們回來後說，我們必須做我們相信的事情，而不是讓投資者告訴我們應該做什麼。我認為這是一個非常重要的一課。幸運的是，我們籌集到一些資金。一些投資者終於非常懷疑地加入了。我還記得我們的A輪融資，我們當時確實請來了高盛。我是說你以為自己去了沙丘路，然後你就得到了高盛。他們當時，高盛像大型私募股權公司一樣進行投資。交易然後他們還有可以說是一個新興科技投資業務，而我恰好知道有個朋友是合夥人，當時在經營那家企業。然後她打電話跟我說，喬，你知道的。我有好消息和壞消息。當時，我已經很沮喪，被各種拒絕搞得心灰意冷。我當時想，再也受不了任何壞消息了，那就先告訴我好消息吧。她說：好吧，我們投資委員會已經決定追加阿里巴巴。我當時想，太棒了。差點都要掛掉電話，她說：等等，等等，你想聽壞消息嗎?我說，好啊，什麼壞消息?她說，“嗯，我們還決定投資另外五家公司，它們的商業模式與你們非常相似，我們就是想看看，讓你們互相競爭一下。主持人傑裡米：結果你們競爭得很好，你們上升到了頂峰，你們從高盛那裡拿到了第一輪，然後又從軟銀那裡拿到了另一輪，然後飛速發展了。在接下來的10年裡，你們推出了支付，即時通訊、消費者對消費者市場、企業對消費者市場，廣告和雲基礎設施產品，該產品目前是世界第四大，在中國擁有40%的市場份額。因此，阿里巴巴成為了一款超級應用。其中一項重大內部成果就是淘系業務。阿里巴巴的消費者對消費者市場，你能帶我們回到淘寶上線的那一刻嗎?當時面對的是來自eBay中國的真正壟斷。Joseph Tsai（蔡崇信）：那是在2000年代初，eBay收購了一家名為易趣的公司，所以他們立刻佔據了很大的市場份額。在最高峰時期，他們在消費者電子商務中佔據了90%的市場份額。而我們是新入場的玩家，話說有一天，我是說當時我正在和這裡的院長交談，她提到了一個叫簡的兒子，他是C-RIP的CEO，而她的丈夫是第一任CTO吳炯。阿里巴巴的第一任CTO他來自雅虎。所以，抱歉，這是個冗長的故事。當傑克決定推出一個消費者市場時，他非常興奮。他曾與軟銀的孫正義談過，孫正義非常支援，因為這是個宏大的構想。Masa會支援任何大項目，任何需要大量資金的事物，他都會支援。而且Jack非常興奮。他說："我從MASA那裡拿到了資金，我要做這件事。然後他去找John Wu，我們的首席技術官，John說，“想都別想，除非我死，否則我們絕不會做這個”，這是因為John的工作，John曾在Yahoo購物部門工作過。而且那並不成功。所以他對我們進入消費者電商領域並不看好。但顯然傑克否決了所有人的意見。我當時...我不知道我是有策略地，我...我無法做出那個判斷。但我決定我們要這樣來架構它，這樣在經濟上我們就不會完全受損。所以實際上從第一天開始，淘寶實際上是一家合資企業，由阿里巴巴和軟銀各持股50%的合資公司，直到大約六歲，七年後雅虎對我們進行了投資，我們抓住機會將淘寶剩餘部分整合進阿里以巴巴實現100%控股。這是我們做過的最明智的決定。否則如今淘寶仍將有一半歸軟銀所有，而我們也不知道它會走向何方。所以，但當我們開始做淘寶時，它是一個秘密項目。如果你們對阿里巴巴有點瞭解的話，幾乎所有的初創企業都曾在這個公寓裡起步，就是馬雲最初的那套公寓，你知道的，那裡有鞋子還有牙刷。所以我們帶了七個人的團隊進入這間公寓，他們都必須簽署保密協議。他們不能告訴同事、配偶或任何人。於是有一天，你知道，你就是那個坐在隔壁隔間的人，旁邊隔間的人消失了，然後他們，你知道，人們都在問，他們去那兒了？所以這最初七個人組建了淘寶業務，他們去了Jack原來的公寓，開展工作，建立那個市場。然後是我們首次推出時的情景，首批在淘寶上售賣的商品都來自每個人的衣櫃。大家只是打開衣櫃然後說:"我們有什麼？能找到什麼在Tabao上賣？我們用大量雜貨填滿了那個市場。這件事就是這樣開始的。據我記得我們把這個項目保密了將近一年，這是個秘密項目。我是說網站上線後人們就知道Taabao的存在了，但沒人知道它和阿里巴巴有關聯。主持人傑裡米：確實，當你想到它是如何將阿里巴巴如此成功的因素結合起來的時，你會發現這是非常了不起的，傑克是一個有遠見的人，他當時是你的首席技術官，有著更保守的心態，然後你組建了合資企業。然後是你們為實現這一目標所關注的焦點。你們在淘寶上取得了很大的成功。顯然，在兩年內，阿里巴巴奪回了市場。你們佔據了ebay 60%的市場份額，最終ebay關閉了中國業務。如今，阿里巴巴是世界上最成功、最重要的公司之一，地球上每六個人中就有一個人定期與阿里巴巴生態系統互動。Joseph Tsai（蔡崇信）：我需要把它放在我們的投資者簡報中。主持人傑裡米：幫我核實一下資料。如果你想讓我加入，我很樂意參與。所以現在你處於不同的位置。你就像ebay曾經一樣，現在正面臨著嶄露頭角的挑戰者。是的，那麼你們是如何從現有職位上繼續創新的呢？Joseph Tsai（蔡崇信）：嗯，我們並非總能做對。我們經歷過一些時期，當我們停止創新時，我們就為此付出了代價。嗯，作為一家大公司。你知道，從定義來說，我的意思是我們有12萬名員工。在這麼大的地方管理員工非常困難，每個人都有自己的角色，對吧？讓人們都去思考關於未來的新事物，創新是很困難的事。因為當你有一個成熟的企業時，你就必須制定商業計畫，你必須設定既定的目標。你有收入，需要達到預期。所有這些事情都會讓你分心，讓你失去創新思維。嗯，然後有人說，你知道為什麼你不成立一個名為創新部的部門，讓這些人創新？這也行不通，因為這些人往往會迷失方向，而且你也不會把最優秀的人才分配到這。你知道因為創新者的困境基本上意味著你的核心業務是最重要的，你會開發並分配最多的資源給它，因為它能帶來收入和利潤以及企業的商業價值。你該如何解決這個問題？你會怎麼處理？我認為這總是沒有完美的答案，但我認為你需要做的一件事是向人們灌輸某種主人翁意識。他們不僅僅是為老闆工作。每個人都應該為他們的客戶工作。你知道很多人都想讓他們的老闆高興。但我認為傑克會說的最重要的事情就是讓你的客戶高興。你如果你是企業的所有者，你就會關心客戶對吧，然後你就必須……如果你關心客戶，那你就會開始擔心了我的客戶正在使用，他們習慣這樣使用產品。但如果有人推出了不同的東西，就像我們這行的情況，一種不同的平台互動方式。你知道我們最初是從列表功能開始的，現在是搜尋功能，現在是短影片，類似不同的平台互動方式。你必須開始考慮這些事情，因為你的客戶不會永遠陪伴你，如果不創新。所以我認為擁有這種主人翁意識很重要，要擔憂顧客未來會需要什麼。你必須領先一步，就像史蒂夫·賈伯斯說的你只需要創造某個產品，然後告訴顧客這就是他們需要的。我認為這是你必須具備的思維方式。另一點是你必須快速做出決策，在我們的行業中，在一個高速發展的科技行業裡，你總是面臨著資訊不足的困境，無法讓你做出完美的決策。因此你必須能夠容忍資訊不全的情況，然後果斷做出決定並堅持執行。如果之後發現錯了，就迅速轉向不同的方向。因此保持這種靈活性也是創新的關鍵。主持人傑裡米：是的。這種敏捷性和主人翁意識顯然對阿里巴巴起到了作用。你們於2014年在美國上市，這是當時歷史上規模最大的首次公開募股。這標誌著阿里巴巴作為一家真正的全球性公司的到來，符合傑克最初的願景。這是一家全球性公司，現在可以接觸到全球投資者。為什麼這很重要？Joseph Tsai（蔡崇信）：全球化為什麼重要？嗯，我認為公司擁有投資者，上市後他們就是你的利益相關方之一，當然你也知道，我開場時就說過，我們只做我們堅信的事，而且我們不會聽從投資者的意見，但如今作為一家大公司，我們擁有非常精明的投資者，他們會就行業動態給我們提供反饋。告訴我們競爭對手的動向，不僅是中國國內的競爭對手，還有全球範圍內的競爭對手，對吧。我們身處一個全球競爭的行業，特別是談到AI，雲端運算，你知道那是全球性業務，所以你能從全球投資者那裡獲得很好的視角。但當然在我們決定的時候，當阿里巴巴2014年進行IPO時，我們原本計畫在香港上市，但最終沒有這麼做，我們去了紐約。這其中有很多需要考慮的因素。我想告訴你，你知道，原本有些複雜的原因，我們分析了所有因素。但簡單的原因是紐約證券交易所你知道很有名，而且市場流動性很高。我認為今天大多數全球投資者仍然是美國人或者有美國血統，所以這對我們來說很自然某種程度上是進入了資本市場在美國。主持人傑裡米：ipo後，你辭去了首席財務官的職務，擔任副董事長。阿里巴巴經歷了你之前討論過的10年，這是10年挑戰中創新性較低的一段。2023年，你以董事長的身份回歸。你談到了快速決策。那麼，你能向我們介紹一下你作為主席回來時做出的快速決定嗎？Joseph Tsai（蔡崇信）：第一，我不會說是我做的決定。我們作為一個新的管理團隊，包括我們的CEO。我們決定了要專注的方向。我們將專注於電子商務這一核心業務，我們將把重點放在人工智慧和雲上，我們要加大投入力度。但它屬於基於CPU的計算業務，我們決定AI將成為推動未來基於雲GPU的雲端運算業務需求的關鍵。我們選擇了這兩條賽道，我們要做到極致優秀的領域，我們將具備極強的競爭力，至於其他資產或業務對我們而言就沒那麼重要了。你必須明確什麼是重點，什麼不是。所以明確重點就是首要決策。第二個決策是關於賽道的選擇。我們要在這些領域做到極致，我們需要什麼樣的資源，不論是資金還是人力都要投入最優質的資源進去。第三個決策是關於其他事務。要麼出售要麼徹底退出或者通過某種方式管理，以免分散管理團隊的注意力。這非常重要，管理團隊精力有限。你一天只有24小時。你不會同時照看八家不同的企業。我們以前去找投資人時，會說阿里巴巴有六個不同的事業部。這個業務類股，那個業務類股，這太蠢了！但你知道，如果我是投資者，開始聽一個傢伙說話，然後他說我有六個生意，那麼他們會說，你到底要專注於什麼？你知道，完全沒有重點。所以，我們我們實際上精簡了很多。我們出售了一些不重要的資產或非戰略性資產。但也有一些業務處於灰色地帶。它們並非核心中的核心。然而，它們具有戰略價值。一個例子是我們的水果配送業務。這是一家名為餓了麼的企業。我們被競爭對手打得落花流水。我們當時只有大約20%的市場份額。我們的競爭對手佔了約70%的市場份額。每個投資者都在問，你們什麼時候退出外賣業務，直接放棄食品配送？我告訴人們，我們就是你們說的那個食品配送本身。我們可能並不想單純做外賣業務，但整個配送基礎設施將會對電子商務非常重要，因為它是即時商業配送基礎設施。如果你想買東西，你點餐，30分鐘內就能送到，甚至更快。但如果未來人們想買衣服，行李箱、手機任何東西，他們都希望30分鐘內送達。如果我賣掉那家公司，我就會失去快速配送的基礎設施。儘管這項業務本身並非我們的核心業務，但它具有戰略意義。所以我們不得不在灰色地帶做出抉擇，該如何處理這些業務？當時存在大量業務，其中許多灰色地帶業務的管理團隊都主動來找我們，聲稱自己具有戰略重要性，我們就像我和你一樣，然後我就覺得，我才不像你那樣。主持人傑裡米：我不想成為那種人！Joseph Tsai（蔡崇信）：但之後你必須稍微直接一點，你必須說，對不起。我知道你工作很努力，做得很好，但你不會成為我們業務的核心。我們得想辦法，要麼為你找到一個合作夥伴，要麼承擔風險，要麼做一些不同的事情。主持人傑裡米：是的，在過去的幾年裡，對戰略的清晰和關注對你很有幫助。你和團隊。你提到了加倍投入雲端運算和人工智慧，所以我想談談人工智慧真的是世界上最大的兩個經濟體——美國和中國——之間競爭的最新舞台。你如何看待比賽的進行？Joseph Tsai（蔡崇信）：我認為AI是公司之間的競賽，但AI不應該是國家之間的競賽。這是一種技術，這是電力。我把ai描述為，擁有AI如同擁有水和空氣。你能想像如果一個國家拒絕另一個國家獲取某種生命必需之物嗎?所以我不認為這是一場兩國之間的競賽。我認為人們可以在很多領域進行合作。並為世界帶來福祉。如今人工智慧在醫療領域已取得諸多成就。我們已經能夠通過AI可以檢測胰腺癌的早期階段。你知道的，很多這些積極效益應該讓每個人都能享受到。所以，我並不認同這種觀點。網友提問廢話比較多，筆者就精選了一條關於阿里職場的問題翻譯給大家！網友提問：我想問你一個關於阿里巴巴文化的問題。人們經常將阿里巴巴文化描述為具有很強的阿里味，阿里巴巴在中文中有自己的語言、制度和期望。這有助於提高性能。這也可能造成壓力和官僚主義。你如何看待在保持強大的企業文化和確保開放性、包容性之間的權衡，Joseph Tsai（蔡崇信）：就像我說的，你必須喜歡和你一起工作的人。所以，如果我要定義阿里味的本質是什麼？那就是下班後你想和同事喝杯啤酒，或者甚至不是下班後，因為這也是工作的一部分，因為因為你正在與同事共度時光。嗯，我認為這很重要，就像那種文化氛圍。如何避免官僚主義？你必須明確界定那些重要那些不重要，並敢於告訴員工他們的部門並不重要。你必須要有勇氣直言不諱，我認為很多公司犯錯就是因為他們不想傷害員工。我是說用"包容性'這個詞，很抱歉，我們是在經營企業。我們無法讓每個人都參與進來。如果他們無法為你的目標做出貢獻，你就必須讓他們知道。https://singjupost.com/joe-tsai-co-founder-and-chairman-alibaba-find-your-people-transcript/TRANSCRIPT:JEREMY TEPPER: Joe, welcome to View from the Top.JOE TSAI: Thank you, Jeremy. Very happy to be here.JEREMY TEPPER: Well, it’s an honor to have you here. And as I was preparing for this interview and trying to wrap my head around just how massive and global Alibaba has become, what really struck me is how you have navigated across cultures throughout your career. East and west, law and business and sports and tech. So I actually want to start with your first cross-cultural major experience. You were born and raised in Taiwan and you moved to the US at age 13 to attend the Lawrenceville school in New Jersey. What was that like for you?Navigating a New CultureJOE TSAI: It was scary. My first, well, first before I went, I never went to seventh grade. I was in elementary school in Taiwan. And then my mother said, “Joe, I’m going to send you to school overseas, but you have to learn how to speak English.” And I didn’t know how to speak English, so I went to an English cram school for one year and skipped seventh grade. So I went into Lawrenceville in the eighth grade.During that cram school year, I basically memorized vocabulary, but I couldn’t put together a sentence in English. And that was sort of—so imagine you didn’t really know the language, you don’t know how to communicate, and then you’re in a completely new cultural environment. Just kind of just land, I mean