公众号记得加星标⭐️，第一时间看推送不会错过。 来源 ：本文编译自CNBC 。 英伟达首席执行官黄仁勋周三表示，他对AMD公司在本周早些时候宣布的数十亿美元合作协议中， 将其10%的股份出售给OpenAI感到意外。 "考虑到他们对下一代产品如此兴奋，这真是富有想象力、独一无二且令人惊讶，"黄仁勋在接受 CNBC旗下Squawk Box节目采访时表示。"我很惊讶他们在公司尚未建成之前就放弃了10%的股份。 不过，我想，这很聪明。" 当被问及OpenAI将如何为与英伟达的交易筹集资金时，黄仁勋表示："他们还没有钱。" "他们首先要通过收入（收入正在呈指数级增长）、股权或债务来筹集资金，"黄仁勋说道，"他们给 了我们机会，让我们在时机成熟时与其他投资者一起投资。" 黄仁勋补充道，在英伟达此前投资过OpenAI之后，"唯一的遗憾就是没有追加投资。" 黄仁勋还证实了英伟达参与了xAI的最新一轮融资。据彭博社报道，埃隆·马斯克旗下的这家人工智能 初创公司正寻求筹集约200亿美元。据该媒体援引知情人士的话报道，英伟达将投资20亿美元。 OpenAI和AMD于周一达成协议，OpenAI承诺在未来几年内购买价值6千兆瓦的芯片， ...

公众号记得加星标⭐️，第一时间看推送不会错过。 来源 ： 内容编译自cnbc 。 本周，英伟达表示将向OpenAI投资1000亿美元。这笔交易凸显出自2022年生成式AI问世以来，这 家芯片制造商的投资组合已变得多么庞大。 就在这笔交易达成的一周前，英伟达刚刚承诺向其曾经的竞争对手英特尔投资50亿美元。此前， 该公司还宣布计划向自动驾驶汽车初创公司Wayve投资5亿美元，并向英国云服务提供商Nscale投 资5亿英镑（约合6.677亿美元）。 英伟达的这轮投资狂潮凸显出，这家芯片制造商已晋升至硅谷的顶端。它提供资金和其备受追捧的 人工智能芯片的使用权，以换取股权和对一些最热门AI初创公司发展方向的洞见。 如果对OpenAI的全部投资完成（预计将在未指明的年限内完成），这将是英伟达有史以来最大的 一笔投资。 英 伟 达 在 8 月 份 的 一 份 财 务 文 件 中 披 露 ， 其 持 有 的 公 开 交 易 证 券 价 值 为 43.3 亿 美 元 ， 其 中 包 括 Applied Digital、Arm、CoreWeave、Nebius Group、Recursion Pharmaceuticals和We ...

Core Insights - Nvidia is investing up to $100 billion in OpenAI and supplying millions of AI chips, raising concerns about an AI bubble [1][2] - The investment creates a closed-loop funding cycle among Nvidia, OpenAI, and Oracle, benefiting all parties involved [3][4] - There are significant risks associated with this collaboration, including OpenAI's ongoing losses and Oracle's high debt levels [7][8] Group 1 - Nvidia's investment in OpenAI is unprecedented in scale, potentially overshadowing other investments in the AI sector [2] - The partnership forms a "perpetual motion machine" where OpenAI buys cloud services from Oracle, which in turn purchases GPUs from Nvidia, creating a cycle of mutual benefit [3] - The collaboration has sparked discussions on social media about the interconnectedness of these major players in the AI space [4] Group 2 - OpenAI is valued at $100 billion but is projected to incur losses exceeding $5 billion by 2025, with annual cloud service expenses reaching $60 billion [7] - Oracle faces challenges with high debt levels, having a debt-to-equity ratio of 427%, which raises concerns about its financial stability [7] - The current AI landscape is compared to the internet bubble of 25 years ago, with analysts warning of potential irrational valuations among AI startups [8]

