Cathie Wood Bought Meta as It Fell 8% — ARK's 30 July Trades in Full

On 30 July 2026, Meta closed down 7.95% and L3Harris closed down 8.61%. They were the two worst performers among everything ARK Invest traded that day — and they were the two things Cathie Wood’s funds bought most of.

Of the fifteen names ARK sold the same day, eleven finished higher.

What did ARK buy on 30 July?

Five names, two of them across multiple funds.

Meta is the one with a story attached. It reported second-quarter results the previous evening in which free cash flow fell to $784 million from $8.5 billion , and the stock spent Thursday down between 7% and 9%. ARK added to it in three funds while that was happening.

The size is modest in absolute terms — 26,509 shares, roughly $14 million using ARK’s own percentages against the funds’ values. What makes it notable is that it lifted an existing position by about a tenth in a single session.

Both the before and after figures are ARK’s own daily holdings files — the one dated 30 July, which is priced at the 29 July close, and the one dated 31 July, priced at the 30 July close. ARKF is absent from that table because ARK’s published ARKF holdings file was still dated 2 January 2026 when we checked, so there is no reliable position on either side of the trade. The purchase itself is disclosed and stands.

Notice that the positions did not move by the full amount bought. ARKK disclosed buying 16,106 shares and its position rose 15,750; ARKW disclosed 7,683 and rose 7,369. That gap is the thing ARK warns about — more on it below.

L3Harris is the larger commitment by weight. The ARKX purchase alone was 0.5942% of that fund, the biggest single trade ARK made all day in percentage terms. In ARKQ, a disclosed purchase of 20,610 shares took the position from 203,241 to 223,224 — an actual increase of 19,983.

What did ARK sell?

Fifteen names across twenty trades, and eleven of them closed higher.

Caterpillar rose 3.38% and was trimmed. Teledyne rose 2.82% and was trimmed twice, in two funds. Iridium rose 7.39% and was trimmed. Bitmine rose 8.98% and was trimmed. Selling into strength is not unusual on its own; doing it on the same day you add to the day’s two worst performers is a decision, not a coincidence of timing.

The part worth noticing: the buys and sells are nearly the same size

Look at the percentages rather than the share counts, and four of the five funds nearly balance. What that means is a separate question, taken up below.

ARKW sold 0.2552% of the fund in Datadog and bought 0.2554% in Meta. ARKF sold 0.2003% in Roku and bought 0.2010% in Meta. Those are not approximate matches — they agree to within two and seven ten-thousandths of a percentage point. One position out, another of the same size in, on the same day.

The industrial funds did the same thing with more pieces. To buy L3Harris, ARKX sold four holdings in near-identical slices: HEICO 0.0995%, Elbit 0.0989%, Teledyne 0.0988%, Deere 0.0986%. ARKQ did it with three: Caterpillar 0.0511%, Elbit 0.0511%, Teledyne 0.0510%. Equal slices out of several positions is what funding a purchase would look like when no single holding is big enough to sell — though the file records sizes and dates, not purposes.

Only ARKK disclosed more buying than selling — 0.1571% in, 0.0177% out. Everywhere else the two sides nearly cancel.

It is worth being careful about what that does and does not show. The file records portfolio adjustments; it does not show cash balances or fund flows, and ARK explicitly excludes creation and redemption activity from it. So ARKK’s positive net says its disclosed purchases exceeded its disclosed sales. It does not establish that new money was put to work, or that the Datadog sale was executed in order to buy Meta. The sizes match to within a rounding error and the trades are same-day, which makes the switch reading a natural one — but it is a reading.

Is this the same as buying the dip?

Directionally yes, with one thing the data cannot tell you.

The two largest buys were the two largest fallers, and most of the sells were in names that rose. That is the shape of a dip purchase. But ARK publishes trades after the close without execution prices or times, so there is no way to know whether it bought Meta at $525 or $538, early or late. “Bought as it fell” describes the session, not the fill.

The buys were not uniformly into weakness either. X-Energy closed up 11.66% and ARK bought it in three funds; Kodiak AI and Pony AI both closed higher and were added in ARKQ. Whatever rule is operating, it is not simply “buy whatever is down.”

What ARK says about these files

Worth reading before drawing conclusions from any single day.

ARK’s own wording on the trade file is that it “only reflects portfolio adjustments made by the ARK investment team”, that the files “are not comprehensive lists of a day’s trades” and “exclude initial/secondary public offering transactions and ETF Creation/Redemption Unit activity”, that “additional files may be posted at a later time”, and that they represent “an UNOFFICIAL, UNRECONCILED account” pending BNY Mellon’s accounting.

ARK also states plainly that its trade notifications “are not an endorsement of any company or a recommendation to buy, sell or hold any security”.

That last exclusion is not a formality, and 30 July gives a clean measurement of it. Comparing ARK’s holdings file for 30 July with the one for 31 July shows what each position actually did, against what the trade file said was bought:

ARKQ’s other two purchases show it more starkly still. Pony AI was disclosed as a 5,000-share buy and the position fell 3,879; Kodiak AI was disclosed as 10,641 and the position fell 1,532. Three of ARKQ’s four disclosed purchases ended the day as smaller positions than they started.

Every gap runs the same way, and that is the shape money leaving a fund makes: redemptions shrink every holding a little, and the trade file does not show them. ARK confirms these files exclude creation and redemption activity — it does not publish the day’s flows, so this is a pattern consistent with redemptions rather than a proven cause. Whatever produced it, the direction is uniform and it is larger than the trades in three cases.

So the trade file answers “what did ARK choose to do” and not “what does ARK now own”. Both are published; they are different questions.

One further consequence: a single day is a single day. ARK trades most sessions, and nothing here establishes a thesis change.

Where this sits in ARK’s recent record

Two earlier pieces give the run-up. In early July we covered Cathie Wood’s AI stock picks during the June selloff — Tesla, Palantir, CoreWeave, and the switch from AMD into Nvidia. In June, ARK sold $163 million of stock, including Roku and AMD, to fund its SpaceX position . Roku appears again in this file, sold to buy Meta.

The Meta purchase reads differently alongside what Meta actually reported : operating cash flow up 25%, capital expenditure up 83%, free cash flow down 91% to $784 million, and $278.99 billion of leases not yet commenced sitting in the 10-Q. ARK bought the quarter the market sold.

Sources

Each link was checked on 31 July 2026.

Like the trade file, all five holdings URLs serve the current file and roll forward daily. The 30 July and 31 July snapshots this article compares were downloaded on those dates and cannot be retrieved from these links later.

Closing prices are from a consumer market-data feed (Yahoo Finance via the yfinance library), for the completed 30 July 2026 session.