literally like fresh off the boat, right. And landing in this very unique environment. Boarding school. It’s an all boys boarding school, so that has its own sort of pitfalls and you just have to fit in.I think I wanted to be one of the guys. I wanted to conform. I wanted to fit in. I look different, I dress different, I speak a different language. So it’s very important for me to feel like, get accepted by the rest of the guys. And then very quickly I went into sports. I got cut from the baseball team, unfortunately. Imagine a kid from Taiwan, Little League baseball. And I didn’t make the team, didn’t make the swimming team.But then I said, the hell with it. I’m going to try completely new things. I played football, American football, and then got into lacrosse because I couldn’t play baseball. It’s a spring sport, lacrosse, and baseball is spring sports. Couldn’t make the baseball team, so I went to play lacrosse.So the funny thing is, yesterday Jeremy and I, I think he was very worried, so he wanted to have like a prep session. We spent—so yesterday we spent an hour to prep for this talk. We talked nothing but sports.JEREMY TEPPER: Yeah, we both love sports. And I’m a huge Lakers fan. And obviously Joe is a big sports fan and a Nets fan and a Liberty fan.JOE TSAI: So we talked about that. Yeah, I won’t blame you for that. But anyway, the Lakers are doing better right now, so congratulations.JEREMY TEPPER: Thank you so much. Thank you.JOE TSAI: And then we found out that we were both football players in high school.JEREMY TEPPER: Yeah, undersized linebackers. Well, Joe, you found your footing. You kind of rebounded from those early sports challenges. You got into lacrosse. You went to Yale for undergrad and law school.JOE TSAI: Right.The Leap to AlibabaJEREMY TEPPER: You found yourself at a tax law firm and ultimately went into private equity. But just four years into Joe’s investing career, he came across a guy named Jack Ma, who was building a company alongside 17 other co-founders. It was an online marketplace. Zero revenue and a few thousand users very early. So what convinced you that was the time to make the leap?JOE TSAI: It was not only zero revenue, Jack didn’t even have a company. It wasn’t incorporated. And so Jack, all Jack had was a domain name and a fresh website. And I still remember the first time I went to see him. I climbed up the stairs of the second floor, second story apartment. There were a bunch of shoes outside. You take off your shoes, you go inside, and we talked for about an hour.And I thought Jack was fascinating. He had just charisma, that leadership quality. Jack used to be a teacher. And I think teachers make very good leaders because you have to communicate well. You also have to be able to identify good talent, good students, and develop them like what teachers do. So all those quality kind of came through. And that was what I was really, I found very appealing.I couldn’t understand the business plan because the Internet was new to everybody at the time. And he was talking about using the Internet to get all the companies in China that makes stuff and trade stuff to overseas and have foreign purchasers and buyers coming to the website to buy stuff from China. It’s a wholesale marketplace. So Alibaba’s first website was actually an English language because it was made for the rest of the world.So an hour into the conversation, I said, “Jack, let’s, you know, I need to have a bathroom break.” So I go into the bathroom and there were like 10 toothbrushes on the sink. So I realized people were just sleeping, living their sleeping bags all over the place.And they were—it was the early days of, you know, kind of the startup culture.JEREMY TEPPER: Yeah.JOE TSAI: And I decided, you know, I would say this to young people when you want to switch jobs or when you find a new job and you want to do something new. Find the people, the people that you want to partner with, the people that you want to follow, the mentors that you want to learn from. And to me, Jack is both a friend, a business partner, and also a mentor. He’s taught me a lot of things, and I felt like I could learn a lot from him.And also at the time, the Internet, if your business had a .com name in it, everybody thought that you were going to become a millionaire in six months. But for me, it was Jack. That personality, that leadership quality that actually came through from having been a teacher.Finding the Right Co-FoundersJEREMY TEPPER: Yeah, you clearly struck gold in that partnership with Jack. And a lot of our classmates here at Stanford are thinking about starting companies and looking for co-founders. So what’s the right formula? What should they be looking for in their co-founders?JOE TSAI: Well, first I think they should—it’s something that you can get along with, like, do you want to go out and have a beer with this person, you know, after work? I think that’s the first test.I think the next thing is complimentary, being complementary. So something that they’re good at, you’re not good at, and something you’re good at, they’re not good at. Having that complementary mentality, whatever that word is, being able to complement each other is really, really important so that the founders are not stepping on each other’s feet.Alibaba started with 18 founders, and I think we all complimented each other. Well, people brought different skill sets to things. And I think as you think about starting companies, it’s, you know, it’s not like you want to go out and find 18 people to found a company with, but I think only two or three founders sometimes is a problem because as you scale the business, your culture gets diluted and there’s no founder around to touch have touch points with employees.And having 18 founders, we had that advantage. You know, we were able to—each founder is able to touch more employees of the company, so I think that’s important. And then just people that you feel you want to spend 24/7 with and have a good time, because as a startup, you’re not going to go home very much or this work-life balance is not going to be something you should be focused on.JEREMY TEPPER: Yeah, I guess I should ask, did you have a toothbrush in the apartment as well?JOE TSAI: Yeah, I brought my toothbrush afterwards.JEREMY TEPPER: Oh, perfect.Bridging Cultural DividesJEREMY TEPPER: So I want to touch on this point of having this many co-founders, 18 in total. It’s very, very unique. And in joining them at Alibaba as CFO in 1999, you were really entering—JOE TSAI: By the way, I was for three months, and then Jack fired me because he said, “Joe, you have no clue how to do operations.” I was like, “All right, fine, I’ll be the CFO.”JEREMY TEPPER: Well, yet again, you rebounded from baseball to lacrosse, COO to CFO. Now, that group of co-founders had some connectivity to each other already, and you were kind of the new