Investment Rating - The report maintains a "Buy" rating for the industry, indicating an expected investment return exceeding the market benchmark by over 15% in the next 6-12 months [16] Core Insights - The NVIDIA GTC conference is pivotal in defining the technological direction of the AI industry, initiating a new growth cycle for the industry chain [1] - The performance of key stocks in the AI computing sector, such as NVIDIA and TSMC, shows a pattern of short-term fluctuations followed by long-term growth driven by strong fundamentals [5][6] - The demand for copper connections and liquid cooling systems is expected to surge due to NVIDIA's NVL72 cabinet, benefiting related industry stocks significantly [8] Summary by Sections AI Computing Platforms - The introduction of the Blackwell and NVL72 systems in 2024 is set to redefine the market, with significant performance improvements and cost reductions [2] - The NVL72 system integrates advanced technologies, including liquid cooling and high-speed copper interconnects, establishing a new industry paradigm [2] High-Speed Interconnects and Network Architecture - The NVLink5 and X800 switch systems will enhance bandwidth capabilities, pushing the industry towards higher performance standards [3] - The ongoing competition between copper and optical technologies is highlighted, with significant advancements expected in optical interconnects by 2026 [3] Stock Performance Analysis - Post-GTC stock performance shows a clear trend: short-term corrections followed by long-term gains, particularly in the computing and optical communication sectors [4][5] - Key stocks like NVIDIA and TSMC have shown substantial price increases over six months, correlating with strong order and delivery performance [5][6] Sector-Specific Insights - The storage and server sectors exhibit mixed performance, with SK Hynix and Micron showing significant growth in 2025 due to rising demand for high-density memory products [9][10] - The optical communication sector is expected to benefit from increased demand for high-bandwidth, low-latency connections, with several companies reporting substantial profit growth [11][12] Recommendations - The report suggests focusing on companies within the optical interconnect industry, such as Lumentum, Coherent, and others, which are poised for growth due to the expanding data center applications [14]

Investment Rating - The report maintains a rating of "Buy" for NVIDIA (NVDA.O) [1] Core Insights - The NVIDIA 2026 GTC conference highlighted the explosive demand for tokens and the evolution of next-generation architectures, enhancing computational efficiency and driving AI into a new phase centered on inference and agents [2] - NVIDIA's revenue guidance has been raised significantly due to the anticipated explosion in token demand, with projections for the Blackwell and Rubin platforms increasing from $500 billion for 2025-2026 to $1 trillion for 2025-2027, surpassing market expectations [11][14] - The report emphasizes NVIDIA's competitive edge in token efficiency, with the GB300 NVL72 achieving approximately 50 times the token output per unit of energy compared to competitors, and a 35 times reduction in token costs [14][17] Financial Summary - Revenue projections for NVIDIA are as follows (in million USD): - 2025: $130,497 - 2026: $215,938 - 2027: $380,098 - 2028: $523,847 - 2029: $637,418 - Year-over-year growth rates are projected at 114.2% for 2025, 65.5% for 2026, 76.0% for 2027, 37.8% for 2028, and 21.7% for 2029 [4] - Non-GAAP net profit projections are as follows (in million USD): - 2025: $74,266 - 2026: $116,996 - 2027: $223,590 - 2028: $306,380 - 2029: $371,027 - Year-over-year growth rates for non-GAAP net profit are projected at 129.8% for 2025, 57.5% for 2026, 91.1% for 2027, 37.0% for 2028, and 21.1% for 2029 [4] Revenue Drivers - The report identifies that approximately 60% of NVIDIA's revenue comes from Cloud Service Providers (CSPs), with the remaining 40% from sovereign AI, enterprise, and emerging cloud demands, indicating a broadening of AI computational needs [11] - The introduction of the LPU (Logic Processing Unit) significantly enhances inference efficiency, with potential revenue from a 1GW data center projected to increase from approximately $300 billion with Blackwell NVL72 to $1,500 billion with Rubin NVL72, and further to $3,000 billion with Rubin+LPX [17][18] System Architecture Evolution - The Rubin large-scale computing cluster features a modular architecture with an increased proportion of LPU and independent CPU cabinets, enhancing system capabilities [23][27] - Future iterations of NVIDIA's architecture include the launch of the Rubin rack in 2026, Rubin Ultra in 2027, and Feynman rack in 2028, which will introduce new CPUs and continue to enhance system bandwidth and computational density [30] Software Ecosystem Development - NVIDIA has launched the NemoClaw platform to expand enterprise-level AI agent development and deployment capabilities, integrating NVIDIA Nemotron models and OpenShell runtime environments for rapid AI agent construction [34][36] Profit Forecast and Investment Advice - The report maintains revenue forecasts for FY2027E-FY2029E at $380.1 billion, $523.8 billion, and $637.4 billion respectively, with non-GAAP net profit forecasts at $223.6 billion, $306.4 billion, and $371.0 billion respectively [39] - The target price is set at $275, maintaining a "Buy" rating based on NVIDIA's position as a leading growth engine in the token economy [39]

