Artificial intelligence is turning the semiconductor industry upside down. For years, memory chips were treated like a commodity business — boom during shortages, bust during gluts, rinse and repeat. But AI may be rewriting the rules. Training large language models and powering hyperscale data centers requires staggering amounts of high-bandwidth memory, or HBM, and suddenly DRAM producers look less like cyclical chipmakers and more like critical infrastructure providers.
That shift has already minted huge winners for investors. The question now is whether buying a basket of the top memory names through the new Roundhill Memory ETF (CBOE:DRAM) offers a smarter way to ride the trend.
Memory Stocks Are Some of 2026’s Biggest Winners
The numbers tell the story better than any hype cycle could. Sandisk (NASDAQ:SNDK | SNDK Price Prediction) has returned more than 4,000% since being spun out of Western Digital (NASDAQ:WDC) in February 2025. Meanwhile:
| Company | 2026 Return |
| Micron Technology (NASDAQ:MU) | 163% |
| Seagate Technology (NASDAQ:STX) | 195% |
| Western Digital | 181% |
Those gains reflect a market finally recognizing that AI servers cannot function without massive amounts of memory bandwidth. Nvidia’s (NASDAQ:NVDA) GPUs may grab headlines, but those accelerators become bottlenecked without HBM and DRAM feeding them data fast enough.
According to Micron’s latest earnings release, HBM revenue is now measured in “multiple billions” annually, while the company expects demand to outstrip supply into 2027. Further, AI servers can require six times more DRAM than traditional cloud servers, changing the math for the entire industry.
DRAM Gives Investors Access to the Whole Memory Ecosystem
Buying Micron, Western Digital, and Seagate only captures part of the opportunity. It misses out on two of the most important memory companies in the world: SK hynix and Samsung Electronics. They are harder for U.S. investors to access directly because neither trades on a major U.S. exchange in the same way many retail investors prefer.
SK hynix has emerged as the dominant HBM supplier for Nvidia’s AI chips, while Samsung Electronics remains one of the largest memory manufacturers on Earth. That’s where DRAM comes in.
The Roundhill Memory ETF launched on April 2 and accumulated $6.5 billion in assets within just 27 trading days. That made it the fastest-growing ETF launch in history, surpassing the previous record held by iShares Bitcoin Trust (NASDAQ:IBIT), which took 30 days to cross the same threshold. Even more eye-catching, DRAM has already climbed 90% since launch.
However, the fund is concentrated — very concentrated. SK hynix, Micron, and Samsung account for roughly 73% of assets. And concentration risk cuts both ways. If memory pricing weakens or AI spending slows, DRAM will feel it quickly.
Still, those three companies arguably control the most critical layer of the AI memory stack.
The Micron Bet Could Magnify Returns — And Risks
DRAM also takes a more aggressive approach than traditional semiconductor ETFs. Roughly 9% of the portfolio is tied to a leveraged long total return swap linked to Micron stock. In other words, the ETF is also using derivatives to amplify gains if Micron rises.
That strategy has worked beautifully so far because Micron has become one of the market’s strongest AI infrastructure plays. But — like concentration — leverage works in both directions. If Micron drops 15%, the derivative exposure could deepen losses inside the ETF.
That said, the managers are clearly signaling conviction. Micron’s HBM leadership, improving pricing power, and expanding margins are central pillars of the AI memory story.
Key Takeaway
In short, the Roundhill Memory ETF is not a low-risk semiconductor ETF. It is a concentrated bet that AI-driven memory demand has fundamentally changed the industry.
Yet, the thesis is compelling. AI data centers are consuming unprecedented amounts of DRAM and HBM, supply remains tight, and the world’s leading memory producers are finally gaining pricing power after decades of brutal cyclicality.
Smart investors should understand the tradeoff. DRAM’s concentration and leveraged Micron exposure introduce volatility. But for shareholders who believe memory chips are evolving from commodity products into foundational AI infrastructure, this ETF may offer one of the purest ways to capture the trend.