person coming in. You’re an American educated lawyer and investor. You’re from Taiwan. They were all Chinese mainlanders. How did you find a way to bridge that cultural—JOE TSAI: I think the first thing is—well, I’m usually more of a listener, and I approach the work just with humility. I’m there to learn. I didn’t know anything about the Internet. I had never worked with a company where it’s majority mainland Chinese in China, because all my prior work experience had been either in the United States or in Hong Kong.So for me, it wasn’t my place to come in and tell people what to do. So I listened a lot and just try to, again, that Lawrenceville experience, try to blend in, try to be one of the crowd, one of the person in the community. And I think that’s helped me a lot.I think you have to bring everything you do, you have to bring a sense of humility to the work, because there’s going to be people, you’re going to find your colleagues, your partners that are smarter than you are that can teach you a lot, and you need to approach your work with that kind of mindset.JEREMY TEPPER: Yeah. Going in and being a listener and being a humble one at that clearly served you well. Now, over time, Jack really became this larger than life figure, the face of the Chinese tech industry. And you were there for the entire journey, but largely behind the scenes. Why is that?JOE TSAI: Because he’s better at being in front on the stage.JEREMY TEPPER: I think you’re doing pretty great.JOE TSAI: He’s a better speaker. He’s a better inspirational leader. Right. I think it’s very, very important. I mean, I think I would like to think I’ve improved over the years in terms of communicating to people, but he’s a natural for him. Just getting on stage, have all the employees in the same room and just say, “Here’s our direction. This is why we’re doing it.” And he’s very persuasive. So he’s better at it.The Sand Hill Road ExperienceJEREMY TEPPER: Well, it’s another great example of finding somebody with a complementary skill set and working together toward the same mission. Now, your first step with Jack, as I understand it, was flying out here to Sand Hill Road, just a few minutes away to pitch investors and raise your seed round. How do you remember that trip?JOE TSAI: That didn’t go very well. I think we struck out. We had something like 15 meetings. We didn’t get a single investor interested in investing in our company. And that’s because I think at Stanford GSB you’re taught to show up with a PowerPoint or something, right? You have to show people something, a demo or something. We didn’t have anything. We came empty handed.And people are like, “Jack, what’s your business plan?” Jack says, “I don’t have a business plan.” He actually said, “I don’t have a business plan.” But the trends at the time in China was, I mean in hindsight you could have seen was the intersection of Internet technology and also China coming into the WTO, which meant trade between China and the rest of the world was going to explode.And today you go to China. China is a manufacturing powerhouse. Everything is made in China and they’re getting better and better, higher and higher quality, higher and higher tech technology. But the seeds were sown back then. That was just on the cusp of entering into the WTO and imagine getting all these manufacturers and trading companies bring all their products online for the rest of the world to see. That was the business idea.But Jack didn’t just talk about the business. He said, “My mission is to make it easy to do business anywhere.” That was our mission at the time. That’s what we told investors. And today that is still our mission. You can go on our website, Alibaba website says to make it easy to do business anywhere.JEREMY TEPPER: That’s incredible. So a mission first leader with no business plan. Did you guys come back from that trip, incorporate feedback from those investors?Lessons from Early FundraisingJOE TSAI: Not really, because the investors just looked very baffled. They didn’t give us a lot of feedback. They were kind of reticent. They were just, I think they just wanted to reject us outright. But that experience was important because we came back and we said we have to do what we believe in rather than have investors tell us what we should do. And I think that was a very important lesson.And then fortunately we were able to raise some capital. Some investors finally very skeptically bought in. I still remember our Series A round. We actually had Goldman Sachs come in. I mean you think, you know, you went to Sand Hill Road and you got Goldman Sachs. At the time, Goldman was investing like large private equity deals and then they had sort of a, you know, emerging technology investment business. And I happen to know, have a friend who is a partner who’s running that business.And then she calls me up and she said, “Joe, you know, I’ve got good news and bad news.” And at the time, you know, by then I was so, you know, disappointed with all the rejections. I was like I can’t stand any bad news. So tell me the good news first. And she said, “Well, we have our investment committee have approved the investment in Alibaba.” I was like that’s great. I was about to hang up. She said, “Wait, wait, you want to hear the bad news?” And I said, “Fine, what’s the bad news?” And she said, “Well, we also decided to back five other companies that’s very similar to your business model and we’re just going to see, let you guys compete.”JEREMY TEPPER: You guys did and you rose to the top. You guys raised the round from Goldman Sachs, then another round from Softbank and then it was really off to the races. Over the next 10 years you guys launched payments, messaging, a consumer to consumer marketplace, a business to consumer marketplace, advertising, and a cloud infrastructure offering which is now the fourth largest in the world and has 40% market share in China. So Alibaba became a super app.Now one of those major in-house wins was Taobao, Alibaba’s consumer to consumer marketplace. Can you take us back to the moment of launching Taobao against what was at the time a real monopoly from eBay China?The Secret Launch of TaobaoJOE TSAI: It was back in the early 2000s. eBay had acquired a company called EachNet. So immediately they had a large market share and at the highest point they had like 90% market share of consumer e-commerce. And we were the new player coming in.And so one day, well going back, I mean I think I was talking to the dean here and she mentioned someone, Jane Sun, who is the CEO of Ctrip, and her husband was the first CTO, John Wu, first CTO at Alibaba. He came from Yahoo. So sorry, this is a long-winded story.And when Jack decided to launch a consumer marketplace, Taobao, he was very excited. He had talked to Masa at Softbank and Masa was very supportive. And because it’s a big idea, Masa will support