History Says Memory Stocks Like Micron and Sandisk Rarely Stay This Profitable for Long. Here's the Case for Why It's Different This Time.

The memory market is having one of its best historical runs ever, helping lift the stocks of both Micron Technology (NASDAQ: MU) and Sandisk (NASDAQ: SNDK). Both have seen their revenue soar and their gross margins balloon, leading to record profits and cash flow.

Micron is one of the big three DRAM makers, along with Korean companies SK Hynix (NASDAQ: SKHY) and Samsung, while it gets nearly a quarter of its revenue from NAND (flash) memory. Sandisk, meanwhile, is a pure-play flash memory maker. The rise in both memory markets is directly linked to the artificial intelligence (AI) infrastructure build-out, although the dynamics behind the current supply-demand imbalances are a bit different.

Flash memory was in a major bull market during the COVID-19 pandemic, as stay-at-home mandates led to a surge in demand for electronics, which are heavy users of flash memory. This led memory makers to increase their NAND capacity, while demand crashed after the pandemic ended. This led to a massive oversupply of NAND, negative gross margins, and to the big three memory makers reducing capacity and turning their attention to more lucrative DRAM.

Image source: The Motley Fool.

However, with the AI boom, data centers soon needed massive enterprise solid-state drives (SSDs) packed with NAND to store training data, greatly increasing demand soon after capacity was cut. Meanwhile, the emergence of high-bandwidth memory (HBM), a specialized form of DRAM used in the AI build-out, became the focus for the big three DRAM makers, and they were in no rush to increase NAND capacity. The result was a supply-constrained flash memory market amid soaring demand.

A similar dynamic was at play in DRAM, as graphics processing units (GPUs) and other AI chips needed to be packaged with HBM to reduce latency and improve power efficiency. As demand for GPUs grew, so did demand for HBM. As DRAM makers focused on increasing HBM capacity, the entire DRAM market became undersupplied, and prices rose sharply.

Historically, the memory market has been highly cyclical, with large boom-and-bust cycles. As mentioned previously, the last NAND memory cycle collapsed, leading memory companies to dump supply at prices below what it cost them to manufacture. This is a big reason why memory stocks trade at single-digit forward P/Es despite seeing rapid growth.

Is this time different?

However, there is an argument to be made that this time is different. HBM has become an integral component in the AI infrastructure build-out, and it is likely to remain supply-constrained well into the future. One of the main reasons is that HBM uses the same EUV (extreme ultraviolet lithography) machines used to manufacture GPUs and other advanced logic devices. ASML is the only company in the world with the technology to make these machines, which limits DRAM makers' ability to dramatically increase capacity. At the same time, HBM requires three times the wafer capacity as run-of-the-mill DRAM, further pressuring capacity increases.

As hyperscalers continue to pour money into AI infrastructure, demand for chips and HBM keeps rising. At the same time, the big three DRAM makers have also started signing long-term contracts for the first time. This should help increase visibility and reduce some of the cyclicality of the memory business.

Meanwhile, the supply of flash memory should also remain constrained for the foreseeable future. The leading memory producers are directing critical cleanroom space toward HBM, and new projects won't bring meaningful capacity online anytime soon. The rise of AI agents will also require more long-term memory storage, which should help keep demand high. Memory companies are also now developing high-bandwidth flash (HBF), which could be another catalyst.

With both the DRAM and NAND memory cycles looking very different from past ones, Micron, SK Hynix, Sandisk, and other memory stocks could have a lot more upside ahead. Their current valuations may not reflect what could be a very elongated cycle. As such, these could be among the top AI stocks to buy.

Geoffrey Seiler has no position in any of the stocks mentioned. The Motley Fool has positions in and recommends ASML and Micron Technology. The Motley Fool has a disclosure policy.