Core Insights - The article discusses the escalating arms race in artificial intelligence (AI) computing capabilities, highlighting a significant wave of global infrastructure investment. However, the AI industry faces a fundamental barrier in data movement between processors due to the limitations of existing copper interconnect technologies, which cannot effectively scale to meet the demands of next-generation AI requiring millions of processors. In this context, photonic chip startups Celestial AI and OpenLight are emerging to provide faster and more energy-efficient solutions to meet the needs of hyperscaler clients like Amazon, Microsoft, and Google [1]. Group 1: Celestial AI Developments - Celestial AI has secured $255 million in investments from TSMC's VentureTech Alliance and Samsung's Catalyst Fund, bringing its total funding to $520 million. The company plans to use this funding to strengthen its supply chain and deepen partnerships with foundries, including TSMC [2]. - Celestial AI expects to begin selling its first chips by mid to late 2026, focusing on server "scale-up" networks that account for 85% of data center traffic. The company anticipates that its scale-up sales will surpass copper-based Ethernet scale-out switches starting in 2025 [5]. - The core technology of Celestial AI, Photonic Fabric, offers optical interconnects from chip-to-chip and server-to-server, significantly outperforming Nvidia's NVLink in terms of power consumption. Celestial AI is positioning its Optical Memory Interface Bridge (OMIB) as an alternative to NVLink [5][7]. Group 2: OpenLight Developments - OpenLight raised $34 million in funding from investors including Juniper Networks and Lam Research. The company plans to use this capital to expand its process design kit (PDK) for active and passive photonic components, with production expected to begin for its first customers by the end of 2025 [4][11]. - OpenLight's competitive advantage lies in its use of indium phosphide technology, which allows for the integration of lasers onto chips. The company has demonstrated a 400 Gbps modulator with low power consumption of approximately 1.5 microjoules per bit [11]. - OpenLight provides a library of components (PDK) to enable customers to design their own chips, collaborating with ASE Holdings' subsidiary, Siliconware Precision Industries [14]. Group 3: Competitive Landscape - The photonic chip sector is becoming increasingly competitive, with players like Ayar Labs, Lightmatter, and Huawei entering the market. Huawei aims to challenge Nvidia's NVLink market with its new SuperPod cluster, which plans to connect up to 15,488 Ascend NPUs by 2026 [15]. - Lightmatter announced the Passage M1000 platform, which offers a total optical bandwidth of 114 Tbps, and is collaborating with GlobalFoundries and Amkor to begin production based on this platform [17][18]. - Quantum computing startup PsiQuantum is also manufacturing photonic chips using standard 45nm silicon nitride processes at GlobalFoundries [21].

Core Viewpoint - Nvidia has entered a strategic partnership with OpenAI to establish AI data centers with a capacity of at least 10 gigawatts (GW), utilizing millions of Nvidia GPUs for training and deploying next-generation AI models [1][2]. Investment Details - Nvidia plans to invest up to $100 billion in OpenAI, marking its largest investment commitment to date. This investment will be phased in as each 1 GW system is deployed, with the first system expected to go live in the second half of 2026 [2][4]. - The investment is expected to generate cumulative revenues of approximately $300 billion to $500 billion for Nvidia in the future, enhancing its competitive barriers against rivals like Broadcom and AMD [4][10]. Strategic Significance - The investment is a strategic equity investment aimed at securing OpenAI as a priority partner, thereby solidifying Nvidia's dominance in AI computing and networking [4][11]. - Nvidia's CEO Jensen Huang emphasized the scale of the partnership, noting that the 10 GW capacity equates to 4 to 5 million GPUs, which is double Nvidia's total shipments for the year [6][7]. Financial Implications - The cost of building a 1 GW data center is estimated between $50 billion and $60 billion, with around $35 billion allocated for Nvidia chips and systems. Thus, the total investment for the 10 GW project could reach several hundred billion dollars [8][10]. - Analysts predict that Nvidia could generate several hundred billion dollars in free cash flow over the next few years, given its revenue scale of up to $200 billion and a free cash flow margin of 40%-50% [10]. Market Positioning - The partnership allows Nvidia to maintain its core position in AI system construction, especially as OpenAI explores hardware development, including self-developed chips [11]. - The collaboration is seen as a strategic move to diversify Nvidia's supply chain and strengthen its position in the AI value chain, moving beyond being just a hardware supplier to a key partner in the AI economic ecosystem [11].

