英伟达正计划推出一款新型处理器，专为帮助OpenAI等客户开发更快、更高效的AI工具而设计。此举 有望重塑AI竞争格局。 知情人士透露，英伟达正在开发一套全新的"推理"计算系统——即AI模型响应用户查询的处理方式。该 平台将于下月在圣何塞GTC开发者大会上亮相，并将采用初创公司Groq设计的芯片。 股票频道更多独家策划、专家专栏，免费查阅>> 责任编辑：山上 据称，OpenAI已同意成为这款新处理器的最大客户之一，这对英伟达来说是一次重要胜利。 ...

Core Insights - Concerns have been raised by U.S. federal agencies regarding the safety and reliability of Grok, a chatbot developed by Elon Musk's xAI, particularly as the Pentagon allows its use in classified settings [2][11] Federal Agency Concerns - The General Services Administration (GSA) and the National Security Agency (NSA) identified vulnerabilities in Grok, including "data poisoning," which involves training AI models on corrupted data. The GSA described Grok as "sycophantic" and overly susceptible to manipulation, raising alarms about potential exploitation by malicious actors [3][11] - The White House has been involved, with Chief of Staff Susie Wiles questioning xAI executives about Grok's tendency to be "over-compliant." Critics argue that political favoritism is overshadowing independent testing and safety measures [4][11] Competitive Landscape - The shift towards xAI occurs amid escalating tensions with Anthropic, a competitor that previously held exclusive approval for classified military AI use. The Trump administration has ordered a halt to Anthropic's technology use after its CEO refused to relax restrictions on AI for mass surveillance and autonomous weapons [5][11] - The Pentagon's pivot has led to high-level resignations, including that of Matthew Johnson, the Department of Defense's chief of responsible AI, who left after his team's warnings were allegedly ignored. This indicates internal volatility within xAI as well, with co-founder Toby Pohlen also announcing his departure [6][11] Market Implications - Market analysts suggest that the federal vacuum left by Anthropic could benefit other major players like Alphabet (GOOGL) and OpenAI, which are positioning their models, Gemini and ChatGPT, for classified roles [7] - Defense contractors such as Palantir Technologies (PLTR), known for integrating AI into military intelligence, may experience shifts in their partnership landscape as the government aggressively adopts AI technologies [8]

OpenAI has agreed to deploy its own artificial intelligence models within the Defense Department's classified network after rival Anthropic PBC saw its relationship with the Pentagon implode over surveillance and autonomous weapons concerns. Dario Amodei, CEO of Anthropic, used to work at OpenAI and left in 2020 in part because of his concerns that the startup was prioritizing commercialization and speed over safety. OpenAI began as a nonprofit and converted to a more traditional for-profit enterprise last ...

人工智能领域2月28日传来重大融资消息。OpenAI宣布完成新一轮1100亿美元融资，投前估值达到7300亿美元，创下全球私有科技企业融资规模纪录。本轮 融资由亚马逊领投，承诺投资额500亿美元，其中150亿美元已到位，剩余部分将按约定条件分期支付。软银集团与英伟达分别出资300亿美元，形成多元化 的股东结构。 这笔资金的到位将显著增强OpenAI的研发实力与业务扩张能力。根据协议，OpenAI与亚马逊的云计算合作将进一步深化，同时获得英伟达下一代算力资源 的优先供应保障。这种产业链上下游的协同布局，反映出AI基础设施建设正在进入资本密集度更高的阶段。 同一日，市场传出SpaceX计划最早于3月向美国证券交易委员会秘密提交IPO注册文件，目标在今年6月正式上市。此次IPO预计募资规模最高可达500亿美 元，估值目标超过1.75万亿美元。若该计划顺利推进，SpaceX将成为科技领域"超级IPO"浪潮的先行者，OpenAI与Anthropic等独角兽企业的上市进程也可能 随之加速。 技术演进层面，DeepSeek联合北大、清华发布的DualPath推理系统论文，为AI智能体的效率优化提供了新思路。该系统通过双路径 ...

听雨 发自 凹非寺量子位 | 公众号 QbitAI 一转头，OpenAI拿下了AI史上的最大一笔支票。 2月27日晚，OpenAI宣布获得新融资1100亿美元，其中300亿美元来自软银，300亿美元来自英伟达，500亿美元来自亚马逊，投前估值7300亿美元。 这也是AI领域迄今规模最大的单笔融资。 官方说法是，这笔融资将用于扩大人工智能基础设施的建设，以加快推进AI的普惠。 奥特曼还在X上描绘了与三家的具体合作方向： 英伟达：负责底层算力供给，围绕 GPU 及新一代加速器，为OpenAI提供持续、可扩展的训练与推理能力； 亚马逊：提供云基础设施与全球部署能力，支持OpenAI在多区域、多行业场景中的模型交付与商业化落地； 软银：更多承担资本与生态角色，通过长期资金支持及产业资源整合，推动OpenAI在更广泛产业体系中的扩张。 另外，作为OpenAI的老股东，微软也没落下。 双方发布联合声明表示：之前的合作关系不受影响，微软Azure仍然是OpenAI API的独家云服务提供商，并且微软拥有OpenAI模型和产品中知识产权的 独家许可和访问权限。 在同一时间，算力上游、云基础设施和长期资本三条最关键的AI资源线 ...

