For those who played games in the past, the graphics processing unit, or GPU, was a unique computer component. For them, the GeForce series, starting with the GeForce 256, represented a splendid history of computer graphics.

It has become common sense that you need to use a good GPU as much as you use a good CPU, and NVIDIA, the maker of the representative GPU, has also become engraved in people's minds.

But now, NVIDIA has become one of the most talked-about companies in artificial intelligence (AI), a future technology trend. At this point, people are left with both questions and surprise. NVIDIA, a company that used to make PC graphics card chipsets, is now a leading company in the AI field? How, and when did that happen?

When did NVIDIA first realize that GPUs, which were originally intended to simply assist CPUs in multimedia tasks like gaming and video editing, could become a key player in AI? Mark Hamilton, vice president of solutions architecture and engineering at NVIDIA, says it was 11 years ago when NVIDIA first shipped CUDA, a parallel computing architecture that improves computing performance based on GPUs. He says that realizing that GPUs could run general-purpose programs, not just graphics codecs, was the first time NVIDIA realized it needed to expand beyond its role as a graphics card manufacturer and into the AI field.

However, it was around 2012 that I became truly confident that we would become an AI company. Geoffrey Hinton, a computer scientist at the University of Toronto, demonstrated outstanding image recognition technology by combining GPU and AI technologies at the ImageNet competition, a computer science event. That same year, Google, in collaboration with Stanford Professor Andrew Ng, conducted a project to identify all objects on YouTube, once again demonstrating the power of GPU-based AI. In the 2013 ImageNet competition, 300 out of 400 participating companies used GPU technology, and in the 2014 competition, all 400 participating companies used GPU technology, proving the power of GPU-based deep learning.

GPUs excel at deep learning compared to CPUs due to their different computational methods. Deep learning, which requires the simultaneous processing of massive amounts of information, is advantageous on GPUs, which contain hundreds or thousands of cores, compared to CPUs, which rely on serial processing.

In an era where computing power increases by 500% every year, CPUs can't keep up.

It is not unrelated that NVIDIA CEO Jensen Huang declared at the NVIDIA GPU Technology Conference (GTC) held on the 10th that “Moore’s Law,” which has been synonymous with increasing integration for decades, should now be applied to GPUs. In his keynote speech, CEO Jensen Huang said, “The advancement of GPU performance redefines Moore’s Law. That is why we exist: to find the path forward after Moore’s Law.”

CPU performance is only increasing by about 10% per year, making Moore's Law seem meaningless. In this era of computing power increasing by 500% every year, a huge gap is emerging, and the explanation is that GPUs can fill that gap.

This doesn't mean NVIDIA wants to compete with CPUs by promoting GPUs. GPUs are, after all, accelerators and always perform best when paired with CPUs. Vice President Mark Hamilton emphasized that NVIDIA's success in the AI field and its yearly performance improvements stem from its position as an AI computing company that integrates hardware and software, rather than simply providing GPU chips.

So, NVIDIA wants to call itself "a platform company enabling GPU-based accelerator computing." As if to prove this point, NVIDIA emphasizes that it invests in "just one" area: GPU computing, specifically accelerator computing, and seems to believe that this can enhance AI capabilities.

Last year alone, $5 billion was reportedly invested in AI startups. Now, the world is talking about AI convergence, as if it began and will end with AI. It's unclear whether NVIDIA's GPU computing will become the core of AI, or whether another chip or software-based technology will become the key. One important point is that GPU-based computing power is currently maximizing AI capabilities, and NVIDIA is at the center of this.