Nvidia(US:NVDA) BOCOM International·2025-05-30 00:23Investment Rating - The report assigns a "Buy" rating for NVIDIA (NVDA US) with a target price of $175.00, indicating a potential upside of 29.8% from the current price of $134.81 [10][11]. Core Insights - The report highlights the significant demand for reasoning applications, which is expected to drive long-term computational resource needs. The management emphasizes the importance of iterative reasoning in enhancing AI performance, with a notable increase in token usage for reasoning applications among major cloud service providers [5][6]. - NVIDIA's Blackwell product line is projected to contribute nearly 70% of data center computing revenue, with major cloud service providers deploying an average of 1,000 NVL72 Blackwell rack systems weekly [5]. - The report anticipates that the impact of export restrictions will be less severe than previously expected, with revenue guidance for FY2Q26 set at a median of $45 billion, and a Non-GAAP gross margin target of 72% [5][6]. Financial Overview - Revenue projections for NVIDIA are as follows: - 2024: $60,922 million - 2025: $130,497 million - 2026E: $210,221 million - 2027E: $267,431 million - 2028E: $285,823 million - Year-on-year growth rates are expected to be 125.9% for 2024 and 114.2% for 2025 [4][12]. - Net profit estimates are: - 2024: $32,312 million - 2025: $74,265 million - 2026E: $115,401 million - 2027E: $155,154 million - 2028E: $162,364 million - Corresponding year-on-year growth rates are 288.2% for 2024 and 131.1% for 2025 [4][12]. - The report also notes an increase in the Non-GAAP EPS forecast for 2025 and 2026 to $4.65 and $6.31, respectively, based on a 32x average P/E ratio for those years [5][6]. Market Position and Trends - The report indicates that NVIDIA's hardware compatibility is improving, aligning with trends in open-source software, which is seen as a positive direction for technological development [5]. - The anticipated demand for AI and reasoning applications is expected to keep computational resource needs high, with a shift from single-instance reasoning to iterative reasoning and eventually to agentic reasoning [5].