Chipmaking giant Nvidia’s (NVDA) closely followed four-day GPU Technology Conference (GTC) 2026 is set to commence later today with a keynote address from CEO Jensen Huang. Analysts and investors alike are looking for updates on several key areas from the annual developer conference. This includes a possible announcement on an inference chip with Groq, a roadmap for the Vera Rubin successor Feynman, a strategy to rival Intel (INTC) and Advanced Micro Devices (AMD) in the central processing unit (CPU) marke ...

Group 1 - NVIDIA CEO Jensen Huang announced significant technological breakthroughs at GTC 2026, including the DLSS5 real-time neural rendering model, which enhances gaming with "cinema-level" lighting and material details, marking the most significant advancement in computer graphics since 2018 [3] - The Feynman next-generation AI chip architecture, utilizing TSMC's 1.6nm A16 process and introducing optical communication technology, is expected to launch in 2028 and reduce AI data center communication energy consumption by over 70% [3] - AI chip revenue forecasts indicate that the Blackwell and next-generation Rubin architectures will accumulate $1 trillion in revenue by the end of 2027, doubling last year's predictions [3] Group 2 - Meta signed a $27 billion AI infrastructure supply agreement with cloud service provider Nebius, planning to purchase $12 billion in dedicated computing power over the next five years, highlighting the urgent need for expanded computing resources for AI product development [4] - Micron Technology announced the completion of an $1.8 billion acquisition of a foundry from Powerchip, planning to build a second large chip manufacturing facility at the site to expand DRAM and HBM production capacity, expected to start by the end of fiscal 2026 to meet growing AI computing demand [4] Group 3 - Alibaba established the Alibaba Token Hub business unit, led by CEO Wu Yongming, covering foundational model research, model service platforms, and AI applications for individuals and enterprises, with a commitment of over $53 billion in AI investments [5] - OpenAI is in discussions with private equity firms including TPG and Bain Capital to establish a joint venture for distributing AI products, with plans for a $4 billion equity investment from these firms [5] Group 4 - Foxconn reported a net profit of NT$45.21 billion for Q4 2025, a 2.4% year-on-year decline, but revenue reached NT$2.61 trillion, exceeding market expectations, with strong revenue growth anticipated for 2026 [7] - Ant Group's acquisition offer for Yao Cai Securities has been approved by Chinese authorities, with the deal expected to complete on March 30, 2025, extending Ant Group's financial services into the Hong Kong securities sector [8] - Beike disclosed a net revenue of 94.6 billion yuan for 2025, a 1% year-on-year increase, with a record 11% growth in second-hand transaction volume and a significant rise in non-real estate business contributions [8] Group 5 - Xiaomi's new SU7 model is set to launch on March 19, alongside other products including a high-performance laptop and a long-lasting smartwatch [9] - Zhipu AI released the world's first general model GLM-5-Turbo optimized for lobster scenarios, raising API prices by 20% and demonstrating significant performance improvements over previous models [9] - Baidu-backed biotech AI company Baitu has submitted a listing application to the Hong Kong Stock Exchange, aiming to raise several hundred million dollars [9] Group 6 - Inditex Group reported a sales revenue of €39.864 billion for the fiscal year ending January 31, 2026, a 3.2% year-on-year increase, with a net profit of €6.22 billion, marking a historical high [19] - Porsche's global revenue for fiscal 2025 was €36.27 billion, a 9.5% decline due to €3.9 billion in special expenses, with a significant drop in sales profit and a 26% decrease in deliveries in the Chinese market [19]

Core Insights - Nvidia CEO Jensen Huang emphasizes that the company is more than just an AI chipmaker, highlighting its growth potential and ongoing innovations [1][6][8] Group 1: Financial Projections - The next-generation Vera Rubin platform, along with its predecessor Blackwell, is projected to generate $1 trillion in revenues through 2027, a significant increase from the previously estimated $500 billion through the end of 2026 [3][6] - Nvidia's stock has seen a rise of nearly 2% following Huang's keynote address, despite a challenging start to the year [8] Group 2: Product Innovations - Huang introduced the Feynman platform, expected to launch in 2028, which will feature a powerful new chip named "Rosa" [4] - The company is also releasing new coding libraries and an open agent framework called NemoClaw, aimed at expanding its influence beyond hardware [5] Group 3: Market Position and Demand - Huang stated that demand for Nvidia's latest offerings is "off the charts," with every major player in the AI space utilizing its technology [6][7] - Nvidia is positioned as a comprehensive computing platform that supports all domains of AI, reinforcing its leadership in the industry [7][8]