Group 1: Gold Market Insights - Gold prices reached a historic high of $3749.27 per ounce, driven by strong expectations for further interest rate cuts by the Federal Reserve and increased investment demand [2][4] - The recent increase in gold prices is attributed to market concerns over the Fed's independence and signs of a cooling U.S. economy, which have heightened expectations for rate cuts and a weaker dollar [4][5] - Analysts predict that the gold bull market will continue, with potential prices reaching $3800 per ounce in the next three months and possibly exceeding $4000 per ounce under certain economic scenarios [6] Group 2: Silver Market Dynamics - Silver prices have also surged, nearing $44 per ounce, with a year-to-date increase of over 40%, supported by rising investment demand and favorable market conditions [7][8] - The silver market is experiencing a dual logic of financial and industrial demand, with its applications in electronics and renewable energy driving further interest [7][8] - Analysts note that the silver market is more volatile than gold, making it more susceptible to sharp price fluctuations, which could impact industrial demand [8] Group 3: Broader Precious Metals Outlook - Citigroup's strategists forecast that the bull market for gold and silver will extend to copper and aluminum by 2026, indicating a new cycle of precious metal price increases [9] - Factors driving this outlook include concerns over the U.S. labor market, trade policy uncertainties, and expectations of a weaker dollar, particularly with a potentially more dovish Fed leadership in 2026 [9][10] - UBS highlights that supply constraints and increased demand are expected to drive copper prices up, with a projected target price of $11,000 per ton by 2026 due to a significant supply-demand gap [10]

2025.09. 23 本文字数：3022，阅读时长大约5分钟 作者 | 第一财经 后歆桐 封面图来源 | 新华社 美联储降息预期、避险需求叠加投资热情，推动金价在今日（23日）亚太交易时段再创3749.27美 元/盎司的历史新高。白银也连续上涨三日，逼近44美元/盎司的逾14年新高。市场正密切关注白银能 否突破1980年1月的历史收盘高点48.70美元/盎司。 金银比翼同飞之际，以大宗商品全球主管马克西米利安·莱顿（Maximilian Layton）为首的花旗策略 师团队预测，"在美联储新鸽派领导层在2026年5~6月上任的前景、美国实际利率下降及美元承压的 推动下，黄金和白银的牛市行情将进一步扩大，并最终在2026年延伸至铜和铝领域。" 美联储上周降息25个基点后，上周五黄金交易所交易基金（ETF）持有量以三年多来最快的速度增 长。美元指数周一下跌0.36%，回吐了上周五全部涨幅，结束三连涨。由于黄金以美元计价，美元的 疲软使得持有其他货币的投资者购买黄金的成本降低，从而扩大了全球范围内的需求。 金价又又又创新高，开启长牛模式 金价今日微涨至3749.27美元/盎司，此前两个交易日连续上涨。 本轮金价上 ...

Core Insights - The article presents a low-cost AI marketing solution designed specifically for small and medium-sized enterprises (SMEs), emphasizing the importance of high cost-performance marketing for their survival and growth [2] - The solution is based on a unique AI marketing five-star model developed by An Zhi Yi, which aims to systematically assist businesses in deploying AI marketing [2][5] - The model focuses on five dimensions: strategy, content, reach, analysis, and optimization, enabling full-process automation and intelligence through AI tools [2][4] Solution Features - The AI marketing solution is characterized by its simplicity, low cost, and quick effectiveness, allowing businesses to implement it without hiring expensive professional teams or purchasing complex systems [2][4] - Key functionalities include: - Intelligent content generation for marketing copy, social media posts, and emails, ensuring high-frequency and high-quality output to attract potential customers [4] - Precise customer targeting through AI analysis of public data and user behavior, enabling personalized communication with likely conversion prospects [4] - Automated operations and follow-ups using AI chatbots to respond to common inquiries 24/7, ensuring no potential opportunity is missed [4] - Data-driven decision-making where AI analyzes marketing activity effectiveness, identifying which channels and content lead to conversions for rapid strategy optimization [4] Implementation and Expertise - The success of this solution is backed by extensive practical experience, with Yang Bi Rui having 17 years of experience in internet marketing training and AI, having served over 1,000 enterprises [5] - Liang Kai, a senior operations consultant, brings a wealth of experience in managing large-scale businesses and ensures the AI marketing solution integrates deeply with overall business operations for maximum resource efficiency [5][7] - Overall, the low-cost AI marketing solution combines innovative methodology, practical validation, and a team of seasoned experts, providing SMEs with a reliable, easy-to-execute, and effective intelligent marketing pathway [7]

更多分析师加入看多更多贵金属行列。 美联储降息预期、避险需求叠加投资热情，推动金价在今日（23日）亚太交易时段再创3749.27美元/盎司的历史 新高。白银也连续上涨三日，逼近44美元/盎司的逾14年新高。市场正密切关注白银能否突破1980年1月的历史收 盘高点48.70美元/盎司。 金银比翼同飞之际，以大宗商品全球主管马克西米利安·莱顿（Maximilian Layton）为首德花旗策略师团队预 测，"在美联储新鸽派领导层在2026年5~6月上任的前景、美国实际利率下降及美元承压的推动下，黄金和白银的 牛市行情将进一步扩大，并最终在2026年延伸至铜和铝领域。" 金价又又又创新高，开启长牛模式 金价今日微涨至3749.27美元/盎司，此前两个交易日连续上涨。 本轮金价上涨最直接的引擎，无疑来自于市场对美联储将进一步降息的强烈预期。虽然美联储主席鲍威尔仍强调 未来降息路径将保持谨慎，但市场仍预计美联储未来将大幅降息。 瑞士宝盛新世纪思维研究主管曼克（Carsten Menke）在降息前后对一财记者表示，过去几周，货币政策无疑是投 资需求增加的主要推动力，短线交易员与市场跟风者担心自己会错过这一波上涨，进一步推 ...

