CUDA almost blew a hole in Nvidia’s finances, according to chief executive Jensen Huang.
Huang told the Lex Fridman podcast that the 2006 push to make GeForce GPUs programmable was a bet that could have cost him the company.
He wanted Nvidia to be a full computing platform outfit, not a graphics specialist flogging silicon to gamers while the serious money went elsewhere.
“The better computing company we become, the worse we become as a specialist. The more specialised we are, the less capacity we have for overall computing. The company has to find that really narrow path, step by step by step, to expand our aperture of computing, but not give up on the most important specialisation that we had.”
Huang traced the early path through programmable pixel shaders, the point at which GPUs started to look useful for more than just pushing polygons and lighting up explosions.
Nvidia’s view was that programmability could move beyond general-purpose CPUs and into GPUs, even if early GPU coding felt clunky and imprecise.
FP32 support in programmable shaders was another breakthrough because it made researchers take GPUs seriously for compute-heavy workloads.
Huang said: “It was the reason why all of the people who were working on stream processors and other types of dataflow processors discovered us.”
Then came the awkward bit, because GeForce customers were paying for frame rates, not developer toolchains, and there were “zero” expectations of payback from CUDA at launch.
Huang said: “It cost the company enormous amounts of our profits, and we couldn’t afford it at the time. But we did it anyway because we wanted to be a computing company. We increased our cost by 50% while we were a 35% gross-margin company. Our market cap went down to like one and a half billion dollars.”
He framed it as a decade-long patience game, keeping the software stack alive long before AI hype turned it into a licence to print money.
Huang said: “You could imagine that someday this would go into workstations and it would go into supercomputers, and in those segments, maybe we can capture more margin. So you could reason your way into being able to afford this. I always say that Nvidia is the house that GeForce built — because it was GeForce that took CUDA out to everybody.”
