Semis Dip-Buy Gets Crowded: $24.7B Chases SOXX, DRAM Lower
$24.7 billion has poured into SOXX, DRAM, SMH, and SOXL since the June 22 peak. Is the semiconductor ETF dip-buy trade already too crowded to work?
Key Points
- Since the June 22 peak, DRAM has dropped nearly 40%, SOXX 24%, SMH 20%, and SOXL 61% — yet investors have deployed $24.7 billion combined into these four funds during the selloff.
- The semiconductor complex peaked simultaneously across all major fund structures, suggesting a sector-wide repricing rather than single-stock idiosyncratic risk.
- The critical question for traders is whether $24.7 billion in fresh dip-buying capital has already priced in the recovery, or whether a second leg lower would trigger forced liquidations in the leveraged products.
Nearly $24.7 billion in fresh money has poured into four semiconductor ETFs since June 22 — while every one of those funds is still in an active double-digit drawdown. The iShares Semiconductor ETF SOXX has lost 24% from its peak. The Roundhill Memory ETF DRAM has dropped close to 40% from its late-June high of 80.72 to an intraday low of 48.64. The VanEck Semiconductor ETF SMH is off 20%. And the Direxion Daily Semiconductor Bull 3X SOXL has been cut by 61%. Investors are not waiting for a bottom to be confirmed — they are buying into the decline with conviction that has not been seen in any other sector this year.
The Scale of the Bet
To understand the magnitude of what is happening in semiconductor ETFs right now, start with the single-day SOXX print from July 8: $5.43 billion in net creations in one session, lifting the fund's AUM by 11.73% to approximately $46.3 billion and accounting for a significant portion of the $18.6 billion that flowed into the entire ETF market that day. That single event is among the largest one-day inflows ever recorded for a sector ETF. For the full week of July 7 through July 10, SOXX pulled in $5.3 billion, SOXL added $2.4 billion, and SMH gathered $2.5 billion. DRAM, the newest entrant in the group — launched in April 2026 — collected $1.7 billion in that same week alone, capping what had already been a $10 billion June for the fund.
The aggregate since the June 22 peak across all four funds: DRAM at $8.8 billion, SOXX at $8.5 billion, SOXL at $5.1 billion, and SMH at $2.3 billion. That $24.7 billion combined figure is the most concentrated dip-buying event in any single sector that the ETF market has produced in 2026, in a year already on pace for $2 trillion in total ETF inflows. The positioning is not subtle, and it is not small.
Why the Trade Is Structurally Complicated
The YTD performance context is essential to understanding why buyers feel emboldened. Through the first half of 2026, SOXX returned 90.03%, outperforming SMH — at 68.78% — by more than 20 percentage points. DRAM, despite launching only in April, gained 166% from launch to its June 22 peak. Triple-leveraged semiconductor products gained 440% in H1 2026. Investors who bought and held through the first six months of this year made generational returns in these funds, and the June 22 drawdown looks, on those charts, like a routine pullback against a historic uptrend. The psychological anchor of those gains is driving the dip-buying behavior as much as any fundamental thesis.
But the structure of the drawdown raises questions that the flow data alone cannot answer. All four funds peaked simultaneously on June 22, which argues for a macro or sector-wide trigger — not stock-specific disappointment — as the driver of the selloff. A sector-wide repricing tied to valuation, rate sensitivity, or a shift in the AI capital expenditure narrative tends to resolve more slowly and more painfully than a single-name earnings miss. The 10-year Treasury yield sitting at 4.63% as of July 21 — versus the 2-year at 4.26% — means the yield curve is steepening in a way that historically pressures high-multiple growth sectors. Semiconductors, which trade at elevated forward earnings multiples on the expectation of sustained AI-driven demand, are directly exposed to that dynamic.
The leveraged product risk adds another layer of complexity. SOXL's 61% drawdown from peak is an arithmetic consequence of 3x daily leverage applied to a 24% index decline — but it also means that any fresh capital entering SOXL is operating with a fund that must recover roughly 156% just to return to its June 22 level. Leveraged ETFs decay through volatility compounding even in flat markets. The $5.1 billion that has entered SOXL during the drawdown is exposed not just to the direction of semiconductors but to the path dependency that destroys leveraged returns in choppy, sideways markets. If the sector does not move decisively higher from current levels, SOXL holders face erosion even if semiconductors eventually recover.
What Needs to Happen Next
DRAM's current AUM of $23.4 billion — only modestly below the late-June peak of $25.9 billion despite the fund losing close to 40% of its price — tells you something specific about the behavior driving these flows. Assets under management fell far less than NAV because new money entered the fund at nearly the same rate as existing holders' NAV declined. That dynamic is only sustainable if the new buyers are right about the timing and depth of the recovery. If the selling pressure extends, DRAM's AUM will compress sharply as price decline outruns new inflows — and at $23.4 billion in a fund that is only three months old, the forced-selling risk on a second leg lower is not trivial.
SMH, by contrast, sits at nearly $69.8 billion in AUM, making it one of the largest semiconductor funds in existence and giving it the liquidity cushion to absorb significant outflows without structural disruption. SOXX at $46.3 billion is similarly positioned. The existential risk in a deeper drawdown belongs to the newer and leveraged products — DRAM and SOXL — where a sustained decline could trigger the kind of redemption spiral that forces portfolio managers to sell underlying holdings into a falling market, accelerating the decline.
For traders, the specific levels to watch are DRAM's intraday low at 48.64 and SOXX's post-peak trough. A decisive break below the July intraday lows on volume that matches or exceeds the dip-buying prints would signal that $24.7 billion in fresh capital is underwater and that the crowded trade is unwinding, not consolidating. Conversely, a recovery that carries SOXX back above its 50-day moving average would validate the dip-buyers and likely trigger a second wave of momentum inflows from trend-following systematic strategies. The next major semiconductor earnings catalysts — including Texas Instruments, which filed an 8-K on July 22 — will be the fundamental data points that determine which scenario plays out. Watch TXN's guidance commentary on AI-related chip demand as the first real test of whether the June 22 peak was a pause or a top.
The Weekly Investor
Daily market analysis for active traders. Free.
