Every year, hundreds of eager buyers flock to Best Buys or constantly refresh online shopping websites, all for one purpose – to get the latest graphics processing unit on launch day. While most casual users incorporate their GPUs into personal PCs for gaming, they serve other purposes as well.
With artificial intelligence rapidly advancing and blockchain soaring in popularity, the demand for processing power has become higher than ever. That’s where GPUs come in, able to train language models and even mine Bitcoin. The need for powerful GPUs is also what drives the success of semiconductor company NVIDIA.
NVIDIA dominates the high-performance GPU market, so they can exert pricing pressure on their customers. Their gross profit margins have risen from just over 60% last January to nearly 80% as of the earnings call on Feb. 21, revealing the extent of their market power.
NVIDIA’s fourth quarter revenue exceeded analyst predictions by over 10 percent, beating Wall Street’s already high expectations for the company. Compared to the fourth quarter of 2022, its net income increased by over 700 percent, owing largely to high demand for its high-performance computing chips. In particular, the GH100, one of the most capable AI chips for servers, significantly contributed to strong sales.
Sophomore and tech enthusiast Sahil Jain has purchased NVIDIA’s graphics cards. He feels that their unique innovations set them apart from rivals like AMD.
“No one else really is competing with NVIDIA because they’re just the highest up and they have the best technology,” Sahil said. “There’s a lack of competition since other companies are so far behind compared to them in the AI game and the GPU game. NVIDIA is really the only good seller, and that’s allowed them to get so big.”
Founded in 1993 by American electrical engineers Jensen Huang, Curtis Priem and Chris Malachowsky, NVIDIA established itself as a key industry player with the release of their RIVA graphics processors and subsequently the GeForce 256, often called the world’s first GPU, a term popularized by NVIDIA. Before this point, scientists and engineers had mainly used the central processing unit (CPU) for performing tasks, designed by companies like Intel, but the GPU soon surpassed the CPU for tasks involving high volumes of instructions.
Three factors make NVIDIA GPUs especially unique. Firstly, the hardware and chips built into each processing unit are extremely fast, with their H100 chip providing 3 terabytes per second of bandwidth. Secondly, users can also customize their GPUs, tailoring the internal settings to specific applications, whether that be gaming or AI. This feature contributes to the general versatility of NVIDIA’s product, adding optimizations that can speed up processing. Finally, NVIDIA GPUs can connect to one another in a seamless manner, allowing them to run in parallel which merges their computation power. Considering how NVIDIA is developing chips specifically for AI functionalities, upper school economics teacher Dean Lizardo believes that AI will drive the decisions of most, if not all, technology companies in the future.
“As AI continues to grow, my guess is that you’re going to see a lot of industries or a lot of businesses crop up to try and fill this gap if there is any,” he said. “And so I think you will see tech find a way to make many, many use cases out of AI. It’s just a matter of time.”
In the past two decades, GPU computing has transformed into the mainstream option for processing. The soaring popularity of gaming, blockchain, and learning models has only increased the demand for NVIDIA’s processing units, snowballing their success.
“I think that this momentum is likely to continue if you look at the revenues from data centers which heavily rely on Nvidia’s GPUs,” Harker Oeconomia president Andy Chung (11) said. “It just looks as if their revenue growth and profit growth is going to continue to increase. I’m very, very excited to see what they can do.”