Nvidia versus AMD: Determining the Superior AI Processor Investment for 2025
Graphics Processing Units (GPUs) serve as a vital component in the development of the world's artificial intelligence (AI) infrastructure. AI models require significant processing power for training and inference, which generates intense computational workloads. Given their capacity for parallel processing, GPUs, initially designed for enhancing video game graphics, prove to be excellent options for delivering such processing power.
Two primary entities control the GPU market - Nvidia (NVDA -2.09%) and Advanced Micro Devices (AMD) (0.10%). Nvidia takes the lead in this sector, while AMD strives to challenge Nvidia and capture a substantial portion of the data center market. By 2024, Nvidia's stock has shown impressive growth, increasing by approximately 175%, contrasting AMD's stock, which has decreased by approximately 15% over the same period.
As we look ahead to 2025, which company seems likelier to prosper?
The GPU market shows promising growth
The forecast for the GPU market remains optimistic. Significant hyperscalers (companies managing vast data centers) have announced intentions to boost their AI-related capital expenditures in 2025. The demand for cloud computing skyrockets due to AI, and prominent cloud computing providers are working diligently to expand their infrastructure to accommodate this demand.
Meanwhile, GPU clusters expand, with leading technology companies and financially well-established start-ups such as OpenAI and Elon Musk-owned xAI developing increasingly complex AI models. Companies like Meta Platforms are projected to utilize 160,000 GPUs to train their upcoming Llama 4 model - a tenfold increase compared to Llama 3. Similarly, xAi's Grok 3 model, which will necessitate 100,000 GPUs, has a projected requirement, following a leap from the 20,000 GPUs needed for Grok 2. Talks of deploying clusters featuring up to 1 million GPUs in the near future persist.
Nvidia has thrived as the primary beneficiary of these companies' insatiable demand for GPUs, with its revenue significantly surpassing AMD's. One key aspect of this success is Nvidia's creation of its free, yet proprietary, CUDA software platform, enabling developers to utilize GPUs for tasks beyond graphics rendering. As such, CUDA became the preferred software platform, generating a competitive edge for the company. Over the years, Nvidia has expanded this lead by providing additional developer tools and AI-specific microlibraries through CUDA X.
Currently, AMD has its GPU software platform and releases equally powerful GPUs - at least on paper. Nonetheless, independent research firm SemiAnalysis has found that AMD's software negatively impacts its GPUs' performance. SemiAnalysis described this as an "unusable" out-of-the-box experience and identified the need for multiple AMD engineering teams to resolve software bugs. Subsequently, the actual performance of AMD's most recent GPU fell short of its projected specifications, while SemiAnalysis lauded the out-of-the-box performance of Nvidia's H100 and H200 GPUs as "amazing."
This helps explain why Nvidia generated data center revenue of $30.8 billion in the last quarter, while AMD generated only $3.5 billion. Notably, both companies demonstrated comparable rates of data center revenue growth: Nvidia's rose by 112%, and AMD's rose by 122%. Even so, Nvidia's impressive performance can be attributed to its larger initial revenue base.
One area where AMD has made headway is AI inference. SemiAnalysis noted that AMD's customers tend to deploy its GPUs for inference, catering to narrow, well-defined scenarios. In such a scenario, AMD could potentially gain market share as the GPU market shifts from training to inference. AMD offers competitively priced GPUs, so a transition in the market could positively impact its performance.
Valuation and conclusion
From a valuation perspective, AMD is a more affordable choice, trading at a forward price-to-earnings (P/E) ratio of 24, compared to nearly 31 for Nvidia. Despite AMD's hardware portfolio, which includes various components, Nvidia's entire revenue has been primarily driven by GPUs, providing faster growth.
In conclusion, I believe that AI training will remain crucial in the foreseeable future as companies continually strive to create more sophisticated AI models. As such, I strongly favor Nvidia's stock as an investment choice for the coming years and believe that investors may still have the opportunity to profitably include it in their portfolios in 2025.
With the increased demand for AI-related investments, major tech companies are prepared to boost their AI-related capital expenditures in 2025. This surge in demand is expected to significantly impact the GPU market, leading to an increase in the utilization of GPUs by companies like Meta Platforms and Elon Musk-owned xAI.
Given Nvidia's dominance in the market, its creation of the CUDA software platform, and the exceptional out-of-the-box performance of its GPUs, it has become the go-to choice for developers seeking to utilize GPUs for tasks beyond graphics rendering. Consequently, Nvidia continues to generate substantial revenue from this market, far exceeding that of AMD.
Investors looking to capitalize on the growth of the GPU market and AI training should consider Nvidia as a potential investment opportunity in 2025, as its forward P/E ratio indicates its strong financial performance. Meanwhile, AMD may benefit from the shift towards inference as its GPUs are competitively priced and perform well in such scenarios. However, its software performance issues may need to be addressed for it to fully compete with Nvidia in the market.
