Cathie Wood's ARK Invest portfolio strategy and ARKK risk

Cathie Wood's ARK Innovation ETF (ARKK) is a concentrated bet on disruptive technology that, on our four tests against MSCI World, is not set up for different market scenarios, has no guard against catastrophic equity falls and trails on volatility-adjusted returns, but beat the index on returns after fees.

At 30 September 2026 its ten largest holdings made up 51.3% of the fund, led by Tesla at 9.4%. From November 2014 to September 2026 it compounded at 14.1% a year in US dollars against 11.1% for MSCI World, by our calculation, but it lost 67.0% in 2022 against 18.1%.

Key takeaways

1. Set up for different market scenarios: no. ARKK amplified the equity cycle: it lost 67.0% in 2022 against 18.1% for MSCI World and gained 185.2% from January 2023 to September 2026 against 98.8%.
2. Guard against catastrophic equity falls: no guard. The fund is almost entirely equities with a beta of 1.77 to MSCI World; its worst month-end fall was 77.1% against 25.4% for the index, and it had not recovered by September 2026.
3. Better volatility-adjusted returns than MSCI World: no. ARKK's Sharpe ratio was 0.51 against 0.73 over 10 years and 0.49 against 0.67 over the full window, and it won on neither Sharpe nor Sortino in any of 24 rolling 10-year windows.
4. Better returns net of fees than MSCI World: yes. ARKK compounded at 15.8% against 12.8% over 10 years and 14.1% against 11.1% over the full window, and led in 23 of 24 rolling 10-year windows.
5. Most investor money arrived near the peak. Morningstar estimated $5.0bn of ARKK shareholder losses over the 10 years to 2025, even though the fund's own 10-year return was positive.

How does ARKK score on our four tests?

We judge ARKK, net of its 0.75% fee, against MSCI World Net in US dollars, with one-month Treasury bills as cash, using monthly returns to September 2026.

Windows: 10 years, Oct 2016 to Sep 2026; full window, Nov 2014 to Sep 2026. ARKK launched on 31 October 2014, so there is no 20-year window and no dotcom or 2008 episode.

What is Cathie Wood's investment strategy?

Cathie Wood, founder of ARK Investment Management, has managed ARKK since it launched on 31 October 2014, according to the fund's summary prospectus dated 30 November 2025. The fund puts at least 65% of its assets into companies relevant to "disruptive innovation", which ARK defines as "the introduction of a technologically enabled new product or service that potentially changes the way the world works." Within themes such as genomics, automation, energy, artificial intelligence and blockchain, ARK selects "its highest-conviction investment ideas", combining top-down sizing of markets with bottom-up company analysis.

ARK's help centre says its research looks for opportunities lasting "at least three to five years". The last of its six scoring metrics is valuation: a "5-Year Return (15% compound annual return hurdle rate)", a rate that would roughly double a holding in five years. ARK's 2026 investment case for ARKK forecasts global real GDP growth of 7.3% by 2030, against an IMF forecast of 3.1%, while noting that forecasts "cannot be relied upon."

The same deck says the fund "may be suited for investors who have a moderate-to-high risk profile and intend to stay invested for the medium-to-long term."

What does ARKK hold in 2026?

At 30 September 2026, ARKK's ten largest holdings made up 51.3% of the $7.92bn fund, according to ARK's fact sheet.

At 30 September 2026: Tesla 9.4%; Space Exploration Technologies 6.4%; Tempus AI 6.1%; Circle Internet Group 4.7%; CRISPR Therapeutics 4.6%; Coinbase Global 4.6%; Twist Bioscience 4.5%; Robinhood Markets 4.0%; 10x Genomics 3.6%; Shopify 3.4%; total 51.3%. Source: ARK, ARKK fact sheet, 30 September 2026, p. 1; chart data (CSV). Weights are shares of the fund as published by ARK.

The same fact sheet puts 29.9% of the fund in health care, 28.9% in information technology and 13.4% in financials, with 94.3% in North American issuers. ARK's daily holdings file dated 30 October 2026 lists 43 companies, two of which, OpenAI Series C (3.46%) and Kalshi (1.33%), carry no exchange ticker.

For scale, MSCI World had 1,249 constituents at 30 September 2026, its ten largest made up 27.85% of the index and Tesla was 1.12% (MSCI factsheet). ARKK held more than eight times the index's weight in Tesla.

What has Cathie Wood said about ARKK's expected returns?

