📊 Ainvest Option Flow Digest — 2026-07-31

Fourteen blocks, ≈$339M net, and not a single sweep — including a ≈$32M crash hedge on the S&P 500 that the feed read backwards

✅ UPDATED August 3, 2026, pre-market — the next-day OI check is in. All 28 checkable legs resolved: 24 opened, 4 closed. Three published reads inverted, each turning out to be a roll rather than the structure we described: MCHP (diagonal → bullish roll, upside not capped), SBUX (long call ladder → roll down; the $110 leg closed a short), and LLY (Sep call BTO → BTC, which strengthens the hedge-restructure thesis). The glance table below and the closing section carry the corrections; every affected article has been updated.

---

🎯 Quick Read

If you remember one thing today: every trade on this board was negotiated — a cross, an auction, or a floor block. There was not one lit sweep. Nobody chased anything. Fourteen desks quietly restructured, financed, or hedged.

• ≈$339M net across 14 names. But ≈$131M of that carries zero directional content — it is delta-one financing, not opinion. Strip it out and the genuinely position-taking flow is ≈$208M.
• ⭐ The most important trade of the day is SPY, and the feed had it backwards. It showed four put sales worth $114M. The tape shows six legs forming two put butterflies — the trader is paying ≈$31.8M, not collecting. In return they get up to ≈$3.0B if the S&P 500 lands near 535 by September 30 or 525 by October 16. That is crash insurance, bought.
• The single biggest ticket is also the emptiest. AAPL printed a "$124M call sale" on the day Apple fell ≈9.5%. It is a conversion: short call, long put at the same strike, plus 3,000,000 shares bought in the same second. The synthetic short priced at $301.21; the stock was bought at $301.22. A one-cent spread on ≈$904M. There is no view in it.
• GXO is the same trick in miniature — a deep-ITM call sold with $0.17 of time value against 700,000 shares in the same second. Also not bearish. Also no signal.
• QQQ looked like a $17M LEAP call buy. The tape shows a second leg — an $11.4M put sale that funds most of it. Real money committed: $5.55M.
• Semis got hedged from both sides. SMH saw two separate defensive trades three hours apart: a $51.7M in-the-money put bought for June 2027, and a $19.0M June-2028 call sold — both on strikes that barely existed before today. Worth noting: they went on into a two-day bounce, not a collapse. Chips V-bottomed on July 29–30.
• Two big rolls, not two new bets. AMZN cashed deep-ITM September calls on Amazon's best day in 11 years and redeployed into 2.5× the contracts, $20 higher. SNDK — up ≈353% year to date — rolled an expiring deep-ITM call out one week, moving it past SanDisk's confirmed August 5 earnings.
• Exactly two trades all day were unhedged and outright — and both were small. CCJ: a $1.69M put on a strike whose open interest was zero. And TIGO: a $1.75M call buy where the market was $2.10 bid / $3.50 offered and they paid the full offer — the closest thing to urgency on the whole board.

Before you copy any of this: eight of fourteen packages printed alongside a stock block that cancelled their delta, and four showed up in the feed with the wrong number of legs (SPY, QQQ, HNI and SNDK). Read the structure, not the headline.

---

🔁 OI Review — Last Session's Provisional Flags, Now Resolved

The July 31 pre-market open-interest snapshot is in, and Thursday's tape was the cleanest run in weeks — with one catch nobody would have found without it. Ten tickers carried ⏳ provisional flags across 16 legs. All 16 resolved: 14 opened, 2 closed, and not a single directional read inverted. But the check did something it usually doesn't — it caught an error of size, not direction: an exchange cancellation shrank SMH's call ticket by 25%, taking Thursday's gross from ≈$298M to ≈$293.6M.

🔄 The Correction — SMH was ≈$44M, not $49M

⚪ SMH — a cancelled clip took 2,000 contracts and ≈$4.4M off the headline.

One of the two 5,000-lot call clips was cancelled by the exchange and re-reported at 2,000 roughly two minutes later. The genuine call sale is 8,000 contracts for ≈$17.6M, not 10,000 for $22M — taking that program from $49M to ≈$44M. The tell: open interest built +8,139, ≈102% of the corrected 8,000 but only ≈81% of the originally-published 10,000. Nothing in the intraday pipeline flagged this — only the OI-versus-size mismatch did.

