DRAM: Inside the Hottest ETF in the Marketplace Today

David Dierking, The Motley Fool

Tue, May 19, 2026 at 10:35 AM EDT 4 min read

The Roundhill Memory ETF (NYSEMKT: DRAM) was launched on April 2. Less than six weeks later, it was up nearly 100%. Today, it already has more than $9 billion in assets.

That kind of performance just doesn't happen in typical exchange-traded funds (ETFs). It happens when the correctly themed fund is launched into a market where the underlying demand is huge, supply is tight, and investors are only starting to realize the opportunity.

The Roundhill Memory ETF is the first fund built exclusively around memory chip companies. Unlike other semiconductor ETFs that have some memory exposure, it's a pure-play portfolio. That's a distinction that matters a lot right now.

How DRAM shapes its portfolio

A lot of AI-themed ETFs are really just large-cap tech ETFs in disguise. They have Nvidia, Microsoft, and Alphabet as their top holdings and add some semiconductor names further down. Investors who buy those funds are essentially getting a broad tech index with an artificial intelligence (AI) label on it.

The Roundhill Memory ETF, however, takes a different approach. Its selection process specifically targets the companies involved in the development or manufacture of semiconductor memory products. That includes DRAM, NAND flash, specialty memory, and embedded memory. By requiring that the companies in the fund get 50% or more of their revenue from this industry, it's a high bar to clear. And only a handful of stocks make the cut.

There are three big names in the memory space: Samsung Electronics (OTC: SSNLF), SK Hynix, and Micron Technology (NASDAQ: MU). Within this portfolio, they account for a combined 75% of total assets. You'll also get some ancillary exposure to names such as Sandisk and Seagate Technology, but this is pretty much a three-stock portfolio.

The memory-demand story

As Roundhill notes on the fund's website, "Memory is a critical bottleneck of the AI revolution, supported by a secular shift toward data-intensive applications and sustained demand growth."

The opportunity is simple. Every company that's building out its AI infrastructure needs memory in order to operate. Every graphics processing unit that goes into a data center needs high-bandwidth memory stacked alongside it. As workloads scaled up in 2025 and capacity requirements grew, the three largest holdings of this fund found themselves with more demand for memory than they could handle.