第一部分：战略篇——AI时代的先知与赌徒 | Part One: Strategy—The Prophet and Gambler of the AI Era

算力觉醒：英伟达的生态战略与AI霸权

第一部分：战略篇——AI时代的先知与赌徒

Part One: Strategy—The Prophet and Gambler of the AI Era

第一章：创世：三个人的赌局与图形革命的序章

Chapter 1: Genesis—The Gambit of Three and the Overture of Graphics Revolution

---

1993年的硅谷正处于一个微妙的转折点。互联网浪潮尚未完全席卷这片创新之地，个人计算机革命刚刚走完第一个十年，而所谓的"多媒体革命"正从概念走向现实。在这样的时代背景下，1993年1月，一个注定改变整个计算产业走向的公司悄然成立。这家公司的名字——NVIDIA——将很快与GPU（图形处理器）这一革命性技术紧密相连，并在三十年后成为AI时代最具影响力的基础设施供应商。

The Silicon Valley of 1993 stood at a delicate turning point. The internet wave had not yet completely swept through this land of innovation; the personal computer revolution had just completed its first decade, and the so-called "multimedia revolution" was transitioning from concept to reality. Against this backdrop, in January 1993, a company that would inevitably reshape the entire computing industry quietly came into being. The name—NVIDIA—would soon become inseparable from GPU (Graphics Processing Unit) technology, and three decades later, it would emerge as the most influential infrastructure provider of the AI era.

然而，回顾英伟达的创立历程，我们发现这段历史远比表面上看到的复杂和戏剧性。在公司正式成立的背后，是三位创始人对于图形计算未来的深刻洞察，是一次对技术发展趋势的大胆押注，更是一场在巨头夹缝中求生存的艰难博弈。本章将深入剖析英伟达创立的历史背景，解读创始人黄仁勋的战略愿景，探索这家日后成为算力代名词的公司如何在创立初期就埋下了成功的种子。

However, looking back at NVIDIA's founding journey, we discover that this history is far more complex and dramatic than it appears on the surface. Behind the company's formal establishment lay the profound insights of three founders into the future of graphics computing, a bold bet on technology development trends, and a fierce struggle for survival in the gaps between giants. This chapter will deeply analyze the historical background of NVIDIA's founding, interpret founder Jensen Huang's strategic vision, and explore how this company, destined to become synonymous with computing power, planted the seeds of success from its very beginning.

---

一、时代的呼唤：1993年的技术图景

I. The Call of the Era: The Technological Landscape of 1993

要理解英伟达的创立，我们必须首先回到1990年代初期的技术环境中去。那时的计算机产业正处于从命令行时代向图形时代过渡的关键阶段，而这一过渡的核心驱动力正是图形用户界面（GUI）的兴起和3D图形技术的萌芽。

To understand NVIDIA's founding, we must first return to the technological environment of the early 1990s. The computer industry at that time was at a critical juncture transitioning from the command-line era to the graphics era, and the core driving force behind this transition was the rise of Graphical User Interfaces (GUI) and the germination of 3D graphics technology.

1.1 从2D到3D：图形革命的暗流涌动

1.1 From 2D to 3D: Undercurrents of the Graphics Revolution

1990年代初的个人计算机市场，2D图形处理已经相对成熟。微软的Windows操作系统从1990年的Windows 3.0开始逐步普及，桌面 metaphors 和窗口化的用户界面已经成为主流。然而，真正的技术革命正在3D图形领域悄然酝酿。

In the early 1990s personal computer market, 2D graphics processing had already matured relatively well. Microsoft's Windows operating system had been gradually popularizing since Windows 3.0 in 1990, and desktop metaphors with windowed user interfaces had become mainstream. However, the true technological revolution was quietly brewing in the field of 3D graphics.

1993年的游戏产业正处于一个特殊的历史节点。家用游戏机领域，任天堂的Super Nintendo和世通的Genesis正在激烈竞争；而在PC领域，id Software的Wolfenstein 3D（1992年发布）刚刚掀起了3D游戏的序幕，DOOM（1993年发布）更是将第一人称射击游戏推向了新的高度。这些游戏对3D图形处理能力提出了前所未有的需求，传统的CPU（中央处理器）已经难以满足日益增长的图形渲染需求。

The gaming industry in 1993 was at a special historical juncture. In the home console market, Nintendo's Super Nintendo and Sega's Genesis were locked in fierce competition; in the PC domain, id Software's Wolfenstein 3D (released in 1992) had just kicked off the 3D gaming era, and DOOM (released in 1993) pushed first-person shooter games to new heights. These games posed unprecedented demands for 3D graphics processing capabilities, and traditional CPUs were already struggling to meet the increasingly demanding graphics rendering requirements.

