Even with new export controls cutting off a vital market in China, demand for advanced chips used to power artificial intelligence (AI) infrastructure remains high. While there is a growing market for custom AI chips, the most commonly used chips for running AI workloads are graphics processing units (GPUs). This name stems from the fact that these chips were originally designed to speed up graphics rendering in video games.
Due to their powerful processing speeds, GPUs are now used for a variety of high-power computing tasks, such as training large language models (LLMs) and running AI inference. The GPU market is basically a duopoly at this point, headed by Nvidia(NASDAQ: NVDA) and Advanced Micro Devices(NASDAQ: AMD). The question many investors ask, though, is: Which stock is the better buy?
The leader versus the challenger
The unquestioned leader in the GPU space is Nvidia, which commands an over 80% market share. Not only is Nvidia larger than AMD, but it's also been growing its data center revenue more quickly. Last quarter, Nvidia grew its data center revenue by 73% to $39.1 billion, while AMD's data center revenue jumped 57% to $3.7 billion.
Nvidia's advantage comes from its software platform, CUDA. It launched the free software platform all the way back in 2006 as a way to let developers program its GPUs for different tasks in an effort to expand beyond the video game market. The company pushed the use of the software to universities and research labs, which made it the software program upon which developers were taught to program GPUs.
While AMD made some half-hearted efforts with software, it didn't launch a true CUDA competitor until around 10 years later with ROCm. By that time, CUDA had already become the default software used to program GPUs, and ROCm was still behind with less hardware support, limited documentation, and more difficulty to install and use. Meanwhile, Nvidia has since expanded upon its software lead through a collection of AI-specific libraries and tools built on top of CUDA, called CUDA X, which helps bolster the performance of its chips for AI tasks.
Ultimately, CUDA has given Nvidia a big network effect advantage. The more CUDA is used, the more tools and libraries are built for it, making Nvidia GPUs all the stickier. While ROCm continues to improve, it still trails CUDA, especially for use in LLM training.
However, where AMD has been able to gain more traction is in AI inference. Training AI models is a much more difficult task, which is why Nvidia has dominated this market and where its CUDA advantage really shines through. Inference, on the other hand, is easier, and there is more of a focus on things such as latency, power consumption, and cost. Due to its competitive positioning, AMD's GPUs tend to be less expensive than those from Nvidia, and while its ROCm software trails CUDA, it is generally considered good enough for running most AI inference workloads.
The good thing for AMD is that the inference market is eventually expected to become the much larger of the two markets. In fact, some pundits, including venture capitalist and former Facebook executive Chamath Palihapitiya, have said the inference market could be up to 100 times larger than the market for training AI models. But whether the inference market becomes 2 times bigger or 100 times than the training market, AMD could have an opportunity to gain some market share.
Image source: Getty Images.
Which stock is the better buy?
When it comes to valuation, both stocks trade in a similar range. Nvidia has a forward price-to-earnings (P/E) ratio of just over 32 times this year's analyst estimates, while AMD is at 28 times. Nvidia, meanwhile, is growing its revenue more quickly.
With their valuations similar, the key factor to which stock will outperform in the coming years will largely come down to growth. Nvidia is the clear leader in the GPU space and should continue to see strong growth as the AI infrastructure buildout continues. However, its AI data center revenue is now 10 times that of AMD. so the law of large numbers can come into play.
As AI infrastructure begins to shift more toward inference, AMD should have a nice opportunity to take some market share. Nvidia still has the lead in inference, but the gap is narrower compared to training. AMD also doesn't need to take a lot a of share in what could be a rapidly growing market to really make a big difference off its much smaller AI data center revenue base. As such, there is a good possibility it could begin to grow more quickly than the much bigger Nvidia. If that happens, I think its stock will outperform.
Ultimately, investors can own both stocks, which is probably a good idea. With AI infrastructure spending still appearing to be in its early days, both should be winners, although I do think AMD has more potential upside. The biggest risk, meanwhile, would be if AI spending unexpectedly starts to slow.
Should you invest $1,000 in Nvidia right now?
Before you buy stock in Nvidia, consider this:
The Motley Fool Stock Advisor analyst team just identified what they believe are the 10 best stocks for investors to buy now… and Nvidia wasn’t one of them. The 10 stocks that made the cut could produce monster returns in the coming years.
Consider whenNetflixmade this list on December 17, 2004... if you invested $1,000 at the time of our recommendation,you’d have $651,049!* Or when Nvidiamade this list on April 15, 2005... if you invested $1,000 at the time of our recommendation,you’d have $828,224!*
Now, it’s worth notingStock Advisor’s total average return is979% — a market-crushing outperformance compared to171%for the S&P 500. Don’t miss out on the latest top 10 list, available when you joinStock Advisor.
Geoffrey Seiler has no position in any of the stocks mentioned. The Motley Fool has positions in and recommends Advanced Micro Devices and Nvidia. The Motley Fool has a disclosure policy.
