ARK Innovation ETF
ARK Invest — Cathie Wood
Flagship actively-managed disruptive-innovation fund — high-conviction, concentrated bets across genomics, AI, fintech, robotics, energy storage and blockchain; ~38 holdings, top 10 ~49% of assets. Expense ratio 0.75%.
Top holdings
Largest position and Wood's signature conviction bet (autonomy/robotaxi thesis)
Crypto-exchange proxy for the blockchain thesis
Consumer/metaverse platform holding
Connected-TV streaming bet
Gene-editing / genomics conviction name
AI-driven precision-medicine name; a recent high-conviction add
Performance
July 2026: ~-11.9% (Jun 30 close $80.82 -> Jul 31 close $71.24); price-derived, ARKK distributes annually in Dec — July was ARKK's entire 2026 loss: entered the month ~+5.1% YTD, exited -7.38% — in a month the S&P 500 was flat (-0.1%) and the AI-hardware complex was what crashed. Drivers: TSLA -26.0% (~-2.6pp at ~10% weight) and the crypto sleeve (COIN -8.03% on Jul 31). 3Y annualized reset to 12.50%, down from ~22% before the July drawdown.
Recent moves
Continued rotating into AI-adjacent healthcare (Tempus AI) and trimming legacy losers; Tesla, Coinbase and Roblox remain the anchor positions through 2026.
Our take
The original iconic 'disruption' vehicle — still the purest expression of Cathie Wood's concentrated, high-beta conviction style, but its multi-theme sprawl makes it a volatility amplifier rather than a clean AI play. Best read as a sentiment gauge for speculative growth, not a core holding.
The critique isn't that Wood is wrong about the future — it's that the fund's return distribution and its ownership distribution are misaligned by design.
A concentrated high-beta book earns in violent, narrow windows; assets arrive after them. Morningstar's investor-return work quantified it on windows ending around 2023: trailing-3y dollar-weighted returns near -28%, more than double the fund's reported loss, and a 5y shortfall above 25 percentage points. Same fund, opposite experience — and no leverage is needed to produce it. After the 2021-22 collapse, regaining the prior peak requires a multi-hundred-percent advance, which is how a positive since-inception CAGR coexists with a shareholder base still underwater. Concentration turns volatility into a permanent tax on anyone who didn't own it from the start. Finally, the six-theme sprawl is a cost, not diversification — genomics, crypto and space share one long-duration rate factor, so the book behaves as a single bet wearing six labels.
Thesis
ARKK's marketing sells conviction — 48 names, top-10 at 49%, Tesla alone 10.32% (Jul 20, 2026). Its record sells something else: the strategy worked while the shareholders didn't. Since-inception return is 12.70% annualized from the Oct 31, 2014 launch, yet assets fell from a ~$25.5B peak (Morningstar, June 2021) to $5.96B (Jul 20, 2026) because most of that return was earned before the money arrived. The honest unit of analysis for ARKK is that gap, not the headline CAGR.
Actively managed, non-diversified 'disruptive innovation' mandate across genomics, AI, fintech, crypto, robotics and space — themes picked by top-down 5-year forecasts, expressed as high-conviction, high-turnover single-stock bets. ARK publishes every trade daily, adds to losers on drawdowns, and funds new conviction by trimming winners. Concentration is deliberate: roughly half the fund sits in ten names. Expense ratio 0.75% (Jul 2026).
Assessment
- Differentiated at the edges — TEM 5.42%, CRSP 4.57%, TWST 3.36% sit outside mainstream growth indexes, though core names (SHOP 4.85%, PLTR 3.17%, AMZN, GOOG) do not.
- Radical transparency: daily disclosure of every trade, rare among active managers and verifiable against the record.
- Holds assets index-bound rivals structurally cannot own — a private OpenAI Group PBC Series C stake at 2.95% plus SPCX at 4.43%.
- Long, unbroken live track record under one decision-maker since Oct 2014 — a genuinely testable process, not a backtest.
- The dollar-weighted vs time-weighted gap is the fund's defining feature, and it is absent from every marketing surface.
- 0.75% fee against a since-inception edge Morningstar measured at roughly 1.6pp/yr over mid-cap growth peers — the fee consumes most of the claimed alpha.
- Top-10 at 49% with Tesla at 10.32% makes this substantially a leveraged expression of one stock's narrative.
- Six themes in one wrapper means investors cannot size or hedge the exposure they actually want.
- AUM down from a ~$25.5B peak (Morningstar, June 2021) to $5.96B; a shrinking base raises the odds of forced selling into weakness.
Record
Attribution splits cleanly by era. The 12.70% since-inception annualized figure (Jul 20, 2026) was overwhelmingly earned 2014-2020, when the fund was small and long-duration growth re-rated; the 2021-22 unwind and the years since added little. 2026 is the sharpest illustration: -0.33% YTD and -3.45% over the trailing year as of Jul 20, 2026, while the AI-compute complex led the market. The reason is compositional, not tactical. ARKK is not absent from AI — OpenAI Series C 2.95%, PLTR 3.17%, AMD 3.97%, CRWV 2.03% — but it is only lightly exposed to the compute-and-memory trade that drove the tape, and its largest position, Tesla at 10.32%, is an AI story the market did not pay for this year. The crypto-linked sleeve (HOOD 4.46%, COIN 4.23%, CRCL 3.39%, ~12% combined) supplied volatility rather than offset.
Risks & fit
- Single-name Tesla risk at 10.32% — one stock drives a disproportionate share of tracking outcome (Jul 20, 2026).
- ~12% aggregate crypto-linked exposure (COIN, HOOD, CRCL) adds a second, correlated high-beta factor.
- Long-duration unprofitable-growth book is acutely rate-sensitive; a higher-for-longer path compresses the whole portfolio at once.
- Key-person concentration — process, forecasts and conviction all trace to one manager.
- SPCX at 4.43% is a recent listing with limited trading history; the OpenAI Series C stake at 2.95% is a private, model-priced mark.
The critique rests on ARKK's structure systematically transferring return away from its own shareholders. It falsifies if Morningstar's dollar-weighted investor return converges on the fund's time-weighted return over a full cycle — post-2023 buyer cohorts earning roughly what the fund reports — while ARKK also beats a plain mid-cap growth index net of the 0.75% fee across that cycle. A closed investor-return gap plus fee-adjusted alpha would show the concentrated, buy-the-drawdown process compounds for real owners. One strong year settles nothing; the gap is cycle-length.
Suits an investor wanting deliberate, disclosed, non-index exposure to long-duration innovation themes, able to tolerate 70%+ drawdowns without changing position size, treating it as a small permanent sleeve rather than something to add to after strength. Sits poorly with anyone seeking AI-compute infrastructure exposure (2026's driver is only lightly represented here), anyone who sizes by recent performance, or anyone needing it to behave independently of Tesla and crypto.
0.75% expense ratio (Jul 2026) — roughly 25x a broad index fund, above passive thematic peers at 0.35-0.68%, with high turnover adding unreported trading costs. Morningstar measured the long-run edge over mid-cap growth peers at ~1.6pp/yr.