From: acquiredfm
Nvidia, founded in 1993, has a history marked by continuous innovation and bold adaptation strategies, enabling it to transition from a highly competitive, low-margin graphics chip market to a dominant position in high-performance computing and artificial intelligence [01:37:37], [01:40:43]. The company’s co-founder and CEO, Jensen Huang, famously “bet the company” three separate times, nearly going bankrupt each instance, but ultimately securing Nvidia’s long-term success [02:35:05].
Early Challenges and First Adaptation (1993-1997)
When Nvidia began, the computer graphics chip market was “brutally competitive and low margin,” with 90 undifferentiated competitors [01:40:43], [01:43:03]. The company’s initial vision, conceived by co-founders Chris Malachowsky, Curtis Priem, and Jensen Huang over dinner at Denny’s, was to create a graphics card to accelerate 3D graphics for consumer PCs, moving beyond expensive professional workstations like those made by Silicon Graphics [02:36:18], [02:22:25].
Despite being the “first dedicated graphics card company” [02:54:06], Nvidia faced a critical misstep. An early deal with Sega to power their arcade and next-generation home consoles (which would become the Sega Saturn) led Nvidia to commit to designing chips using quadrilaterals as their polygon primitive [03:31:31], [03:42:01]. This proved problematic when Microsoft introduced DirectX, a standard API for 3D graphics on Windows, which standardized on triangles – the more fundamental building block for 3D shapes [03:48:51], [03:55:00].
By 1996, Nvidia was “up a creek,” with its Sega deal crumbling and competitors aligning with Microsoft’s standard [03:59:59], [04:02:18]. Compounding the issue, the cost of memory had dropped significantly due to Moore’s Law, making Nvidia’s memory-tight chip designs (costing 50 [04:12:00].
Facing just nine months of runway, Jensen Huang demonstrated “intellectual honesty” [04:26:00] by deciding to scrap everything and align with Microsoft’s DirectX architecture [04:21:03]. This required laying off 70% of the company, reducing staff to about 35 people [04:18:21]. To ship a new chip within this tight timeline, they made a radical decision: they would use an unproven software emulator to design and debug their chip without physical prototypes, even though the emulator ran at one frame every 30 seconds [04:47:01]. This “lunacy” was born of necessity [04:59:59].
The resulting chip, the Riva 128, launched in 1997, became a massive success, selling 1 million units within four months [05:49:59]. It was “way more powerful than any customers are telling them they want” [05:14:00]. This demanding timeline forced Nvidia to develop a unique process, allowing them to design and ship new hardware generations every six months, while the rest of the industry operated on 18-to-24-month cycles [05:59:59]. This rapid iteration became a core competency, enabling them to double performance at a given price point every six months, far outpacing Moore’s Law [05:59:59].
Second Transformation: Pioneering the GPU (1999-2006)
With the success of the Riva 128, Nvidia gained traction, leading to a partnership with TSMC, a foundry that had previously ignored Jensen Huang’s calls [01:00:59]. This “landmark deal” solidified Nvidia’s manufacturing capabilities [01:17:00].
In 1999, Nvidia rebranded its products, choosing “GeForce” for its gaming cards, a brand still in use today [01:04:11]. The launch of the GeForce 256 marked a pivotal moment: Nvidia formally coined the term “GPU” (Graphics Processing Unit) [01:04:49]. This wasn’t mere marketing bravado; it signified a fundamental shift. Prior graphics cards were “fixed-function Graphics accelerators” [01:11:11], lacking the ability to handle dynamic elements like lighting and shading, which were hardcoded at the CPU level [01:00:03].
The next major innovation was the GeForce 3, launched in 2001, which introduced “programmable shaders and lighting on the GPU” for the first time [01:09:37]. This was another “bet the company” move, requiring enormous capital investment [01:11:50], but it allowed game developers unprecedented artistic control and “meaningfully different experiences” [01:10:00], [01:08:00]. Nvidia developed its own programming language, CG (an extension of C), to allow direct GPU programming [01:11:00]. This cemented the GPU as an intelligent, programmable processing unit, distinguishing it from commodity sound or networking cards [01:11:32].
A crucial strategic partnership during this period was with Microsoft for the original Xbox, a deal worth 200 million advance [01:08:43], [01:11:00]. While financially challenging due to low margins [01:32:00], it was “absolutely the right strategic decision” for Nvidia, leveraging Microsoft’s support to further develop programmable shaders and combat Intel’s “embrace, extend, extinguish” strategy [01:31:00], [01:05:27]. Intel’s own attempts at dedicated graphics cards around this time “suck[ed],” reinforcing Nvidia’s specialized approach [01:15:32].
This era saw Nvidia achieve meteoric revenue growth, becoming the “fastest semiconductor ever to reach a billion in revenue” by 2001 [01:23:15]. By 2006, however, revenue growth had flattened, and gross margins were just 29%, compared to today’s 66% [01:30:33]. Competitors like ATI (later acquired by AMD) had caught up on programmable shaders [01:24:51].
Diversification and Future Bets (2006 Onwards)
Around 2006, Nvidia began to explore opportunities beyond gaming. An anecdote, possibly a composite of experiences but “true in spirit,” describes a quantum chemistry researcher using off-the-shelf GeForce cards and CG to complete computations in hours that previously took weeks on supercomputers [01:26:16]. This illustrated the potential of Nvidia’s transition from gaming to enterprise and scientific computing and massively parallel computing for scientific computing [01:27:40], [01:41:00].
Nvidia’s strategic vision continued to evolve towards the power of simulation, recognizing that anything that could be simulated digitally — from aerospace parts to drug discovery — could be done faster and more efficiently [01:38:40]. This foresight, born from their own near-bankruptcy experience where they relied on chip emulation, positioned them for future markets like the growth of artificial intelligence and deep learning [01:39:42].
A notable example of this foresight was Nvidia’s early investment in Keyhole (later acquired by Google to become Google Earth), a company that created a graphical model of the Earth [01:46:58], [01:51:00]. Jensen Huang saw this as a “simulation” company, demonstrating the broader relevance of computer graphics beyond video games [01:51:00].
Nvidia’s ability to “Reinventing yourself,” as Jensen Huang states, is crucial when “technology moves this fast” [01:14:01]. This continuous self-examination and re-betting of the business, initially driven by the need to survive fierce competition, became the bedrock of Nvidia’s business strategy and growth and its later dominance in AI and data centers [01:14:01].