Leveraged ETFs: growth, crash risk and a portfolio you can maintain

Leverage can improve compounded growth, but the useful retail decision is how much exposure the whole portfolio can sustain after costs, crashes and rebalancing. From the Nasdaq 100's record close on 6 October 2026 to 9 November, the index fell 10.6% and ProShares' 3x TQQQ 31.6%; the index needs an 11.9% gain to get back, TQQQ needs 46.2%.

In this fall the 3x funds did not lose more than three times their benchmark; they lost slightly less. The damage sits in the recovery, which a daily-reset fund has to earn on a smaller base with every swing working against it. The fund's 3x label alone does not tell you your portfolio's exposure. That also depends on its weight, your other holdings and how you maintain them.

Key takeaways

1. The loss was about three times, the recovery hurdle about four. TQQQ needs 46.2% to regain its 6 October NAV, 3.9 times the 11.9% the Nasdaq 100 needs; SOXL needs 5.7 times what the semiconductor ETF needs.
2. Kelly estimates are sensitive to the period. In a simplified single-index model, the fitted growth-maximising multiple was 3.8 for 2010-2026 and below zero for 2000-2009; neither is a portfolio allocation to copy.
3. A crash can end the position. Simulated from the March 2000 record after fees and financing, $100 in a 3x Nasdaq 100 fund fell to 1.6 cents by March 2009 and was worth $14.66 on 9 November 2026.
4. Count exposure across the whole portfolio. A smaller leveraged holding alongside cash has different risks from putting everything into the fund; the multiplier itself adds no independent return source.
5. Write the maintenance rule before investing. The fund resets its own leverage each day, but it does not rebalance your portfolio or protect your spending money; compare costs and monitoring demands with unleveraged exposure.

How much did 2x and 3x Nasdaq and chip funds lose from 6 October?

To 9 November, ProShares' 2x QLD lost 21.4% and 3x TQQQ 31.6% of NAV while the Nasdaq 100 lost 10.6%; Direxion's 3x SOXL lost 65.7% while the iShares Semiconductor ETF lost 25.3%. The Nasdaq 100 closed at its record of 31,224.69 on 6 October and at its lowest since then, 26,629.81, on 5 November, after falling 6.4% that day. TQQQ's NAV fell 19.1% on 5 November and SOXL's 38.7%.

6 October to 9 November 2026: Nasdaq 100 −10.62%, QLD −21.36%, TQQQ −31.58%; iShares Semiconductor ETF −25.31%, SOXL −65.74%. Source: FRED NASDAQ100, ProShares TQQQ NAV history and QLD NAV history, Direxion SOXL product page, Nasdaq SOXX quotes; episode summary (CSV), daily levels (CSV).

At the 5 November low the funds had lost 1.92, 2.74 and 2.29 times their benchmark's fall.

SOXL tracks the NYSE Semiconductor Index, the benchmark of the iShares Semiconductor ETF, so we use that ETF's closing prices as its unleveraged comparison. For semiconductors the record dates from 22 June 2026: from that close to 9 November, the iShares ETF lost 32.8% and SOXL 81.3% on closing prices. A three-times loss of 98.6% would have left almost nothing; SOXL kept a quarter of its value.

Why did a 3x fund lose less than three times in the fall?

Because the fund resets to three times its current value every evening, so after each down day its next loss is taken on a smaller amount. In a fall that runs mostly in one direction this works in the holder's favour. Our daily simulation of a 3x Nasdaq 100 fund, after TQQQ's 0.78% fee and financing at the federal funds rate, lost 31.4% over the period; TQQQ's NAV lost 31.6%.

The same rule costs money when the index goes back and forth. From 14 to 29 October, 11 trading days, the Nasdaq 100 slipped 1.8% while TQQQ lost 6.5%, 3.5 times as much. The iShares Semiconductor ETF lost 1.1% over those days and SOXL 7.2%, 6.7 times as much. A swing of +10% and −10% costs the index 1%; at three times it costs 9%.

The general statement is a formula, not a forecast. For a fund that holds a constant multiple L of an index with expected return μ and volatility σ, financed at the cash rate r, expected log growth is roughly r + L(μ − r) − L²σ²/2. Tripling the multiple triples the excess-return term and multiplies the variance term by nine.

Between 6 October and 9 November the Nasdaq 100's daily moves annualised to a volatility of 32.5%, and the iShares Semiconductor ETF's to 70.1%. At those levels the extra variance cost of 3x, beyond three times the index's own, would be about 32 percentage points of log growth a year on the Nasdaq 100 and well over 100 points on semiconductors. Those are rates for a volatile month, not annual expectations.

How much do the funds now need to recover?

From the 9 November close, the Nasdaq 100 needs 11.9% to regain its record, QLD 27.2%, TQQQ 46.2%, the iShares Semiconductor ETF 33.9% and SOXL 191.9%. The required gain is the old level divided by the current one, minus one, the same arithmetic as in our guide to recovery time and loss maths. The difference is that the leveraged fund cannot earn that gain directly; it has to earn three times each day's index move.

