黄仁勋「拯救」韩国股市:锁定SK海力士内存,芯片短缺还将持续

marsbitPublicado em 2026-06-08Última atualização em 2026-06-08

Resumo

6月5日,韩国股市经历“黑色星期五”,KOSPI指数暴跌。6月8日盘中跌幅一度超8%,触发熔断,三星与SK海力士股价重挫。在此市场动荡之际,英伟达CEO黄仁勋的访韩起到了“救市”作用。 6月7日晚,黄仁勋与SK集团及SK海力士高层共进晚餐。餐后,他确认英伟达新款Vera CPU将采用SK海力士的DRAM,双方正为下半年及明年的“超大规模合作”做准备,并预计当前的内存芯片短缺将持续数年。随后,双方官宣了一份多年期技术合作协议,涉及AI超级计算机、机器人、数字孪生和半导体制造等领域。黄仁勋更公开表示,当前AI公司股价极具吸引力。 合作具体包括:英伟达将Vera CPU等产品线的内存供应锚定SK海力士,共同开发下一代内存以支持AI基础设施。同时,SK海力士将利用英伟达的CUDA-X库和AI技术加速其芯片设计与制造流程,如计算光刻和构建晶圆厂数字孪生,目标是实现完全自主的工厂运营。 此次合作早有铺垫。2025年10月,双方曾宣布在韩国合作建设大型AI工厂。尽管与SK海力士深化绑定,但英伟达在下一代HBM4内存上并未独家采购,三星、SK海力士和美光均已获得认证并将共同供货。黄仁勋指出,由于AI工厂建设导致需求激增,从晶圆到封装等整个半导体供应链的短缺状况预计将持续数年。 黄仁勋此行还会见了现代、LG、三星等韩国企业,表明英伟达正系统性深化与韩国科技产业的联系,SK集团是其中的关键一环。

6 月 5 日,韩国股市遭遇「黑色星期五」,KOSPI 指数收盘暴跌 5.54%。6 月 8 日开盘后,盘中跌幅一度扩大至 8% 以上,交易平台触发熔断机制,三星与 SK 海力士双双跌近 10%。

市场风声鹤唳之际,黄仁勋的到访戏剧性地承担了「救市」的角色。

此前,韩国当地时间 6 月 7 日周日晚,黄仁勋与 SK 集团会长崔泰源、SK 海力士 CEO 郭鲁正等举办「晚餐会」。

餐后,黄仁勋向现场媒体确认:英伟达新推出的 Vera CPU 将采用 SK 海力士 DRAM;双方正在为今年下半年和明年「超大规模的合作」做准备;关于眼下的内存芯片短缺,他认为会持续好几年。

随后,英伟达与 SK 海力士官宣一份多年期技术合作协议,涉及 AI 超级计算机延伸到机器人、数字孪生和半导体制造。

黄仁勋更是在新闻发布会上直接吹票,「如果你是 AI 公司股东,你会感到高兴,目前它们的价格非常低。」

01 锁定 SK 海力士内存

Vera 是英伟达推出的首款独立数据中心专用 CPU,它所面对的竞争对手包括英特尔至强产品线、AMD 的 Epyc 芯片,以及亚马逊 Graviton 这类大型云服务商的自研项目。

在这个新战场上,英伟达从一开始就将内存供应锚定在 SK 海力士身上。

6 月 7 日,英伟达与 SK 海力士正式宣布建立多年期技术合作伙伴关系,围绕英伟达 AI 基础设施路线图,共同开发与之匹配的下一代内存。

据了解,双方合作覆盖了英伟达 Vera Rubin AI 超级计算机、Vera CPU、搭载 RTX Spark 的 PC 以及 Jetson Thor 机器人计算平台等一系列个人端侧与云端的产品线。

公告指出,合作旨在保障先进内存的供应,以应对此类产品较长的开发周期、复杂的制造工艺和高额的资本投入,进而支撑全球 AI 工厂的持续建设。

公告还列举了 SK 海力士将拓展进入英伟达正在开创的几个新市场,包括 AI 基础设施、个人 AI 和物理 AI。

02 AI 反哺芯片制造

除了供应内存,SK 海力士开始把英伟达的 AI 技术用到自芯片设计和制造流程里。

类似的合作此前已经在台积电落地,最典型的就是「计算光刻」。

根据公告,SK 海力士正在使用英伟达 CUDA-X 库和 AI 来加速半导体仿真,覆盖技术计算机辅助设计(TCAD)和计算光刻等环节。

双方还在推动将这些工具扩展到半导体电子设计自动化(EDA)和仿真生态系统中去,为芯片制造商、英伟达和 EDA 软件供应商之间的三方合作铺路。

这意味着,双方的合作已经不限于 SK 海力士自用,而是在摸索一套可以推向整个半导体行业的模式。

在制造环节,SK 海力士正在推进晶圆厂数字孪生功能的开发,目标是实现完全自主的工厂运营。这项工作的基础是英伟达 Omniverse 平台。借助 Omniverse 库和 OpenUSD 流程,SK 海力士得以构建 3D 工厂场景,用于可视化、模拟和优化复杂的半导体制造环境。