Investment Rating - The industry investment rating is "Overweight" [5]. Core Insights - The GTC 2026 conference is expected to focus on AI inference evolving into system-level infrastructure, with key components including Rubin Ultra and Feynman architectures [2][3]. - The year 2026 is anticipated to be a pivotal year for Agentic AI, with significant developments in CPO and LPU technologies [2]. - The integration of Groq into NVIDIA's ecosystem is expected to enhance LPU capabilities, transitioning from a standalone LPX rack to a more integrated approach within NVIDIA's GPU roadmap [3][4]. Summary by Sections GTC 2026 Conference Highlights - NVIDIA's GTC 2026 will feature a keynote by CEO Jensen Huang, focusing on AI infrastructure advancements [2]. - Key technologies discussed will include Rubin Ultra, Feynman, CPX, and LPU, with expectations for significant updates on these platforms [2]. Rubin Ultra and Feynman Architecture - Rubin Ultra is expected to integrate 144 GPUs in a Kyber rack design, with a total power consumption of approximately 600 kW [2]. - Feynman may adopt TSMC's A16 process technology and is projected for a 2028 launch, with potential outsourcing of I/O die to Intel [2][3]. LPU and LPX Developments - The LPX rack is seen as a transitional solution, with LPU capabilities expected to be integrated into NVIDIA's GPU roadmap starting from the Feynman architecture [3]. - The LPU version of the LPX rack is being evaluated for a 256 LPU configuration, indicating a significant increase from the initial 64 LPU version [3]. CPO and Optical Interconnects - CPO and optical interconnects are anticipated to be central themes at GTC 2026, with a focus on the evolution from Scale-Out to Scale-Up architectures [4]. - The introduction of Scale-Up CPO switches is expected to complement the Rubin Ultra architecture, with significant bandwidth capabilities [4][10].

Investment Rating - The report maintains a "Recommended" rating for key companies in the electronic industry, including Huadian Co., Pengding Holdings, and Shenghong Technology [3][4]. Core Insights - The 2026 GTC conference is expected to unveil significant advancements in AI computing power, including new chip architectures like Rubin Ultra and Feynman, which will enhance performance and efficiency [13][14]. - The introduction of the LPU (Language Processing Unit) by Nvidia aims to address the growing demand for efficient AI inference, leveraging Groq's technology to improve performance and reduce costs [10][22]. Summary by Sections 1. GTC 2026 Outlook: Multi-Dimensional Technological Innovations - The GTC 2026 conference will showcase advancements in chip architectures, power supply systems, liquid cooling technologies, CPO (Co-Packaged Optics), and storage solutions [13]. - The Rubin Ultra chip, featuring 88 custom Olympus cores and two GPUs with 288GB HBM4 memory, is set to support large model training and inference [14]. - The transition to liquid cooling and the introduction of 800V HVDC power supply solutions are critical for managing the increased power demands of new chip architectures [16][18]. - CPO technology is expected to move from testing to large-scale commercialization, with Nvidia investing significantly in partnerships to enhance its capabilities [19]. - HBM4 technology is anticipated to provide over 11Gbps speeds and bandwidth exceeding 3.0TB/s, becoming a key differentiator for the Vera Rubin architecture [20]. 2. Nvidia's Introduction of the LPU - The LPU is designed specifically for AI inference, offering significant advantages in speed and energy efficiency compared to traditional GPUs [22][27]. - Groq's LPU architecture emphasizes software prioritization, programmable pipelines, deterministic computing, and on-chip storage, which collectively enhance performance for AI applications [29][30][33]. - The LPU's SRAM bandwidth exceeds 80TB/s, providing a substantial performance edge over GPU memory systems [37]. 3. Market Performance Review - The electronic sector experienced a decline of 5.00% in the past week, with a year-to-date increase of 10.73% compared to the Shanghai Composite Index [39]. - Sub-sectors such as LED and consumer electronics showed varied performance, with LED gaining 0.9% while consumer electronics fell by 2.4% [41].