| will history of the | | 10. HTT 1/5 = 1/1 | | --- | --- | --- | | 心 被共コスモ世界 - LEE (14 13) 11 11 11 11 11 11 | | 日落平 | | 148.32-33 | | 45680.25 | | 1.665 | | +584.42 | | +0.491 +0.025 | | | | 编 6:55 9:42 | 1.650 | 14791 @ 9:00 45193.// | | 148.37 雷 9:42 高 12:00 | 1.650 | 雷 13:03 45757.74 | | 安 13:49 安 7:02 | 1.665 | 147.87 @ 9:00 45193.77 | | 2 2 2 8 中文网 | | | 受到长期利率上涨的影响，日本的银行股明显上涨（9月22日，东京都中央区） 日本央行9与19日发表了出售ETF的方针，不过由于出售速度缓慢，有分析认为这对市场的 影响轻微。有观点认为，"脱离对央行的依赖"反而有可能促进日本企业的公司治理改革，日 经平均股指22日再次创出新高…… 即使日本银行（央行 ...

当地时间周一，英伟达首席执行官黄仁勋宣布，英伟达计划向投资1000亿美元，双方共建大规模数据中 心。 "其次，也许更重要的是，这项合作将使英伟达成为OpenAI首选的战略计算和网络合作伙伴。虽然没有 明确的份额，但从表面上看，这项协议增加了包括博通和AMD在内的其它供应商的竞争风险。"美银分 析师表示。 Hargreaves Lansdown的高级股票分析师Matt Britzman也发表了类似的观点。他表示，"对于英伟达来 说，回报是巨大的——每千兆瓦的人工智能数据中心容量能带来大约500亿美元的营收，这意味着该项 目的价值可能高达5000亿美元。" "通过锁定OpenAI作为战略合作伙伴，并共同优化硬件和软件路线图，英伟达正在确保其图形处理器 （GPU）依然是下一代AI基础设施的支柱。" Britzman表示。 这标志着全球人工智能（AI）竞赛中两家最引人注目的参与者之间的合作，提振了投资者对AI板块的 信心。甚至有分析认为，这表明AI热潮尚处早期阶段。 受此影响，英伟达股价周一大涨 3.97%，报183.61美元，创下收盘历史新高。这则消息还带动了美股科 技股上涨，并推动美股三大指数齐创历史新高。 华尔 ...

Group 1 - The U.S. Air Force Chief of Staff, David Alvin, announced the construction of the first F-47 fighter jet, expected to make its maiden flight in 2028 [1] - Boeing has begun manufacturing the first F-47, with a commitment to achieve flight by 2028 [1][3] - The U.S. Air Force plans to procure at least 185 F-47 jets, which will replace the F-22 Raptor [3] Group 2 - The estimated unit price of the F-47 is projected to be as high as $300 million, significantly more than the F-22's cost of approximately $143 million [3] - The "Next Generation Air Dominance" (NGAD) program, which includes the F-47, has faced criticism regarding its costs and technological innovation [4] - The F-47 is part of a broader strategy that includes the B-21 stealth bomber and a naval version of the sixth-generation fighter, F/A-XX, raising concerns about the financial burden on the Pentagon [4]

Group 1 - The A-share market indices opened higher on September 23, with the electronic sector showing significant gains, particularly the electronic ETF (159997) which rose by 0.47% and had a trading volume exceeding 15 million yuan [1] - The electronic ETF (159997) tracks the CSI Electronic Index, which includes companies involved in semiconductor production, computer storage, retail of electronic products, and consumer electronics manufacturing, reflecting the overall performance of the electronic sector [1] - On September 22, the electronic ETF (159997) experienced a net inflow of 23.47 million yuan, indicating strong investor interest [2] Group 2 - Nvidia and OpenAI announced a partnership to build a large-scale data center, with Nvidia potentially investing up to 100 billion dollars, which will utilize 10 gigawatts of power for AI model training, equivalent to the electricity consumption of 8 million American households [2] - Wedbush raised Apple's target price from 270 dollars to 310 dollars due to strong demand for the iPhone 17 series, suggesting a significant upgrade cycle for Apple after years of low product launches [2] - Zheshang Securities highlighted that the semiconductor sector is experiencing an upward demand cycle, driven by AI as a core growth engine, and remains optimistic about the performance of related sectors in the second half of the year [3]