Key Points Summary of NVIDIA FY4Q26 Earnings Call Company Overview - **Company**: NVIDIA - **Fiscal Year**: FY4Q26 Financial Performance - **Revenue**: $68.127 billion, up 73% year-over-year and 20% quarter-over-quarter, exceeding market expectations of $65.912 billion [1] - **Net Profit**: $42.960 billion, up 94% year-over-year and 35% quarter-over-quarter, surpassing market expectations of $36.302 billion [1] - **Earnings Per Share (EPS)**: $1.76, exceeding market expectations of $1.48 [1] - **Guidance for FY1Q27**: Expected revenue of $78 billion (±2%), up 77% year-over-year and 14% quarter-over-quarter, exceeding market expectations of $72.778 billion [1][2] Gross Margin - **Gross Margin for FY4Q26**: 75.0%, slightly above company guidance of 74.8% and in line with market expectations [3] - **Guidance for FY1Q27 Gross Margin**: 74.9% (GAAP) and 75.0% (Non-GAAP), consistent with the current quarter [3] Business Segment Performance - **Data Center Business**: $62.314 billion, up 75% year-over-year and 22% quarter-over-quarter, exceeding market expectations of $60.360 billion [4] - **Compute Segment**: $51.334 billion, up 58% year-over-year and 19% quarter-over-quarter, meeting market expectations of $51.609 billion [4] - **Networking Segment**: $10.980 billion, up 263% year-over-year and 34% quarter-over-quarter, exceeding market expectations of $9.019 billion [4] - **Gaming Business**: $3.727 billion, up 47% year-over-year but down 13% quarter-over-quarter, below market expectations of $4.011 billion [4] - **Professional Visualization**: $1.321 billion, up 159% year-over-year and 74% quarter-over-quarter, exceeding market expectations of $0.771 billion [4] - **Automotive Business**: $0.604 billion, up 6% year-over-year and 2% quarter-over-quarter, below market expectations of $0.643 billion [4] - **OEM and Other**: $0.161 billion, up 28% year-over-year but down 7% quarter-over-quarter [4] Additional Insights - **Revenue Outlook for 2026**: Optimistic, with expected growth exceeding the previously shared $500 billion target for Blackwell and Rubin [5] - **Sales in China**: H200 chip has not generated any revenue yet due to regulatory conditions [5] - **Rubin Progress**: First batch of Vera Rubin samples delivered to customers, with mass delivery expected in H2 2026 [5] - **Capital Expenditure**: Major companies are investing heavily in Capex, indicating strong cash flow growth and the potential for revenue from Agentic AI [5] - **Space Data Centers**: NVIDIA's Hopper is the first GPU in space, with economic viability expected to improve over time [5] - **Token and Revenue**: Emphasis on "inference" as a revenue driver, focusing on higher speed inference and optimal performance per watt [5][6] - **AI Infrastructure Definition**: NVIDIA positions itself as an AI infrastructure company, integrating GPU, CPU, and networking capabilities [7]

Core Insights - Nvidia's Q4 and full-year fiscal 2026 results exceeded expectations across revenue, profit, and gross margin metrics, indicating strong performance and growth potential [1] - The company reported Q4 total revenue of $68.13 billion, a year-over-year increase of 73% and a quarter-over-quarter increase of 20% [1] - Full-year revenue reached $215.9 billion, up 65% year-over-year, with data center revenue contributing significantly [1] Financial Performance - Q4 net profit was $42.96 billion, reflecting a 94% year-over-year increase [1] - Non-GAAP EPS for Q4 was $1.62, significantly surpassing expectations [1] - Q4 GAAP gross margin was 75.0%, and Non-GAAP gross margin was 75.2%, both showing improvements compared to previous periods [1] Revenue Breakdown - Data center revenue for the full year was $193.7 billion, up 68% year-over-year, demonstrating strong demand for high-performance computing [1] - Gaming revenue for Q4 was $3.727 billion, a 47% year-over-year increase but fell short of market expectations [2] - Professional visualization revenue surged to $1.321 billion, a 159% year-over-year increase, benefiting from the Blackwell architecture [2] Strategic Developments - Nvidia's network business saw quarterly sales of $10.98 billion, a remarkable 263% year-over-year increase, establishing it as a significant growth driver [1] - The company is focusing on integrating computing power with networking and complete machine solutions, enhancing customer stickiness and average transaction value [2] - The introduction of advanced architectures like GB300 NVL72 offers significant performance improvements, emphasizing the importance of performance per watt in the industry [2] Future Guidance - For Q1 of fiscal 2027, Nvidia projects revenue of $78 billion (±2%), excluding data center revenue from China [2] - The company expects GAAP and Non-GAAP gross margins to remain high at 74.9% and 75.0%, respectively [2]

