A $27bn memory ETF with a quarter in Micron is not diversified exposure

The Roundhill Memory ETF, which launched on April 2 under the ticker DRAM, now manages $27.44bn and has returned roughly 161.51% since inception, according to figures cited by Cryptobriefing ^[1][2][3]. As of mid-August, 25.42% of the portfolio sat in Micron Technology, which works out to about $6.97bn riding on one chipmaker ^[4][1].

The fund's stated remit is companies that derive at least half their revenue from memory products: DRAM, NAND, solid-state drives, hard disk drives and high-bandwidth memory ^[5]. It holds 24 names, with heavy weight in the three producers that Cryptobriefing describes as effectively oligopolising memory: Micron, Samsung and SK Hynix ^[6][7]. Do the arithmetic on 24 holdings and an equal-weight portfolio would put 4.17% in each name, so Micron's slug is about 6.1 times a neutral weight ^[2]. The remaining 23 positions share 74.58%, averaging 3.24% each, which makes Micron close to eight times the size of the average other holding ^[3].

That is the whole structure of the risk. A 10% fall in Micron alone, with every other holding flat, removes roughly 2.54% of net asset value, or about $697m at current assets ^[4]. And the other holdings are not a counterweight to that move: they are the same memory pricing cycle in different jurisdictions ^[7]. Buyers looking for exposure to memory as a sector get something closer to a Micron position with a satellite sleeve attached.

Micron's weight is not accidental. Cryptobriefing notes it is positioned as a primary beneficiary of AI memory demand, particularly in HBM, the premium modules sitting inside Nvidia data centre GPUs, and that demand has been strong enough for memory makers to shift capacity away from conventional DRAM to fill HBM orders ^[8][9]. It is also the only US-headquartered company among the three major producers, from Boise, Idaho, which the publisher frames as a strategic advantage in the current geopolitical climate ^[10][11].

The economics of the wrapper are worth noting. At a 0.65% expense ratio, $27.44bn implies roughly $178m in annual fee revenue ^[12][5]. Roundhill also launched a leveraged counterpart, RAM, on June 24, designed to amplify the daily returns of the same memory basket ^[13] - daily leverage stacked on a basket that is already a quarter one stock.

Cryptobriefing also raises the reflexivity problem: when one ETF becomes a significant buyer of a concentrated basket, its own inflows can contribute to the price appreciation that attracts further inflows ^[14]. Note too that a 161.51% return means a large share of the $27.44bn is appreciation rather than fresh money, so the headline asset figure would deflate quickly on a drawdown ^[2][3].

Watch three things. First, Micron's next quarter, specifically HBM yields and conventional DRAM pricing, which the publisher identifies as the disproportionate risks to the fund ^[15]. Second, whether Micron's weight drifts further above 25.42% and how the index handles it; the source material does not say whether there is a single-name cap. Third, RAM's asset growth, because the leveraged sibling adds daily rebalancing flow on top of an already concentrated book ^[13].