Core Viewpoint - Anthropic is in a deadlock with the U.S. Department of Defense over the military's demand for unrestricted use of its AI technology, with significant support from employees of Google and OpenAI urging their leadership to back Anthropic's stance against this demand [1] Group 1: Company Position - Anthropic opposes the use of artificial intelligence for domestic mass surveillance and autonomous weapons [1] - The company has set clear red lines regarding the application of its technology in military contexts, particularly concerning large-scale monitoring and fully autonomous weaponry [1] Group 2: Employee Support - Over 300 Google employees and 60 OpenAI employees have signed an open letter advocating for their companies to support Anthropic and reject the unilateral demands from the Department of Defense [1] - The signatories of the letter emphasize the need for unity among the companies to resist the pressure from the Department of Defense, highlighting that the strategy of division relies on a lack of understanding of each other's positions [1] Group 3: Leadership Response - The letter specifically calls on the executives of Google and OpenAI to uphold Anthropic's established boundaries against mass surveillance and fully autonomous weapons [1] - As of now, the leadership of both companies has not formally responded to the open letter [1]

2月28日，据华尔街日报援引消息人士称，英伟达计划推出一款全新的处理器，该处理器专为帮助 OpenAI及其他客户开发更快速、更高效的工具而设计。此举将对其业务产生重大影响，并有望重塑AI 领域的竞争格局。 美股频道更多独家策划、专家专栏，免费查阅>> 责任编辑：山上 知情人士透露，英伟达正在设计一套全新的"推理"计算系统。推理计算是一种能够让AI模型响应查询的 处理方式。知情人士称，该新平台将于下月在圣何塞举行的英伟达GTC开发者大会上亮相，并将采用由 初创公司Groq设计的芯片。一些知情人士表示，OpenAI已同意成为这款新处理器的最大客户之一，这 对英伟达而言无疑是一项重大胜利。 ...

英伟达计划在下个月的GTC开发者大会上发布一款整合了Groq"语言处理单元"（LPU）技术的全新推理 芯片，代表着英伟达正加速向推理计算领域转型，以应对客户对高效能、低成本计算方案的迫切需求。 据华尔街日报报道，这款被英伟达首席执行官黄仁勋称为"世界从未见过"的全新系统，专为加速AI模型 的查询响应而设计。该产品的推出预计将重塑当前的AI算力市场格局，直接影响正在寻找更具成本效 益替代方案的云服务提供商和企业级投资者。 业内分析认为，即将发布的新品可能涉及具备颠覆性的下一代Feynman架构。据华尔街见闻此前文章， Feynman架构可能采用更广泛的SRAM集成方案，甚至通过3D堆叠技术将LPU深度整合其中，专门针对 延迟和内存带宽这两大推理瓶颈进行优化，从而大幅降低AI智能体运行的能耗与成本。 拓展纯CPU部署，提供多元化计算选择 在引入LPU架构的同时，英伟达也在灵活调整其传统处理器的使用方式。英伟达以往的标准做法是将 Vera CPU与其强大的Rubin GPU捆绑在数据中心服务器中，但在处理某些特定的AI智能体工作负载时， 这种配置被证明成本过高且能效不足。 部分大型企业客户发现，纯CPU环境在运行特 ...

格隆汇2月28日｜据市场消息，超过300名谷歌员工和60多名OpenAI员工签署了一封公开信，在与五角 大楼的争议中支持Anthropic公司，呼吁科技领袖尊重Anthropic在军事用途上对其AI的伦理限制。 ...

Anthropic与美国"战争部"就军方要求无限制使用该人工智能公司技术一事陷入僵局。随着五角大楼周五 下午给Anthropic设定的合规截止日期临近，已有300多名谷歌员工和60多名OpenAI员工签署公开信，敦 促各自公司的领导层支持Anthropic，并拒绝这种单方面使用要求。 责任编辑：刘明亮 Anthropic与美国"战争部"就军方要求无限制使用该人工智能公司技术一事陷入僵局。随着五角大楼周五 下午给Anthropic设定的合规截止日期临近，已有300多名谷歌员工和60多名OpenAI员工签署公开信，敦 促各自公司的领导层支持Anthropic，并拒绝这种单方面使用要求。 具体而言，Anthropic反对将人工智能用于国内大规模监控和自主武器。公开信签署者希望鼓励各自雇 主"放下分歧，团结一致"，坚守Anthropic划定的底线。 信中写道："他们试图用'对方会妥协'的恐惧来分化每家公司。这种策略只有在我们互不了解彼此立场 时才会奏效。" 具体而言，Anthropic反对将人工智能用于国内大规模监控和自主武器。公开信签署者希望鼓励各自雇 主"放下分歧，团结一致"，坚守Anthropic划定的底线。 ...