Hyperscaler Stocks vs. AI Infrastructure Stocks: Which Are the Better Buys?

Artificial intelligence (AI) continues to be one of the driving themes in the stock market. However, the stocks of hyperscalers (owners of large data centers) have often traded very differently from those of AI infrastructure players during this tech bull market.

AI infrastructure stocks have received the most love from investors, as makers of chips, servers, and networking hardware have been huge beneficiaries of the spending to build data centers. At the same time, the hyperscalers that are doing much of that construction often get punished for spending so much on this AI infrastructure. However, the relationship between AI hardware companies and hyperscalers is ultimately symbiotic. If hyperscalers don't get strong returns on their AI infrastructure spending, they will cut back on it. If that happened, the AI infrastructure stocks would take a big hit, as it would greatly impact their revenue and earnings.

With that in mind, let's consider which group of AI stocks looks like the better buy now.

Hyperscalers: The big spenders

The category of hyperscalers is led by the big three cloud computing companies: Amazon, Microsoft, and Alphabet. All three share some common traits. In addition to their fast-growing cloud computing units, they also have other strong businesses, such as e-commerce (Amazon), search (Alphabet), and enterprise software (Microsoft).

They also all generate boatloads of operating cash flow that helps them pay for their AI infrastructure spending. When they cut back on their AI data center spending, they will start generating strong free cash flow again as well.

Meta Platforms, which is spending big on AI infrastructure mostly for its own internal use, is also considered a hyperscaler. It has been seeing strong revenue growth in its core social media advertising business, thanks in part to its AI spending, and it recently announced it may enter the cloud computing business by renting out any data center capacity it has that it's not utilizing.

Image source: Getty Images.

AI infrastructure players: The pick-and-shovel plays

Hyperscalers' spending is driving strong revenue growth among AI infrastructure companies. Nvidia has been the biggest beneficiary, as its graphics processing units (GPUs) are the most commonly used parallel processors for AI model training. Advanced Micro Devices, meanwhile, has started to gain momentum. The rising volume of inference workloads has lifted demand for its chiplet GPUs, and the outlook for agentic AI demand has data center operators scooping up its CPUs, too. Cerebras Systems, which just went public in May, also has a unique chip offering for inference that could shake up the industry, although right now it's more of a premium, niche solution. Intel and Arm Holdings, on the other hand, are looking toward the data center CPU market to drive strong growth in the coming years.

Both Broadcom and Marvell are benefiting by helping big tech clients design their own custom AI chips, as well as by supplying data center networking and interconnects. These are fast-growing markets that are becoming more important as the sizes of AI chip clusters grow.

Then, of course, there are the memory makers. Micron has been a big winner from the AI build-out, as high-end processors need to be packaged with a special form of dynamic random-access memory (DRAM) called high-bandwidth memory (HBM), and soaring demand for HBM has caused a DRAM shortage, lifting prices and driving strong revenue growth and margin expansion for the companies that sell it. A similar dynamic is also playing out with flash memory (NAND) -- the AI infrastructure build-out has sent demand for massive flash-based solid-state drives (SSDs) soaring. This has led to Sandisk's revenue and gross margins surging.

The verdict

While AI infrastructure stocks have been getting most of the love from investors so far, ultimately, it is the hyperscalers that have two ways to win. If they keep getting nice returns on their capex spending, then they will continue to spend and see rapid growth. Meanwhile, they have the option to cut back on their AI infrastructure outlays, which would lead to them generating massive free cash flow. This is why I actually prefer this group.

Among this group, I really like Alphabet and Amazon, since both also enjoy cost advantages from using their own custom AI chips. That said, Meta Platforms and Microsoft also look undervalued.

At the same time, I still think the AI infrastructure players can perform well. Nvidia remains the king of the AI chip space, and I really like the setups for AMD and Broadcom. I also think the market may be underestimating the longevity of the memory supercycle that Micron is enjoying.

Geoffrey Seiler has positions in Advanced Micro Devices, Alphabet, Amazon, Broadcom, and Meta Platforms. The Motley Fool has positions in and recommends Advanced Micro Devices, Alphabet, Amazon, Arm Holdings, Broadcom, Intel, Marvell Technology, Meta Platforms, Micron Technology, Microsoft, and Nvidia. The Motley Fool has a disclosure policy.