anything that’s big, anything that costs a lot of money he will support. And Jack’s very excited. He said, “I got the money from Masa, I’m going to do this.” And he goes to John Wu, our CTO, and John said, “No way in hell, over my dead body are we going to do this.”And that’s because John had worked on Yahoo Shopping and it was not successful. So he didn’t have a very positive view of us going into the consumer e-commerce space. But then obviously Jack overruled everybody and I was, I didn’t know strategically, I couldn’t make that judgment. But I decided that we were going to structure it so that financially we weren’t going to get totally hurt.So actually from day one, Taobao actually was a joint venture, 50/50 JV between Alibaba and SoftBank. And it wasn’t until like six, seven years later, when Yahoo invested in us, we took the opportunity to fold in the rest of Taobao into Alibaba to own 100%. And that was the big, the best decision we’ve ever made. Otherwise today, Taobao will still be half owned by SoftBank. Right? And we wouldn’t know where it would go.So, but when we started Taobao, it was a secret project. And if you guys know Alibaba a little bit, pretty much almost all the startup new businesses went through this apartment, the Jack original apartment, with the shoes and the toothbrushes. So we took away a team of seven people into this apartment. They all had to sign NDAs. They couldn’t tell their colleagues or spouses or anybody. So one day, you know, the person that you’re sitting next to, the cubicle next door, at the cubicle next door disappeared. And they, you know, people are like, where do they go?So the first seven people seeded the Taobao business. They went to the Jack, the Jack original apartment to, you know, do the work, create that marketplace. And then our first, when we launched it, the first items being sold on Taobao came from everybody’s closet. Everybody just went to their closet and said, “What do we have? What can we find to sell it on Taobao?” We seeded that marketplace with a lot of junk. That’s how this thing started.And I remember it was for almost a year, we kept it a secret project. I mean, the website was launched so people knew Taobao existed, but nobody knew it was connected to Alibaba.JEREMY TEPPER: It’s really quite remarkable when you think about how it combines what has made Alibaba so successful. The complementary skill sets between Jack the visionary, your CTO at the time, who had a more conservative mindset, and then you structuring the joint venture and then the focus that you guys applied to get this off the ground.You guys had a lot of success with Taobao. Obviously. Within two years, Alibaba took back the market, you guys took 60% market share. And eBay China ultimately closed down. Now today, Alibaba is one of the most successful and important companies in the world. One in six people on this planet interacts with Alibaba’s ecosystem regularly now. I think that’s right.JOE TSAI: I need to put that into our investor presentation.JEREMY TEPPER: Fact check me on that. Yeah. If you want me to join, I’m happy to hop in. So now you’re in a different position. You are the incumbent, like eBay once was, and you’re facing up and coming challengers.JOE TSAI: Yes.JEREMY TEPPER: So how do you guys continue to innovate from that incumbent position?The Challenge of Innovation at ScaleJOE TSAI: Well, we don’t always do it right. We’ve gone through periods where, when we stopped innovating and we suffer from it. As a large company, by definition, we have 120,000 employees. It’s very difficult to have people in such a large place. Everybody has their role. Right. It’s very difficult to get people to think about new things, about the future, innovate.Because when you have an established business, then you have to set business plan, you have to set established goals, you have revenue projection that you need to meet, and all those things get preoccupied and take you away from an innovative mindset. And then some people will say, “Well, why don’t you just set up a division called the innovation division to have those people innovate?” That doesn’t work either because those people tend to get lost. And also, you’re not going to allocate the best talent to that, you know, because the innovator’s dilemma basically, by definition means that what your core business is the most important thing and you’re going to allocate the most resources to it because it generates revenues and profits and value, enterprise value for the business.How do you address that? How do you deal with it? I think it’s always, there’s no perfect answer. But I think the one thing that you need to do is to instill in people some sense of ownership. They’re not just working for their boss. Everybody should work for their customers. And, you know, a lot of people want to make their boss happy. But I think the most important thing Jack would say is to make your customer happy. You are, if you’re the owner of the business, then you care about the customers, right?And then you have to, then if you care about the customers, then you start to worry, “All right, my customers are using, they’re used to using the product this way. But what if someone else comes in with something different?” Like you know, in our business, a different way to engage with the platform. You know, we started with a listing, then a search, now it’s short video. Like different ways of engaging with the platform. You have to start thinking about these things because your customer is not going to stay with you forever if you don’t innovate.So I think sense of ownership, worrying about what the customers will want in the future, and you have to be a step ahead. Almost like what Steve Jobs says, you need to just invent something and then tell the customers that’s what they need. I think that’s the mindset you need to have.The other thing is you have to make decisions very quick. And in our business, in a high growth technology business, you are always living with a deficiency of information that allows you to make the perfect decision. And so you have to be able to tolerate not having full information and then just making a decision and commit to it. And then if you find out you’re wrong, pivot fast in a different direction. So having that agility is also key to being innovative.Going Global: The 2014 IPOJEREMY TEPPER: Yeah, that agility, that sense of ownership has clearly worked for Alibaba. You guys went public in the US in 2014 in what was at the time the largest IPO in history. And it marked Alibaba’s arrival as a truly global company in line with Jack’s original vision. And this was a global company with access to now global investors. Why was that important, being global?JOE TSAI: Why is that important, access to global investors specifically?JEREMY TEPPER: Why is that important, access to global investors specifically?JOE TSAI: