ARK Next Generation Internet ETF
ARK Invest — Cathie Wood
Active fund in cloud computing, AI, e-commerce, big data, mobile, social platforms and digital assets; ~45 holdings. Holds the ARK Bitcoin ETF as a position for crypto exposure. Expense ratio 0.88%.
Top holdings
AI-compute exposure, recently the top weight
AI/autonomy + robotaxi thesis
In-house spot-Bitcoin ETF used for crypto exposure inside the fund
Digital-brokerage / crypto-trading platform
Connected-TV / streaming-internet bet
Performance
July 2026: −5.54% (stockscan monthly table, Jul 31 2026); June 2026 was −5.31% — July 2026 was an AI-hardware rout (SOX −20.6%, SMH −17.6%) but ARKW fell only −5.54% — its AI-compute sleeve is AMD/accelerators, not the memory/WFE/connectivity epicentre. Against the S&P's −0.1%, though, ARKW's July was a heavy idiosyncratic loss. ARK bought the dip on Jul 28: NVDA initiated across all five ETFs (ARKW 11,984 sh, now 1.68%), plus SpaceX and Tesla adds.
Recent moves
Pushed AMD to the top weight as the AI-compute thesis took over from social/e-commerce names; maintains a ~5% allocation to its own ARK Bitcoin ETF for digital-asset exposure.
Our take
Effectively ARKK's software/internet sleeve concentrated — now tilted hard toward AI-compute (AMD) and crypto (ARKB, HOOD). A reasonable 'next-gen internet + AI + digital assets' wrapper, but overlaps heavily with ARKK and carries the highest fee in the suite.
Three things fund marketing won't foreground.
First, the return record is a path story: +35.9% annualized over 3 years sits on top of −0.4% annualized over 5 years (Jun 30 2026). The 3-year figure starts after the 2022 collapse — a recovery slope, not compounding. Full-cycle holders lost money while paying 0.76% a year. Second, the portfolio has quietly become a crypto-beta vehicle wearing an internet label: the ARK Bitcoin holdco (4.68%) + HOOD (5.88%) + COIN (3.56%) + Circle (3.13%) + Block (3.43%) is 20.7% tied to digital-asset prices — though HOOD earns a majority of revenue from non-crypto brokerage, so the pure-factor share sits below that headline. Third, ~5.7% is in two private, unlisted positions — SpaceX (3.24%) and OpenAI Group PBC Series C (2.50%) — carried at ARK's own marks rather than a market price: a real illiquidity and NAV-integrity wrinkle inside a daily-redeemable wrapper.
Thesis
ARKW sells conviction: ~45 names, 49.3% in the top 10 (Jul 17 2026), expressing Cathie Wood's view that cloud, AI-compute and digital assets compound faster than the market prices. It is a single-manager, high-beta (1.96) bet on disruptive-innovation forecasting — not a diversified internet index. The thesis lives or dies on whether ARK's research edge survives contact with a market where AI compute is now the consensus trade, not a contrarian one.
Actively managed, high-turnover concentration in 'next-generation internet': cloud, AI infrastructure, fintech, digital assets, e-commerce. Positions are sized top-down from ARK's 5-year price targets, so winners are trimmed and losers averaged into — the opposite of momentum. Crypto exposure runs through equities plus a ~4.7% position in an ARK holdco that owns its own spot Bitcoin ETF. Holdings are published daily.
- 2026-07July's loss was idiosyncratic rather than market beta: the S&P 500 fell -0.1% while ARKW fell -5.5%, a move a 1.96 beta would put near -0.2%. AAII prints beta at 2.34 (30 Jun 2026).
Assessment
- Genuine active differentiation: 45 names, 49.3% top-10, nothing resembling a benchmark hug — you are buying a distinct view, not closet indexing.
- Daily full holdings disclosure and public research make the thesis auditable in a way most active funds are not.
- Access to two pre-IPO positions (SpaceX 3.24%, OpenAI Series C 2.50%) that retail cannot easily buy directly.