Group 1 - Nvidia's Q4 FY2026 earnings report shows revenue, net profit, and guidance exceeding market expectations, highlighting the strong demand for AI computing power [1][3] - Customers have deployed 9GW of Blackwell infrastructure, indicating a significant increase in AI computing power translating into real electricity demand, putting unprecedented pressure on the North American power grid [1][3] - The data center business generated $62.314 billion in revenue, a 75% year-over-year increase, with networking revenue reaching $10.980 billion, up 263% year-over-year and 34% quarter-over-quarter, driven by NVLink 72 scale-up switches [3] Group 2 - The focus on "performance/watt" is critical, with Nvidia's GB300 NVL72 offering 50 times the performance per watt and reducing costs by 35 times per token compared to market alternatives [4] - North America is experiencing a real electricity shortage due to the rapid expansion of AI, with PJM planning to increase grid investments to support data center demands, indicating a long-term planning shift [5] - The supply chain constraints are limiting natural gas power generation growth, with delivery cycles extending from 2 years to 5 years, necessitating alternative power supply methods [5] Group 3 - During the 14th Five-Year Plan, the State Grid's fixed asset investment is expected to reach 4 trillion yuan, a 40% increase compared to the previous plan, with a focus on enhancing cross-province transmission capacity by over 30% [6] - The establishment of a unified national electricity market by 2030 aims to increase market-based electricity trading to about 70%, providing a structured foundation for electricity asset revenue [8] - The electric grid ETF (561380) offers a diversified investment opportunity across the electric power industry, covering key components such as transformers and communication cables, benefiting from the strong demand driven by AI [9][10]

Core Insights - The focus in the AI chip market is shifting from performance to cost efficiency, as commercial pressures mount and the cost of inference becomes a critical factor in determining competitive advantage [1][2][3] Group 1: Shift in Evaluation Criteria - The evaluation criteria for AI chips are transitioning from "who computes faster" to "who computes cheaper and more sustainably" as inference becomes a significant source of long-term cash flow [2][3] - High costs associated with inference are becoming more pronounced as deployment and commercialization of large models progress, leading to a reevaluation of chip performance metrics [3] Group 2: TPU's Cost Reduction - Google/Broadcom's TPU has significantly reduced its inference cost, with the transition from TPU v6 to TPU v7 resulting in a 70% decrease in unit token inference cost, making it competitive with NVIDIA's GB200 NVL72 [1][4] - The cost reduction in TPU v7 is attributed to system-level optimizations rather than a single technological breakthrough, indicating that future cost reductions will depend on advancements in adjacent technologies [4] Group 3: Competitive Landscape - Despite TPU's advancements, NVIDIA maintains a time-to-market advantage with ongoing product iterations, which are crucial for customer retention [5][6] - The investment outlook remains positive for both NVIDIA and Broadcom, with Broadcom's earnings forecast for FY2026 raised to $10.87 per share, reflecting its strong position in AI networking and custom computing [7] Group 4: Industry Dynamics - The report suggests a clearer division of labor within the industry, where GPUs continue to dominate training and general computing markets, while custom ASICs penetrate predictable inference workloads [7][8] - The significant drop in TPU costs serves as a critical stress test for the viability of AI business models, highlighting the importance of economic considerations in the ongoing GPU vs. ASIC competition [8]

