英伟达最新财报再度背书AI繁荣? 高管透露明年目标、回应AI“循环交易”

Core Viewpoint - Nvidia's Q3 FY2026 financial report shows record revenue of $57 billion, reinforcing the growth of the AI market amid skepticism about an AI bubble [1][2]. Financial Performance - Nvidia reported Q3 revenue of $57 billion, a 22% increase from the previous quarter and a 62% increase year-over-year [2]. - Net profit reached $31.91 billion, up 65% year-over-year, marking the highest growth rate for the year [2]. - Data center revenue hit a record $51.2 billion, growing 66% year-over-year, while gaming revenue was $4.3 billion, up 30% [2]. Future Projections - Nvidia's CFO projected that revenue from the Blackwell and Rubin platforms could reach $500 billion by December 2026, with potential for further growth [2]. - The company plans to use the anticipated cash flow for growth support, stock buybacks, and investments to expand the CUDA ecosystem [2]. Profitability Metrics - The GAAP gross margin for Q3 was 73.4%, a decrease of 1.2 percentage points year-over-year, attributed to a shift in business model and a $4.5 billion expense related to H20 [3]. - Nvidia aims to improve gross margin to around 75% by the end of the fiscal year through cost improvements and business structure optimization [3]. Regional Revenue Insights - Nvidia's revenue from the U.S. nearly doubled to $39.177 billion year-over-year, while revenue from China fell over 60% to $2.973 billion due to geopolitical issues and market competition [4][6]. - The company is committed to engaging with the Chinese government to maintain a competitive position in the AI computing sector [6]. Strategic Partnerships - Nvidia has established significant partnerships with AI leaders, including a $100 billion investment agreement with OpenAI and collaboration with Microsoft on Anthropic [4]. - These partnerships are part of a "circular" AI investment strategy, enhancing Nvidia's platform capabilities across various AI models [4]. Challenges and Market Dynamics - Nvidia faces challenges from the need to develop comprehensive systems rather than just GPUs, as the complexity of AI models and memory requirements increases [5]. - The company emphasizes that its CUDA architecture can adapt to diverse workloads, unlike ASICs, which struggle to meet the evolving demands of AI [5].