Well, I think having investors in your company and being a public company, they’re one of your constituencies. And you know, of course I started this talk by saying, you know, we do what we believe in, we don’t listen to investors. But now as a large company, we have very smart investors that give us feedback about the industry, about what our competitors are doing. Not just competitors in China, but competitors around the world. Right. We were in a globally competitive industry, especially talking about AI cloud. You know, that’s, you know, that’s a global business. So you get a very good perspective from global investors.But of course, at the time when we decided, when Alibaba went IPO in 2014, we were going to go list in Hong Kong, but we ended up not doing that. We went to New York. There were a lot of considerations that went into it. I would like to tell you there was some intricate reason and we analyze everything, but the simple reason was the New York Stock Exchange is, you know, well known, and there was a lot of market liquidity. Most global investors, I think still is today are American or have American roots. So it was natural for us to sort of tap into the capital markets in the United States.Returning as ChairmanJEREMY TEPPER: Well after the IPO you stepped down as CFO and stepped into the role of vice chairman. And Alibaba went through a 10 year period which you discussed earlier was on the less innovative side, a 10 year period of challenge. And you returned as chairman in 2023. You talked about quick decision making. So can you walk us through the quick decisions you made when you returned as chairman?Strategic Focus and Resource AllocationJOE TSAI: Number one, I wouldn’t say I made the decision. We collectively, as a new management team, including our CEO Eddie Wu, we decided what we were going to focus on. We were going to focus on our core business of e-commerce and we were going to focus on AI and cloud. We were doubling down. I mean we had a cloud business, but it’s a CPU-based compute business. We decided that AI is going to drive future need for cloud GPU-based cloud computing business.And so you pick those two lanes where we’re going to be extremely good at, we’re going to be extremely competitive and then the rest of the assets or businesses that we have are just less important. And you have to decide what’s important, what’s not. So deciding what’s important, first decision, the second decision is okay in the lanes that we want to be very good at, what kind of resources we want to allocate to, whether it’s capital or people, you want to put the best resources into it.The third decision is on the rest of the stuff. Let’s sell them or get rid of an exit or whatever. Somehow manage them in a way so that it’s not distracting to the management team. It’s very important. Management team has limited bandwidth. You only have 24 hours a day. You’re not going to be able to watch eight different businesses.I mean we used to go to investors and say Alibaba has six different divisions and this here, these segment and that segment, that’s stupid. You know, if I was an investor, start listening to a guy you know that says I have six businesses, then they’re going to say what are you going to really focus on? There’s no focus. So we actually streamlined a lot of that. We sold unimportant assets or non-strategic assets.But then there are some businesses that are in the gray area. They are not core core. However, they have strategic value. One example is our food delivery business. It’s a business called Element. We were getting killed by our competition. We had, you know, had like, only 20% market share, our competition, like 70% market share. And every investor is like, when are you going to exit this business? Just get out of food delivery.And I told people we are. You’re right. Food delivery itself, we may not want to be in the food delivery business per se, but the whole delivery infrastructure was going to be important for e-commerce because it’s an instant commerce delivery infrastructure. If you want to buy stuff, you order meals, they’ll come to you in 30 minutes or even shorter period of time. But what if in the future people want to buy clothing, luggage, their cell phones, whatever. They want it in 30 minutes.If I sold that business, I would have lost that quick delivery infrastructure. And so even though that business itself was not core to us, it has strategic importance. So we had to sort of decide in the gray area what to do with those. There were a lot of businesses where a lot of the management teams of those gray area businesses came to us and said, oh, we’re strategic and important. And we’re like, I’m like, you’re not.JEREMY TEPPER: I would hate to be one of those folks.JOE TSAI: But then you have to be a little bit direct. You have to say, I’m sorry, I know you’ve worked very hard, you’ve done a very good job, but you’re just not going to be core to our business. And we kind of have to find some way to either find a partner for you to take on the risk or do something different.The AI Race: Companies vs. CountriesJEREMY TEPPER: Yeah. That clarity and focus on the strategy has served you well over the last few years. You and the team you brought up doubling down on cloud computing and AI. So I want to talk about that. AI is really the newest arena for a race between the world’s two largest economies, the United States and China. How do you see the race playing out?JOE TSAI: I think AI is a race among companies. But AI shouldn’t be a race between countries. It’s a technology. It’s electricity. I described AI as having access to AI as having access to water and air. Can you imagine if a country denies another country something that’s essential to life?So I don’t see this race as a race between two countries. I think there is a lot of areas where people can cooperate and benefit the world. AI today has achieved a lot in the medical area. We’ve been able to have AI tools that could detect early stages of pancreatic cancer. And I mean, a lot of these positive benefits should be available to everybody.So I don’t subscribe to this view that there should be a race between two countries. I can understand that there’s a school of thought that says if the military gets a hold of AI, that whoever runs faster or develops AI faster for military use will have supremacy. But let’s just limit that to the military field and let AI capabilities proliferate around the world. And that’s the way to go, I think.JEREMY TEPPER: Yeah, I think most people today would agree that AI has this potential to be hugely transformative globally, but yet there are two very different strategies between Chinese companies and American companies. China, Chinese companies have access to very low cost of energy. They’re pursuing mass distribution open source, like Alibaba is. And American companies have a hardware