这背后的技术逻辑是清晰的：3D图形渲染涉及大量的并行计算任务，包括顶点变换、三角形光栅化、纹理映射、光照计算等。这些任务具有天然的并行性——每个像素的计算彼此独立，非常适合通过专用硬件来加速。相比之下，传统CPU的架构设计更适合处理串行任务，其有限的晶体管预算需要权衡缓存、分支预测、执行单元等多种资源，难以针对图形渲染进行专门优化。

The technical logic behind this was clear: 3D graphics rendering involves massive parallel computing tasks, including vertex transformation, triangle rasterization, texture mapping, lighting calculations, and more. These tasks have natural parallelism—each pixel's calculation is independent of others, making them ideal candidates for specialized hardware acceleration. In contrast, traditional CPU architectures are better suited for serial tasks; their limited transistor budgets must balance caches, branch prediction, execution units, and various other resources, making it difficult to optimize specifically for graphics rendering.

然而，1993年的图形处理器市场几乎是一片空白。当时的图形处理主要依赖两种方案：一是使用专门的图形扩展卡，如IBM的8514和S3的ViRGE，但这些芯片功能有限，主要用于2D加速；二是使用软件渲染，将图形计算完全交给CPU处理，典型的如Windows的GDI（图形设备接口）和各种游戏引擎的软渲染模块。市场上缺乏专门为3D图形设计的高性能硬件，这既是挑战，也是巨大的机遇。

However, the graphics processor market in 1993 was almost a blank slate. At that time, graphics processing primarily relied on two solutions: using specialized graphics accelerator cards like IBM's 8514 and S3's ViRGE, but these chips had limited functionality and mainly handled 2D acceleration; or using software rendering, where all graphics calculations were handled entirely by the CPU, typical examples being Windows' GDI (Graphics Device Interface) and various game engines' software rendering modules. The market lacked high-performance hardware specifically designed for 3D graphics—this was both a challenge and a tremendous opportunity.

1.2 技术生态的成熟：图形革命的基础设施条件

1.2 Maturation of the Technology Ecosystem: Infrastructure Conditions for the Graphics Revolution

英伟达的创立并非空中楼阁，其背后是整个技术生态系统的逐步成熟。在1990年代初期，几个关键的技术条件已经就绪，为图形处理器的发展奠定了基础。

NVIDIA's founding was not built on nothing—it was underpinned by the gradual maturation of the entire technology ecosystem. In the early 1990s, several key technical conditions had fallen into place, laying the foundation for graphics processor development.

首先，半导体制造工艺的进步使得在单个芯片上集成大量晶体管成为可能。1993年，英特尔刚刚推出486DX2处理器，采用0.8微米工艺制造，集成约120万个晶体管。而到了1990年代中期，0.5微米乃至0.35微米的工艺开始普及，这为复杂图形处理器的设计提供了必要的硬件基础。晶体管密度的提升意味着可以在单个芯片上实现更多的渲染管线、更多的纹理单元和更大的帧缓冲。

First, advances in semiconductor manufacturing processes made it possible to integrate vast numbers of transistors onto a single chip. In 1993, Intel had just released the 486DX2 processor, manufactured using 0.8 micron process technology and integrating approximately 1.2 million transistors. By the mid-1990s, 0.5 micron and even 0.35 micron processes began to proliferate, providing the necessary hardware foundation for complex graphics processor designs. Increased transistor density meant more rendering pipelines, more texture units, and larger frame buffers could be implemented on a single chip.

其次，PC架构的标准化降低了图形卡设计的门槛。IBM PC/AT兼容机成为事实上的行业标准，PCI（外设组件互连标准）总线在1993年已经开始普及，取代了之前的ISA和VLB（VESA本地总线）标准。PCI总线提供了更高的带宽（132MB/s的理论峰值），能够更有效地在CPU和图形卡之间传输数据。这种标准化的PC架构为图形处理器供应商提供了稳定的开发目标，降低了硬件和软件适配的复杂度。

Second, the standardization of PC architecture lowered the barrier to entry for graphics card design. IBM PC/AT compatible machines had become the de facto industry standard, and the PCI (Peripheral Component Interconnect) bus had begun proliferating in 1993, replacing the earlier ISA and VLB (VESA Local Bus) standards. The PCI bus provided higher bandwidth (132MB/s theoretical peak), enabling more efficient data transfer between CPU and graphics card. This standardized PC architecture provided graphics processor vendors with stable development targets and reduced the complexity of hardware and software adaptation.

第三，图形API的演进为软件与硬件的解耦创造了条件。1992年，Silicon Graphics推出了OpenGL（开放图形库），这是一种跨平台的图形API，很快成为专业图形领域的标准。OpenGL定义了从应用程序到图形硬件的软件接口抽象层，使得开发者可以编写与具体硬件无关的图形代码。这意味着图形芯片供应商可以在硬件层面进行创新，而不需要等待软件开发商为每种新硬件单独适配。

Third, the evolution of graphics APIs created conditions for software-hardware decoupling. In 1992, Silicon Graphics released OpenGL (Open Graphics Library), a cross-platform graphics API that quickly became the standard in professional graphics. OpenGL defined an abstraction layer for s