[新聞] 輝達50億鎂押注intel，建立晶片合作關係

原文標題： Nvidia bets big on Intel with $5 billion stake and chip partnership 原文連結：https://tinyurl.com/42uefatb 發布時間：September 18, 2025 7:52 PM GMT+8 記者署名： Reporting by Stephen Nellis, Jeffery Dastin and Max Cherney in San Francisco; Editing by Sam Holmes 原文內容： SAN FRANCISCO, Sept 18 (Reuters) - Nvidia (NVDA.O) on Thursday said it will invest $5 billion in Intel (INTC.O), throwing its heft behind the struggling U .S. chipmaker just weeks after the White House engineered an extraordinary deal for the U.S. government to take a massive stake in the company. 舊金山，9 月 18 日（路透社）——輝達（Nvidia, NVDA.O）週四表示，將投資 50 億美 元於英特爾（Intel, INTC.O），在這家陷入困境的美國晶片製造商背後下注。這距離白 宮策劃美國政府取得該公司大量股權的特殊交易僅幾週時間。 The stake instantly will make Nvidia one of Intel's largest shareholders, giving it roughly 4% or more of the company after new shares are issued to complete the deal. Nvidia's support represents a new opening for Intel after years of turnaround efforts at the famed U.S. manufacturer failed to pay off, and it sent shares soaring 30% in premarket action. 此舉將立即使輝達成為英特爾最大股東之一，交易完成並發行新股後，輝達將持有約 4% 以上股份。輝達的支持為英特爾帶來新的轉機，此前多年重整努力未見成效，而該消息也 使英特爾盤前股價飆升 30%。 The company – once the chip industry’s flagbearer that claimed to put the " silicon" in Silicon Valley – appointed a new CEO, Lip-Bu Tan, in March. He came under fire from U.S. elected officials including U.S. President Trump, who called for him to resign due to concerns about his connections with China. That led to a swiftly arranged meeting in Washington that ended with Intel's unusual arrangement to give the United States a 10% stake in the company. 這家曾被譽為「把矽帶進矽谷」的晶片業旗手，今年 3 月任命新執行長陳立武（Lip-Bu Tan）。他因與中國的關係而受到美國政界（包括總統川普）的批評，川普甚至呼籲他下 台。這導致華府緊急召開會議，最終促成英特爾將 10% 股權交給美國政府的特殊安排。 The pact includes a plan for Intel and Nvidia to jointly develop PC and data center chips, but crucially will not involve Intel’s contract manufacturing business, known as a "foundry" in the chip industry, making chips for Nvidia. Most analysts believe that for Intel’s foundry to survive, it would need to eventually win a large customer such as Nvidia, Apple, Qualcomm or Broadcom. 協議包括英特爾與輝達共同開發 PC 與資料中心晶片，但關鍵在於，不涉及英特爾的代工 業務（業界稱為 foundry）為輝達製造晶片。大多數分析師認為，若英特爾的代工業務要 存活，終將需要贏得像輝達、蘋果、高通或博通這樣的大客戶。 Nvidia, whose must-have chips are powering a global artificial intelligence boom, said in a statement it will pay $23.28 per share for Intel common stock, a price slightly below the $24.90 at which Intel shares closed on Wednesday. 輝達表示，其 AI 晶片正推動全球人工智慧浪潮，公司將以每股 23.28 美元收購英特爾 普通股，略低於週三英特爾收盤價 24.90 美元。 However, that is higher than the $20.47 price per share that the United States government paid for an extraordinary 10% stake it took in Intel last month. 然而，這一價格高於美國政府上個月為取得 10% 股權所支付的每股 20.47 美元。 "It's a reflection of Nvidia looking to diversify to an extent its investment within the US and as well to gain some brownie points with the U.S. government ," said Chris Beauchamp, chief market analyst at IG Group in London. 倫敦 IG 集團首席市場分析師 Chris Beauchamp 表示：「這反映了輝達希望在美國多元 化其投資，同時也藉此在美國政府面前加分。」 The pact represents a potential risk to Taiwan’s TSMC (2330.TW). TSMC currently manufactures Nvidia’s flagship processors, business that the world ’s most valuable company could one day extend to Intel. AMD (AMD.O), which competes with Intel for supplying chips to data centers, also stands to lose thanks to Nvidia’s backing. 這項協議對台灣的台積電（TSMC, 2330.TW）構成潛在風險。台積電目前為輝達製造旗艦 處理器，而這門生意未來可能部分轉移至英特爾。與英特爾競爭資料中心晶片的 AMD（ AMD.O）也因輝達的支持而可能受損。 Shares of Nvidia rose more than 3%. AMD slipped nearly 4%, while U.S.-listed shares of TSMC slid 2%. TSMC and AMD didn’t immediately respond to a comment request. 輝達股價上漲逾 3%。AMD 下跌近 4%，台積電美國掛牌股票下跌 2%。台積電與 AMD 尚未 回應評論請求。 Tan has vowed to make Intel’s operation lean and build factory capacity only when there’s demand to match it. 陳立武承諾將精簡英特爾營運，並只在需求匹配時才擴建產能。 