【文/观察者网 王一】9月18日，美国芯片巨头英伟达在其官网上发公告宣布，将投资50亿美元入股美 国芯片制造商英特尔公司，并共同开发用于个人电脑和数据中心的芯片。 英国路透社称，交易完成后，英伟达将成为英特尔最大的股东之一，这对近几年挣扎着试图扭亏为盈的 英特尔来说，是一个新的机遇。美国彭博社指出，这一出人意料的投资似乎意在帮助陷入困境的竞争对 手，曾经被英特尔边缘化的英伟达，如今不仅在人工智能（AI）计算领域占据主导地位，还反过来向 老牌巨头提供资金并带来关键技术支持。 消息公布后，英特尔股价在盘前交易中一度猛涨超30%，英伟达股价也上涨3%左右。 9月18日，英特 尔股价一度上涨超30%。 截图 根据公告，英伟达将以每股23.28美元的价格认购英特尔普通股。这一价格略低于英特尔17日的收盘价 每股24.90美元，但高于美国政府上个月收购英特尔股份时支付的每股20.47美元的价格。17日收盘时， 英特尔市值为1160亿美元，这意味着英伟达投资50亿美元持有的英特尔股份在4%左右。 作为合作的一部分，英特尔还将在新一代个人电脑芯片中采用英伟达的技术，并为基于英伟达硬件打造 的数据中心产品提供处理器。但双方并未 ...

每经记者｜王晶 每经编辑｜文多 封面图片来源：视觉中国-VCG211478193393 9月16日，联发科方面宣布，公司首款采用台积电2纳米制程的旗舰系统单芯片（SoC）已成功完成设计 流片，预计将于明年底进入量产。"流片"是半导体研发过程中一个关键节点，它标志着芯片设计阶段基 本完成，进入到真正的制造验证环节。 在半导体产业中，纳米数越小，代表晶体管的体积更小、密度更高，芯片性能越强。目前，台积电、三 星、英特尔等均在竞逐更小制程，因为2纳米制程不仅意味着手机处理器性能和能效的进一步跃升，它 还能为端侧大模型、生成式AI、高性能计算等应用提供保障。 具体来看，台积电的2纳米制程技术首次采用纳米片电晶体结构，能够带来更优异的性能、功耗与良 率。联发科方面表示："台积电增强版2纳米制程技术与现有的N3E（台积电3纳米制程工艺升级版）制 程相比，逻辑密度增加1.2倍，在相同功耗下性能提升高达18%，并能在相同速度下功耗减少约36%。" 尽管台积电在技术方面取得突破，但高昂的制造成本仍然是行业面临的一大考验。有消息称，一枚2纳 米制程晶圆的成本约为3万美元，包括苹果、英伟达等在内的厂商需要在"性能"与"成本"之间进 ...

Core Viewpoints - CoreWeave has secured at least $6.3 billion in orders from NVIDIA, based on an agreement reached in 2023, which obligates NVIDIA to purchase CoreWeave's remaining unsold computing capacity until April 2023 [3][10] - Since its IPO in March, CoreWeave's stock price surged by 390% to $187, although it has since experienced significant volatility, dropping below $90 before recovering to close at $120.47, with a market capitalization nearing $59 billion [4][10] Business Model and Market Position - CoreWeave's business model heavily relies on NVIDIA, as it purchases tens of thousands of NVIDIA GPUs and leases the computing power to clients, creating a dependency that was noted as a risk factor in its IPO prospectus [6][8] - The company has established a large customer base in the computing rental market, despite its reliance on NVIDIA chips [8] Strategic Partnerships - CoreWeave has formed a strong partnership with NVIDIA, which is also an early supporter holding approximately 7% of CoreWeave's Class A shares as of June 30 [6] - This close relationship has fostered a mutual interest in the rapidly growing AI computing market [7] Financial Performance - CoreWeave reported explosive revenue growth, with Q2 revenues reaching $1.21 billion, a 207% increase year-over-year, although the company recorded a net loss of $290.5 million during the same period [10] - The long-term order from NVIDIA alleviates concerns regarding CoreWeave's future revenue stability and underscores its essential role in AI infrastructure [10] Contracts and Demand - Earlier this year, CoreWeave entered into a five-year contract worth up to $11.9 billion with OpenAI, and major cloud companies like Google and Microsoft are also seeking additional computing capacity from CoreWeave to meet their growing AI demands [9]

Core Insights - CoreWeave has secured a $6.3 billion order from NVIDIA, which obligates NVIDIA to purchase CoreWeave's remaining unsold computing capacity until April 2023 [2] - CoreWeave's business model heavily relies on NVIDIA, as it procures hundreds of thousands of NVIDIA GPUs and leases the computing power to clients [2] - In Q2 of this year, CoreWeave reported revenue of $1.21 billion, representing a year-over-year increase of 207%, but also recorded a net loss of $290.5 million during the same period [2]