Gain needed from 9 November 2026: Nasdaq 100 11.89%, QLD 27.17%, TQQQ 46.15%, iShares Semiconductor ETF 33.88%, SOXL 191.85%. Source: ProShares NAV histories, Direxion SOXL product page, FRED NASDAQ100, Nasdaq SOXX quotes; episode summary (CSV).

Suppose, as a scenario, that each unleveraged benchmark climbs back to its 6 October level over 20 trading days. We ran two paths, before fees and financing: 20 equal daily gains, and the same net gain in alternating up and down days, with swings the size of the average daily move between 6 October and 9 November.

In both paths the benchmark is back at its 6 October level and the leveraged funds are not. A round trip that leaves the index where it started leaves a daily-reset fund lower, and the larger the swings along the way, such as SOXL's 38.7% fall on 5 November, the larger the shortfall. Source: recovery scenarios (CSV).

New money kept arriving during the fall. TQQQ's shares outstanding rose from 482.65 million on 6 October to 563.0 million on 9 November; valued at each day's NAV, that is about $5.1 billion of net creations. Its assets still fell from $40.65 billion to $32.44 billion.

Counting the same way, TQQQ's net creations in the calendar month of October come to $3.25 billion, against $3.49 billion reported by 247 Wall St on 3 November. 247 Wall St also reported $7.27 billion of October inflows into SOXL, a press figure we could not check against Direxion's data. Flow totals record money arriving, not what each buyer paid or whether they still hold.

What can Kelly sizing tell you about long-run growth?

It turns the fund into a constant-leverage portfolio, so its long-run growth depends on how the index's excess return compares with its variance over the years you hold it. That can work either way. Applying the formula above to Nasdaq 100 daily returns, with the federal funds rate as the cost of borrowing, gives very different answers depending on the sample.

These are historical inputs to a single-asset Kelly approximation: maximising expected log wealth by choosing leverage. The model assumes returns independent from day to day, constant mean and volatility, continuous rebalancing, no fees and borrowing at the federal funds rate. None of those holds exactly, and the estimates of the mean are the weakest input. Source: FRED NASDAQ100 and DFF; formula inputs (CSV).

The table does not identify a correct multiple. It shows that the multiple that maximises growth is set by inputs nobody observes in advance, and that being above it costs growth quickly because the variance term grows with the square of leverage. For a positive growth-maximising multiple, leverage above twice that level puts the formula's expected growth below the cash rate.

If short selling is excluded, a negative fitted optimum becomes zero exposure in this model, not a short position to copy. Full Kelly can still suffer deep drawdowns. Using a fraction of an estimate reduces exposure but does not establish crash protection or remove the risk of estimating returns incorrectly.

For a household portfolio, compare allocations and leverage together, including no borrowing, under the same spending needs and financing constraints. Test lower expected returns, joint crashes and more expensive funding. The single-index formula above cannot choose a holding's weight alongside stocks, bonds and cash without their joint return assumptions.

Our comparison of the Sharpe ratio and CAGR explains why an arithmetic average does not describe the growth of a balance; a leveraged fund widens that gap. In this sell-off the simple multiple misled in both directions: the 3x loss was below three times the index's, and the recovery hurdle is far above three times.

What happened to a 3x Nasdaq fund held through 2000-2002 and 2022?

Simulated from the Nasdaq 100's March 2000 record, a 3x daily fund lost 99.98% by March 2009 and had not regained its starting value on 9 November 2026; started in March 2010, the same rule compounded at 35.9% a year. TQQQ was launched in February 2010, so the earlier periods are simulations, not fund records.

Annualised to 9 November 2026, index, 2x and 3x simulated: from January 1986 14.02%, 16.28%, 11.69%; from 27 March 2000 6.92%, 3.05%, −6.96%; from 31 March 2010 17.34%, 28.53%, 35.93%; from 19 November 2021 11.05%, 11.22%, 6.44%. Source: FRED NASDAQ100 and DFF; simulation results (CSV).

The simulation applies two or three times each day's index price return, subtracts the federal funds rate on the borrowed part and a yearly fee of 0.89% or 0.78%, ProShares' current net expense ratios for QLD and TQQQ. Checked against TQQQ's actual NAV from March 2010, it compounded at 35.9% a year against the fund's 37.5%, and differed by 2.8 percentage points in an average calendar year. Source: simulation against TQQQ (CSV).

In the 2000-2002 crash the index itself fell 82.9% and took 15.6 years to regain its record. From the same record the simulated 2x fund lost 99.0% by March 2009 and took 21.3 years to get back; the 3x fund lost 99.98%, after which it needed a gain of more than 600,000%. Even from January 1986, which includes the whole 1990s boom, the simulated 3x fund grew more slowly than the index.

The fund's own record shows the same mechanism. In calendar 2022 the Nasdaq 100 lost 33.0%, QLD 60.6% and TQQQ 79.1%. From the November 2021 record, the simulated 3x fund regained its starting value only in December 2024, and by 9 November 2026 had grown more slowly than the index. The crash study across asset classes puts the 2000-2002 and 2022 falls in context.

