Half the hottest AI ETF is shut for the day before New York opens. What are we really buying?

Roughly 49% of 2026's hottest AI fund, Roundhill's DRAM, sits in two Seoul-listed stocks that stay closed through the whole New York session, so the creation and redemption arbitrage that normally pins an ETF to its holdings cannot run, and half the fund trades on a market-maker's guess about the next Seoul open.
The only force keeping an ETF priced near what it actually owns is the creation and redemption arbitrage run by a handful of large broker-dealers, and it can only run while the underlying stocks are open. Roughly 49% of 2026's hottest AI fund, Roundhill's DRAM, sits in two Seoul-listed names that stay shut through the entire New York session, so for half the fund the intraday price is a market-maker's forecast of the next Seoul open rather than an arbitraged value. That gap is an execution-quality and risk-surface question worth understanding before trading any cross-border single-theme ETF.
The most talked-about exchange-traded fund of 2026 is not an S&P 500 tracker or a single chip giant. It is a memory-chip fund, Roundhill’s DRAM, which has pulled in close to $10 billion since it launched on April 2 and run up roughly 90% in under two months. I went looking at what it actually holds, and the number that stopped me was geographic, not financial. About 49% of the fund sits in two South Korean stocks, Samsung Electronics and SK hynix, with Micron making up most of the rest of a top three that is 73% of the portfolio. Which means that for the entire New York trading day, roughly half of this fund is closed for business.
Here is the plumbing we all assume is always running. An ETF trades all day on an exchange, but the only thing tethering its price to the value of the stocks inside it is a small set of large broker-dealers called authorized participants (the firms approved to create and redeem fund shares directly with the issuer). When the fund trades richer than its holdings, an authorized participant buys the underlying basket, hands it to the issuer in a block of at least 25,000 shares (a creation unit), receives new fund shares, and sells them, pocketing the difference. When the fund trades cheap, they run the loop in reverse. That arbitrage is the whole reason an ETF normally tracks its net asset value (NAV, the per-share value of what the fund actually owns) within a penny or two.
Now remove half the basket. Samsung and SK hynix trade in Seoul, which closes at 3 p.m. local time, roughly 14 hours before the New York bell, and stays shut for the whole US session. An authorized participant cannot buy or hedge those shares in real time, so the arbitrage that pins the price cannot run on that half of the fund. The published NAV is struck off the last stale Seoul close. And the intraday price of DRAM in New York becomes a market-maker’s estimate of where Samsung and SK hynix will open in Korea tonight.
For roughly half of this fund, the New York price is not arbitraged to the value of the holdings. It is a forecast of tomorrow's Seoul open, quoted as if it were a settled fact.
This stopped being theoretical on June 5. SK hynix fell 9.92% and Samsung dropped 6.40% in Seoul, and the single-stock leveraged funds Korean retail traders had crowded into fell about 20% in a day, both closing at lows since launch, enough that Korea’s Financial Services Commission (the FSC, the country’s top markets regulator) called a meeting to study the products. Those leveraged funds are a different animal from DRAM, with their own daily-reset mechanics. But they are a clean illustration of the timing problem: when the Seoul session does the damage, anyone holding a New York wrapper on those names finds out at the open, not during the move.
As 24/7 Wall St put it when it mapped the fund’s holdings:
"49% South Korean equities, meaning roughly half of what you own stops trading at 3 p.m. Seoul time."
The same piece described the New York quote during those hours bluntly, as “a guess dressed up as a spread.” On a calm day the guess is good and the gap is a rounding error. On a day when something breaks in Seoul after its close, the gap is exactly where the surprise lives.
So what does this change for those of us holding cross-border single-theme funds. Three practical things. The premium or discount to fair value tends to be widest in the first and last minutes of the US session and around the Seoul close and open, so a market order into those windows is the most expensive way in or out; a limit order against a sober fair-value estimate is cheaper. The fund’s intraday indicative value, published every 15 seconds, is itself stale on the closed half, so it is a rough guide, not a referee. And a 49% single-country weight means currency counts too: the won can reprice those holdings overnight before a single share changes hands.
The creation and redemption machine is one of the quiet triumphs of modern markets. It works so well that most of us have stopped noticing it is there. DRAM is a useful reminder that it only works while the underlying is open. I keep thinking about how many of the funds we treat as continuously and fairly priced are, for part of every day, just a well-informed guess about a market that has already gone home.
This is editorial analysis, not investment advice. Cerevisor does not hold or recommend the named positions, and information here can become stale within hours of publication.
Sources
- Memory Chip Prices Will Make or Break DRAM ETF Before Year-End 2026 - 24/7 Wall St, 2026-06-05
- Samsung-SK hynix Slide Sends '2x ETFs' Down 20% as Regulators Review Impact - Seoul Economic Daily, 2026-06-05
- DRAM's 49 Percent Korea Weighting Means Half the Fund Trades While New York Is Closed - 24/7 Wall St, 2026-05-26
- This ETF Is Seeing a Surge of Inflows Right Now: Is It Too Late to Buy In? - The Motley Fool, 2026-05-27