NVIDIA released the GeForce 256 on October 11, 1999, and marketed it as “the world’s first GPU,” a graphics processing unit. NVIDIA defined the term as a single-chip processor that integrated transform, lighting, triangle setup and rendering engines and could process at least 10 million polygons per second. By moving the geometry and lighting math off the CPU and onto a dedicated chip, the GeForce 256 let game developers use far more polygons and produce much more detailed scenes.
The card was built for gaming, but its architecture - many small processing units running the same calculation in parallel - turned out to be the key to something much larger. NVIDIA’s own retrospective notes that the GeForce 256’s parallel-processing design “later proved ideal for deep learning,” and that researchers began using NVIDIA GPUs around 2011 to accelerate AI, eventually powering systems like ChatGPT.
The lineage runs straight from this chip to modern AI hardware. The 2012 AlexNet result that launched the deep learning era was trained on two consumer NVIDIA GPUs, and the parallel design first sold to gamers in 1999 became the substrate for the entire AI industry.
Why business readers should care: the most consequential property of the GPU - massively parallel arithmetic - was a side effect of making games look good. The GeForce 256 shows how general-purpose value can hide inside a product built for a narrow consumer market, and why NVIDIA was positioned to dominate AI a decade before AI mattered.