Nvidia and Competing Stocks in AI Infrastructure

Since the artificial intelligence (AI) infrastructure boom began, Nvidia has been the clear frontrunner. But the landscape is shifting as companies like Advanced Micro Devices (AMD) and Broadcom are positioning themselves as significant contenders, particularly in areas like inference and agent AI.

Let’s delve into what each of these companies brings to the table and identify which stocks are worth considering right now.

Nvidia

Nvidia has firmly established itself as the leader in training large-scale language models (LLMs), and that dominance isn’t likely to wane. They built a substantial competitive edge with their CUDA software platform, which has become standard in many AI research environments, ensuring that most basic AI programming is optimized for their graphics processing units (GPUs).

However, Nvidia is not just resting on its laurels. Their recent acquisition of Groq brought in a Language Processing Unit (LPU) specifically for inference, now integrated into the CUDA ecosystem. They are also stepping up in the agent AI realm, unveiling the Vera Rubin central processing unit (CPU), which the company views as a new $200 billion market—a hefty revenue boost of potentially $20 billion this year.

Overall, Nvidia is evolving into a comprehensive AI infrastructure solutions provider that goes well beyond GPUs; indeed, their rapidly growing networking portfolio suggests a promising future.

AMD

AMD, once overshadowed in the GPU sector, is now emerging as a formidable alternative to Nvidia, especially in the inference market. Unlike its competitor, AMD’s memory-efficient chiplet design allows for better packaging, which is critical given the constraints of the inference market. The company has enhanced its ROCm software and partnered with major players to drive future growth. Talk is circulating that Anthropic may adopt AMD’s latest GPUs for inference tasks.

In addition, AMD holds a strong position in the data center CPU market, which is crucial for agent AI. As the ratio of GPUs to CPUs in data centers shifts from 8:1 for training down to 1:1 for agent AI, the demand for high-performance CPUs is expected to surge. Even though CPUs traditionally cost less than GPUs, this presents a compelling market opportunity for AMD, which is still smaller in revenue than Nvidia.

Broadcom

With Nvidia’s latest Rubin GPUs priced at about $55,000 each, it’s understandable that many hyperscalers—who own large data centers—are now opting to design custom chips to save costs. Notably, Alphabet has achieved success with its Tensor Processing Unit (TPU) and seems eager to replicate that success.

This shift has led hyperscalers to increasingly seek Broadcom’s assistance. Renowned for its expertise in application-specific integrated circuits (ASICs), Broadcom played a pivotal role in developing Alphabet’s TPU. As the demand for custom chips grows, Broadcom has also seen an uptick in TPU adoption, enabling some customers to purchase TPUs directly. Projections suggest ASIC revenue could exceed $100 billion by 2027.

Broadcom’s custom chip business also bolsters its data center networking endeavors, positioning the company for significant growth in the years ahead.

Conclusion

At this stage, investing in any of these three AI stocks seems like a sound decision. I would consider all three—Nvidia, AMD, and Broadcom—as potential buys. Among these, if I had to choose just one, I’d lean towards AMD. It has hefty market opportunities still in their infancy, making it particularly attractive in a growth-focused investment landscape.

Should You Buy Nvidia Stock Now?

Before making a decision on Nvidia stock, it’s prudent to weigh several factors:

According to some analysts, while Nvidia remains a strong option, a selection of ten other stocks currently show impressive potential that are worth exploring. These stocks could yield remarkable returns over the next few years, putting them on investors’ radars.