Does holding a 3x fund make your portfolio three times leveraged?

No: the portfolio's exposure depends on the fund's weight and on leverage in the other holdings. For one day's index move, effective exposure is the fund's portfolio weight multiplied by its daily leverage, plus exposure in other holdings. The following hypothetical mixes begin with the same index exposure; the weights illustrate mechanics and are not suggested allocations.

Source: our exposure and drift calculation (CSV). The scenario assumes exact daily multiples, unchanged cash, no fees, financing, taxes or currency moves. The fund resets internally, but its shrinking weight reduces the portfolio's next-day exposure. Keeping exposure matched across days would require portfolio rebalancing, with its costs and use of cash.

Real products also differ in cost. At an illustrative half weight, Amundi's 2x fund's 0.60% ongoing charge contributes 0.30% a year to portfolio costs, before financing and other costs. At a one-third weight, QQQ3's 0.75% management charge contributes 0.25%, and its KID's 2.34% transaction-cost estimate contributes another 0.78%. These figures do not establish which implementation is cheaper overall.

Compare the chosen unleveraged fund's fee with those costs, the yield actually earned on retained cash, spreads, taxes and the cost of rebalancing. Check whether quoted transaction costs already include financing before adding it again. The product documents below establish available structures, not a common net return or a best allocation.

Retained cash can fund spending or rebalancing, but the same money cannot be fully committed to both. If the leveraged fund goes to zero, the separately held cash remains, assuming it was not pledged or invested in another risky holding. That limits the loss to the funded position; it does not make the strategy growth-optimal or insure the remaining portfolio.

Which leveraged products can a European investor buy, and what do their documents say?

The products we checked all state a recommended holding period of one day in their key information documents, and two of the three sit in the highest PRIIPs risk class. TQQQ and SOXL are US funds; European retail investors usually reach the same exposure through exchange-traded products or UCITS funds listed in Europe.

WisdomTree's KID for QQQ3, dated 5 October 2026 like the one for 3SEM, states that "For periods of longer than one day, the return of the product is not the same as the return of the Index multiplied by a factor of 3". Its one-day stress scenario shows a loss of 100%. Amundi's KID is dated 12 June 2026. ProShares' page for TQQQ says: "For any holding period other than a day, your return may be higher or lower than the Daily Target. These differences may be significant."

The ETPs are debt securities of WisdomTree Multi Asset Issuer PLC backed by swaps and collateral, which adds issuer and counterparty risk; Amundi's KID describes its fund's assets as held by the depositary, separate from the management company.

QQQ3 tracks three times the Nasdaq 100 net total return index, with a dollar base currency and no currency hedge, so a euro holder also carries the exchange rate. Amundi's KID says its 2x fund invests at least 75% of its assets in shares of EU companies, which makes it PEA-eligible, and swaps that portfolio's performance for the leveraged index's.

What can you check before holding a leveraged ETF?

These checks use your broker statement and the product documents; none of them tells you whether leverage suits your plans.

1. Read the holding period and risk class in the KID. All three European documents above assume one day; held for months, you are outside the case the risk indicator describes.
2. Compute your own recovery hurdle. Divide the NAV at your purchase, or at the last peak, by today's NAV and subtract one, then compare it with the same figure for the index.
3. Count your exposure with the multiple. Multiply the fund's weight by its leverage and add the index exposure in your other ETFs; 10% in a 3x Nasdaq fund moves like 30% of the portfolio in the Nasdaq 100 for a day.
4. Replay 2000-2002 and 2022 on the position. The simulated 3x fund lost 99.98% and 81.7%; ask what the rest of the portfolio would still fund if the position went close to zero.
5. Add the costs the TER leaves out. QQQ3's KID adds 2.34% of transaction costs a year to its 0.75% fee, and its return is "adjusted to reflect fees and costs inherent to maintaining a leveraged position in stocks".
6. Check wrapper, issuer and currency. A debt-security ETP carries issuer and swap-counterparty risk; a product with a dollar base currency and no hedge exposes a euro investor to the exchange rate.
7. Write down your rule after a large fall. A position that has halved needs a 100% gain to recover before further costs; define the exposure or cash shortfall that would prompt a review.

The 10% figure in check 3 illustrates the arithmetic and is not a suggested weight. The same thinking applies to withdrawals during a crash: selling after a deep fall leaves fewer units for the rebound. For a broader view of the episode, our report on the November 2026 AI sell-off measures what it did to world-index investors.

What would maintaining it involve?

Record your effective exposure, cash reserved for spending, financing costs and the rule for rebalancing the portfolio. Specify a review interval and the changes in exposure, product terms or household needs that would trigger an earlier check. The fund handles its daily reset; its one-day recommended holding period is not an instruction that you must trade every day.

A portfolio designed to hold constant daily exposure needs much more frequent attention than one that allows it to drift. An annual review cannot enforce a daily leverage limit. If you cannot monitor the chosen rule, compare a simpler unleveraged implementation under the same growth and spending objectives before taking the extra exposure.