电子行业周报 英伟达 GTC 将至，Feynman 及 LPU 或将登场 glmszqdatemark 新芯片架构及配套多维度技术革新。随着 Vera Rubin 平台在 CES 2026 确认量 产后，GTC 2026 有望迎来其强化版 Rubin Ultra 的正式亮相，并可能提前揭晓 下一代 Feynman 架构的初步技术路线。 重点公司盈利预测、估值与评级 | 代码 | 简称 | 股价 | | EPS（元） | | | PE（X） | | 评级 | | --- | --- | --- | --- | --- | --- | --- | --- | --- | --- | | | | （元） | 2025E | 2026E | 2027E | 2025E | 2026E | 2027E | | | 002463 | 沪电股份 | 74.19 | 1.99 | 2.84 | 4.48 | 37 | 26 | 17 | 推荐 | | 002938 | 鹏鼎控股 | 50.37 | 1.96 | 2.40 | 2.84 | 26 | 21 | 18 | 推荐 | | 330476 | 胜宏科技 | 270.0 ...

Key Points Summary Company Overview - **Company**: NVIDIA - **Fiscal Year**: FY26 Financial Performance - **Q4 Revenue**: $68.127 billion, a year-over-year increase of 73%, exceeding Bloomberg consensus by 3.24% [1] - **Gross Margin**: 75.2%, surpassing the consensus estimate of 74.7% [1] - **Net Profit**: $39.552 billion, a year-over-year increase of 79%, exceeding Bloomberg consensus by 5.48% [1] - **Adjusted EPS**: $1.62, surpassing Bloomberg consensus by 5.77% [1] - **Q1 FY27 Revenue Guidance**: Expected total revenue of $78 billion (±2%), exceeding Bloomberg consensus [1] - **Gross Margin Guidance**: Expected GAAP and non-GAAP gross margins of 74.9% and 75% (±50bps) [1] Business Segment Performance - **Data Center Revenue**: $62.31 billion, a year-over-year increase of 75%, exceeding Bloomberg consensus of $60.36 billion [2] - **Gaming Revenue**: $3.73 billion, a year-over-year increase of 47%, below consensus of $4.01 billion [2] - **Professional Visualization Revenue**: $1.321 billion, a year-over-year increase of 159%, significantly exceeding consensus of $0.771 billion [2] - **Automotive Revenue**: $604 million, a year-over-year increase of 6% [2] - **Other Revenue**: $16.1 million [2] Strategic Insights - **Blackwell System Deployment**: Nearly 9 GW of infrastructure deployed, contributing approximately two-thirds of data center revenue [3] - **Top Five Cloud Providers**: Account for 50% of data center revenue, with non-hyperscaler growth outpacing hyperscalers [3] - **CapEx Increase**: 2026 CapEx for top five cloud providers raised to nearly $700 billion, an increase of $120 billion from the beginning of the year [3] - **AI Business Growth**: Sovereign AI business grew over threefold year-over-year, surpassing $30 billion in scale [3] - **Management Emphasis**: "Compute equals revenues," highlighting that inference performance/power ratio directly impacts cloud providers' revenue capabilities [3] - **Rubin Platform**: Launched at CES and samples are being provided to customers [3] - **Upcoming GTC 2026**: Anticipation of new chip releases, with all technologies nearing their limits; Feynman with LPU solution expected to significantly enhance overall inference performance [3] Disclaimer - Information is based on publicly available data and may be subject to delays or updates; does not constitute investment advice [3]