在工厂运营层面,这些数字孪生功能还能接入英伟达 cuOpt 决策优化引擎和 Metropolis 平台,用来调度自主移动机器人和晶圆厂内的其他资产。

公告还透露,两家公司正在探索将数字孪生与现有传统软件及智能体 AI 工作流打通,让 AI 系统能够基于晶圆厂数据进行推理、自动执行任务并改进制造决策。

03 提前半年铺垫

2025 年 10 月,英伟达和 SK 海力士就宣布过一项大规模的基础设施合作。

彼时,SK 集团正在建设一座配备超过 5 万颗英伟达 GPU 的 AI 工厂,第一阶段计划于 2027 年底前完工。建成后,这预计将成为韩国规模最大的 AI 工厂之一。

该工厂采用「GPU 即服务」的模式,面向 SK 集团旗下子公司及外部组织开放,目标是加速韩国产业的数字化转型和工业创新。

SK 电讯也在其中承担具体的部署工作。

作为英伟达的云合作伙伴,SK 电讯计划在亚洲使用英伟达 RTX PRO 6000 Blackwell 服务器版 GPU 构建工业 AI 云,初始部署规模超过 2000 颗 GPU,专门运行 Omniverse 工作负载,为 SK 海力士的半导体制造、晶圆厂数字孪生和内部 AI 智能体提供算力支撑。

黄仁勋在此次访韩期间还透露了一个信息:他正在与电信公司展开讨论,因为未来的 AI 将使用电信网络。这与 SK 电讯参与合作的方向是一致的。

04 三家分走 HBM4 订单

尽管英伟达与 SK 海力士签下了多年技术合作协议,但在 HBM4 的供应上,英伟达并没有把鸡蛋放在一个篮子里。

黄仁勋在抵达首尔时就向记者明确表态:「三家供应商均已获得资格认证。三家供应商都已投入生产,他们都在竞相支持 Vera Rubin。」

这三家供应商对应三星电子、SK 海力士和美光科技。

在台北电脑展的主题演讲中,黄仁勋已确认 Vera Rubin 全面投产,计划于今年第三季度交付。这套系统围绕英伟达 Vera CPU 和 Rubin 图形核心集群构建,每个服务器机架系统都搭配 TB 级别的 HBM4 内存。

从 HBM4 的实际进度来看,SK 海力士仍处于领跑位置。

路透社去年 9 月曾报道,SK 海力士当时已完成 HBM4 芯片的内部认证,并为客户建立了生产体系,目标是在 2025 年下半年完成 12 层 HBM4 产品的量产准备。Meritz 证券高级分析师 Kim Sunwoo 当时预测,得益于向关键客户的早期供应以及由此带来的先发优势,SK 海力士 2026 年的 HBM 市场份额将保持在六成出头。

05 芯片短缺将持续数年

在 HBM4 供应上铺开三家的局面,并不意味着供应压力就此缓解。

黄仁勋在周日晚餐后,还给出了一个不太乐观的判断。他对现场媒体表示,存储芯片短缺不会很快结束。「整个行业供应链——从晶圆到封装再到硅光子......一切都在短缺,因为需求如此之高。这种情况将持续数年。」

这段表态的背景,是全球 AI 工厂建设对先进内存近乎无休止的消耗。

黄仁勋所说的短缺,不是缺某一种料,而是产业链上几乎每个环节都供应紧张。英伟达推出 Vera Rubin、推广 AI 工厂、进入个人 AI 和物理 AI 领域,每一件事都在拉高对内存的需求。这也是为什么他会说三家 HBM4 供应商都在竞相支持 Vera Rubin。

没人愿意在供不应求的局面里掉队。

这次韩国之行,SK 集团虽然是重点,但并不是黄仁勋全部日程。他在抵达时曾透露,自己安排了与现代汽车、LG、SK、三星以及 Naver 的会议。他还透露,英伟达正在积极为韩国新研发中心招聘人员。从这些动向来看,英伟达正在系统性地加深与整个韩国科技产业的联系,SK 集团是其中的关键一环,但并非唯一一环。

Perguntas relacionadas

Q韩国股市在6月5日至8日期间经历了什么?

