(Bloomberg) — Brian Emes manages a retail store in Lethbridge, Alberta, a small city on the Canadian prairie about a two-hour drive from the US border. On the morning of May 11, before opening the shop, the 43-year-old launched his brokerage app and bought 55 shares of an exchange-traded fund that hadn’t existed until early April.
Most Read from Bloomberg
The fund is the Roundhill Memory ETF or DRAM. It now accounts for roughly 7% of his portfolio — a concentrated wager, made on the strength of Reddit threads and YouTube videos, that artificial intelligence has turned memory chips into the trade he could not afford to miss.
Emes had never owned a memory stock. His investments largely sit, as they always have, in broad index products: mostly the iShares Core Equity ETF Portfolio fund. DRAM is his first real thematic bet.
“I’m thrilled I can directly invest in memory producers,” said Emes.
He is among a slew of retail investors who have piled into the fund — up some 84% since its April 2 launch — pushing it to about $10 billion in assets and into the top 10 US ETFs by year-to-date inflows, out of more than 5,000 listed products. By some measures, no ETF in history has grown faster.
DRAM is what fearless risk-taking looks like in the spring of 2026. In a year when nothing has been allowed to fall for long — when prediction contracts and zero-day options have all found their willing bidders — a six-week-old ETF built around three Asian chipmakers has become a mainstream retail trade.
Specialist products that once sat at the edges of the ETF industry are being pulled into its center, even as the lopsided options buying around DRAM gives some traders pause about how much further the wager can run. On Friday, DRAM fell 5% as rising bond yields threaten to knock the AI trade off its upward course.
Built by boutique firm Roundhill, the vehicle provides concentrated access to a handful of memory companies, including Samsung Electronics and SK Hynix, which trade primarily on the Korean exchange and have historically been difficult for North American investors to own directly. About half of DRAM’s portfolio is in SK Hynix and Samsung — and it is exactly this condensed exposure that makes it appealing, Emes said.
“The ETF reduces risk of relying on one company like Micron,” he said. At the same time, “it is direct exposure to the AI bottleneck.”
High-bandwidth memory is essential for the GPUs running AI training and inference workloads in data centers, and the market is dominated by a handful of producers, said Aniket Ullal, senior vice president and head of ETF research and analytics at CFRA. While funds like the tech-heavy Invesco QQQ Trust Series 1 provide exposure to AI hyperscalers, gaining access to the South Korean companies that dominate the memory-chip industry is harder to come by. That makes DRAM “the most targeted vehicle for the memory trade,” Ullal said.
DRAM took 27 trading sessions to cross $6.5 billion in assets, which the previous top-spot-holding fund — BlackRock Inc.’s blockbuster Bitcoin ETF — took 30 days to achieve, according to data compiled by Bloomberg. Over the past two weeks, it has steadily climbed up the ranks of most-traded ETFs, going from having the 34th-most volume at the start of the month to top 20 currently. It is already among the most-bought securities on retail-trader platforms like Interactive Brokers.
The memory fund is also commandeering the attention of options traders. It’s been one of the busiest ETF options markets over the past week, with more than $1 billion worth of options traded each day, according to Asym Research’s Rocky Fishman. That figure has grown so quickly that it has surpassed long-standing active option underlyings in some of the most-active full-sector ETFs, including the State Street Energy Select Sector SPDR ETF (XLE) and the State Street Financial Select Sector SPDR ETF (XLF). Both launched back in 1998.
“Activity has been very lopsided toward call options, presumably with investors buying calls to chase the rally,” he said. Heavily skewed call buying is a classic late-cycle signal in single-name and thematic trades, often preceding sharp reversals when sentiment turns.
The bull case is that AI has structurally rewired memory demand, locking in capacity contracts the cyclical industry never enjoyed before. The bear case is that the same story has been told at the top of every tech cycle since the 1990s, and that retail-driven flows tend to unwind faster than they accumulate.
“All thematic ETFs should be thought of as hot sauce,” said Eric Balchunas, senior ETF analyst at Bloomberg Intelligence. But “even if you have a selloff, it’s not like this trade will go away. AI isn’t going away.”
DRAM’s success is already catching the attention of other ETF issuers. In recent days, T-Rex has filed for a 2x inverse DRAM fund. Leverage Shares has filed for a long 2x memory ETF. Defiance has submitted paperwork for an options-income vehicle based on DRAM.
“The monumental success of DRAM has surprised even me,” said Jane Edmondson, head of index product strategy at TMX VettaFi and an 11-year veteran of the ETFs space. “Now we are seeing levered, inverse, and income-enhanced versions of DRAM,” she said, adding “how quickly this has all occurred is unprecedented.”
Most Read from Bloomberg Businessweek
©2026 Bloomberg L.P.
