Graphics Processing Units (GPUs) made Nvidia and its investors a fortune, but why are they important and are there other stocks in that space?

10-15 years only ago, only professional techheads would know what Graphics Processing Units (GPUs) are, but everyone does now. In this article, we’ll look at what they are, how they made Nvidia a fortune and other players in the space.

What are the Best ASX AI Stocks to invest in right now?

Check our buy/sell tips

What are Graphics Processing Units?

To understand GPUs, it is first important to recognise CPUs (Central Processing Units). It is essentially the main processor of a computer or device driven by a computer. GPUs were originally designed to complement them, specifically to render images and video for gaming and graphic applications.

As gaming and graphics applications need to perform massive numbers of parallel computations, that is just what GPUs can do. They can handle thousands of simple tasks at once, ideal for parallelizable workloads like image rendering or matrix calculations.

The AI boom has led to GPUs going from a niche application to the mainstream. Modern AI, especially deep learning models like ChatGPT or autonomous driving systems, require huge amounts of matrix math. Training large models (like GPT) requires thousands of units. Inference (running models once trained) is also faster with them.

Cloud computing has been an important application too as well as certain types of crypto mining (specifically Ethereum). Gaming and graphics are still a major market, but it is AI that has investors excited.

How did they make Nvidia a fortune?

GPUs single-handedly made Nvidia the biggest in the world. Or specifically, CUDA which is a proprietary programming platform Nvidia launched in 2006 that let developers use GPUs for general-purpose computing, not just graphics. This was a turning point for the entire technology as CUDA made Nvidia GPUs programmable like CPUs. Suddenly, GPUs weren’t just for gaming — they were supercomputers.

Having had early success in the PC gaming market, early 2010s AI researchers found Nvidia’s GPUs were perfect for training neural networks. And so Nvidia built an entire system, not just chips but entire AI platforms and its customers were (and still are) buying thousands at a time.

It launched the A100 and H100 GPU lines, purpose-built for AI training and inference. These chips are used in supercomputers, cloud platforms (AWS, Azure, Google Cloud), and AI labs. Then it launched the Nvidia AI Enterprise software suite (for enterprise AI workloads) as well as Omniverse, a platform for simulation, robotics and digital twins.

Is anyone else in this space?

Yes, there are plenty of companies exposed. The key competitors to Nvidia are AMD (Advanced Micro Devices) and Intel. It is the former that is the bigger threat, but Intel’s Habana Labs acquisition for AI chips has indicated that Intel may become a bigger threat in the future. Cloud providers are a possibility, some like Amazon just buy Nvidia’s chips but others make their own (Google being one).

Also look to chip makers like TSMC, ASML and Candence. These companies make chips, chip-making machines and semiconductor design software tools respectively.

Can’t decide on any one player but want broader exposure? Investing in Semiconductor or Robotics ETFs might be worth considering.

Conclusion

There is an exciting opportunity in this space, but investors should not mistakenly think Nvidia is the one and only opportunity here – there are other companies with direct and indirect exposure to consider.

The challenge for investors is to find the right company and have patience to stick it out for the long-haul. Because Nvidia has not built its dominance in the short-term, as much as the contrary may appear to be the case. But whichever companies win or how much longer Nvidia reigns at the top, it is clear that GPUs are here to stay.