🔄 The Read That Changed — EWY was NOT a roll

🛡️ EWY — OI 4,553 → 11,768 (+7,215). We published a real asterisk: the multi-day OI trend had been slipping, and the print was a negotiated cross with no readable aggressor, so a roll of an existing Korea hedge was equally plausible. Open interest more than doubled — ≈96% of the print created new contracts. This is brand-new insurance, not recycled.

✅ Confirmations

• 🐂 MSFT — the flagship flag, now proven. OI 10,736 → 21,511 (+10,775) against a 9,953-lot print. We said plainly that flat-or-falling OI would flip the bullish story. It doubled, and the build exceeded our buyer's own print — other desks opened alongside.
• ⚪ LRCX — the $45M put was an EXIT, confirmed to within 0.8%. OI 16,783 → 9,344 (−7,439) against a 7,500-lot print. Nobody put on $45M of new bearish risk — somebody retired an old hedge.
• ⚪ AMGN — roll confirmed on both sides. August $310 fell −1,726; October $310 rose +2,194. Opposite moves on the same strike are the signature of a roll.
• ⚪ QQQ — all four legs opened. The Sep-2026 $710 call sale rose +10,968. ⚠️ One asterisk: the Jun-2027 $600 put built only ≈41% new contracts, so ≈59% was existing holders transferring.
• 🐂 FLY 143 → 5,762, ≈164% of the print · 🔵 AVTR 138 → 23,190, matched to 0.3% · 🛡️ CORZ 3,105 → 8,163, a full open · ⚪ EQIX near leg a genuine fresh buy (3,181 → 6,022), far leg matched exactly (1 → 2,501).

The standing lesson, with a new wrinkle: a big headline is provisional until next-day open interest confirms it — and it can be wrong in size as well as in direction. Open interest is the only record that cannot be spun.

---

📋 At a Glance

Net vs gross, and why it matters twice today. The ≈$339M headline is net — long premiums paid minus short premiums collected. On top of that, ≈$131M of it (AAPL + GXO) is delta-one financing with no directional content whatsoever. Genuine position-taking flow: ≈$208M.

---

🔍 The Three Worth Your Time

🦋 SPY — ≈$32M of crash insurance, and the feed showed it upside-down

Two floor-negotiated packages, 26 minutes apart, with the S&P 500 ETF near $746. The feed reported four legs, all sales, ≈$114M collected. The tape shows six legs — and a completely different trade.

September 30: buy 150,000 of the $635 put, sell 300,000 of the $535 put, buy 150,000 of the $435 put. October 16: buy 150,000 of the $625 put, sell 300,000 of the $525 put, buy 150,000 of the $425 put.

That 1 × 2 × 1 ratio is the signature of a butterfly. Net cost: $0.99 and $1.13 per share — ≈$31.8M in total, paid, not collected. Maximum value if the index settles right at a body strike: $100 per share, or ≈$1.5B per package.

Why we read it as bought rather than sold: every leg printed at the bid, but on a floor combo the exchange spreads one net price across the legs, so per-leg side flags mean little. The economics decide it. Selling this structure means risking ≈$1.485B to collect $14.85M — about 100:1 against you. Nobody does that. Paying ≈$32M for deep crash convexity is a standard pension and tail-fund hedge.

Now the honest part, because the payoff ratio is seductive. The body strikes sit at 535 and 525 — roughly 28–30% below spot. A butterfly pays best if the index lands near those strikes at expiry. A crash straight through 435/425 pays very little. No crash at all and the entire $31.8M is gone. This is insurance with a narrow strike zone and a hard deadline — not a forecast, and not something to copy.

All six legs are proven fresh opens: 150,000 and 300,000 contracts against prior open interest of just 425 to 5,778.

One more piece of context that matters: the S&P 500 itself barely flinched in July. While the Nasdaq-100 fell ≈10% and semiconductors ≈29% peak-to-trough, SPY's own drawdown was only ≈3.4%, and it sits under 2% below its high. This insurance was bought when the index was calm and the strikes were cheap — which is when insurance is supposed to be bought, and is the opposite of panic. Full breakdown →

🧊 AAPL — ≈$904M that means absolutely nothing

A desk sold 30,000 Apple $260 calls expiring today for $41.22 while the stock sat near $301.20. On its own that reads as a $124M bearish bet on a day Apple dropped ≈9.5%.