semiconductor advantage. So maybe we don’t call it a race, but which approach do you think will win?JOE TSAI: We have to define what winning means. I think for me, winning means it gets proliferated. I hate the word diffusion. I wanted to say proliferation. But then people bring in kind of the analogy of nuclear weapons, whatever. So AI should be used by the most number of people and in whichever society where they have most people use AI and benefit from AI is going to be the winner.Building Qwen: Scaling AI Through ExperienceJEREMY TEPPER: Well, Alibaba has pursued that strategy. Qwen Alibaba’s model recently surpassed 700 million downloads globally. It’s the leading open source model in the world. How much of Qwen’s success, your success in building Qwen as a product comes from prior successes scaling other products like Taobao?JOE TSAI: A lot. If you think about these large scale Internet companies with hundreds of millions or even billions of users, the problems that we try to solve is the problem of scale, the problem of concurrent users coming in. So the problems of parallel computing. So having that experience really helped us in developing the right algorithms to, you know, ultimately that turned into AI.And our effort in, you know, the modern AI as we know it is based on the transformer architecture. So we had development efforts dating back to 2019. And remember, this is like three years before OpenAI publicly came out with ChatGPT. But we didn’t put enough resources into it. Right back then, people were still debating whether that’s the right technology path versus something different.But once back, you know, in 2022, when ChatGPT came out, I think everybody woke up and said, we need to double down, we need to put those investments in. But by then we would have, we had already, you know, years of three, at least three years of development, development in transformers, but also large scale computing.That’s all we had invested in our cloud infrastructure back 17 years ago. When we decided that our e-commerce business was getting so large in terms of users and also the data that we have to manage, we have to develop proprietary software to manage computing clusters. And basically my layman’s way of thinking about it is your software enables clusters of 50,000 computers, 100,000 computers, to work and think like one brain, even though they’re located in different data centers. Right.So that’s the technology that we develop, proprietary that formed the basis of our cloud computing business today. So we went into cloud computing out of necessity, not because it was a hot thing back 17 years ago. Nobody has heard about cloud computing. Nobody really thought it was a business. So, yeah. So basically we had a lot of good foundational work that’s done before sort of the current craze about the current version of AI.Sports Investments and Following Your PassionJEREMY TEPPER: Yeah. Well, it will be interesting to see how this competition plays out. I want to return to a lifelong passion of yours on this point of competition, sports. You acquired majority ownership of both the NBA’s Brooklyn Nets and the WNBA’s New York Liberty in 2019, and together they were recently valued at nearly double what you acquired them for. How did you know these were great investments and not just passion projects?JOE TSAI: I didn’t. I thought they were safe investments in the sense that the value was not going to go down. I think a few years ago I gave the analogy of if you live in New York City, you see a, you know, rooftop apartment, Park Avenue apartment, that’s. You buy into it, you’re not going to lose money because it’s in a good location. It’s a prize asset that everybody wants.So I thought the downside was going to be protected. But I did not anticipate how much value the valuation of these professional sports teams have increased over the last few years. There are two leagues in the United States, the NFL and the NBA, where there’s tremendous amount of value being created. And that’s because of the media rights, meaning broadcast rights, streamers and TV networks pay a lot for these media rights to be able to broadcast your games. And that’s supported the increase in valuation.JEREMY TEPPER: Yeah, you applied that investor mindset, assessing risk in becoming an investor in these two amazing franchises. Now, you turned a true passion sports into an investment strategy and business school. Advice on passion is somewhat mixed. Some say avoid turning it into a career, and others say you got to do what you love. Where do you stand?JOE TSAI: I can’t imagine if you woke up every morning, you went to work and you hated your job. That’s, you know, that’s not a recipe for success, but you don’t, you know, you could be passionate about a lot of things, but at least you have to be interested in what you do and you want to do it better. Whether you call it passion or something different or dedication or interest, you know, there’s gradations.You know, if one of you says, “I’m passionate about cooking” and then you go out and try to open up a restaurant, that’s great. But that’s probably not why you went to Stanford Business School for, you know. So my advice is find the people first. Find the people that you want to spend time with. And then obviously the mission of the company and the stuff that you do is something that you have an inherent interest in. Whether you call that passion or not. You should be passionate about your colleagues, about your co-workers. That’s what I call passion.Q&A: Capital Allocation and Business StructuringJEREMY TEPPER: Well, with that, Joe, I’d like to turn it over to a few of my classmates who have prepared questions for you. Hi Joe, thank you so much for coming over and speaking with us today. My name is Katherine. I’m a current MBA 2 student and before GSB I actually spent 3 years working at JP Morgan Hong Kong as a coverage banker for Alibaba Group, working across many business units and then ecosystem players such as Alicloud, Huma, Alipicture, Ali Health, et cetera.What really amazed me is Alibaba’s capability of pioneering very sophisticated business structuring and capital allocation. Looking back, what are some decisions that you’ve made or the group have made are misunderstood by the market, but it’s really critical for the growth of the company?JOE TSAI: Well, I would like to think we’re very sophisticated in technology, in products, this and this. You know, I mean I think we do a pretty decent job in terms of managing the capital markets and, and I guess so-called financial engineering. I’m sorry, your question is what is…JEREMY TEPPER: Yeah, what are some decisions on business structuring and capital allocation that are potentially misunderstood by the market yet to be very critical for Alibaba’s growth today?The AI Investment ChallengeJOE TSAI: Yeah, yeah, look, I think today