A韩国股市遭遇了剧烈下跌。6月5日,KOSPI指数收盘暴跌5.54%,市场称之为“黑色星期五”。6月8日开盘后,盘中跌幅一度扩大至8%以上,触发了交易平台的熔断机制,三星与SK海力士股价双双跌近10%。

Q英伟达与SK海力士达成了怎样的核心合作?

A双方宣布了一份多年期的技术合作协议。核心内容包括:1) 英伟达新推出的Vera CPU将采用SK海力士的DRAM;2) 双方将为今年下半年和明年的“超大规模合作”做准备;3) 合作范围覆盖AI超级计算机、机器人、数字孪生和半导体制造,旨在共同开发下一代内存以支持英伟达的AI基础设施路线图。

Q黄仁勋对芯片短缺问题有何判断?

A黄仁勋认为存储芯片的短缺情况将持续数年。他指出,整个行业的供应链——从晶圆到封装再到硅光子技术等几乎所有环节都处于短缺状态,原因是市场需求异常高涨。这意味着芯片供应紧张的局面不会很快结束。

QSK海力士如何将英伟达的AI技术应用于自身业务?

ASK海力士正将英伟达的AI技术应用于其芯片设计和制造流程中。具体包括:1) 使用英伟达CUDA-X库和AI加速半导体仿真,覆盖技术计算机辅助设计和计算光刻等环节;2) 利用英伟达Omniverse平台开发晶圆厂数字孪生功能,目标是实现完全自主的工厂运营,并可调度自主移动机器人;3) 探索打通数字孪生与现有软件及AI工作流,以改进制造决策。

Q在下一代HBM4内存的供应上,英伟达采取了什么策略?

A英伟达采取了多元化供应商策略,并未将订单集中于SK海力士一家。黄仁勋确认,三星电子、SK海力士和美光科技这三家供应商均已获得英伟达HBM4的资格认证并投入生产,他们正在竞相支持英伟达的Vera Rubin AI系统。其中,SK海力士在HBM4的开发进度上仍处于领跑地位。

Leituras Relacionadas

7 Key Questions Regarding 'DeepSeek Completes Over 50 Billion Yuan in Financing'

DeepSeek, the prominent Chinese AI company, has reportedly completed its first external funding round, raising over 50 billion RMB (approximately $6.9 billion USD), according to foreign media reports. This financing values the company at over $50 billion USD, a significant increase from a $10 billion valuation when the round began in April 2026. Founder Liang Wenfeng personally contributed 20 billion RMB, with other investors including Tencent, CATL (Contemporary Amperex Technology), JD.com, NetEase, and IDG Capital. Notably, the National Artificial Intelligence Industry Investment Fund also participated with a direct investment of 1 billion RMB. The funding round is distinctive for its structure. Most investors' capital is channeled into a limited partnership controlled by CEO Liang Wenfeng, rather than directly into DeepSeek. These investors face a five-year lock-up period, have no voting rights, but receive priority financial information and future investment rights. This arrangement is seen as a measure to ensure Liang's absolute control over the company's direction, aligning with its stated principles of prioritizing groundbreaking AGI (Artificial General Intelligence) research over short-term profitability. The report highlights strategic reasons for key investors. Tencent's involvement is viewed as a strategic alignment, building on existing collaborations. CATL's investment is linked to the critical energy and power infrastructure needed for AI data centers, a key growth area for the battery giant. The participation of a state-backed fund underscores the national strategic importance of AI development. Post-funding, DeepSeek is expected to accelerate infrastructure development, including building its own data centers, hire more talent, and release new model versions with enhanced capabilities like image and audio support. While this marks a major step, the company's long-term journey towards AGI remains a challenging path that will require balancing its foundational ideals with the expectations of its new investors.

