Graphics processing units (GPUs) serve as a critical component in the foundation of building the world’s artificial intelligence (AI) infrastructure. Training AI models and performing AI inference requires high-speed processing power and generates computational workloads that are best handled using parallel processing. This has positioned GPUs, which were originally developed to speed up the rendering of video game graphics, as one of the best hardware options for providing the right kind of processing power.
Two companies dominate the GPU market. Nvidia (NVDA -2.09%) We are by far the leader in this field. Rival Advanced Micro Devices (AMD 0.10%)Meanwhile, it is trying to take on Nvidia and capture a significant share of the data center market. Nvidia stock is the clear winner in 2024, up about 175% as of this writing. Meanwhile, AMD stock is down about 15% year over year.
But which stocks are likely to perform better in 2025?
GPU market continues to grow
The outlook for the GPU market remains positive. Most major hyperscalers (companies that operate extremely large data centers) have already indicated plans to increase AI-related capital spending in 2025. AI is rapidly increasing the demand for cloud computing, and all major cloud computing companies are expanding their infrastructure to meet that demand.
Meanwhile, GPU clusters are getting bigger and bigger as big tech companies and well-funded startups like OpenAI and Elon Musk-owned xAI race to develop ever more sophisticated AI models.
Meta Platforms is expected to use 160,000 GPUs to train its upcoming Llama 4 models, which is 10 times the number it used for Llama 3. This case is similar to xAi’s Grok 3 model and is expected to require between 20,000 and 100,000 pieces. Used for Grok 2 training. Meanwhile, there are talks of companies deploying clusters with as many as 1 million GPUs. In the near future.
Nvidia is the biggest beneficiary of these companies’ insatiable demand for GPUs, with its revenue far exceeding AMD’s. A big reason for that is that Nvidia developed a free (but proprietary) CUDA software platform a long time ago. This allows developers to program purchased GPUs for tasks other than graphics rendering. Therefore, CUDA became the software platform for many developers to learn to program GPUs, creating a wide moat for the company. Since then, Nvidia has expanded its software lead by adding developer tools and AI-specific micro-libraries through CUDA X.
AMD now has its own GPU software platform and makes GPUs that are as powerful, if not more powerful, than Nvidia’s, at least on paper. However, tests by independent research and analysis firm SemiAnaracy concluded that AMD’s software was throttling GPU performance. The company said in a report that AMD’s out-of-the-box experience is “unusable” and that “multiple AMD engineering teams” are needed to help fix bugs in the software. As it turns out, the paper specs for AMD’s latest GPUs didn’t match real-world performance. In contrast, SemiAnalysis described the out-of-the-box performance of Nvidia’s H100 and H200 GPUs as “amazing.”
This helps explain why Nvidia generated $30.8 billion in data center revenue last quarter, while AMD generated just $3.5 billion. However, what’s notable is that both have similar data center revenue growth rates. Nvidia increased by 112% and AMD by 122%. Of course, Nvidia doubled its revenue from a much larger base, making its performance even more impressive.
One area where AMD has been able to carve a niche for itself is in AI inference. SemiAnalysis noted that AMD customers tend to deploy GPUs for inference that presents narrow and well-defined use cases. So one way AMD could start to gain market share over the next few years is if much of the GPU market shifts from training to inference. AMD’s GPUs are cheaper than Nvidia’s, so such a move could potentially be profitable.
Image source: Getty Images
Rating and verdict
From a valuation perspective, AMD is a slightly cheaper stock, trading at 24 times forward earnings, compared to Nvidia’s nearly 31 times. However, because GPUs are only part of AMD’s larger hardware portfolio, Nvidia is growing its total revenue more rapidly (94% versus 18% in the previous quarter).
AMD PE Ratio (Forward 1y) data by YCharts.
For now, I think AI training will remain important in the coming years as companies continue to strive to develop more advanced AI models. That’s why I very much like holding Nvidia stock going forward, and think investors can still profitably add Nvidia stock to their portfolios well into 2025.
Randi Zuckerberg is a former head of market development and spokesperson at Facebook, sister of Meta Platforms CEO Mark Zuckerberg, and a member of the Motley Fool’s board of directors. Jeffrey Seiler has no position in any stocks mentioned. The Motley Fool has positions in and recommends Advanced Micro Devices, Meta Platforms, and Nvidia. The Motley Fool has a disclosure policy.
