SOXL vs SOXS and Why Owning Both 3x Semiconductor ETFs Lost 3.31 Percent

On SOXL vs SOXS, a 50/50 book in both 3x semiconductor ETFs lost 3.31 percent over the 27 daily bars the two Phemex perpetuals share, from 26 August 2026 to the Monday 21 September close. SOXL gained 15.83 percent across that window and SOXS lost 22.45 percent.

Rebalance that book to 50/50 every session and the loss shrinks to 0.71 percent. Fees account for 0.062 points of it. The remaining 0.644 points is the daily reset, the mechanic both funds spell out on the first page of their own summary prospectuses. SOXS listed on 26 August 2026, so those 27 bars are the entire shared history and no paired 30-session window exists.

What Do SOXL and SOXS Do?

Both funds come from one issuer and both reset at every US close. The SOXL ETF seeks 300 percent of the daily performance of the NYSE Semiconductor Index, and the SOXS ETF seeks 300 percent of the inverse of the same index. That index is a float-adjusted, market-cap-weighted basket of the thirty largest US-listed semiconductor companies, defined by ICE's own sector classification.

Traders get the benchmark wrong on this pair more often than they get the leverage wrong. The ticker SOX belongs to the PHLX Semiconductor Sector Index, and neither of these funds tracks it. A supplement filed on 3 November 2023updated the name of the underlying index to the NYSE Semiconductor Index for both funds, and the 27 February 2026 summary prospectuses carry that name throughout. Quote a SOX level next to a SOXL move and you are comparing two different baskets.

Our 2026 explainer on the SOXL ETF teaches the daily reset with an invented index that drops 10 percent and rebounds 11.1 percent. This page runs the same arithmetic on bars that printed.

SOXL vs SOXS Decay Over the 27 Bars They Share

Anyone measuring SOXL vs SOXS decay over 30, 90 or 180 sessions is measuring something that doesn't exist. SOXS listed on Phemex at 10:00 UTC on 26 August 2026 and SOXL listed on 18 May, so the overlap runs 27 daily bars to the 21 September 2026 close. That is every session the two contracts have ever shared, and a 30-session pairing would mean inventing three more.

Those 27 bars split 19 weekday and 8 weekend, and the split changes what the numbers mean. The US cash market is shut on Saturday and Sunday while the perpetual keeps printing a mark, so the weekend bars carry thinner turnover and smaller moves. SOXL's median absolute move across the 18 weekday return periods is 5.66 percent against 2.35 percent across the 8 weekend ones. How TradFi perps trade when Wall Street is shut covers what sets the price on those bars.

The anchor bar shows the pair doing its job. SOXL closed 144.77 on 21 September 2026, up 16.19 percent on the Sunday bar and up 17.55 percent measured from the Friday 18 September close. SOXS closed 34.62 on the same session, down 15.99 percent and down 16.90 percent against those two references. A 50/50 book gained 0.10 percent on the wildest day in the window, which is the whole argument for the pair and the reason the slow loss goes unnoticed.

What Owning Both 3x Semiconductor ETFs Cost

Put $5,000 into each leg on 26 August 2026, leave both untouched, and the 21 September close values the book at $9,669. The SOXL side is worth $5,792 and the SOXS side $3,878. That is 3.31 percent gone from a position holding no directional view at all. The chip cycle had nothing to do with it. You paid for the privilege of owning both answers at once.

Rebalance back to equal weights at every daily close and the same 26 return periods leave $9,929, a loss of 0.71 percent. The held version loses more because one leg ran and the other collapsed, so by mid-September the book was mostly SOXL and no longer market-neutral in any useful sense. The rebalanced figure is the cleaner measure of what the structure costs, because it pins the weights and lets only the compounding move.

Neither number includes what you pay to hold the exposure. How SOXL/USDT trading stacks two layers of leverage is the other half of this problem. A perpetual position adds its own multiple on top of the fund's three, and funding lands on it three times a day.

How Much of the SOXL vs SOXS Loss Is the Expense Ratio?

The obvious suspect for a 0.71 percent loss over 26 return periods is fees. The arithmetic says no.

The SOXL expense ratio in the fund's 27 February 2026 summary prospectus runs 0.91 percent gross, cut to 0.75 percent by a contractual advisory fee waiver that holds through 1 September 2027. SOXS carries no waiver at all and its table totals 1.00 percent, built from a 0.75 percent management fee, 0.12 percent of other expenses and 0.13 percent of acquired fund fees. Average the two and spread them across 26 of 365 days and you get 0.062 points of drag.

Some fund data pages quote SOXS at 0.87 percent, which is the same table with acquired fund fees stripped out. The total is the number your return carries, and the 27 February 2026 filing puts that total at 1.00 percent.

