听完，我觉得，他像在讲人类的下一种生产方式。现在，请允许我，把理解后的内容，汇报给你。 黄仁勋这次演讲，质量有点高。 10月6日，他出现在纽约，美国城堡证券（Citadel Securities）举办的一场闭门对话，对话在10天后，也 就是昨天，被公布。 台下坐着华尔街最敏锐的一群人，掌控着全球数万亿美金的资金流；台上，黄仁勋穿着那件标志性的黑 皮夹克，讲了一个横跨30年的故事。 从显卡、到加速计算、再到AI工厂，他几乎重述了整部「人工智能的演化史」。 这场对话密度，像在听一位哲学家回顾工业革命，只不过他谈是算力。最让我印象深的，是他那句几乎 带点预言意味的话： The future of computation is 100% generated.；未来的计算，将是百分之百的生成式。 01 先说说他都说了什么吧；回到了1993年，那个互联网还没普及的年代。 那时所有投资都在押CPU，因为摩尔定律还在，晶体管越做越小，性能就能翻倍。所有人都在追「更通 用、更强大的处理器」。 但他看到的了极限，他说： 通用技术的最大问题，是它往往对「极难的问题'，没那么好用」。 所以，他干了一件「反主流」的事，造一个专门为「难 ...

Core Insights - The conversation between Huang Renxun and Sequoia Capital highlights NVIDIA's evolution from a 3D graphics chip startup to a cornerstone of global AI computing [1][3] - Huang emphasizes the need to invent both technology and market simultaneously, stating that the future of AI will reshape industries worth trillions of dollars [4][10] Group 1: Founding NVIDIA - NVIDIA was founded in 1993, driven by the insight that general-purpose technology struggles with complex problems, leading to the need for accelerated computing [4][18] - The company faced a "chicken or egg" dilemma, needing a large market that did not exist at the time, which led to the creation of the modern 3D graphics video game market as a "killer application" for its technology [5][24] Group 2: Birth of CUDA - The introduction of the CUDA platform marked a pivotal shift from a hardware company to an ecosystem platform, allowing scientists to leverage GPU power for various complex problems [7][28] - CUDA served as a bridge for researchers to utilize GPU capabilities, alleviating computational bottlenecks caused by the slowing of Moore's Law [7][28] Group 3: AI Revolution - The launch of AlexNet in 2012, which achieved significant breakthroughs in computer vision using NVIDIA GPUs, marked a turning point for the company, leading to a full commitment to deep learning [8][29] - NVIDIA's development of the DGX-1, the first supercomputer designed for AI, solidified its role as a core infrastructure builder in the AI revolution [8][33] Group 4: AI Factory Concept - Huang describes the future data center as an "AI factory," where the value is measured by the computational throughput per unit of energy, transforming how infrastructure is perceived [9][37] - This new paradigm explains why major companies invest heavily in NVIDIA's infrastructure, as it serves as a direct revenue engine rather than a cost center [9][37] Group 5: Future Waves of AI - The next wave of AI will involve "digital labor" (agent AI) and "physical AI" (robotics), which will reshape industries worth trillions [10][41] - Huang envisions a future where human and digital workers coexist, enhancing productivity across various sectors [10][41] Group 6: Paradigm Shift to Generative Computing - Huang predicts a fundamental shift from "retrieval-based" to "generative" computing, where information is generated in real-time rather than retrieved [11][41] - This transformation will redefine human-computer interaction, moving towards collaborative creation rather than simple command execution [11][41] Group 7: AI Investment and Opportunities - Huang notes that AI is not just about new companies but is transforming existing large-scale enterprises, with significant revenue implications [39][40] - The emergence of AI-native companies and the shift towards AI-driven operations in major firms represent a new market opportunity worth trillions [40][41] Group 8: Robotics and Physical AI - Huang discusses the potential of robotics, suggesting that if AI can generate actions in a virtual environment, it can also control physical robots [50][51] - The future of robotics will involve multi-modal AI that can operate across various physical forms, enhancing capabilities in numerous applications [55][56]