Summary of Key Points from the Conference Call Transcript Industry Overview - **Industry**: China Data Center Equipment - **Key Focus**: Power and cooling upgrades, particularly related to NVIDIA's Vera Rubin platform launch in 2026 [2][3] Core Insights and Arguments - **NVIDIA's Rubin Platform**: - The Rubin platform is set to double the rack power compared to the previous Blackwell platform, moving from 80% to 100% liquid cooling [2] - Rubin is currently in full production [2] - **VR200 NVL72 Specifications**: - Delivers approximately 3.5x to 5x the training/inference AI computing power compared to GB300 NVL72, significantly increasing rack-level power demand [3] - Upgrades power shelves to a 3*3U 110 kW configuration, compared to the GB300's 8*1U 33 kW shelves [3] - Introduces a 3+1 redundancy design for power shelves [3] - **Future Developments**: - Anticipation of transitioning to the next-generation Kyber rack design, which will support 800V HVDC to meet rising power requirements driven by AI compute scaling [3] - Potential launch of Rubin Ultra in 2027 could further unleash HVDC demand [3] Investment Opportunities - **Stock Picks**: - **Kstar and Kehua**: Expected to benefit from stronger UPS demand and potential upgrades to HVDC and SST [4] - Kstar is anticipated to have strong order intake from US hyperscalers due to strategic partnerships [4] - Kehua's domestic GPU ramp-up and H200 shipment could enhance order intake from domestic hyperscalers [4] Risks and Valuation - **Downside Risks for Data Center Equipment Sector**: - Slower-than-expected AI data center capacity growth [6] - Slower penetration of high-power density products [6] - Challenges in gaining market share in the overseas AIDC equipment supply chain [6] - **Kehua's Price Target and Risks**: - Price target based on DCF methodology; downside risks include slower IDC capacity expansion and lower overseas shipment expectations [7] - Upside risks include faster IDC capacity expansion and stronger relationships with hyperscalers [7] - **Kstar's Price Target and Risks**: - Similar DCF methodology for price target; downside risks include slower IDC capacity expansion and new entrants in the market [8] - Upside risks include faster IDC capacity expansion and higher overseas shipments [8] Additional Important Information - **Analyst Team**: The report was prepared by UBS Securities Asia Limited, with analysts Yishu Yan, Anna Yuan, and Ken Liu involved [5] - **Valuation Methodology**: The report emphasizes the use of DCF methodology for price targets and highlights the importance of understanding risks before making investment decisions [7][8] This summary encapsulates the critical insights and data from the conference call, focusing on the developments in the China data center equipment industry, particularly regarding NVIDIA's advancements and the implications for investment opportunities in Kstar and Kehua.

AI server assembler Wiwynn’s recent 4Q25 gross margin miss (7.2% vs. consensus estimates of 8–8.3%) has triggered a share price correction and reignited investor concerns. While rising component costs can mechanically dilute reported gross margins, it is more critical to examine the underlying profitability trends in server assembly through the lens of structural changes in AI server design.Nvidia continues to increase the level of design integration in AI servers to boost token output per unit of space and ...

近期AI伺服器組裝廠商緯穎公布低於預期的4Q25毛利率 (7.2% vs. 市場預期的8-8.3%)，導致股價下跌，此事又引起投資人關注。雖零組件漲價也會稀釋帳面上的毛利率，但更重要的是從AI伺服器設計的本質上去檢視組裝的獲利能力趨勢。Nvidia持續提升AI伺服器的設計整合程度，以提升每單位空間與電力的Token產出，來因應資料中心有限空間與稀缺電力挑戰。此外，高度整合的設計也有助於改善生產效率，進而有利供應鏈管理與降低維修成本。VR200 NVL72就是設計高度整合的例子，我的產業調查顯示，其Compute tray導入無線化設計（Cable-less），零組件項目顯著減少約40% (vs. GB300 NVL72)。但隨整合程度提高，伴隨而來的是Nvidia指定用料比重提升，與客製化空間也遭到壓縮，這些都不利組裝廠商的獲利能力。我的產業調查顯示，VR200 NVL72 compute tray「不允許客製化規格」的零組件項目佔比提升至20-22%，遠高於GB300 NVL72的5-7%，且VR200 NVL72更直接取消「ODM設計」的零組件項目。事實上，提高整合程度的設計趨勢自GB300 NVL72就 ...

Key Updates on NVIDIA Vera Rubin/VR200 NVL72(Integrating my latest supply chain checks and NVIDIA CEO Jensen Huang’s CES 2026 keynote)1. NVIDIA has renamed the AI server VR200 NVL144 (die-based) to VR200 NVL72 (package-based). My supply chain checks indicate VR200 NVL72 will be offered in two power profiles: Max Q and Max P.➢ Jensen referenced NVL72, rather than the previously used NVL144 naming.➢ Max Q and Max P share the same hardware design.➢ Max Q GPU/rack power (TGP/TDP): ~1.8/190 (kW).➢ Max P GPU/rack ...