Core Insights - NVIDIA's CEO Jensen Huang highlighted the rapid advancements in AI technology, particularly with the introduction of Alpamayo, the world's first thinking and reasoning model for autonomous vehicles, which utilizes a 10 billion-parameter vision-language-action model [1][2] - The demand for AI computing is surging, with model sizes increasing by 10 times annually and reasoning generating 5 times more tokens per year, while the cost of tokens is decreasing by 10 times [3] - NVIDIA's new Vera Rubin GPU system features significant innovations, including a design stripped of cables, hoses, and fans, utilizing hot-water liquid cooling at 45°C, which could disrupt the data center cooling market [6][7] NVIDIA Innovations - The Vera Rubin system is expected to be a game-changer in the robotics industry, with a projected cost of approximately $5 million per supercomputer rack-scale system [5] - The system's design allows AI factories to eliminate mechanical water chillers, relying instead on simpler cooling solutions, which has raised concerns for companies like Modine Manufacturing that specialize in data center chillers [7][9] Market Reactions - Modine's stock experienced a significant drop of over 20% following the announcement of the Vera Rubin system, reflecting investor concerns about the future demand for traditional cooling solutions [8][9] - Credo Technology Group also faced stock volatility, with shares falling due to fears that the new systems would replace copper cabling with optical interconnects, although a subsequent recovery was noted as investors reassessed the situation [10][11] Credo Technology Developments - Credo is diversifying its product offerings, focusing on both copper-based Active Electrical Cables and optical interconnect solutions, with recent innovations including ZeroFlap optical transceivers and a strategic acquisition of Hyperlume, Inc. [14][15][16] - The company aims to expand its Total Addressable Market (TAM) by prioritizing optical solutions while maintaining its leadership in the AEC market [16]

Summary of NVIDIA's Optical Interconnection Developments Company Overview - **Company**: NVIDIA (Ticker: NVDA US) - **Industry**: AI Networking and Optical Interconnection Key Points Industry Developments - NVIDIA is expected to incorporate CPO (Co-Packaged Optics) into its 2027 Rubin Ultra architecture, following Google's adoption of OCS (Optical Circuit Switching) for its TPU V7, which interconnects over 9,000 chips, surpassing NVIDIA's projected deployment of 576 dies in 2027 [1][2] - The transition to CPO for rack-to-rack interconnects is anticipated to enhance power consumption, latency, density, and cost efficiency compared to AOC (Active Optical Cable) [2] Product and Technology Insights - NVIDIA's scale-up optical solutions may arrive sooner than expected, with CPO being considered for the 576-die architecture starting in the second half of 2027 [2] - Compute trays and switch trays will continue to use PCB backplane connectivity, while rack-to-rack interconnects are likely to adopt CPO-based optical interconnects [2] - The scale-up CPO presents an incremental opportunity for the optical interconnect supply chain, with key beneficiaries including companies like LITE, Sumitomo, and Browave [3] Market Expectations - NVIDIA's scale-out CPO switch is projected to have deployment figures of 2,000, 20,000, and 35,000 units in 2025, 2026, and 2027 respectively [4] - The anticipated demand for NVIDIA's OIO (Optical Interconnection) solution is expected to coincide with the Feynman architecture, driving demand for CW lasers, FAUs, and optical engines [4] Risks - Potential risks include deceleration in AI demand, geopolitical uncertainties, and increased competition within the industry [5] Rating and Performance Expectations - NVIDIA is rated as a "Buy," indicating an expectation to outperform the benchmark by more than 15% [6] Additional Considerations - The report emphasizes that the CPO penetration rate may not be significant in the 1.6T generation due to factors such as maturity and technical reliability [4] - The supply chain for scale-up CPO is similar to that of scale-out CPO, indicating a consistent market structure [3] This summary encapsulates the critical insights and projections regarding NVIDIA's advancements in optical interconnection technology and its implications for the industry.

Group 1 - Nvidia is likely to become the sole customer for TSMC's A16 process (1.6 nm), with plans to integrate this technology into its next-generation GPU "Feynman" [1] - TSMC is accelerating the construction of its Kaohsiung P3 plant to start mass production for Nvidia by 2027, in response to Nvidia's significant demand for 3 nm chips and to prepare for A16 [1] - The A16 process utilizes nanosheet transistor architecture and SPR backside power delivery technology, which enhances logic density and reduces power drop, achieving an 8-10% speed increase and a 15-20% reduction in power consumption at the same voltage compared to N2P [1] Group 2 - Other major foundries are also advancing backside power delivery technology, with Samsung announcing plans to mass-produce BSPDN (backside power delivery network) process SF2Z by 2027, and Intel's PowerVia architecture set for the 2026 18A process node [2] - According to Nvidia's product roadmap, the Vera Rubin is expected to launch in the second half of 2026, Rubin Ultra in the second half of 2027, and Feynman in 2028, with A16 anticipated to be a key process for this generation of GPUs [2]