In a conversational interview, Kevin Surace explains how Nvidia, originally a graphics‑processing unit (GPU) maker for video games, unintentionally became the dominant hardware platform for artificial intelligence. He outlines the technical reason: GPUs excel at vector‑math operations, which are also the core of deep‑learning workloads. Engineers building AI models turned to GPUs because CPUs were too slow, and Nvidia’s CUDA software stack quickly became the de‑facto standard for training and inference. Surace notes that competitors like AMD, Intel, and newer AI‑specific chips such as Google’s TPUs arrived later, but Nvidia’s early foothold turned into a massive market shift. The company’s valuation exploded to $3‑$4 trillion, surpassing the combined worth of all other semiconductor firms, a growth Surace describes as both accidental and strategic.
Chapters
Nvidia is now worth more than every other chip company on Earth combined. And the funny part? The chips were never built for the reason they became so valuable.
So being in Silicon Valley, we knew the company, right? And the thing was, and I even bet against them a couple of times, the thing was is they're in the GPU business and GPUs were needed to put graphics on a screen. That's the business they're in. Here's the background on that. So they had that, AMD had GPUs, Intel to a lesser extent. And what happened is as AI came along, the math to do AI, which is vector math, is similar enough to the way a GPU A CPU processes vectors for the screen, basically math for the screen, right? And so engineers developing AI and GenAI and transformers and deep learning said, well, what silicon is out there that we can use? CPUs are dead slow. They don't process that math in the right way. Well, there's these GPUs that are putting graphics on screen. You know what? We can adapt what we do to that and it'll work pretty well. And so then that became the standard through CUDA, which is a software interface that NVIDIA has, became the standard way to do training and inference because there was nothing else out there. Now there's TPUs coming, there's other stuff coming. You know, Nvidia fell into this. They didn't come out with a GPU and say, this is for AI. They said, this is for graphics. And video games. And video games and those kinds of things. And then they realized over in this corner, it's all these weird people doing this odds of deep learning or something, and they're using our chips. And they didn't pay much attention because they used 10 chips one year, then they used 10,000, and then they used a million. They go, hold it, that is the market. The heck with graphics, right? So they fell into it smartly. He's a great manager, great CEO. But no one thought that that company would be worth $3 or $4 trillion, which is worth more than the entire rest of the semiconductor industry combined.
