Nvidia's Biggest Risk Isn't Custom AI Chips From Broadcom or AMD -- It's Something That's Hidden in Plain Sight

It's no secret that artificial intelligence (AI) has been Wall Street's hottest trend for years. Empowering software and systems with the tools to make split-second, autonomous decisions is a technological leap forward that PwC believes can create $15.7 trillion in global economic value by 2030.

Chipmakers have been at the forefront of what makes AI data centers tick, with Nvidia's (NASDAQ: NVDA) graphics processing units (GPUs) dominating the landscape. However, hyperscalers will also be relying on custom application-specific integrated circuits (ASICs) from Broadcom (NASDAQ: AVGO) and specialized variants of Advanced Micro Devices' (NASDAQ: AMD) core lineup of GPUs.

On paper, Broadcom and AMD look to be the biggest risk to Nvidia's AI infrastructure dominance. But dig a bit deeper, and you'll discover that Nvidia's top competitive risk is something nearer and dearer.

Image source: Nvidia.

Nvidia holds a clear market share edge in AI data centers

In addition to its first-mover advantages in the AI arena, no AI chipmakers have come particularly close to matching the compute capabilities of Nvidia's four generations of GPUs (Hopper, Blackwell, Blackwell Ultra, and now Vera Rubin). Having the preferred hardware to power generative AI and train large language models has made Nvidia's GPUs the undisputed top choice by businesses.

Nvidia CEO Jensen Huang has also made it incredibly difficult for competitors to siphon away market share. Huang is attempting to bring an advanced AI GPU to market annually, ensuring that its newest hardware is always superior.

AMD can still thrive as a lower-cost option to Nvidia's GPUs. Although AMD's pricing power has improved for its Instinct GPUs, thanks to strong demand for AI-accelerating hardware, it's the company's lower price point and smaller backlog that can make it an attractive alternative.

Meanwhile, Broadcom's ASICs are specialized for hyperscalers and not used for the general-purpose workloads that Nvidia's GPUs are designed for. Broadcom might be garnering plenty of headlines, but it's not at risk of toppling Nvidia's AI data center dominance.

Image source: Getty Images.

Nvidia's top customers can be its undoing

If you want to uncover the leading risk for Wall Street's largest publicly traded company, look no further than its top customers by net sales.

Although Nvidia has racked up some eye-popping orders from members of the "Magnificent Seven," many of these same companies are also internally developing chips for use in their data centers. While these internally developed chips aren't faster than Nvidia's and won't be sold externally, they still pose serious problems for the face of the AI revolution.

For starters, some of Nvidia's pricing power and otherworldly gross margin originates from persistent GPU supply shortages. If the company's top customers are developing chips that take up valuable data center real estate, Nvidia can quickly lose its biggest tailwind.

Secondly, these internally developed chips are notably less costly than Nvidia's hardware and, presumably, more readily available (a function of demand creating a sizable backlog for Nvidia's GPUs).

Lastly, AI chips developed by Nvidia's top customers may delay future upgrade cycles. Between Nvidia's accelerated innovation cycle, which is depreciating prior-generation GPUs, and the company's top customers developing cost-effective in-house solutions, hyperscalers may push out AI data center upgrades.

There's no question that internal competition is the greatest threat to Nvidia's ongoing success.

Sean Williams has no position in any of the stocks mentioned. The Motley Fool has positions in and recommends Advanced Micro Devices, Broadcom, and Nvidia. The Motley Fool has a disclosure policy.