A 2x daily leveraged ETF that owns nothing but swaps on a single AI server stock did what 2x daily leveraged ETFs on single AI server stocks do when the AI server trade breaks. Defiance Daily Target 2X Long SMCI ETF (NASDAQ:SMCX) fell 22% on June 5, 2026, closing at $23.22 after opening near $29.95. The underlying, Super Micro Computer (NASDAQ:SMCI | SMCI Price Prediction), dropped 11% the same day, from $46.90 to $41.64. Two times eleven is twenty two. That is, on the daily reset, exactly the contract the fund signed.
What looks dramatic on a one-day chart looks worse on a one-year chart. A holder who bought SMCX a year ago at $69.42 is now sitting on a 67% loss, while the underlying SMCI is up 2% over the same window. That gap is the article. Everything else is detail.
What blew up the week
The selloff was not random. On Wednesday, June 3, Broadcom (NASDAQ:AVGO) guided Q3 AI semiconductor revenue to $16.0 billion against a consensus near $17.2 billion, a roughly 7% miss. On the same call, CEO Hock Tan suggested Google may use multiple chip suppliers, which is the kind of throwaway sentence that reprices an order book. Broadcom shares fell 13% to 15% on Thursday. Then Friday morning brought the second leg. May payrolls came in at 172,000 against an 80,000 estimate, the 2-year yield spiked to 4.16%, and the PHLX Semiconductor Index dropped 10.26%, erasing about $1.2 trillion in market value.
SMCI amplified the move rather than leading it. The stock is one of the purest expressions of hyperscaler AI server demand in the public market, and any guidance that suggests AI capital spending is not growing as fast as the order books implied hits it harder than the broad semi tape. Add a rate scare on top, and a name with $8.8 billion in total debt and convertibles on its balance sheet (per the most recent Q3 FY26 filing) gets repriced quickly.
How a 2x daily fund eats itself
SMCX does not own SMCI shares. It owns swaps. The fund’s most recent semi-annual report shows 42.7% of net assets in Super Micro Computer total return swaps spread across maturities from March 2026 through September 2028, with the largest single position, a 23.8% swap maturing March 10, 2026, doing most of the daily work. The rest sits in Treasury bills and government money market funds, about 11.3% combined, as collateral and liquidity. The gross expense ratio is 1.31%.
The mechanism that produced Friday’s print is the same mechanism that produced the one-year loss. The fund delivers 2x the daily return of SMCI, then resets. In a sustained uptrend with low volatility, that compounds beautifully. In a downtrend, it compounds the wrong way. In a choppy tape, it just bleeds, because a down-then-up sequence of equal percentages on the underlying leaves the underlying flat but leaves the 2x fund in a hole. That is how the product is designed to behave. The prospectus says so. The shareholder report says so. The math is unkind even when the salesperson is friendly.
To see the decay, line up the windows. Over the last month, SMCI is up 20% and SMCX is up 32%, less than two times the underlying despite the leverage label. Year to date, SMCI is up 42% and SMCX is up 2%. That is what volatility drag looks like in plain dollars. The leverage does its job each day. The compounding does its job each month.
What is actually scaring people about SMCI
The earnings record behind the AI hardware trade has been louder than its growth rate suggests. Q3 FY26 revenue came in at $10.24 billion, up 122.7% year over year, but missed the $12.45 billion consensus by 18%. Non-GAAP EPS of $0.84 beat the $0.62 estimate, and GAAP gross margin recovered to 9.9% from 6.3% sequentially, but the print was filed as preliminary pending Board review of export control matters. The previous quarter showed the opposite shape, with revenue beating handily and GAAP gross margin compressed to 6.3% from 11.8% a year earlier. Margins on AI server boxes are tight to begin with, and they get tighter when hyperscale and neocloud customers dominate the mix.
The longer-term concern centers on allocation. SMCI’s revenue depends on receiving tight allocations of NVIDIA Blackwell GPUs and reselling them inside complete systems. If hyperscalers diversify chip suppliers, as Tan suggested Google may, the dollar amount of NVIDIA silicon flowing through SMCI’s order book becomes a smaller share of total AI compute. The system integrator is geared to one customer mix. Change the mix and the cascade changes.
The macro tape is not helping. The 10Y-2Y Treasury spread sat at 0.38% on June 5, the lowest reading of the trailing twelve months, down from a 0.74% peak on February 9, 2026. A flattening curve driven by short rates going up is the exact wrong macro for a leveraged, capital-intensive name with debt to service.
What to actually watch from here
For SMCI itself, the next data points are knowable and dated. Any 8-K filing on customer wins or order book restatements matters more than the next earnings beat, because the market is no longer pricing on near-term EPS. Analyst target price sits at $37.62 against a current $41.64, with 5 buys, 9 holds, and 4 sells, and 39 recent insider transactions skewing toward selling. NVIDIA commentary on Blackwell allocation to system integrators and the enterprise server color from Dell and HP earnings will tell you whether the AI hardware trade is consolidating to a few winners or fragmenting. SMCI does fine in the first scenario and poorly in the second.
For SMCX specifically, the question is narrower. The fund will do exactly what it says on the tin on any given day. If SMCI rallies 10%, SMCX will rally about 20%, less the daily borrowing cost embedded in the swaps and the 1.31% expense ratio. If SMCI chops sideways for six months around current levels while the AI capex narrative gets re-underwritten, SMCX will lose money even if the underlying ends flat. The product is a directional, short-duration trading vehicle. Held longer than a few days, it is a bet on both direction and low realized volatility. Right now SMCI has neither in its favor. Reddit sentiment on the day of the drop scored 68 on wallstreetbets, which is bullish, which is fine, but bullish sentiment on a 2x fund in a chop tape is how the volatility drag eats people who were technically right about direction.
The honest read is that Friday’s 22% move was the product working as designed against a stock that lost an eighth of its market value in one session because the order book story behind AI server demand got a small but real downgrade. The conditions that produced the decline, a sustained downtrend in the underlying paired with elevated realized volatility, are exactly the conditions under which 2x daily-reset products do their worst work. Until SMCI either stabilizes into a low-volatility uptrend or the AI capex order book gets reaffirmed by NVIDIA and the hyperscalers, the leveraged wrapper carries unfavorable mechanics for the underlying thesis. The prospectus itself frames SMCX as a short-duration trading vehicle, not a buy-and-hold instrument.
Top Gaining Stocks
Top Losing Stocks