Two Baskets, Not One Big Trade: What Actually Happened in the iShares Book on August 27

Here's the headline number, and here's why you should immediately distrust it: net position flow across 365 iShares funds and 4,701 names came in at -$96.9M on $2.72B of gross two-way activity. That looks like a quiet, forgettable day. It is not quiet. It's just that two enormous, offsetting baskets did almost all the work, and if you only read the top-line lists — NVDA getting hammered, software getting bought — you'd walk away with a completely wrong story about what BlackRock's funds actually did.

Start with the arithmetic. IVV shed $1.00B, which is 36.9% of the day's entire gross flow. IGV built $655.0M, which is 24.1%. Add them up and two funds account for roughly 61% of everything that moved in a 365-fund complex. Everything else — the top-in list, the top-out list, most of the sector table — is downstream of those two baskets.

The redemption that looks like a sell call, and isn't

IVV's -$1.00B is the S&P 500 tracker shedding shares pro-rata across its book, and the top-out list reads exactly like the S&P's largest weights: NVDA -$114.6M, AAPL -$64.8M, AVGO -$45.1M, GOOGL -$38.6M, GOOG -$33.9M, AMZN -$31.2M, XOM -$23.3M, JNJ -$21.6M. That's not a house view on chips or Apple. It's an authorized-participant redemption basket, and the breadth numbers prove it: NVDA had 14 funds adding against 11 trimming, AAPL had 16 adding against 9 trimming. Most funds were net buyers of these names. One very large redemption basket simply swamped everyone else in the dollar tally. Against IVV's $889.8B in AUM, this $1.00B is about 0.11% — routine plumbing, not a signal.

The mirror image is IGV, the software-sector ETF, building $655.0M — about 4.4% of its $14.88B AUM — and that creation basket is essentially the entire top-in list: PLTR +$63.5M, PANW +$58.3M, CRWD +$43.7M, CRM +$39.7M, ORCL +$32.2M, TWLO +$31.4M, ADBE +$31.3M, FTNT +$28.7M. Sum those up and you get most of IGV's build. Which means the sector table's flashy Information Technology +$222.7M print is mostly one ETF's creation basket net of the semiconductor drag from SOXX (-$157.5M, feeding NVDA, AVGO, MU -$33.4M, AMD -$29.9M, AMAT -$14.0M). Call it "software over semis" if you like the narrative, but it's two ETFs' ledgers, not a manager rotating the portfolio.

The best joke in the data: NVDA

If you only looked at market value, you'd think the desk went all-in on Nvidia. NVDA's market value rose +$11.50B in this file — 78% of the entire complex's +$14.68B equity value change — while its actual share-count flow was -$114.6M. Position count went down slightly; the price did everything else. That single divergence is the whole lesson of this dataset: flow and market value tell different stories, and reading mvDelta alone would have you conclude exactly the opposite of what happened.

The stuff that's genuinely odd — and the stuff that's genuinely fake

Two names deserve a flag for very different reasons. EWN, the Netherlands ETF, shed $95.3M against just $728M of AUM — roughly 13% of the fund in one file, the single most anomalous line in the whole dataset. No sourced catalyst, but it explains ASML's -$29.8M move (only 8 funds hold it, 5 of them selling) without needing any ASML-specific story at all.

Predictive Discovery DEF (PDIDB AU) +$29.3M is not a materials call — it's plumbing. Six funds, all six brand new to the position, and the flow figure matches the market-value change to the dollar. That's the signature of a deferred-settlement corporate-action line, not conviction.

SomniGroup International +$52.2M landed in 25 funds with 8 of them opening the position from scratch and 19 buying versus 2 selling. Eight simultaneous new holdings across a passive complex is an index add or an identifier change, full stop — not the discretionary story floated elsewhere.

Global Unichip +$64.3M, the single largest dollar add of the day, sits in exactly two funds, both buying. That's too narrow to be an index-wide event, whatever the MSCI-review theory suggests.

And BAI, the actively managed AI/tech fund, +$124.8M is the one line in this entire file that's honestly a portfolio manager's decision rather than a basket mechanic — worth noting precisely because it's rare.

What to watch next

Whether the software creation trend in IGV persists past a single file, whether EWN's redemption repeats or was one-off, and whether Global Unichip picks up more than two funds' worth of buyers — that last one would confirm or kill the index-inclusion theory. Also worth remembering: the prior two publication days were nearly empty (5 names each), so this file likely compresses more than one session's worth of real change. Treat every "one-day" number here with that discount in mind.