everybody is worried about the amount of capital that’s being allocated to AI. Different at the different layers. Right. In our company, AI means a full stack of things. You know, how much money we put into developing our LLM, how much CapEx we make into our cloud infrastructure. So there’s, and also we have, in order to sort of test how good your AI is, you have a consumer application. Right.So we now have the Chengwen app which is our consumer app. So how much money do you allocate to promoting marketing, to promote the consumer application? So along the full stack you have to allocate capital. And I think every company that’s serious about AI is doing all these things and perhaps that’s the part that investors are trying to figure out whether you’re allocating correctly to different parts of the stack.I don’t think there is the right, I don’t even have the right answer for you, but I do think that these three things that I’ve mentioned, a consumer AI application, the large language model that’s backing up, that’s supporting the application, and also the infrastructure that’s supporting everything, those are equally important. And you kind of have to anticipate that. You just have to believe that there’s going to be demand like infrastructure, you know, the infrastructure investments.You have to be very forward looking, securing data center capacity, securing the energy, buying the equipment. There is a lead time to it. But I think that your judgment of whether there’s going to be demand should be based on what ultimately the users, the enterprises that are going to need a demand as opposed to some philosophical pursuit of AGI. There is a lot of philosophical pursuit of AGI right now that’s driving the demand. But then at the end of the day, you need to kind of look at what the market can bear.Maintaining Alibaba’s CultureJEREMY TEPPER: Hi Joe, my name is Hong, I’m MBA1 currently at GSB. I was also a coverage banker at Credit Suisse in Hong Kong. And I wanted to ask you a question on Alibaba’s culture. People often describe Alibaba’s culture as having a very strong Ali flavor, AKA Alibur in Chinese, with its own language systems and expectations. While this has helped drive performance, it can potentially also create pressure and bureaucracy. How do you think about the trade off between maintaining a strong corporate culture and ensuring openness, inclusiveness and innovation as Alibaba continues to mature in the future?JOE TSAI: Like I said, you have to like the people you work with. So I think if I were to define what is the essence of Ali, it is, you want to have a beer with your coworker after work or it’s not even after work because it’s also part of work because you’re spending time with your coworkers and I think that’s important.And you know, the culture. How do you avoid bureaucracy? Right. Well, you have to define what’s important, what is not, and be able to tell people that their division is not important. You have to have the guts to tell them. I think a lot of companies make mistakes because they don’t want to hurt the people. I mean use the word inclusive. I’m sorry, we’re running a business. We can’t include everybody in the enterprise. If they’re not going to be contributing to where you want to go, you have to let them know.The Most Challenging Step in AI TransformationJEREMY TEPPER: Thanks Joe for all the great sharing. So I’m Musea, second year MBA. Sadly I’m now coverage banker for Alibaba, but I actually work as a product manager in China and also build partnership with Alibaba cloud and broader ecosystem. So I was wondering since Alibaba’s ecosystem and also full stack AI strategy as you mentioned, really fascinated me. So from your perspective, what has been the most challenging and most difficult step to made but is really crucial for the company in terms of AI transformation?JOE TSAI: I think what is difficult is every part of the stack that I mentioned that we have to invest in costs a lot of money and then you invariably people ask, what’s the ROI? But like I said, you have to have a certain belief that it is strategically important and you’re not going to focus today on ROI.And the other thing is it’s not like we came to this with a well designed grand plan that these are layers of the AI stack that we have to. They all come sort of in different moments and if they didn’t have their own individual success at their layer, then they wouldn’t have gotten attention. Right. So it was a very bottoms up process. Then when you look back in hindsight, you think, well that’s kind of haphazard because there was no grand design to begin with.What if your LLM development was unsuccessful? Then you would not have put any more resources into it. And it turns out that having a really good large language model, a large foundational model is really, really important in the AI race. I’m not talking about race between countries, I’m talking about race among companies. So those are some of the difficulties and in a way there’s a little bit of luck involved. But then at the end of the day I go back to the people you have to work, you know, identify the right people to work on these projects.Rapid Fire RoundJEREMY TEPPER: Thank you, Joe. And thank you to my classmates for those thoughtful questions. Joe, before we wrap up, we’re going to do a View from the Top tradition, the Rapid Fire segment. Are you ready?JOE TSAI: Yeah.JEREMY TEPPER: All right, let’s do it. City that feels most like home to you.JOE TSAI: Taipei. That’s where I was born.JEREMY TEPPER: Favorite sports arena food.JOE TSAI: Barclays center for sure.JEREMY TEPPER: Ok, but what item at Barkley Center?JOE TSAI: I can know for some reason. I’m thinking of the Vietnamese Banh Mi pork sandwich.JEREMY TEPPER: All right. Yeah, I’m a hot dog guy, but yeah, that’s great, too. Best purchase you’ve ever made through Taobao.JOE TSAI: Oh, a pair of gym shorts. Under Armour. Nice. I think I bought them 10 years ago. I’m still wearing them when I exercise.JEREMY TEPPER: It’s a great endorsement for Under Armour, the app on your phone that you use most.JOE TSAI: I have two apps that use a lot most. One is a Quinn app because that’s now it’s an assistant for me when I do research, you know, whatever. I want to know something, I go to the Quinn app. The other one is Twitter. I get all my news. If I want to know what’s going on in the world, I can get it instantaneously from serving. I mean, now the algorithm is so good, so they know what I’m looking for. It’s basically subjects on AI and sports. Those are the two things.JEREMY TEPPER: Well, last rapid fire question on sports. Would you rather win an NBA title with the Brooklyn Nets or become a professional lacrosse player?JOE TSAI: Oh, that’s a tough one. Both have low probability, but I’ll take the NBA title.JEREMY TEPPER: All right.JOE TSAI: NBA title it is.Final Advice: Win Locally FirstJEREMY TEPPER: And to close one final question, and this one isn’t rapid fire. What’s your best piece of advice for those in the crowd today hoping to build a global company?JOE TSAI: I don’t think you should think about global. Building a global company from day one, because you have to win local. It’s if you have a grand plan. I mean, the world is a large place. To build a global company, you have to have a lot of infrastructure that starts with small pieces. So you have to win. Have small wins. So think about small wins and win locally where you start. You have to win the market where you started.And then you can think about going overseas, going global, because with those winning local battles, you’re training your team, you’re developing talent that enables you to be a global player. So you got to start somewhere.JEREMY TEPPER: Ladies and gentlemen, Joe Tsai. Thanks, Joe.JOE TSAI: Thank you. (瑞恩資本RyanbenCapital)

