Last Updated on January 23, 2026 by Dave Farquhar
On October 15, 2002, 3dfx went out of business. It was really a technicality, however. 3dfx hadn’t been a viable company for about 2 years, having sold the majority of its intellectual property to rival Nvidia on December 15, 2000 after its creditors forced it into bankruptcy. In this blog post, we’ll explore what happened to 3dfx and why 3dfx failed.
Given that the rise of AI pushed Nvidia past Google in 2024 to become the third most valuable company in the world and it rose to #1 in 2025, it seems hard to believe that the two pioneering makers of what we today call a GPU struggled to make it through their first decade. Nvidia had a brush with bankruptcy too, back in 1996. What happened to 3dfx could have happened to Nvidia instead. It could have just as easily been 3dfx GPUs powering AI today.
3dfx Voodoo and Glide: Inventing a product category
3dfx was out of the gate first with the Glide API in its original Voodoo cards being supported by about 150 popular games, including Quake II. Not every gamer had a 3dfx card like the Orchid Righteous 3D, but that didn’t mean they didn’t want one. When I bought my first Nvidia-based card, I knew the 3dfx cards were better. But since first person shooter games all look the same to me–sorry not sorry–I couldn’t justify the extra cost of a 3dfx card.
The idea of accelerating 3D graphics was something multiple companies had been working on since the early 1990s. Some of them started even before 3dfx’s founding in 1994. If I say 3dfx invented the product category of 3D acceleration, I’m sure someone would get upset with me. But I think it’s safe to say 3dfx’s Voodoo cards were the first commercially successful 3D gaming accelerators.
Owning the early 3dfx cards required some dedication too. They were strictly 3D cards, with no 2D capability built in. You still needed a conventional 2D card, and the 3dfx card worked together with it. Later 3dfx cards included 2D capability, eliminating the need for a separate card for 2D.
Why 3dfx failed: Losing its way
It’s pretty clear where the turning point was. 3dfx decided to play the vertical integration game. In December 1998, 3dfx paid $141 million to purchase STB Systems Inc., one of the more highly regarded graphics card makers. This put 3dfx into the business of producing graphics cards, not just the chips that went on them.
This was key to what happened to 3dfx.
3dfx left every other maker of graphics cards with nowhere to turn except to start buying chips from Nvidia. The other maker of viable 3D accelerated chipsets at the time was ATI, and ATI also made graphics cards.
This meant Nvidia’s chips didn’t have to be the best. If Nvidia’s chips were competitive at all, every other maker of graphics cards would rather buy chips from them rather than from a competitor. And if they did end up being the fastest at any given generation, so much the better. As it turned out, the GeForce 256 was rather good.
Nvidia also didn’t have to worry about any supply chain issues beyond its own chips. If any of the graphics card makers had difficulty producing cards, that wasn’t Nvidia’s problem, as long as someone was able to produce cards.
3dfx’s move into graphics card production made life much simpler for Nvidia, and potentially much harder for itself. The gamble had upside, but it was a gamble.
How vertical integration failed 3dfx
If all went well, 3dfx stood to enjoy higher profit margins owning STB. On the other hand, if the former STB had difficulty producing cards, 3dfx was going to be in a world of hurt.
And that was what happened. STB was used to producing Nvidia-based cards, and had been Nvidia’s largest customer. The two companies didn’t integrate smoothly. Making matters worse, the former STB had quality control problems in its plant in Juarez, Mexico. The quality control issues caused some system integrators to switch to Nvidia-based cards.
3dfx and STB were together for two generations, the Voodoo 3 and Voodoo 5. Nvidia and ATI were better in some regards but not all during those generations. The Voodoo 3 and Voodoo 5 sold reasonably well, but not in high enough numbers to ensure profitability. In desperation, 3dfx discontinued card production November 16, 2000, looking to reverse the decision and sell chips to all takers again. But it was too late. What happened to 3dfx next wasn’t good.
December 15, 2000, when its creditors initiated bankruptcy proceedings, 3dfx agreed to sell its assets to Nvidia and wind down the company once the sale completed. Here’s the blog post I wrote at the time.
And that proved to be the end of the game, apart from some legal wrangling over the sale price. For 3dfx, the end came quite a bit sooner than anyone would have expected.
What might have been for 3dfx and the GPU industry
If 3dfx had not merged with STB, there is every reason to believe they would have survived at least two more chip generations. And there’s a very real possibility they would still be in business today, challenging Nvidia and AMD, the company who bought ATI in 2006.
STB had also approached Nvidia. It’s fair to speculate that an STB-Nvidia tieup would have gone better, since STB was used to producing Nvidia-based cards. But based on Nvidia’s and ATI’s success selling chips to other companies who make the graphics cards, it seems picking one line of business or the other wins out over vertical integration in this case. ATI used to make its own graphics cards, but got out of that business itself after 3dfx failed.
David Farquhar is a computer security professional, entrepreneur, and author. He has written professionally about computers since 1991, so he was writing about retro computers when they were still new. He has been working in IT professionally since 1994 and has specialized in vulnerability management since 2013. He holds Security+ and CISSP certifications. Today he blogs five times a week, mostly about retro computers and retro gaming covering the time period from 1975 to 2000.
I bought a Voodoo2 in the late 90s when 3d accelerated FPS games were all the rage and 3dfx was the market leader. Games magazines of the time were obsessed with frame rates on the likes of Quake and Quake 2, as was the young me.
Looking back, the notion of a separate card for 3d in addition to your regular SVGA card was awkward; integrated 2d/3d cards were clearly the logical way forward. (Although I do remember a brief period from 2005-10ish when the PhysX card was on the market, another attempt to get PC gamers to add an additional card for extra graphics effects.)
3dfx did produce an integrated 2d/3d card eventually but by that stage some of its competitors offered better performance. Another case of the first-mover in the market not always retaining its advantage.