Forget AMD. Here's Who Nvidia Really Needs to Be Worried About.

The question of which AI hardware company is best seems to always circle around Advanced Micro Devices (NASDAQ: AMD) and Nvidia (NASDAQ: NVDA). However, I don't think that's who Nvidia shareholders need to be concerned about.

Instead, I think investors need to look outside the GPU realm and start looking at custom AI chips that could steal market share. The biggest name in this space is Broadcom (NASDAQ: AVGO), and I think it is actually Nvidia's No. 1 competitor right now.

This shifts the narrative completely, and once you see the growth Broadcom is projecting, it's easy to understand why Nvidia has cause for concern.

Image source: Getty Images.

AMD's results aren't as impressive as Nvidia's

First, let's figure out why AMD isn't a threat. At the start of the AI arms race, Nvidia captured a lot of the data center computing market share because its products were flat-out better than AMD's. The biggest advantage Nvidia had was its CUDA software, which controls how its GPUs process calculations. AMD didn't have a great counterpart to this software, so every company naturally gravitated toward Nvidia's GPUs. While there could be some arguments made for whether AMD's or Nvidia's hardware is better today, the reality is that a vast majority of existing infrastructure is already set up for Nvidia GPUs, so a wholesale switch is highly unlikely due to switching costs.

Additionally, Nvidia is still growing faster despite AMD's smaller size (which should ideally make it easier to grow faster). During Nvidia's Q2, ended July, its data center division grew at a 117% pace to $89 billion. AMD's data center division generated a mere $6.7 billion and grew at a 107% pace. That clearly indicates Nvidia is still growing faster and has a huge size advantage as well.

However, another competitor is rising.

Broadcom forecasts monster growth over the next few years

GPU-based computing is great for all sorts of workloads, and the flexibility it provides is unparalleled. However, GPUs aren't always used for their full capabilities, and sometimes end up running one workload over their entire lifespan. That wastes many features, making it not optimized for the task.

As the AI build-out progresses, the AI hyperscalers are looking to optimize computing costs, and that means developing custom AI hardware specifically built for workloads they will see. There will always be a place for GPUs in AI, as there may be odd data sets or other workloads that aren't suited for custom AI chips, but a healthy amount of workloads can be run for cheaper on optimized hardware.

None of the AI hyperscalers has the expertise to design and build its own computing chip, so they bring on a partner like Broadcom to handle this. This creates a great relationship, as Broadcom brings the manufacturing capabilities and design knowledge, while the AI hyperscalers bring massive computing demand.

Broadcom has four major clients: Alphabet, Meta Platforms, OpenAI, and Anthropic. Currently, Alphabet's Tensor Processing Unit (TPU) is the only computing unit being used in large numbers. Still, the other three are ramping up their orders, which is causing Broadcom's revenue to grow rapidly. During the fiscal third quarter (ended Aug. 2), Broadcom's AI semiconductor revenue totaled $16.7 billion, up 221% year over year.

While that's smaller than Nvidia by a long shot, it's growing far faster than either Nvidia or AMD. Furthermore, Broadcom expects its AI semiconductor revenue to double in 2027 to $115 billion and again in 2028 to $230 billion. That makes it a force to be reckoned with in the computing space, and I think Nvidia has more to worry about with Broadcom than it does with AMD.

I think the best way to invest in AI computing is to take positions in both Nvidia and Broadcom, as I have done. That way, you can benefit from whichever one ends up being the winner. While I think both of these companies will end up sharing a large portion of market share, Broadcom is certainly growing faster right now and is well-positioned to continue capturing additional market share.

Keithen Drury has positions in Alphabet, Broadcom, Meta Platforms, and Nvidia. The Motley Fool has positions in and recommends Advanced Micro Devices, Alphabet, Broadcom, Meta Platforms, and Nvidia. The Motley Fool has a disclosure policy.