The best-performing way to play the chip boom in 2026 has not been Nvidia (NVDA +1.70%), Advanced Micro Devices (AMD +7.45%), or Broadcom (AVGO +1.91%). It has been a small, brand-new fund built around the one corner of semiconductors that giants can't live without: memory. The Roundhill Memory ETF (DRAM +10.55%) has roughly doubled since it launched this spring, outrunning all three AI chip stars, and the reason comes down to a technology most investors ignore.

NYSEMKT: DRAM
Key Data Points
What DRAM is, and why it became the AI bottleneck
DRAM, or dynamic random-access memory, is a computer's short-term working memory, the place a chip keeps the data it is actively crunching. For decades, it was a sleepy commodity. AI flipped that on its head. Running and training large models means moving colossal amounts of data at high speed, which has created frantic demand for a premium form of DRAM called high-bandwidth memory, or HBM. HBM is DRAM stacked in layers and placed right beside the AI processor so information flows almost instantly.
Here's the crux: Each new generation of AI hardware needs far more of this memory than the last, and only a few companies can make it. That has turned memory, not the processors themselves, into the tightest bottleneck in the entire AI supply chain. When a critical ingredient is scarce, its makers gain enormous pricing power, and their stocks can run even harder than the chip designers everyone talks about.
Image source: Getty Images.
Inside the Roundhill Memory ETF
The Roundhill Memory ETF has taken the DRAM moniker for its ticker, and it's the first exchange-traded fund built purely around memory-chip makers. It holds about 20 companies that each derive at least half their revenue from memory products, spanning DRAM, HBM, NAND flash, and solid-state drives. That gives you one-ticker access to the whole memory complex rather than betting on a single name.
The concentration, though, is extreme. Roughly three-quarters of the fund sits in just three stocks: Samsung, SK Hynix (SKHY +12.10%), and Micron Technology (MU +11.80%), the trio that dominates global memory production. It also holds stocks such as Sandisk (SNDK +11.58%). Since its April debut, the fund has surged well past 100%, trouncing Nvidia, AMD, and Broadcom over the same stretch. The engine is simple: Memory companies have been the hottest part of the chip market, with HBM sold out and SK Hynix reportedly claiming the lion's share of the memory going into Nvidia's next-generation systems.

NASDAQ: SKHY
Key Data Points
Why it has beaten the AI chip giants
The AI story is usually told through processors, but a processor is useless without enough memory to feed it. Because HBM eats up far more factory space per chip than ordinary memory, supply has stayed painfully tight even as demand explodes. That scarcity has let memory makers raise prices and lock in multiyear contracts, driving their shares higher than the chip designers. Owning the memory basket, in other words, has been like owning the shovel sellers during a gold rush.
The catch investors should weigh
I wouldn't mistake this for a safe, diversified fund. With about 75% in three stocks, the DRAM ETF is really a concentrated bet dressed as an ETF, and a stumble at any one of them would sting. Two of those three are foreign companies, which adds currency and geopolitical exposure. The fund charges 0.65% a year, on the pricey side, and it is brutally volatile, with single-day swings above 14% in both directions. Most importantly, memory is famously cyclical. It already dipped into a bear market this summer, a reminder that today's shortage can flip to glut, and prices can fall as fast as they rose.
How investors should think of Roundhill Memory ETF
The Roundhill Memory ETF is a clever, convenient way to ride the memory supercycle that has quietly outrun the AI chip elite. But convenience is not the same as safety. You are buying a concentrated, high-fee, high-volatility bet on one of the most cyclical industries in tech. If you believe the memory shortage has legs, DRAM offers exposure in a single ticker. Just size it small, and never forget that in memory, the boom and the bust are always close cousins.



