DRAM ETF Review 2026: Inside the $24.35B Roundhill Memory Fund (0.65% Fee, 75% in 3 Stocks)

What DRAM Is: The Fund Facts

DRAM is the Roundhill Memory ETF — the first ETF built purely around memory-chip manufacturers, the companies making the HBM, DRAM, and NAND that have become the supply bottleneck of the AI buildout. It launched April 2, 2026 on Cboe BZX, and the money arrived fast: $24.35 billion in assets within three months, on the back of a since-inception return of 161.51% at market price through June 30, 2026. Everything below is from Roundhill's own fund page and holdings file dated July 1, 2026.

The pitch is simple: broad semiconductor funds dilute the memory trade with foundries, designers, and equipment makers. DRAM does not. It is a concentrated bet that memory manufacturers keep pricing power as AI datacentres soak up every HBM wafer the industry can produce. Whether that concentration is a feature or a flaw is most of this review.

What DRAM Actually Holds — Two Very Different Tables

Roundhill publishes a combined-exposure table and a full holdings file, and you need both. The combined table, as of July 1, 2026: Samsung Electronics 25.18%, SK hynix 24.76%, Micron 24.76%, Sandisk 4.99%, Kioxia 4.60%, Western Digital 4.31%, Seagate 4.26%, GigaDevice 3.09%, Nanya 1.81%, Winbond 1.09%. Top three: 74.7% of the fund.

The full file shows how that exposure is actually built:

Look at the Micron line closely. Micron is billed at 24.76% of the fund, but only 2.14% is actual Micron stock — the other 22.62% is swap contracts. Roundhill is transparent about why: US Regulated Investment Company rules cap how concentrated a fund's direct positions can be, and with three names at 74.7%, physical shares alone would break the fund's tax status. The swaps are honest engineering. They still change what you own: a swap carries bank counterparty risk and an embedded financing cost that never appears in the 0.65% expense ratio, and the $3.75 billion money market position exists to collateralize that book. Since the June 9 holdings file — the one we screened in our DRAM Shariah ruling — swap exposure has actually grown, from 37.06% to 41.37%.

Performance: A 2.6x in Three Months, Then the July 1 Air Pocket

The raw numbers are extraordinary. From launch to June 30, 2026, DRAM returned 161.51% at market price and 156.30% at NAV. Memory pricing has been the hottest corner of the AI trade, and a fund holding nothing else captured all of it.

The 5-point gap between the market-price return and the NAV return is the part most buyers miss. It exists because investors kept paying more than the underlying stocks were worth:

Then July 1 happened: the market price fell 10.82% in a single session while NAV fell only 6.76%, and DRAM closed at $65.86 — a 2.40% discount to its $67.48 NAV. Anyone who bought at a 3% premium in June and sold into that discount surrendered over 5% to fund plumbing alone, before memory prices did anything. The 0.01% bid/ask spread makes DRAM cheap to trade; the premium/discount whipsaw makes it expensive to trade badly. If you buy this fund, check the live premium/discount on Roundhill's page first and use limit orders.

What It Costs a Canadian Investor

Three layers of cost, in order of size:

• The fee: 0.65% gross — $650 per year on $100K. That is 3x a plain index fund and above every screened halal ETF (SPUS 0.45%, HLAL 0.50%, WSHR 0.50%).
• The currency: DRAM trades in USD, so you convert on the way in and the way out. Questrade adds 100 pips — a full cent per dollar — to the exchange rate on USD conversions, and big-bank brokerages are rarely better — call it roughly $700 to $1,000 round-trip on a $50K position unless you use Norbert's gambit or a USD account you already hold.
• The timing risk: the premium/discount swing documented above, which in Q2 2026 was worth more than the annual fee.

What barely matters here: withholding tax. The US takes 15% of distributions under the Canada-US treaty — waived in an RRSP, unrecoverable in a TFSA — but DRAM's 30-day SEC yield was 0.56% as of June 30, 2026, paid once a year. On a $10,000 TFSA position that is about $8 of withholding annually. The account-placement rule that matters more: US-listed ETFs are specified foreign property for Form T1135, so a non-registered position counts toward the $100,000 CAD cost threshold; TFSA and RRSP holdings are exempt from that filing. And if you sell at a gain in a non-registered account, the 50% inclusion rate under section 38(a) of the Income Tax Act applies — on a $20,000 gain, $10,000 is taxable, roughly $5,353 at Ontario's top 53.53% combined rate.

The Shariah Question: Verdict Already Issued

Memory chips are a halal business — the AAOIFI business-activity screen has no problem with semiconductor manufacturing. The wrapper is the problem. Applying AAOIFI Shari'ah Standard No. 21 to the July 1, 2026 file: 41.37% of the fund is conventional total return swaps (interest-referenced derivatives, not permitted), and 15.24% is an interest-yielding money market fund (riba by definition). Over half the fund's weight sits in non-compliant instruments — a structural fail that no quarterly ratio drift can fix. Zoya rates DRAM non-compliant under AAOIFI guidelines (listing last updated July 2026), and the swap sleeve has only grown since our June screen.

The full DRAM Shariah ruling walks the screen line by line and ranks the compliant routes to the same trade: individually screened Micron stock (24.76% of DRAM anyway, and zero fund fee), or a purpose-built halal ETF. This is the same structural lesson as XEQT, which fails on banks and insurers — a fund can fail the screen even when its theme looks clean. Our halal ETF guide for Canada covers every screened fund Canadians can actually buy, and if you want the exposure inside a registered account, the halal TFSA guide maps the $7,000 of 2026 room (and $109,000 cumulative since 2009) onto compliant funds.

Who DRAM Suits — and Who It Doesn't

DRAM makes sense for an investor with a specific, high-conviction view that memory pricing stays tight through the AI capex cycle, who wants all three global manufacturers in one USD ticket — including SK hynix and Samsung, which most Canadian brokerages cannot buy directly on the Korea Exchange. That single-ticket access to Seoul-listed giants is DRAM's most defensible feature, and the 0.65% fee is partly the price of it. Even then, the honest sizing is a satellite position: 74.7% in three stocks, in one boom-bust industry, after a 161% three-month run, is not a core holding under any allocation framework I would sign my name to.

DRAM does not make sense for anyone who screens for Shariah compliance (structural fail — see the ruling), anyone whose thesis is really just "Micron goes up" (buy Micron, save the fee, skip the swap book), or anyone who cannot stomach the arithmetic of a memory downcycle — the honest stress test on a fund like this is a 40-50% drawdown when supply catches demand. If what you actually want is screened growth exposure with semiconductor weight as a by-product, Wealthsimple's halal portfolio built on WSHR or the US-listed screened funds get you there without the concentration or the compliance problem.

The Honest Bottom Line

As a product, DRAM executed its job almost perfectly: first mover on a real theme, $24.35 billion raised in three months, a 161.51% since-inception return, tight spreads, an options chain, and full daily transparency about a swap structure most issuers would bury. As an investment sitting in a Canadian portfolio in July 2026, it is a leveraged-feeling, three-stock, one-industry bet trading within days of a violent premium unwind — and for Muslim investors, a fund whose wrapper fails the screen even though its theme passes. Buy it small and eyes-open, replicate it with screened Micron if compliance matters, or skip it. The memory cycle will not ask your permission either way.