The hottest ETF since bitcoin-mania just added $1 billion in a day

Roundhill's Memory ETF (DRAM) has gathered more than $5 billion since its April 2 launch, including $1.1 billion on Thursday alone, and has risen 70% since inception. The fund has seen inflows every day for 23 straight sessions, with options activity surging above 90,000 contracts traded Thursday as investors target AI memory-chip bottlenecks. The move highlights strong demand for AI-linked chip exposure, especially names such as Micron, SK Hynix, and Samsung Electronics.

Analysis

This flow is less about a single ETF launch and more about a forced rerating of the entire memory complex: when a new product captures multi-billion dollar inflows in days, it creates a reflexive loop where primary issuance, dealer hedging, and headline momentum all reinforce each other. The second-order effect is that the liquid U.S.-listed names become the translation layer for global memory scarcity, so the market is likely to keep paying up for the most “accessible” proxies even if fundamentals merely stay tight rather than improve further.

The bigger implication is that the winner set is broader than the obvious U.S. chip names. Korea exposure is a structural gap in most domestic portfolios, so any vehicle that packages SK Hynix/Samsung-like purity gets a scarcity premium; that tends to compress the relative valuation discount of the most direct U.S. substitutes. Conversely, suppliers and adjacent equipment vendors may see mixed effects: the memory producers benefit from pricing power, but the capital cycle will eventually pull forward tool orders and raise the probability of a supply response 6-12 months out.

Short-term, the most important risk is positioning, not fundamentals. A 70% ETF rally plus concentrated call demand means the next air pocket likely comes from volatility reversion or a pause in inflows, not from a change in the AI narrative; that makes the trade vulnerable to a 1-2 week digestion phase even if the medium-term thesis remains intact. Over a 3-9 month horizon, the main bearish catalyst is capacity additions and margin normalization if pricing remains elevated long enough to justify aggressive wafer starts.

The contrarian view is that the market may be overestimating how much of the AI spend becomes durable memory demand versus cyclical inventory restocking. If hyperscaler capex growth slows, memory names can underperform quickly because they are the cleanest expression of the AI bottleneck trade and therefore the first to de-rate when sentiment cracks. That argues for owning the secular theme but financing it with short volatility or a relative-value hedge instead of outright chasing beta after the move.

Key Decisions for Investors

• Buy SNDK on a 2-3 week pullback rather than strength; use a 10-15% retracement as entry, targeting a continued momentum leg over 3-6 months with the stop below the prior breakout level.
• Pair trade: long SNDK / short CBOE on a 1-2 month horizon; monetizes the divergence between speculative AI flow and the exchange’s ability to capture volume, with better upside if retail call demand stays hot.
• Take profits on momentum longs into the next 5-10% ETF-level extension; use call-heavy sentiment as a signal to trim 25-33% of exposure because the trade is increasingly reflexive and vulnerable to a volatility reset.
• Structure a call spread on SNDK or a basket proxy for a 3-6 month view; prefer defined-risk upside participation because implied vol is likely inflated by the same call buying driving the tape.
• Monitor for supply-response signals from Korea/U.S. foundry commentary over the next quarter; if capex guides up materially, rotate from outright long memory beta into a relative-value long/short rather than staying naked long.