In the spring of 1993, three engineers met at a Denny's restaurant in San Jose, California, and decided to start a company. Jensen Huang, Chris Malachowsky, and Curtis Priem had a shared conviction: that visual computing was about to become central to everything, and that a specialised chip for graphics processing would define the next era of computing. They founded NVIDIA with approximately $40,000 in seed capital.
Thirty years later, that company crossed a market capitalisation of one trillion dollars — the first chipmaker in history to do so. The path between that Denny's booth and the trillion-dollar milestone was not a straight line.
NVIDIA's early years included a near-fatal product decision. The company's first chip, the NV1 (1995), supported a gaming format — Sega's quadratic texture mapping standard — that the market rejected in favour of a competing approach. NVIDIA had to pivot the entire product architecture to the emerging Direct3D standard from Microsoft, requiring a rebuild that consumed the company's remaining resources. The RIVA 128, launched in 1997, succeeded. The company survived. A pattern was established: identify the inflection point before others see it, and be willing to reorganise everything to catch it.
The most consequential pivot came not in gaming but in artificial intelligence. In 2006, NVIDIA launched CUDA — a parallel computing platform that allowed developers to use GPU cores for general-purpose computation, not just graphics. At launch, it was largely a research curiosity. In 2012, a team of researchers at the University of Toronto used CUDA-powered GPUs to train AlexNet, the neural network that won the ImageNet competition by a margin that shocked the machine learning community. That result is widely cited as the beginning of the modern deep learning era.
It also established NVIDIA's GPUs as the default hardware substrate for AI research — a position the company had not explicitly designed for, but was structurally positioned to dominate. The CUDA platform meant that developers who had learned to write AI code for NVIDIA GPUs had no straightforward path to competing hardware. The moat was not the chip — it was the software ecosystem built around it over more than a decade.
By 2022, the explosion of large language models and generative AI had transformed NVIDIA from a well-regarded chip company into a critical piece of global technology infrastructure. The H100 GPU became arguably the most sought-after component in the technology supply chain, with demand outstripping supply across virtually every major cloud provider and AI research institution.
Huang has managed NVIDIA with an unusually flat management structure, reportedly maintaining direct reports across many organisational levels. He has also spoken publicly, in various executive interviews and at company events, about the importance of moving at the speed that physics and available resources allow — a philosophy that shaped NVIDIA's consistent reinvestment in architecture research even during periods when the commercial applications were not yet clear.
The NVIDIA story is, at its core, a story about recognising technological transitions before they are obvious and making irreversible commitments to them. The gaming bet in the 1990s. The CUDA platform in the 2000s. The AI infrastructure position in the 2010s. Each looked premature from the outside. Each proved to be correctly timed from the inside.