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techpowerup 。 在与 OpenAI 意外宣布合作后不久，AMD 首席执行官苏姿丰在接受雅虎财经采访时确认，下一代 Instinct MI450 图形加速器将采用 2 纳米制程技术。"我们对 MI450 感到非常兴奋。它拥有 2 纳米 技术，拥有最先进的制造能力，"AMD 首席执行官苏姿丰表示。 | | Instinct™ MI400 Series AMD "Helios" | vs. Vera Rubin | | --- | --- | --- | | | | Oberon | | GPU DOMAIN | 72 | 1x | | SCALE UP BANDWIDTH | 260 TB/s | 1x | | FP4 · FP8 FLOPS | 2.9 EF · 1.4 EF | ~1x | | HBM4 MEMORY CAPACITY | 31 TB | 1.5x | | MEMORY BANDWIDTH | 1.4 PB/s | 1.5x | | SCALE OUT BANDWIDTH 27 | 43 TB/s See ende ...

Meanwhile, uh the other big story of the morning, uh maybe of the week, we'll see. OpenAI making a deal with AMD on data centers that's going to run on AMD processors. The stock is up big this morning.Of course, this comes two weeks after Nvidia made a massive deal uh with Open AI. Jim Kramer, who of course has been following the AI story at Nvidia specifically for so very long, is at the New York Stock Exchange, head of his interview with AMD CEO Lisa Sue, and Greg Brockman, president of OpenAI. Want to uh ...

Core Viewpoint - NVIDIA and AMD are competing to develop superior AI architectures, with significant upgrades planned for their next-generation products in terms of power consumption, memory bandwidth, and process node utilization [2][3]. Group 1: AI Product Competition - AMD's Instinct MI450 AI series is expected to compete fiercely with NVIDIA's Vera Rubin, with both companies making substantial modifications to their designs [2][5]. - AMD executive Forrest Norrod expressed optimism about the MI450 product line, likening it to AMD's transformative "Milan moment" with the EPYC 7003 series [3]. - The MI450 is projected to be more competitive than NVIDIA's Vera Rubin, with AMD planning to leverage its own technology stack for future products [3]. Group 2: Technical Specifications - The MI450X's Total Graphics Power (TGP) has increased by 200W, while the TGP for Rubin has risen by 500W to 2300W, indicating a response to market competition [5]. - Memory bandwidth for Rubin has improved from 13 TB/s to 20 TB/s per GPU, showcasing the enhancements made in both product lines [5]. - AMD's MI450 is rumored to feature HBM4 memory with up to 432 GB per GPU, while NVIDIA's Rubin is expected to have around 288 GB per GPU [6]. Group 3: Interconnect Technology - AMD plans to significantly enhance its chip-to-chip (D2D) interconnect technology with the upcoming Zen 6 processors, as evidenced by developments in the Strix Halo APU [8][10]. - The new D2D interconnect method reduces power consumption and latency by eliminating the need for serialization/deserialization, thus improving overall bandwidth [12][15]. - The Strix Halo's design utilizes TSMC's InFO-oS technology and redistribution layers (RDL) to facilitate efficient communication between chips [10][15].

Just when you think Nvidia NVDA can’t go any bigger, it does. In Nvidia's latest AI flex, the tech behemoth announced a whopping $100 billion investment in OpenAI. The deal anchors the rollout of 10+ gigawatts of Nvidia systems to efficiently train and run the next wave of models, with deliveries set to begin in the second half of 2026. Think utility-scale compute, but for artificial intelligence. Nvidia's OpenAI investment at a glance: Staged LOI: Letter of intent to deploy ≥10 GW of NVIDIA systems; wi ...