- Fee is 0.76% net (Jul 20 2026) — moderate for an actively managed, concentrated mandate that also sources private positions.
- ~20.7% of the portfolio (Bitcoin holdco, HOOD, COIN, CRCL, XYZ) moves largely with one crypto-price factor — though HOOD's non-crypto brokerage revenue softens that attribution.
- Heavy overlap with sibling ARKK — owning both concentrates the same manager, same theses, twice, at two fees.
- ~5.7% in private, ARK-marked holdings (SpaceX + OpenAI) inside a daily-liquidity ETF: valuation is an estimate, and neither can be sold to meet redemptions.
- Buy-the-dip sizing off 5-year price targets means the fund adds to falling positions — devastating when a thesis is simply wrong rather than early.
- AI compute (AMD 7.6%) is now a crowded consensus trade; the contrarian premium ARK's process depends on has largely been arbitraged away.
- 2026-07Holdings at 31 Jul 2026: AMD 7.48%, TSLA 7.32%, with SHOP 4.71% now fourth and ROKU outside the top 15. ARK initiated NVDA across all five ETFs on 28 Jul — 11,984 shares here, 1.68% of the fund.
Record
The record decomposes cleanly by regime. 5-year annualized −0.37% against 3-year annualized +35.90% (both Jun 30 2026): essentially all of the 'good' number is the rebound off the 2022 trough, and none of it is durable compounding — five-year holders are below water before fees. The trailing year is highly date-sensitive: −0.35% to Jun 30 vs −9.39% to Jul 20 2026, a three-week swing that is itself a 1.96-beta demonstration. Since-inception ~20.6% annualized is real but was earned overwhelmingly in 2017 and 2020, when the fund was small — Morningstar's work on the ARK complex shows dollar-weighted investor returns running far below reported time-weighted returns, because assets arrived after the gains. Max drawdown −75.4%, 53 months to recover. Recent attribution mixes AI-compute (AMD) and the crypto sleeve; the trailing drawdown is that beta in reverse.
- 2026-07Two consecutive down months, not a three-week window: June -5.31% and July -5.54%, about -10.6% cumulative, leaving the trailing year at -11.8% and 2026 at -7.4%.
Risks & fit
- Manager/key-person risk is undiversifiable — the process is one person's forecast framework, not a repeatable rules set.
- Beta 1.96 with 45 names: max drawdown of −75.4% took 53 months to recover, and a repeat is structurally available.
- Crypto drawdown hits ~20% of the book near-simultaneously; these five names are not independent bets.
- Private SpaceX + OpenAI marks (~5.7%) create NAV-staleness and redemption-pressure risk if flows reverse hard.
- Redemption spiral: AUM at $1.69B is well off peak, and forced selling in illiquid mid-caps amplifies losses.
The research-edge thesis fails if, measured from a start date that does not cherry-pick the 2022 trough, ARKW trails a passive internet/software benchmark net of the 0.76% fee over a full cycle, and attribution shows the wins came from broad AI/crypto beta rather than security selection. Concretely: if the next peak-to-peak cycle again leaves 5-year annualized returns at or below zero while a plain internet index compounds, the active premium is not there. ARK is vindicated instead if the fund out-earns those benchmarks through a drawdown, not just off one.
Suits an investor who explicitly wants concentrated, single-manager exposure to AI-compute plus digital assets, understands roughly a fifth of the portfolio tracks crypto prices and ~5.7% is privately marked, can hold through 70%+ drawdowns, and is sizing this as a small satellite rather than a core holding. Poorly matched to anyone reading 'next generation internet' as a diversified tech allocation, who already owns ARKK, or who needs every line item market-priced.
0.76% net (stockanalysis.com, Jul 20 2026) — roughly 19x a broad-market index fund, charged on a portfolio whose largest public holdings (TSLA, AMD, AMZN, GOOG) are freely investable at near-zero cost.