marsbitHá 5m

7 Key Questions Regarding 'DeepSeek Completes Over 50 Billion Yuan in Financing'

marsbitHá 5m

The World Cup is Here: The Battle for Entry into Prediction Markets Has Begun

The 2026 FIFA World Cup has begun, and alongside the on-field competition, a new off-field battleground is emerging: prediction markets. These blockchain-based platforms, which convert crowd wisdom into tradable probabilities, are gaining significant traction. However, their complexity—involving wallets, gas fees, and smart contracts—has historically limited participation to crypto-native users. Centralized exchanges (CEXs), like Gate, are tackling this adoption barrier. By integrating with leading prediction market protocol Polymarket, Gate simplifies the user experience. Users can participate directly with their exchange account and USDT, bypassing complex Web3 steps. Gate offers a streamlined "Prediction Mode" for casual users and a professional "Trading Mode" with advanced tools. Key features include two-way trading (allowing users to buy or sell positions before event resolution), support for diverse markets (sports, crypto, macroeconomics), and a suite of information tools like a "Smart Money" leaderboard, wallet tracking, and AI-powered insights. For the World Cup, Gate launched a dedicated hub aggregating schedules, standings, and relevant markets. This allows fans to seamlessly follow games and trade on outcomes, transforming passive viewing into active participation where they can monetize their predictions and trade on shifting consensus throughout a match. The article argues that prediction markets have proven their value in event forecasting. The next challenge is mass adoption. The competition is shifting from building effective protocols to creating accessible user entry points. By lowering technical barriers and building a complete ecosystem for information and trading, platforms like Gate aim to transition prediction markets from a niche crypto tool to a mainstream platform for expressing and trading on collective intelligence.

Odaily星球日报Há 15m

The World Cup is Here: The Battle for Entry into Prediction Markets Has Begun

Odaily星球日报Há 15m

Spain Held to a Draw by Cape Verde, Jucom Prediction Market Witnesses Historic Upset

In a major upset at the 2026 FIFA World Cup, tournament favorites Spain were held to a surprising 0-0 draw by debutants Cape Verde in their Group H opener on June 16, Beijing time. Despite dominating possession (74%) and recording 27 shots with an expected goals figure of 2.16, Spain failed to break down a resilient Cape Verde defense, with their 40-year-old goalkeeper Vozinha making 7 saves to earn Man of the Match. Pre-match predictions on the Jucom prediction market had heavily favored Spain, assigning them a 92% win probability. The actual result, a goalless draw, triggered significant volatility across related prediction markets. This outcome forces a market-wide reassessment of several key probabilities, including Spain's likelihood of winning the group and the tournament itself, while Cape Verde's previously near-zero chance of advancing is now being re-evaluated. The event highlights both the efficiency and the inherent limitations of prediction markets. While prices aggregate known information, football's low-scoring, high-variance nature means unquantifiable in-game factors can lead to unlikely results. The core value of such markets lies not in perfect foresight but in their ability to dynamically reflect how new information is incorporated into collective expectations. Platforms like Jucom, which track outcomes from single matches to the final champion, provide a real-time lens into how global consensus evolves with each game.

链捕手Há 19m

Spain Held to a Draw by Cape Verde, Jucom Prediction Market Witnesses Historic Upset

链捕手Há 19m

Trading

Spot
Futuros

Artigos em Destaque

Como comprar CHIP

Bem-vindo à HTX.com!Tornámos a compra de USD.AI (CHIP) simples e conveniente.Segue o nosso guia passo a passo para iniciar a tua jornada no mundo das criptos.Passo 1: cria a tua conta HTXUtiliza o teu e-mail ou número de telefone para te inscreveres numa conta gratuita na HTX.Desfruta de um processo de inscrição sem complicações e desbloqueia todas as funcionalidades.Obter a minha contaPasso 2: vai para Comprar Cripto e escolhe o teu método de pagamentoCartão de crédito/débito: usa o teu visa ou mastercard para comprar USD.AI (CHIP) instantaneamente.Saldo: usa os fundos da tua conta HTX para transacionar sem problemas.Terceiros: adicionamos métodos de pagamento populares, como Google Pay e Apple Pay, para aumentar a conveniência.P2P: transaciona diretamente com outros utilizadores na HTX.Mercado de balcão (OTC): oferecemos serviços personalizados e taxas de câmbio competitivas para os traders.Passo 3: armazena teu USD.AI (CHIP)Depois de comprar o teu USD.AI (CHIP), armazena-o na tua conta HTX.Alternativamente, podes enviá-lo para outro lugar através de transferência blockchain ou usá-lo para transacionar outras criptomoedas.Passo 4: transaciona USD.AI (CHIP)Transaciona facilmente USD.AI (CHIP) no mercado à vista da HTX.Acede simplesmente à tua conta, seleciona o teu par de trading, executa as tuas transações e monitoriza em tempo real.Oferecemos uma experiência de fácil utilização tanto para principiantes como para traders experientes.

348 Visualizações TotaisPublicado em {updateTime}Atualizado em 2026.06.02

Como comprar CHIP

Discussões

Bem-vindo à Comunidade HTX. Aqui, pode manter-se informado sobre os mais recentes desenvolvimentos da plataforma e obter acesso a análises profissionais de mercado. As opiniões dos utilizadores sobre o preço de CHIP (CHIP) são apresentadas abaixo.

活动图片