CoreWeave披露获得来自英伟达的至少63亿美元订单，这份新订单是基于双方在2023年达成的一项协议。 9月15日，CoreWeave在一份文件中表示，根据协议条款，英伟达有义务购买CoreWeave截至2032年4月的剩余未售算力容 量。CoreWeave表示将在公布第三季度财报时披露完整协议副本。 据媒体报道，CoreWeave发言人在一封电子邮件中表示： 该协议反映了CoreWeave在加速全球人工智能创新方面发挥的规模、信任和关键作用。 今年3月以来，CoreWeave从IPO价格40美元一度飙涨390%至187美元，随后股价大幅回落、曾跌至90美元下方。周一股价 盘中涨逾8.4%，最终收涨7.6%，公司市值接近590亿美元。 英伟达扮演双重角色 CoreWeave的商业模式在很大程度上依赖英伟达，它通过采购数十万个英伟达图形处理单元（GPU），再将其算力租赁给 客户。 这种依赖关系在其IPO招股说明书中曾被列为风险因素，当时其基础设施中的所有GPU均来自英伟达。 与此同时，英伟达也是CoreWeave的早期支持者。截至6月30日，英伟达持有CoreWeave约7%的A类股。这种紧密的伙伴关 系也 ...

Core Viewpoint - Nvidia's recent financial forecast indicates a potential slowdown in growth after two years of rapid expansion in the artificial intelligence sector, despite a strong overall performance in the market [1][2]. Financial Performance - Nvidia expects third-quarter sales to reach approximately $54 billion, aligning with Wall Street expectations but falling short of some analysts' optimistic projections exceeding $60 billion [1]. - In the previous quarter, Nvidia's sales grew by 56% to $46.7 billion, slightly above market expectations of $46.2 billion, but this marked the smallest percentage increase in over two years [2]. Market Position and Future Outlook - Nvidia has become one of the largest chip manufacturers globally, with annual sales projected to reach $200 billion and expected to exceed $300 billion by 2028, capturing about one-third of total revenue in the chip industry [9]. - CEO Jensen Huang emphasized the vast future opportunities in artificial intelligence, predicting infrastructure spending could reach $3 trillion to $4 trillion by the end of the century [1]. - Nvidia's growth heavily relies on spending plans from a few large data center operators, such as Microsoft and Amazon, which account for about half of its sales [1].

Core Insights - The report by Goldman Sachs titled "Powering the AI Era" emphasizes that the most pressing bottleneck for the current AI revolution is not capital or technology, but rather the power infrastructure needed to support it [2] - The future of AI will be built not only on code and large language models but also on concrete, steel, and silicon, highlighting the immense energy demand required [2] Group 1: Paradigm Shift in Infrastructure - The rise of generative AI is fundamentally changing digital infrastructure, with AI workloads relying heavily on energy-intensive GPUs, leading to an exponential increase in power demand [3] - It is predicted that by 2030, global data center power demand will surge by 160% [3] - The cost structure of AI data centers has fundamentally changed, with internal computing devices like GPUs potentially costing 3 to 4 times more than the physical buildings themselves, disrupting traditional real estate financing models [3] - Despite these challenges, demand for data centers remains strong, with vacancy rates dropping to a historical low of 3% [3] - Hyperscalers are expected to invest over $1 trillion in AI by 2027 to meet this demand [3] Group 2: Urgent Power Challenges - The report identifies power supply as the current major obstacle, with the average age of the U.S. power grid infrastructure being 40 years, not designed to accommodate the explosive demand growth from AI [4] - After a decade of stability, power demand has suddenly surged, while new generation capacity faces significant challenges [4] - The approval and construction cycle for natural gas power plants can take 5 to 7 years, and renewable energy sources like wind and solar currently cannot provide stable base-load power [4] - Nuclear energy is viewed as a long-term solution, with companies like Microsoft signing agreements to restart closed nuclear reactors and exploring small modular reactors (SMRs) as reliable carbon-free power sources [4] - Some companies are adopting "behind the meter" solutions to ensure power supply by building microgrids on-site or near power plants [4] Group 3: Geopolitical and Capital Demand - The report discusses the geopolitical implications of AI infrastructure, with data centers becoming strategic tools for nations, similar to embassies [5] - Establishing partnerships globally will be crucial as the U.S. may face bottlenecks in data center expansion [5] - An unprecedented capital investment of approximately $5 trillion will be required in the digital infrastructure and power sectors by 2030 [5] - Innovative financing solutions are emerging to meet this demand, including joint ventures, private credit, and broader public-private partnerships to attract long-term capital from pension funds, insurance companies, and sovereign wealth funds [5] - The report concludes that addressing the "power challenge" is key to unlocking the full potential of AI, necessitating technological innovation and cross-industry strategic collaboration [5]