New Memory ETFs Line Up to Challenge Runaway DRAM

The Roundhill Memory ETF (DRAM) is one of the most successful fund launches of all time. Since coming to market on April 2, it has pulled in more than $21 billion of net inflows while its share price has nearly tripled, pushing assets close to $26 billion.

All of that happened in roughly three months, which makes DRAM the fastest-growing ETF on record.

The timing could not have been better. DRAM launched just as memory stocks were going vertical, driven by one of the sharpest supply/demand imbalances the industry has ever seen.

Before DRAM, it wasn't easy for U.S. investors to play the memory theme. Two of the biggest names in the space, SK Hynix and Samsung, do not trade on U.S. exchanges (the former is set to list ADRs on the Nasdaq this Friday), so investors who wanted the exposure were buying South Korea funds like the iShares MSCI South Korea ETF (EWY), which included the memory giants along with a host of unrelated stocks.

DRAM gave them a pure-play alternative aimed squarely at memory.

But given the enormous inflows DRAM has seen, it was only a matter of time before other issuers tried to peel off a piece for themselves. Three have shown up so far, but interestingly, none is competing on price.

DRAM charges 0.65%, and the newcomers run from a matching 0.65% up to 0.95%. Instead, each is trying to stake out a different slice of the memory theme.

What These Funds Own

It helps to understand the memory industry before comparing the funds. Memory chips come in two broad flavors. DRAM (the type of memory, not the ETF) is the fast, volatile working memory that loses its contents the moment the power goes off, and high-bandwidth memory, or HBM, is a premium version of it, built by stacking DRAM chips vertically and wiring them together so data can move at very high speeds.

HBM is the component that sits right next to the GPUs in an AI server, and it is the biggest bottleneck in the current build-out.

NAND flash is the other category, the non-volatile storage that holds data whether the power is on or not, and the stuff inside solid-state drives.

The big three, SK Hynix, Samsung and Micron, dominate DRAM and HBM. They make NAND too, and Samsung is in fact the biggest NAND producer, but their profits come mostly from the DRAM and HBM side right now. 

Kioxia and SanDisk are the pure NAND plays, with no DRAM or HBM businesses of their own. 

DRAM, the ETF, focuses on, well, DRAM. SK Hynix, Samsung and Micron—the three companies that dominate HBM—each make up roughly a quarter of the portfolio, about three-quarters of the fund between them, with SanDisk, Seagate, Western Digital, Kioxia and a handful of others filling out the rest.

HBMX Reaches Beyond the Chipmakers

The first challenger to DRAM was the Tuttle Capital Concentrated Memory Stack ETF (HBMX), which launched June 2 and charges 0.95%. Tuttle casts a wider net, targeting the whole "memory semiconductor ecosystem," which means not just the chipmakers but the companies that supply the equipment, materials and services used to build memory.

Micron sits around 9% and SanDisk around 5%, but the fund also holds Applied Materials near 8%, ASML at 6% and Lam Research at 6%. Those equipment makers do supply the memory manufacturers, but they also sell to logic customers like TSMC, so their fortunes track overall semiconductor capex rather than memory specifically.

That makes HBMX less of a pure memory bet and more of a memory-plus-semicap play.

KMEM Tilts Hard Toward SK Hynix

The Kurv Memory Select ETF (KMEM) went the other way. It launched July 1, matches DRAM's 0.65% fee, and doubles down on the big three. SK Hynix alone is about 42% of the portfolio, with Micron near 20% and Samsung around 19%.

So like DRAM, roughly three-quarters of the fund sits in the HBM trio, only with a much heavier tilt toward SK Hynix, which holds the largest share of the HBM market and, in Kurv's telling, trades cheaper than its peers.

It is almost an attempt to out-DRAM DRAM. If you are more bullish on SK Hynix in particular, this is one way to express it.

DISK Bets on Flash Instead

The Tema Memory ETF (DISK), which launched June 30 at 0.75%, is the one doing something genuinely interesting. It stays inside the memory theme but deliberately leans away from HBM.

Its top holdings are Kioxia at about 17% and SanDisk at 16%, with Samsung around 9%, SK Hynix near 8% and Micron further down the list at 5%.

Kioxia and SanDisk are storage and NAND-flash names rather than HBM producers, so DISK is effectively betting on the parts of the memory market that the HBM-heavy funds underweight.

Of the three, it is the most differentiated from DRAM while still being unmistakably a memory fund.

Tema's Case

DISK’s tilt is a deliberate call on where memory demand is heading, and Tema's chief investment officer, Yuri Khodjamirian, laid out the case for overweighting NAND in an interview with ETF.com.

On the demand side, memory is eating up a growing share of what hyperscalers spend, by the firm's estimate somewhere around 30% of the bill of materials this year and potentially closer to half within a year or two.

DRAM is the expensive part of that bill, and as agentic AI widens context windows, with agents spinning up other agents and each one needing to hold its own instructions in working memory, keeping all of it in DRAM and HBM starts to get prohibitively expensive.

Tema's bet is that data centers increasingly offload some of that context onto cheaper flash, which plays straight to the NAND names.

Meanwhile, on the supply side, because DRAM and HBM carry much fatter margins right now, the manufacturers that make both are steering fab capacity toward them and away from NAND, which tightens the flash market and pushes prices up.

Of course, there is a risk to this bet. DRAM and HBM are where the fattest margins and the clearest AI demand sit today, so leaning away from them means tilting toward a more commodity-like and more cyclical corner of memory. 

NAND has historically been more volatile on pricing and quicker to see its margins compress when the cycle turns, and the context-offload thesis is a forecast rather than a fact. 

If HBM demand keeps surging and the shift toward flash arrives slowly, DISK's NAND overweight could cause it to lag the HBM-heavy funds.

Early Traction

The flows for the three DRAM ETF competitors have so far been modest, but it's early days. Each of the three has taken in somewhere around $30 million since launch.

For HBMX, which has had roughly a month to gather assets, that isn't much to write home about. For DISK and KMEM, both barely a week old, it is a solid start.

The more important question for investors is whether they are worth owning. I won't make an investment call here, but to me, DISK appears the most differentiated versus DRAM.

The ETF gives you memory without the massive overweight in the HBM names, which is smart product positioning on the part of Tema, but also potentially compelling for investors who are bullish on NAND.

HBMX is the one I would question. Reaching into equipment makers and the broader ecosystem waters down the very thing that made DRAM a phenomenon—a clean and concentrated bet on memory.