The deal adds to a growing reserve of capital that Intel has accumulated weeks after it announced a $2 billion investment from Softbank and received $5.7 billion from the U.S. government. David Zinsner, Intel’s chief financial officer, told investors at a Deutsche Bank conference last month that the company was in a "good cash position" and would not require much more capital until it saw significant demand for 14A, a next-generation manufacturing process that it expects to invest heavily in building. 該交易為英特爾不斷累積的資本再添一筆。幾週前，英特爾已獲得軟銀 20 億美元投資， 以及美國政府 57 億美元資金。英特爾 CFO David Zinsner 上月在德銀會議上對投資者 表示，公司目前「現金狀況良好」，在看到對下一代製程 14A 的重大需求之前，不需要 更多資本。 Under the deal announced Thursday, Intel is planning to design custom data center central processors that Nvidia will package with its AI chips, known as GPUs. A proprietary Nvidia technology will let the Intel and Nvidia chips communicate at higher speeds than before. 根據週四宣布的協議，英特爾計劃設計客製化資料中心 CPU，並由輝達將其與 GPU（AI 晶片）進行封裝。輝達的專有技術將讓英特爾與輝達晶片之間的通訊速度更快。 Those speedy links are a key differentiator in the AI market because many chips must be strung together to act as one to chew through massive amounts of data. 這些高速連結是 AI 市場的關鍵差異點，因為大量晶片必須串聯運作，才能處理龐大數據 。 At present, Nvidia's best-selling AI servers with those speedy links are only available using Nvidia’s own chips, but the deal would now put Intel on equal footing, giving it a chance to make money off each Nvidia server. The combined Nvidia-Intel chips could provide a major competitive challenge to AMD , which is developing its own AI servers, and Broadcom (AVGO.O), which also has chip-to-chip connection technology and helps companies such as Google develop AI chips. 目前，輝達最暢銷、配備這些高速連結的 AI 伺服器僅使用自家晶片。但此協議將使英特 爾站上同一起跑線，並有機會從每台輝達伺服器中獲利。這些 Nvidia-Intel 聯合晶片將 對 AMD（正在開發自家 AI 伺服器）與 Broadcom（擁有晶片互連技術並協助 Google 等 公司開發 AI 晶片）構成重大挑戰。 "Anything that NVIDIA decides to endorse just by association will make that stock of Intel appear attractive because it implies that Nvidia sees value in Intel," said Peter Andersen, founder of Andersen Capital Management in Boston. 波士頓 Andersen 資本管理公司創辦人 Peter Andersen 表示：「只要輝達選擇支持的對 象，就會使該公司股票看起來具吸引力，因為這意味著輝達認為它有價值。」 For consumer markets, Nvidia will provide Intel with a custom graphics chip that Intel can package with its PC central processors with the same speedy links, potentially giving it an edge against rivals such as AMD. 在消費市場方面，輝達將提供英特爾客製化的顯示晶片，英特爾可將其與 PC CPU 一起封 裝，並具備相同的高速連結，可能讓英特爾在與 AMD 的競爭中取得優勢。 While Intel’s x86 computing architecture has lost ground in both data centers and PCs to chips with technology from Arm Ltd, it still has majority market share. 雖然英特爾的 x86 架構在資料中心與 PC 市場逐漸輸給 Arm 技術的晶片，但仍保持多數 市場份額。 "This historic collaboration tightly couples Nvidia's AI and accelerated computing stack with Intel’s CPUs and the vast x86 ecosystem—a fusion of two world-class platforms," Nvidia CEO Jensen Huang said in a press release. " Together, we will expand our ecosystems and lay the foundation for the next era of computing." 輝達執行長黃仁勳在新聞稿中表示：「這項歷史性合作將輝達的 AI 與加速運算堆疊，與 英特爾 CPU 及龐大的 x86 生態系緊密結合——這是兩大世界級平台的融合。我們將共同 擴展生態系，為下一個運算時代奠基。」 The two companies did not disclose financial terms of their technical collaboration but said they would make "multiple generations" of future products. Nvidia and Intel officials described the collaboration as a commercial arrangement under which they will provide chips to one another to create products and said there was no licensing component to the deal. 兩家公司未披露技術合作的財務條款，但表示將推出「多代」未來產品。輝達與英特爾官 員將此合作描述為一種商業安排，雙方將互相提供晶片來共同創建產品，並表示協議中不 涉及授權。 The companies declined to give a date for when the first joint products will come to market but said that their product plans prior to the joint deal have not changed. 雙方未透露第一款聯合產品上市日期，但表示此前的產品計畫未受影響。 Nvidia in recent years has entered both the PC central processor market and the data center central processor market. Intel meanwhile has tried to sell several AI chips that compete with Nvidia and has said it plans to develop an AI data center server that would compete with Nvidia. 近年來，輝達已進軍 PC 與資料中心 CPU 市場；同時，英特爾也嘗試銷售多款與輝達競 爭的 AI 晶片，並表示計劃開發可與輝達競爭的 AI 資料中心伺服器。 心得/評論： Nvidia 投資Intel是 政治 + 技術 + 商業的三重策略： 政治上取悅美國政府，減少政策風險； 技術上培養第二供應鏈，避免台積電單點風險； 商業上與 Intel 綁定，鞏固 AI 平台優勢，壓縮 AMD 與 Broadcom 的生存空間。