Investors Are Flocking to This ETF, but Its Holdings Look Overvalued

If you were trying to predict which exchange-traded fund would have the most successful debut in fund history, my guess is you wouldn’t have chosen the one that invests in a basket of “memory” semiconductor stocks. Yet, the Roundhill Memory ETF DRAM has surged to nearly $10 billion in net assets since being launched on April 2, 2026, smashing the record for asset growth in the first 45 days of existence.

Roundhill Memory ETF: Daily Net Assets Since April 2, 2026, Inception

How’d it get so big so fast? Unlike many other ETFs that also came out of the gate quickly, thanks to strong early inflows, the Roundhill ETF has not only seen robust investor demand but also notched a very high return—77.9% from inception through May 15, 2026. I estimate that translated to around $951 million in market appreciation the ETF racked up through May 15, 2026.

Roundhill Memory ETF: Growth of $10,000 Since April 2, 2026, Inception

It’s also seen torrential inflows from investors seeking a way to get in on the artificial intelligence infrastructure buildout theme, with memory products (the “memory” in the ETF’s name) being key inputs to graphics processing units essential to the process of training and running large models. All told, the ETF gathered around $8.7 billion in flows from inception through May 15, 2026, explaining its meteoric growth.

Roundhill Memory ETF: Cumulative Market Appreciation and Flows Since April 2, 2026, Inception

Don’t Forget: Price Matters

Convincing as the AI buildout narrative might sound, and impressive as the ETF’s early returns have been, I think investors should tread very carefully here.

The portfolio is highly concentrated, spread across fewer than 10 names. In fact, the top three holdings alone recently accounted for over 75% of net assets. This reflects the memory industry’s extreme concentration, meaning the ETF is likely to live and die with the stocks that dominate the portfolio.

One of those stocks is the ETF’s top weighting, SK Hynix 000660, which is the leading supplier of dynamic random access memory and “not-and” chips—also known as DRAM and NAND. It’s soared to a $1.3 trillion market cap behind insatiable demand from AI labs racing to build massive data centers. That propelled the stock to a 125% gain over the Roundhill ETF’s roughly seven-week existence through May 15, 2026.

SK Hynix: Growth of $10,000 Since Roundhill Memory ETF's April 2, 2026, Inception

The ETF’s other top holdings—Samsung Electronics 005930, Micron Technology MU, Kloxia Holdings 285A, and SanDisk SNDK—have also surged higher over this period amid a DRAM and NAND supply crunch and buyers’ strong willingness to pay up for what they need, which has pushed DRAM prices sharply higher in recent years.

Average DRAM Price

Much of that has fallen to the bottom line and, thus, these firms have reported soaring profits, which have further stoked investor enthusiasm.

Year-Over-Year Growth in Diluted Earnings per Share From Continuing Operations

Notwithstanding that, these stocks look anything but cheap: Here’s each holding’s Morningstar stock rating, our analysts’

per share compared with the stock’s price, as well as and fair value .

(Limited to stocks covered by Morningstar equity analysts; as of May 18, 2026. Source: Morningstar Direct; Roundhill)

Every stock is trading at a premium to our analysts’ estimate of its intrinsic value, explaining their relatively low star ratings. Moreover, these are not high-quality businesses: None possesses an economic moat in our analysts’ view, and there’s considerable uncertainty about what each is fundamentally worth; this is owing to the commodity-like nature of the memory business.

Why the disconnect between our analysts’ assessment and what has investors agog about these stocks’ prospects? Time horizon. It appears the market is focused on the near-term outlook, which our analysts agree is bright given the likelihood that demand will outstrip supply this year and next, as well as possibly 2028. But beyond that, they foresee moderating revenue growth as more capacity is brought online and memory-semiconductor pricing normalizes, with their revenue forecast for the big three manufacturers shown below.

This is likely to put a lid on these stocks’ returns, making the future less likely to resemble the past few years of heady profit growth and gains.

Bottom Line

The Roundhill Memory ETF has been a smashing success so far, growing faster than virtually any other fund in history at this stage of its life. That growth has been fueled by blistering demand from investors, as well as big gains from the memory stocks it’s bought.

But even by the standards of thematic ETFs, this portfolio is very concentrated, with investments bunched up in a small handful of fast-growing but ultimately commodity-like businesses that don’t possess durable competitive advantages. What’s more, those stocks aren’t cheap, with virtually every holding trading above our analysts’ estimate of what it’s intrinsically worth, tempering the outlook for their future returns.

Beyond the stocks’ fundamentals, we’ve also done research finding investors have struggled to succeed with thematic strategies and funds that have surged in popularity. In these cases, subsequent time- and dollar-weighted returns tended to erode. This ETF, which levies a 0.65% expense ratio, need not meet that fate, but it’s worth keeping in mind before taking the plunge.

Lastly, given the massive influx of assets, the portfolio’s heavy concentration, and the quick amassing of gains, it’s advisable for investors to also keep taxable distributions in mind. Normally, equity ETFs are able to sidestep capital gains distributions, thanks to the in-kind creation/redemption mechanism they utilize to push out low-basis shares and by harvesting losses to offset gains. But this ETF may employ cash redemptions, not in-kind, and as mentioned, holds a very small number of highly correlated stocks (which can make loss-harvesting more challenging).

Switched On

Here are other things I’m reading or listening to:

• “Err in the direction of kindness”: George Saunders’s 2013 commencement speech
• An inspiring, moving interview with the Artemis II crew
• Econtalk podcast: “The Man Who Built Nvidia”
• “London Falling” by Patrick Radden Keefe

Don’t Be a Stranger

I love hearing from you. Have some feedback? An angle for an article? Email me at jeffrey.ptak@morningstar.com. If you’re so inclined, you can also follow me on Twitter/X at @syouth1, and I do some odds-and-ends writing on a Substack called Basis Pointing.