When I was a teenage computer geek building my own desktop machines to play multiplayer video games, there was one brand of graphics card I always bought – Nvidia. Intel was the big name in central processing units (CPUs) in the 1990s, making the chips that ran the bulk of the world’s PCs, including my well-used Pentium 486 machine. But Nvidia, founded in California in 1993, won acclaim developing a separate graphics processing unit (GPU) to handle complex animation and 3D graphics.
Its technology made video games look slick. By the early 2000s, Nvidia was the go-to name in graphics chips, fuelling many late-night gaming sessions all over the world. Last month, Nvidia, which is publicly listed on the Nasdaq exchange, became just the fifth company in history to surpass US$2 trillion in market capitalisation, alongside the likes of Apple and Microsoft.
Imagine: one computer component maker is worth eight years of New Zealand’s entire gross domestic product. But it wasn’t video games that got Nvidia into the $2 trillion club. It turns out that the speed with which its GPUs can render 3D graphics on screen can be applied to other useful tasks.
Unlike a CPU, which runs numerous complex operations at once to keep your computer functioning, GPUs focus on doing simpler computational tasks, like figuring out how to display an image on a computer screen. This single-mindedness has, in the past decade or so, been applied to mining the Bitcoin cryptocurrency, which involves solving complex mathematical equations at high speed to unlock fractions of a Bitcoin, a process essential to the system’s operation and stability.
Nvidia’s chips turned out to be highly suited to the job. At one stage, Chinese Bitcoin miners were flying plane loads of Nvidia GPUs to mainland China to set up new mining operations. The huge power consumption of mining Bitcoin unnerved the Chinese Government, which ended up banning the practice.
Nvidia’s sales nevertheless soared on growing global demand. Then the artificial intelligence revolution cranked up a gear in late-2022 with the arrival of ChatGPT and other intelligent chatbots. Nvidia’s GPUs proved adept at rapidly training the large language models (LLMs) that underpin services like ChatGPT. Since then, tech giants such as AWS, Microsoft and Google have been buying up Nvidia GPUs in vast quantities.
Meta, the owner of Facebook, is in the process of building the world’s most powerful AI supercomputers for its AI operations. It is buying 350,000 of Nvidia’s most powerful H100 graphics cards as part of the multibillion-dollar project.
Nvidia now controls more than 80% of the market for GPUs for AI applications. That’s a concerning concentration of power, which is why Microsoft and Amazon are now building their own silicon AI chips to avoid being totally reliant on Nvidia to power their AI services.
Nvidia’s founder and chief executive, Jensen Huang, is now being recognised for the shrewd businessman that he is, having steered the company along a 30-year journey to computer chip supremacy.
Last month, he urged nations not to rely on Big Tech companies such as Microsoft and Google, or AI pioneer OpenAI, to provide large language models. He suggested they develop sovereign AI, owned by the people and reflecting the culture and priorities of each country, rather than those of Silicon Valley. That’s a goal worth pursuing, even for a small country like New Zealand.
Can Huang continue Nvidia’s stellar run? His company has the technical edge for now and a backlog of orders as owners of data centres upgrade their equipment for AI tasks. But fellow members of the trillion-dollar club are hot on his tail.
