Nvidia stock rallied on the tech giant's rise as the undisputed king of AI.
Its powerful processors designed for video games and Hollywood blockbusters provided the heavy-duty computing necessary for AI. Nvidia has leaned on another advantage: CUDA. The computing platform has enabled Nvidia to build a community of developers key to maintaining its lead in AI.
Chips And CUDA Power Nvidia
Nvidia blazed the trail by developing chips for computer graphics. These turned out to be ideal for artificial intelligence applications. But a key to Nvidia's success is the Compute Unified Device Architecture, or CUDA, the computing platform that made it easier for developers to create complex programs.
Nvidia's graphics chips solve multiple, relatively simple equations at the same time. They were ideal for creating and rendering images in video games and visual effects in movies. The processors were nearly a perfect match for AI.
Nvidia played a major role in the development of graphics processors, which were designed for heavy-duty calculations.
"When we say they are simpler, they are not complex, there's nuance to it," Anand Raghunathan, Silicon Valley chair professor of electrical and computer engineering at Purdue University, told Investor's Business Daily. "They are complex in the sense that the amount of computation needed is massive, it's tremendous."
Power Of GPUs
Graphics processors, also known as graphic processing units or GPUs, "boil down to accessing lots of data in parallel from memory and performing these parallel vectors in a matrix," he said.
"You're just multiplying and adding lots of numbers in parallel," Raghunathan added. In computing, the concept is called parallel processing, and it's something that Nvidia's GPUs are designed to handle.
"If you're trying to plow a farm, one oxen is much better than a thousand chickens," he said. "But if what you're trying to do is, you have a lot of seeds scattered all over the place and you need them to be eaten up quickly, a thousand chickens is better."
Nvidia's Developer Investments
Nvidia's lead in parallel processing helped propel its rise in AI. "This happened to be a perfect match, a match made in heaven for AI and most recently generative AI," Raghunathan said. Generative AI is built on parallel processing: accessing data from memory and processing multiple, relatively simple math equations at the same time.
But it wasn't just about the hardware architecture. "What was fortuitous and I think in hindsight looks like an absolute genius move by Nvidia was that they came up with this program framework called CUDA," Raghunathan said.
CUDA makes it much easier for developers to take full advantage of parallel processing in GPUs. Parallel processing is notoriously tricky to program, says Raghunathan. Nvidia recognized early on the need for a more programmable architecture, he said.
CUDA, which launched in 2006, simplified the difficulties of coding for parallel processing, including reducing the need for in-depth knowledge. Early versions were focused on streamlining video game physics, for example helping create realistic smoke effects or building debris.
The move was a boon for developers, who in turn built libraries of code for specific and general purposes around CUDA frameworks — libraries that weren't easy to port to competing frameworks. Meanwhile, CUDA helped Nvidia build a moat around its products, with developers who preferred building on existing CUDA libraries rather than starting from scratch.
CUDA Factor Boosts Nvidia Stock
CUDA effectively solidified the company's dominance in AI, pushing Nvidia stock to record highs. Nvidia stock retains a composite rating of 92 out of 99 and is ranked fourth out of 39 stocks in IBD's fabless semiconductor industry group, according to IBD Stock Checkup.
"If you're looking for a long-term idea, I can't think of anything better than Nvidia right now," Jay Woods, chief global strategist at Freedom Capital Markets, told Investor's Business Daily's "Investing with IBD" podcast. "Long term, you buy it, you put it away, you don't look at it."
Working With Developers
Nvidia has helped developers quickly deploy generative AI applications to accelerate their workflows.
Pinar Seyhan Demirdag, co-founder and CEO of creative AI platform Cuebric, says generative AI powered by CUDA and Nvidia's chipsets have drastically cut down on production time. The company develops software used by TV shows and movie productions to render backgrounds and effects. Its clients and licensees include several large Hollywood studios.
"So imagine I'm sitting at a breakfast table, there's a cup in front of me, a milk container and a jug of water," Demirdag told IBD. "To simulate that in a virtual environment, you need to build it all from scratch, as if you're a sculptor."
"All this requires a lot of time, I mean weeks, and thousands of dollars per scene," she says.
Generative AI flattens this process. "It identifies the time consuming and tedious parts of this production cycle, and using AI, it streamlines and accelerates that production time," Demirdag said. "You push a button, it's perfect 3D, as if you modeled it by hand. I'm pretty sure a large percentage of professionals would take that over modeling it by hand."
The Benefits Of Building A Community
Nvidia leapfrogged rivals by steadily releasing new products. Its Blackwell platform is set to release its B200 and GB200 superchips. Meanwhile, Nvidia developers say the company has continued to invest in its community, reinforcing its moat of programmers already familiar with CUDA and Nvidia's technologies.
Developers and programmers credit Nvidia's community development for why they continue to use their ecosystem.
Demirdag recalled how she asked Nvidia for help with a scene-rendering issue that would help "make my software more widespread and adopted."
"Do you have research around it?" Demirdag asked.
"It's very expensive to find an answer," she recalled. "We couldn't have done that research internally. Nvidia said, 'Actually, we have a solution for that, we can share it with you.'"
Follow Mike Juang on X at @mikejuangnews and on Threads at @namedvillage.