Wood frames ARK's prospects over five-year spans, and has given specific numbers. On 24 November 2021, with ARKK down nearly 15% for the year, she told CNBC's Squawk Box: "we now see a quadrupling over the next five years in our portfolio" (CNBC). Quadrupling in five years implies about 32% a year. From the end of November 2021 to 30 September 2026, ARKK's total return at net asset value was -15.0%, by our calculation from ARK's daily NAV with distributions reinvested; MSCI World gained 69.7% over the same months.

On 7 March 2022, Wood told CNBC she expected "spectacular returns" over five years and that "a lot of our investor base is averaging down". She added: "You'd be amazed if you average down over time, how quickly a strategy can come back above that average." Averaging down lowers an investor's average purchase price, but it adds money to the same concentrated exposure and does not change how the fund compounds.

Is ARKK set up for different market scenarios?

No: over its life ARKK amplified MSCI World's moves in both directions, with one exception in the Covid crash.

ARKK against MSCI World, in USD: Covid crash (February to March 2020) -15.1% and -20.6%; inflation and rate shock (2022) -67.0% and -18.1%; rally (January 2023 to September 2026) +185.2% and +98.8%. Source: our calculation from monthly total returns (data (CSV)); chart data (CSV).

In the 18 months when MSCI World rose 5% or more, ARKK gained 13.9% on average against 7.6%; in the 12 months when it fell 5% or more, ARKK lost 12.7% against 7.7%. It rose in only one of those 12 months: February 2020, when it gained 2.0%. That month explains the Covid result; in March 2020 ARKK fell 16.7% against 13.2%.

Over November 2014 to September 2026, ARKK's monthly returns had a correlation of 0.71 with MSCI World and a beta of 1.77, so on average it moved about 1.8% for each 1% move in the index.

ARK's fact sheet offers ARKK as "a tool for diversification due to little overlap with traditional indices." Low overlap in holdings is a label test; the return record shows a fund that amplified an equity portfolio's main driver rather than offsetting it.

Morningstar's Arnott wrote in 2021 that the portfolio "courts extreme levels of risk on almost every type of fundamental metric, including concentration, momentum, liquidity, valuation, and financial health." Its 15 themes across six sectors rest on one premise, which ARK's prospectus names as a principal risk: companies it believes are capitalising on disruptive innovation "may not in fact do so."

Our explainer on how correlation drives portfolio volatility shows how a highly correlated holding takes a larger share of a portfolio's swings than its capital weight. For contrasting approaches, compare how Stanley Druckenmiller concentrates his bets with Cliff Asness's case for diversifying across return sources.

Does ARKK guard against catastrophic equity falls at all?

No: ARKK holds almost nothing that would hold value or pay out in an equity crash. Its holdings file for 30 October 2026 lists 43 companies, a Treasury money market fund at 0.45%, ARK's own autocallable income ETF and a warrant, with no put options or short positions. There is no explicit protection and no asset with a different driver to cushion a fall, and the beta of 1.77 means the equity exposure is larger than the index's.

On month-end data, ARKK fell 77.1% from January 2021 to December 2022, a loss that needs a 336% gain to recover (the maths of recovering from large losses). At 30 September 2026 it was still about a third below its January 2021 level.

On ARK's daily NAV with distributions reinvested, our calculation puts the peak on 12 February 2021, the trough on 28 December 2022 and the fall at 80.9%; the fund was 42.6% below that peak on 30 September 2026. MSCI World's worst month-end fall in the same window was 25.4%, from December 2021 to September 2022, and it had recovered by December 2023. ARKK's smaller loss in the Covid crash came from its holdings rising in February 2020, not from any protection.

Are ARKK's volatility-adjusted returns better than MSCI World?

No: in every window we measured, MSCI World earned more return per unit of volatility. The Sharpe ratio divides the average monthly return above cash by the volatility of that excess return, annualised. The Sortino ratio divides the same excess return by downside deviation, so only months that fell short of cash enter the denominator.

Over the full window ARKK's annualised volatility was 36.1% against 14.5%, and its downside deviation 21.7% against 9.4%; the 24 rolling windows start between November 2014 and October 2016 and overlap heavily, so they are not independent tests.

ARKK's monthly excess returns were positively skewed, with six months above +20% against two below -20%, so a ratio that ignores upside swings should flatter it most. It did, lifting ARKK's ratio by two-thirds against about half for the index, yet ARKK still trailed on Sortino in every window.