2026年一開年，高德北京總部洋溢著喜悅之情。高德員工宋磊高興地說，過去一年大家幹勁十足，年底能拿到豐厚年終獎了。員工士氣高漲背後，高德地圖最新月活躍使用者數已達9.96億，高出第二名一大截。宋磊更開心的是，團隊過去一年的努力，得到了阿里巴巴集團CEO吳泳銘的認可，他不光替高德的活動站台，還在新年家書裡，稱讚高德去年推出的掃街榜能讓商家專心經營、讓消費者放心消費。高德掃街榜是阿里生態協同的新典範。它在去年9月發佈後，帶動高德地圖月活躍使用者數持續攀升。曾經游離於阿里主航道的高德，是怎麼成為集團打通線上線下的關鍵棋子的？這背後藏著一個阿里技術派蟄伏與崛起的故事。01 技術派掌門的蟄伏與崛起“他是個被低估的阿里合夥人。”阿里前員工李笠告訴《財經天下》。高德現任董事長劉振飛為人低調，畢業於電腦專業，擁有北京科技大學學士和北京大學碩士學位。在職業生涯早期，劉振飛曾在微軟Office組工作，2006年加入阿里，2014年便成為阿里合夥人。想要成為阿里合夥人並不容易：至少在阿里體系內工作滿5年；具備突出的領導能力，對公司發展有實在的貢獻，高度認同並願意全力傳承阿里的企業文化和價值觀；需要由現任合夥人提名，然後經過全體現任合夥人投票，並且必須獲得超過75%的同意票才能當選。這些年阿里合夥人幾經變遷，但劉振飛一直穩居其列。李笠認為，他擔任合夥人靠的不是前台曝光度，是實打實的技術。他很符合吳泳銘心中年輕化、技術派導向的人才，有點像阿里的首席消防員兼建築師。搭建阿里媽媽廣告系統是劉振飛在阿里的第一個重任，2009年組建淘寶技術保障部，是他職業生涯的另一個關鍵節點。彼時，淘寶網頻繁當機讓管理層頭疼不已，劉振飛臨危受命組建淘寶技術保障部，合併阿里媽媽與淘寶的維運體系。合併後，團隊面臨的首次大考是“雙11”。這場“硬仗”被劉振飛形容為“痛苦合併”——兩套技術團隊、兩種工作習慣，衝突不斷。他頂著壓力，在“雙11”的流量洪水中，練出一支技術鐵軍。“要完成所有鏈路的壓測，一秒鐘就有潑天的訂單進來，要保證訂單不能丟失，付款不能中斷。”李笠說，在一秒鐘零風險零故障上，劉振飛是真正的大功臣之一。也正是他創立的技術保障總指揮部、全鏈路壓測機制，支撐了淘寶交易額從9億元躍升至362億元。劉振飛啃下的“硬骨頭”，還有阿里“去IOE”戰略，即擺脫對企業級IT系統IBM、Oracle、EMC的依賴。2010年，他在向時任集團首席架構師王堅匯報預算時，提出“不再購買小型機”，卻被王堅追問：“為何不徹底斷後路？”此後，他成為這一戰略最堅定的執行者，甚至不惜得罪業務部門，強行要求阿里金融、聚石塔等項目使用尚不成熟的阿里雲體系。這種近乎執拗的堅持，最終為阿里技術自主權打開了突破口。因在奠定阿里技術基礎方面的關鍵作用，劉振飛於2014年成為阿里合夥人，直到今天他也沒下牌桌。2017年，劉振飛正式加入高德。彼時，俞永福作為高德的掌舵人，負責戰略規劃，作為技術出身的劉振飛，將俞永福制定的戰略藍圖，通過技術管理和產品執行力變為現實，解決“怎麼做”的問題。例如，將“做好一張地圖”和“成為出行平台”的戰略，落實為體驗提升、技術升級和使用者增長。在高德發展的特定階段，“戰略—執行”模式是高效且成功的。在俞永福時代，高德做了不少微創新，比如紅綠燈、車道級導航、會車提醒等功能的第一時間發佈等。其中，皆有劉振飛的身影。鮮為人知的是，劉振飛還主導了高德“上雲”，即將高德地圖全部核心業務系統，從自建的傳統資料中心，遷移到阿里雲平台。這是高德的重要戰役。“上雲雖然是一個技術的必然，但那時候上，負擔還是很重的。”高德前員工趙清向《財經天下》回憶道。彼時，趙清從其他大廠入職高德沒多久，他發現這仍是一個相對傳統的技術公司，“不上雲，很多業務跑不起來，生產成本也會居高不下。”在劉振飛的主導下，高德成功“上雲”後，系統的可用性得以大幅提升。02 阿里合夥人下了一盤大棋2024年3月，劉振飛從高德總裁升任董事長。這之後不到兩年，他帶領高德成功融入阿里業務的主航道，得到吳泳銘的稱讚。劉振飛下的這盤大棋，正是高德掃街榜。高德掃街榜基於導航到店人數、復購率等300多個維度的真實使用者行為資料，通過AI演算法每日動態生成，是使用者用腳投票出來的榜單。劉振飛在內部網路發帖闡釋了推出的初衷：“技術不僅要追求星辰大海，更要呵護人間煙火；要用AI賦能情懷，也要用情懷賦能AI。”業內對掃街榜發佈後的資料感到震驚：上線不到1個月，使用者破4億；上線100天，使用者規模突破6.6億。QuestMobile資料顯示，在掃街榜的帶動下，高德地圖單月新增4600萬月活使用者，月活躍使用者數增至9.96億。這也是吳泳銘點贊掃街榜的直接原因。更深層次的原因在於：掃街榜是阿里“AI+大消費”新戰略協同的關鍵支點，直接體現了吳泳銘的核心管理思想：以AI技術為驅動，實現爆發式使用者增長；通過生態協同，形成獨特競爭優勢。趙清告訴《財經天下》，掃街榜的核心壁壘是高德自研的世界模型技術。生態協同的邏輯是這樣的：以前，高德指路，淘寶賣東西，支付寶付錢，各幹各的。現在，它們合夥開店，互相帶客，消費者通過高德地圖導航去吃飯，去看電影，去逛景點，可以與阿里系其他App互動起來。高德能實現產品的創新性突破，與劉振飛帶領的高德“新班委”有關。除靠技術坐穩阿里合夥人位置的董事長劉振飛，晉陞為CEO的郭寧是高德內部培養起來的頂尖技術專家，是唯一在基礎平台、中台、業務都擔任過負責人的高管。“這種技術+技術的搭檔，確保技術視角在最高決策中的核心地位。”某網際網路公司技術總監王維告訴《財經天下》。多名高德前員工告訴《財經天下》，劉振飛的柔性管理與親民風格，在高德內部是重要的凝聚力來源。他被員工親切地稱為“大飛哥”，其激勵方式頗為獨特：常在釘釘群發佈個人伏地挺身視訊，帶動全員運動；除夕夜主動發紅包活躍氣氛。在業務推進中，他更注重即時激勵與公開表彰，讓團隊成員在高效協作中獲得強烈成就感，從而持續提升士氣。在劉振飛的帶領下，高德不僅突破產品創新，還實現技術飛躍，去年將“高德地圖2025”升級為基於空間智能的AI原生應用；宣佈入局Robotaxi賽道，與小鵬合作提供L4級自動駕駛出行服務等。從“地圖導航工具”升級為“智能出行服務平台”的高德，打破了阿里的收購“魔咒”。03 本地生活賽道中“進攻的尖刀”在阿里的收購歷史上，高德是少見的成功案例。“背後的原因在於，高德不追求短期流量變現，而是通過持續試錯、打磨產品體驗，在關鍵節點提升核心能力。”趙清總結道，這與許多急於求成整合的被收購公司形成對比。將視線拉回至2013年，阿里豪擲近15億美元收購高德。彼時，高德正處於從傳統地圖服務商向網際網路轉型的關鍵階段，嘗試將團購、打車、購物等業務都整合到高德地圖中。阿里的收購，改變了高德的轉型方向。在阿里接手半年後，高德宣佈暫停O2O佈局，專注做好地圖服務，並提出了“三年不商業化”的目標。阿里對高德展現出極大的耐心：在持續虧損的階段，每年仍投入超過20億元用於技術研發。時任高德董事長的俞永福也不負眾望，僅用兩年就將百度等競爭對手甩在身後，讓高德成為阿里在移動網際網路領域的巨大流量入口和底層支撐。因此，高德在阿里系一直有很高的地位。“每年的績效都是3.75，屬於非常靠前的第二梯隊。”李笠向《財經天下》透露。在穩住地圖服務基本盤的同時，高德也在探索商業化。2020年9月，高德發佈“高德指南”，對標大眾點評的餐飲團購和酒旅業。2021年，高德、餓了麼和飛豬整合為生活服務類股“飛高了”。高德業務不斷擴展，但由於高昂的研發投入和激烈的市場競爭，長期處於虧損狀態。2023年6月，吳泳銘成為阿里集團新掌門時，高德的壓力更是陡增。此時，阿里的戰略方向發生根本性改變：聚焦核心戰略電商和雲，進一步推動戰略聚焦與業務協同。高德的主營業務地圖、導航、出行，與阿里明確的兩大核心業務“電商”與“雲+AI”的直接配合相對較弱。在去年的業務架構調整中，高德也因此從之前重要的“生活服務類股”中被剝離出來，和菜鳥、優酷等業務一起，被重新歸類到“所有其他”業務類股。這意味著，高德在阿里內部的地位變了。這些被劃分到其他業務類股裡的業務，最核心的任務是盈利，不能再靠集團輸血，如果持續虧損，阿里會將其擺上售出的貨架，步高鑫零售的後塵。好消息是，經過一番努力，高德終於賺錢了，2025財年第三季度（截至2024年12月31日），高德地圖首次實現盈利。進入2026財年（2025年4月1日～2026年3月31日），高德一方面在使用者增長和產品創新上保持了強勁勢頭，另一方面在使用者規模和產品創新上繼續進取。去年推出的掃街榜，更是讓高德找到了與集團主航道互動的可能性，成了阿里本地生活賽道中那把“進攻的尖刀”。今年一開年，高德掃街榜推出全球首個“飛行街景”功能，成為高德持續技術創新的最新例證。不過，在多名高德前員工看來，劉振飛帶領下的高德，更具想像力的方向是B端“空間智能底座”：讓AI看懂現實世界，並幫你提前預判路況的技術。高德每天海量的導航資料，不斷“喂養”AI模型，使得發生事故時其能提前預警風險，進而吸引更多人使用高德，這樣會產生更多資料，AI因此變得更聰明。只不過，這條路需要長期投入。如何度過高投入、低產出的起步階段，是阿里合夥人、技術派掌門劉振飛及團隊要闖過的另一道難關。 (品牌頭版)

阿里，正在主動瘦身回籠資金300億，阿里，準備幹大事了。01 阿里，再度減持！阿里的減持動作，還在繼續。就在近日，三江購物發佈公告：阿里巴巴旗下的杭州阿里巴巴澤泰資訊技術有限公司（以下簡稱“阿里澤泰”），已經完成了最新的股份減持。從公告中披露的資料來看，在去年11月27日至今年2月5日期間，阿里澤泰累計減持了三江購物的股份約1643萬股，減持比例達到3%。圖源：三江購物從公開資訊來看，這並非一拍腦袋的“清倉離場”，而是一場提前規劃、分階段推進的資本操作。要知道，早在2025年4月，阿里方面就已經啟動了首次減持計畫。當時的理由並不複雜，屬於常見的“商業安排需要”，但這一動作本身已經釋放出一個清晰訊號：三江購物不再是阿里必須長期重倉的資產。隨後，第二輪減持在當年8月完成，減持方式同時覆蓋集中競價與大宗交易，持股比例進一步下調；到了2025年11月，第三輪減持計畫被正式披露，並在今年2月如期完成。三次減持，時間跨度不到一年，但每一步都踩在公告窗口期內，穩步且合規地推進。在電商天下看來，這恰恰說明減持不是倉促撤退，而是有計畫地回籠資金、降低持股權重。從結果來看，阿里在這一輪操作中累計回籠資金超過3.6億元，目前對三江購物的持股比例降至不足三成。這也就意味著，它依然保留重要股東身份，但已不再承擔“戰略繫結”的角色。值得注意的是，三江購物並非阿里減持的個例。近年來，阿里密集出售了銀泰百貨、高鑫零售、紅星美凱龍等多個傳統零售資產；而對於個別非核心參股公司如圓通速遞、漱玉平民等，阿里也選擇了戰略撤退。圖源：圓通速遞單以圓通為例，阿里旗下杭州灝月企業就在一年時間內完成了兩次減持，套現約19.36億元——第二次套現超10億，才剛於1月19日被圓通公示。據不完全統計，2025年以來，阿里通過出售、減持非核心資產等方式，已累計套現超過300億元。電商天下必須指出，這一系列操作背後並非缺錢，而是戰略主動轉向的訊號：阿里，正在從“多點開花”走向“聚焦核心”，從“資產堆砌”走向“精益營運”。而且，阿里並沒有對這些“告別動作”遮掩太多。財報會上，蔡崇信就直言：“傳統實體零售不再是我們的核心關注點，退出是合理之選。”一句話，道出了阿里對“輕資產化”的堅定態度，也為後續的戰略轉向埋下了伏筆。02 阿里的主戰場，已經明確如果只看到“減持”“剝離”“退出”，很容易誤判阿里是在收縮戰線；但事實上，阿里並沒有變得保守，而是變得更集中。過去一年，阿里的資源投向已經非常明確：即時零售與AI，正在成為新的戰略軸心。一方面，即時零售被提升到前所未有的高度。圖源：淘寶閃購過去一年，我們可以看見，業務入口升級、流量資源傾斜、補貼規模擴大、體系協同加速，幾乎是在同步進行：從餓了麼正式併入淘寶閃購，再到多業務線資源打通，阿里正在建構一個覆蓋“吃住行遊購”的大消費服務網路。在電商天下看來，這是一種典型的平台化打法。不再執著於單一業態的控制權，而是通過技術、流量與履約能力，成為整個即時消費生態的中樞。另一方面，AI被明確列為下一階段的核心變數。無論是通用模型，還是面向具體場景的應用產品，阿里都在試圖將AI從“技術能力”，轉化為“使用者工具”。圖源：千問而且，就在近日，通義千問APP才以“30億春節免單”活動風靡全網。一個“千問請你喝奶茶”的免單活動，將千問與淘寶閃購打通，讓千問登頂App Store免費榜首，大量湧入的使用者甚至一度讓點單系統崩潰。我們必須承認，當AI開始真正走向大眾市場，對算力、資料、資金的長期投入，就成為不可迴避的現實需求。圖源：千問更何況，這場“技術押注”已經開始在財報上顯現成果。2026財年第二季度財報顯示，阿里總營收同比增長5%，其中AI驅動的雲端運算業務表現突出，季度營收增長高達34%，AI相關產品實現9個季度連續三位數增長。可以說，從出售線下資產到加碼AI，阿里已完成了戰略重心的徹底遷移，而轉向的回報，市場有目共睹。03 阿里下一階段能跑多遠？站在今天回望，阿里的這場“戰略轉向”並不是被動求生，而是為“再出發”打地基。從新零售“去重”到AI“加碼”，從舊生態的清理到新體系的搭建，這一整套組合拳，構成了阿里重塑自我、輕裝上陣的樣本路徑。更重要的是，阿里的“斷舍離”並未傷筋動骨。無論是對圓通、三江購物的減持，還是對銀泰、高鑫的出售，阿里仍保留一定持股比例或合作介面。這種“有控制的退出”既避免傷害生態夥伴，也為日後潛在協同留下空間。在電商天下看來，阿里式的“瘦身計畫”或許也會成為未來一批平台型企業的共同選擇。在AI成為新一輪流量與基礎設施入口的大背景下，傳統電商、線下零售、平台化營運將逐步讓位於算力、模型、晶片的全端式競爭。接下來，不再只是比拚誰的門店多、版圖大，而是誰能在即時零售和AI重構的賽道上，更快形成規模效應、更深建立使用者黏性。在這樣的新周期裡，卸下包袱、輕裝前行，或許才是通往下一個高品質增長階段的必要動作。在電商天下看來，三江購物乃至300億的套現，都只是這場戰略轉向中的清晰註腳，真正重要的是，阿里已經邁出第一步，並且展示出了轉型的勇氣和執行力。而它是否能真正從“多而重”走向“專而強”，是否能在AI時代重新奪回話語權，我們不妨拭目以待。 (電商天下)