The GPU revolution

A failed first chip, a rescued company, and the product that named an industry.

NV1 (1995): the elegant mistake

NVIDIA’s first product, the NV1 (1995), was a technical achievement pointed at the wrong target. It rendered 3D graphics using quadratic texture mapping — mathematically elegant, but incompatible with the triangle-based polygons that Microsoft’s DirectX was establishing as the industry standard, and that rival 3dfx was riding to dominance.

The market chose triangles. The NV1 died commercially. It was, as Huang later put it in retellings, a bet on being right about the technology and wrong about the standard — a distinction that nearly ended the company.

The Sega contract (1996–1997)

NVIDIA’s lifeline was a contract to build the graphics chip for Sega’s next game console. But the same standards problem applied: NVIDIA couldn’t deliver a DirectX-compatible part on Sega’s timeline.

What happened next is one of the most-retold stories in Silicon Valley. Huang went to Sega, admitted NVIDIA couldn’t fulfill the contract, and asked to be released from it. Sega’s CEO reportedly agreed — and, according to Huang’s retellings, Sega paid NVIDIA roughly $5 million anyway for work done, effectively funding the company’s survival. The figure is reported, not confirmed in filings; the release from the contract is the documented core.

Huang tells the story as a lesson in radical honesty with partners — and in the value of a reputation worth more than a contract.

RIVA 128 (1997): the rescue

With Sega’s money and a skeleton crew — the company made painful layoffs — NVIDIA bet everything on one chip: the RIVA 128, built for the DirectX triangle standard this time.

It shipped in August 1997 and reportedly sold over one million units in its first four months. The revenue funded the next generation and ended the existential crisis. From this era comes the internal mantra Huang still quotes: “Our company is thirty days from going out of business” — a slogan, not an audited fact, but one he says kept the company paranoid and hungry.

GeForce 256 (1999): naming the category

In October 1999, NVIDIA launched the GeForce 256, which the company marketed as “the world’s first GPU” — the first chip to integrate transform, lighting, and rendering on a single processor.

The claim is debated: historians of computing note that earlier chips performed graphics acceleration, and “GPU” as a term has fuzzy boundaries (see The myth machine). But the product’s impact isn’t debated. The GeForce line made NVIDIA the dominant force in PC graphics, outlasting 3dfx (whose assets NVIDIA acquired in 2000) and ATI (later bought by AMD in 2006).

Going public (1999)

NVIDIA went public in January 1999, giving it the currency — cash and stock — to survive the capital-intensive chip business. In semiconductors, each new chip generation costs more than the last; the IPO funded the treadmill.

What the revolution established

1. The dedicated GPU as a product category. Before NVIDIA, graphics was a feature of other chips. After, it was an industry.
2. The annual cadence. NVIDIA learned to ship a new architecture on a relentless schedule — the operational muscle that later let it ride the AI wave.
3. The TSMC partnership. NVIDIA was “fabless” — it designed chips and had Taiwan’s TSMC manufacture them. The Huang–TSMC relationship, and Huang’s personal rapport with TSMC founder Morris Chang, became one of the most important alliances in tech.
4. Paranoia as culture. “Thirty days from going out of business” outlived the crisis that coined it.

Source notes

Product history draws on NVIDIA’s corporate records, contemporaneous press coverage of the NV1, RIVA 128, and GeForce 256 launches, and Huang’s interviews (including the Sequoia oral history). The $5 million Sega figure and RIVA 128 unit numbers are reported in profiles and retellings, not confirmed in filings. The “world’s first GPU” debate is covered on The myth machine.