Core Viewpoint - Nvidia and OpenAI have announced a strategic partnership to deploy up to 10 gigawatts of Nvidia systems, significantly enhancing AI infrastructure and capabilities [1][4]. Group 1: Partnership Details - OpenAI will utilize Nvidia's systems to build and deploy at least 10 gigawatts of AI data centers, which will consist of millions of GPUs for training and running advanced AI models [5][6]. - Nvidia's CEO Jensen Huang stated that the 10 gigawatts of power equates to approximately 4 to 5 million GPUs, doubling the number Nvidia is expected to ship this year [6]. - Nvidia plans to invest up to $100 billion in total to support the deployment, with investments made in phases based on the gigawatt deployment progress [6]. Group 2: Technological Advancements - The first phase of the system is expected to be operational by the second half of 2026, based on Nvidia's Vera Rubin platform, which integrates CPUs, GPUs, and dedicated accelerators for complex AI tasks [6]. - The collaboration aims to push the boundaries of AI technology, with both companies emphasizing the importance of computational infrastructure for future economic growth [7][8]. Group 3: Market Impact - Following the announcement, Nvidia's stock price rose nearly 4%, adding approximately $170 billion to its market capitalization, which is now close to $4.5 trillion [9].

国海证券发布研报称，大模型训推带动AI算力需求增长，GB300、Vera Rubin等新一代算力架构将推 出，算力产业链中的AI芯片、服务器整机、铜连接、HBM、液冷、光模块、IDC等环节有望持续受 益。维持对计算机行业"推荐"评级。 国海证券主要观点如下： GPU核心：GB300计算能力大幅提升，Rubin预期乐观 B300基于Blackwell Ultra架构，采用TSMC 4NP工艺与CoWoS-L封装，FP4浮点算力可达15PFLOPS，是 B200的1.5倍。https://www.hibor.com.cn(慧博投研资讯)GB300搭载288GB的HBM3E显存。Vera Rubin NVL144性能是GB300 NVL72的3.3倍。Rubin Ultra NVL576性能较GB300 NVL72提升14倍，内存是 GB300的8倍。FY2026Q2，英伟达营收467亿美元，同比+56%。 CPO将硅光子器件与ASIC封装，取代传统的可插拔光模块，与传统网络相比，提升能效3.5倍、部署速 度1.3倍。Quantum-X和Spectrum-X交换机减少了对传统光收发器的依赖，为超大规模人工智能工厂提 ...

Core Viewpoint - Nvidia has launched the Rubin CPX, a CUDA GPU designed for processing large-scale context AI, capable of handling millions of tokens efficiently and quickly [5][4]. Group 1: Product Overview - Rubin CPX is the first CUDA GPU specifically built for processing millions of tokens, featuring 30 petaflops (NVFP4) computing power and 128 GB GDDR7 memory [5][6]. - The GPU can complete million-token level inference in just 1 second, significantly enhancing performance for AI applications [5][4]. - The architecture allows for a division of labor between GPUs, optimizing cost and performance by using GDDR7 instead of HBM [9][12]. Group 2: Performance and Cost Efficiency - The Rubin CPX offers a cost-effective solution, with a single chip costing only 1/4 of the R200 while delivering 80% of its computing power [12][13]. - The total cost of ownership (TCO) in scenarios with long prompts and large batches can drop from $0.6 to $0.06 per hour, representing a tenfold reduction [13]. - Companies investing in Rubin CPX can expect a 50x return on investment, significantly higher than the 10x return from previous models [14]. Group 3: Competitive Landscape - Nvidia's strategy of splitting a general-purpose chip into specialized chips positions it favorably against competitors like AMD, Google, and AWS [15][20]. - The architecture of the Rubin CPX allows for a significant increase in performance, with the potential to outperform existing flagship systems by up to 6.5 times [14][20]. Group 4: Industry Implications - The introduction of Rubin CPX is expected to benefit the PCB industry, as new designs and materials will be required to support the GPU's architecture [24][29]. - The demand for optical modules is anticipated to rise significantly due to the increased bandwidth requirements of the new architecture [30][38]. - The overall power consumption of systems using Rubin CPX is projected to increase, leading to advancements in power supply and cooling solutions [39][40].