That leaves 0.644 points, more than ten times the fee drag, produced by the reset itself. Both documents say so before they say anything else. The SOXS prospectus warns that the fund will lose money if the index is flat over any period longer than a trading day. It also says you could lose the full principal value inside a single session if the index gains more than 33 percent.

What the Bootstrap Shows and What It Does Not

Resampling says the 0.71 percent is real and it also says the window is too short to call it proven.

Method: draw 26 paired daily returns with replacement from the 26 the window produced, then compound a daily-rebalanced 50/50 book across them. Repeat that 1,000 times under each of 40 seeds numbered 1 to 40. The test is one-sided, because the only question is how often the book ends below zero.

The median draw comes back at -0.701 percent and the seed-to-seed spread of those medians is 0.105 points, so the point estimate barely moves when the randomness changes. What doesn't come back is significance. Only 74.4 percent of draws land below zero, which leaves better than one draw in four above it. That result is consistent and it is not significant, and reading it as proof that the pair always bleeds overstates what 26 return periods can carry.

One check sits underneath all of it. Divide each leg's daily return by its own multiple and you get two independent readings of the same index, one from the bull fund and one from the bear. Across the 26 periods those readings agree to a median 7.94 basis points, and the worst single session is a 50.0 basis point gap on 14 September 2026. The two contracts are pricing the same basket, so the loss is structural and not a quoting artefact.

How Does Leveraged ETF Decay Look in SOXL Alone?

SOXL holds 127 daily bars from its 18 May 2026 listing, which covers 30 and 90 sessions and still falls short of 180. Measuring leveraged ETF decay on it needs a benchmark series, and the same division supplies one. Take SOXL's daily return and divide by three to get the implied index. Compound that across the window, multiply by three, and the gap to what SOXL actually returned is the decay.

The gap runs 2.27 points across the 30 bars from 22 August to 21 September 2026, nine of them weekend marks. Stretch it to 90 bars back to 23 June and the gap is 13.93 points, and across all 126 return periods in the file it is 38.87 points. Every one of those readings goes against the holder. Over those 126 periods the implied index gained 33.63 percent while SOXL lost 5.24 percent. Three times a 33.63 percent gain is a long way from minus 5.24, and the distance is the reset compounding against a violent tape.

Widen it to every rolling 30-session window the file supports and 84 of 97 come back negative, which is 86.6 percent. What leverage trading does to a position covers the mechanics from your side of the screen. The fund version automates the same move at every close and never asks the holder.

Can You Trade SOXL and SOXS on Phemex?

Both perpetuals are Listed. SOXLUSDT caps at 10x and SOXSUSDT at 20x, read from the product record and never from the page copy, and each contract is worth one share of its fund. Funding settles on an eight-hour interval, so a position held a full day pays or collects three times.

Turnover on the anchor bar ran $1,157,856 on the SOXL contract against $1,860,886 on SOXS, so the bear fund traded the heavier book on the session it fell 15.99 percent. Trading stocks and precious metals around the clock sets out what those weekend sessions are for and who is on the other side of them.

Frequently Asked Questions

Where did SOXL and SOXS close at their window extremes?

SOXL's lowest close of the 27 bars was 102.53 on 15 September 2026 and its highest was the 144.77 anchor. SOXS peaked at 52.63 on 1 September 2026 and bottomed on the anchor at 34.62.

What do the filings say a $10,000 position costs to hold?

The SOXS summary prospectus works a $10,000 investment at an assumed 5 percent annual return out to $102 of expenses over one year and $1,225 over ten. Both funds were also renamed in the February 2026 documents, from Shares to ETF.

What portfolio turnover rate do the filings report?

The SOXS summary prospectus reports 0 percent for the most recent fiscal year, then adds that the calculation ignores cash instruments and derivative transactions and would be significantly higher if they counted.

How thin are the weekend bars on SOXL?

Median quote turnover across the 8 weekend bars in the common window is $863,883 against $1,166,128 across the 19 weekday bars.

Bottom Line

Owning SOXL and SOXS together is a bet on nothing, and it still costs money. Held untouched from 26 August to 21 September 2026 the pair took 3.31 percent. Rebalanced daily it took 0.71 percent, and only 0.062 of those points reached the two issuers as fees.

The rest is the product doing exactly what its prospectus says it does. Both documents warn on page one that any period longer than a trading day compounds, and that a flat index still loses money for the holder. Twenty-six return periods is a thin base to prove that on, and 74.4 percent of resampled draws below zero is a direction, not a verdict.

What the 27 bars do settle is the size. The structure charges roughly seven tenths of a point a month for holding both sides of a semiconductor view, on top of fees that cost a twentieth of that, and it charges it whichever way the chips go. A trader who wants the exposure without paying the reset can take the view directly on the contract and close it the same day.

Disclaimer: This article is for informational purposes only and is not financial advice. Cryptocurrency trading involves substantial risk. Always do your own research before making investment decisions.