At MSCI World's 14.5% volatility, a monthly-rebalanced mix of 40% ARKK and 60% Treasury bills compounded at 8.3% a year with a worst drawdown of 41.8%, against 11.1% and 25.4% for the index.

These ratios cover one fund over one period, monthly data understates intra-month swings, and the cash rate used changes the Sharpe ratio. Choosing a famous fund after the fact is itself a selection effect.

Are ARKK's returns net of fees better than MSCI World?

Yes: after its 0.75% fee, ARKK compounded faster than MSCI World over the trailing 10 years, the full window and 23 of 24 rolling 10-year windows.

Fund returns are net of fees; MSCI World Net is an index and carries no costs, while a fund that tracks it does (the iShares Core MSCI World UCITS ETF charges 0.20% a year), so a shortfall smaller than such a fee counts as close rather than a clear loss. ARKK's lead of about 3 points a year is far larger than that.

$10,000 invested in November 2014 grew to $48,070 by September 2026, against $35,056 in the index. ARK's fact sheet reports 14.26% a year at NAV since inception and 15.78% over 10 years to 30 September 2026, against 11.19% and 12.81% for MSCI World. Across the rolling windows the gap ranged from -0.3 points (November 2014 to October 2024) to +5.5 points a year, so the lead is large but depends on dates.

The lead came early. From November 2014 to December 2020, ARKK compounded at 36.0% a year against 9.7% for MSCI World; from January 2021 to September 2026 it lost 5.5% a year while the index gained 12.6% a year. Its arithmetic mean return was 19.6% against a compound rate of 14.1%, a 5.5-point gap from volatility drag, against 0.5 points for MSCI World (why Sharpe ratios and CAGR rank portfolios differently).

Did ARKK investors earn the fund's return?

Morningstar's estimates suggest most did not, because the money arrived after the gains. A fund's reported return is time-weighted, assuming one investment held throughout; an investor return is dollar-weighted, reflecting when money actually went in and out.

In December 2021, Morningstar's Amy Arnott estimated that ARKK returned 41.3% a year over the five years to 30 November 2021, while investors earned about 9.9% a year, "less than a fourth of that return" (Morningstar).

She found that the fund averaged about $116m of assets in 2017, when it gained 87.4%, and that 2020 and 2021 accounted for about 90% of cumulative net inflows since launch.

On 21 July 2026, Arnott and Jeffrey Ptak ranked ARKK fourth among US funds by estimated shareholder losses over the 10 years to 2025, at $5.0bn, with an estimated $14.1bn of inflows in 2020 and 2021 (Morningstar). They note that the fund "earned a positive total return over the 10-year period" and that many shareholders were still underwater "because they bought in after performance had already peaked." Morningstar calls its ETF investor-return figures estimates, because secondary-market trading complicates flow data.

Weighing the four results

ARKK gave investors who held from launch to September 2026 a higher compound return than MSCI World after fees. In exchange it took a far deeper fall with nothing to guard against it, earned less return per unit of volatility, and amplified rather than balanced the market's swings.

The higher return also depended on holding through a 77% fall that was unrecovered in September 2026, which Morningstar's estimates suggest most investor money did not do. How to weigh those results depends on the reader's horizon, spending needs and capacity for loss.

What can an individual investor learn from Cathie Wood?

ARK states its hurdle rate, horizon and themes openly; the record adds the costs of concentration when themes fall together and of arriving late. Questions to examine in your own portfolio:

• If your largest theme or fund fell 77%, what would your whole portfolio lose, and would you need to sell during the fall?
• Does a fund you hold add a different return driver, or a higher-beta version of the equities you already own?
• How did your holdings behave together in months when global shares fell 5% or more?
• Did you add money after a run of exceptional returns, and what is your own dollar-weighted return?

How can copio help you examine your own allocation?

Copio is in private beta across Europe, with early access by application. It looks through funds to the holdings and asset classes underneath, so a thematic ETF's overlap with your other equities becomes visible. Its concentration analysis models a fall in your largest position and compares the effect on your portfolio with a broad fund.

A return-correlation map places your asset classes as main driver, true diversifiers, dead weight or shock absorbers, alongside a correlation matrix. A crash test shows how a mix like yours would have fared in past crises, and an efficient frontier plots your portfolio against a modelled frontier of assumed 10-year growth and volatility, labelled as not a forecast.

Copio provides educational analysis, not personalised advice; it does not find "the optimal allocation" or place trades, and shows where your portfolio sits and what drives its returns and losses. You can start with the free portfolio checker.