Core Insights - The rise of artificial intelligence (AI) is projected to significantly boost global GDP, with PwC forecasting a $15.7 trillion increase by 2030, leading to a surge in AI-related stocks [2] - Nvidia has emerged as a leader in the AI revolution, adding approximately $3.8 trillion in market value since the beginning of 2023, with its stock price increasing by 1,070% [4] - Despite Nvidia's strong market position, insider trading activity raises concerns, as insiders have sold a net of $4.7 billion worth of stock over the past five years, with minimal buying activity [15][18] Company Overview - Nvidia is synonymous with AI due to its GPUs, which are essential for enterprise data centers, with its Hopper (H100) and Blackwell chips dominating the market [6] - The company is expected to maintain its competitive edge with annual next-gen chip launches, including the upcoming Blackwell Ultra [7][8] - Nvidia benefits from a scarcity of AI GPUs, allowing it to sustain premium pricing and improve gross margins [9] Insider Trading Activity - Insiders are required to report their trading activities, and Nvidia's insider selling has been persistent, with no significant buying since December 2020 [12][13][18] - The lack of insider purchases raises questions about the company's future performance, especially given the stock's high price-to-sales ratio of over 25, which suggests potential overvaluation [19]

Core Insights - Nvidia's latest earnings report, while showcasing strong financial performance, contains subtle warnings that may indicate potential challenges ahead for the company and its investors [3][10]. Financial Performance - Nvidia reported $46.7 billion in net sales for the fiscal second quarter, representing a 56% increase year-over-year, and earnings per share (EPS) of $1.05, exceeding Wall Street expectations for the 11th consecutive quarter [5]. - The data center segment was the primary driver of sales, accounting for over 88% of total revenue, with strong demand for the Blackwell and Blackwell Ultra chips [6]. Gross Margin and Pricing - The GAAP gross margin for Nvidia was 72.4%, down 270 basis points from the previous year, but it marked the first sequential improvement in over a year, indicating strong pricing power for its AI hardware [7][8]. Revenue Concentration Risks - Nvidia's revenue concentration is a concern, with two customers accounting for 39% of total revenue in the latest quarter, highlighting increasing reliance on a narrow customer base [11]. - The two major customers are likely Meta Platforms and Microsoft, both of which are developing their own AI-GPUs, potentially impacting Nvidia's future sales [12][13]. Future Product Development - Nvidia's strategy to introduce a new advanced AI chip annually may lead to rapid depreciation of existing GPUs, which could affect upgrade cycles and gross margins if customers opt for cheaper alternatives [14]. Share Buyback Program - Nvidia's board approved an additional $60 billion share repurchase program, raising concerns about the company's ability to find high-growth investment opportunities, especially after a significant stock price increase [16][17]. - The lack of insider purchases since December 2020 and significant stock sales by executives over the past five years further complicate the narrative around the company's financial health [18].

Core Viewpoint - The company anticipates over $1 trillion in sovereign AI investment projects in the next five years, which will drive significant growth in the computing power market [1][6][13]. Group 1: Sovereign AI Investment Projects - Major projects include the U.S. Stargate ($500 billion), the EU InvestAI (€200 billion), and Saudi Arabia's Humain AI ($1 trillion) [1][14]. - Some projects are already accelerating, with expected revenue contributions starting in 2026 [1][15]. Group 2: AI Revenue Growth - The company expects a 170% quarter-over-quarter growth in AI revenue for Q3, with rack growth projected at 300% [3]. - AI server revenue is expected to exceed NT$1 trillion by 2025, capturing 40% of the market share [3][8]. Group 3: Product Transition and Market Share - The company does not foresee major transition issues with the GB200 and GB300 products, with GB300 expected to dominate shipments in the second half of 2025 [2][4]. - The number of cloud service provider clients is expected to increase from 2 for GB200 to 3 for GB300, with additional clients for Vera Rubin products [4]. Group 4: Profitability and Cost Management - Despite potential short-term gross margin dilution due to increased GPU costs, the company aims to stabilize operating profit margins around 3% through improved operational efficiency [9][11]. - The company is extending its value chain in AI servers to capture a larger share of capital expenditures, aiming to increase its capture ratio from 40% to 60% [12]. Group 5: Manufacturing and Capacity Expansion - The company operates three factories in the U.S., with a fourth under development, and expects significant growth in U.S. operations by 2026 [16]. - The Ohio factory is likely to become a manufacturing base for modular data centers related to the Stargate project [15].