It isn't. Two other things printed in the same breath:

• They bought 30,000 of the same-strike $260 puts for one cent.
• 3,000,000 shares crossed at $301.22 in the same second.

Short call + long put at one strike = a synthetic short at 260 + (41.22 − 0.01) = $301.21. They bought the stock at $301.22. That is a one-cent spread on ≈$904 million of notional.

Walk it to this afternoon's expiry: the $260 call gets assigned, they hand over the 3,000,000 shares just bought, collect $260 a share, and keep the $41.22 — $301.22 total, exactly what they paid. Flat by construction. The penny put is insurance for the few hours in between.

This is a conversion — a financing and stock-loan trade. Desks do them for carry, never for a view. Learn the shape: deep-ITM short call + same-strike long put + a same-second stock block = no opinion whatsoever. Full breakdown →

🛡️ SMH — the one place somebody actually changed their mind

Two separate crosses, three hours apart, both leaning the same way on semiconductors:

• 10:38 — sold 2,000 June-2028 $730 calls for $19.0M. Prior open interest on that strike: 3 contracts.
• 13:45 — bought 3,900 June-2027 $625 puts for $51.7M. Prior open interest: 49. A 195,000-share block printed 1.5 seconds later, almost exactly matching the position's delta.

Both are proven fresh opens on strikes that barely existed. Together: buy downside, sell away upside, out to 2027 and 2028.

The context is a sector already in a drawdown — the SOX index down ≈24%, more than $1T of chip value erased, and the Nasdaq-100 entering a ≈10% correction on July 29. And note the shift: yesterday the largest SMH ticket was somebody selling volatility. Today somebody is paying up for protection — an in-the-money put with ≈$54 of time value in it is an expensive hedge, which says something about conviction on timing.

The honest limit: each trade was delta-hedged with stock, so the dealer is flat and the customer carries the negative delta. Whether that customer is protecting a semiconductor book they already own or pressing a bearish view, the tape cannot tell us. Full breakdown →

☢️ CCJ and 📞 TIGO — the only two unhedged bets on the board

15,085 September $74 puts at $1.12, bought at 09:49. Prior open interest: 0. The contract did not exist before this morning — the cleanest open/close read of the entire day.

It has one companion. TIGO (Millicom, the Latin American telecom, ≈$95) saw 5,000 August $100 calls bought for $1.75M against prior open interest of just 96 — and Millicom has confirmed its Q2 report for August 6, which lands inside that expiry — so this is squarely an earnings-window bet. The detail worth noting: the market was $2.10 bid, $3.50 offered — and they paid the full $3.50. On a $1.40-wide market that is real urgency, and it is the only urgency anywhere today. The catch is that the stock is already up ≈69% year to date, so this is pressing a winner rather than catching a falling knife. Its option chain is also thin enough that a retail trader would give up a painful amount just crossing the spread.

The timing: Cameco reported that same morning and missed badly (adjusted EPS $0.18 vs $0.38 expected) while raising revenue guidance. The stock popped to $92.02 pre-market, then round-tripped to ≈$87. This buyer stepped in during the fade.

The catch: the $74 strike sits below the entire published sell-side range — the lowest street target is $84.42 — and no earnings lands inside the contract's life (Cameco next reports October 30, 49 days after expiry). At $1.12 and ≈15% out of the money, most options like this expire worthless. Full breakdown →

---

📅 Upcoming Catalysts — and the Expiration Each One Actually Affects

⚠️ A catalyst only matters to a contract if it lands before that contract expires. Several of today's expirations miss their own company's next earnings entirely.

---

👥 How Four Different Traders Should Read This Board

🎲 The YOLO Trader

SPY's butterflies will call to you — a ≈100:1 payoff ratio is exactly the kind of number that ends badly. Read it carefully: that ratio is the payoff at one exact price on one exact day, and the body sits ≈28% below spot. The realistic outcome is a total loss of premium. Institutions can wear that as a rounding-error hedge line; a retail account cannot.

CCJ is the only clean directional ticket, and the cheapest at $1.12. Respect what that price means: ≈15% out of the money, six weeks, and no earnings catalyst inside it. The market prices a low probability because it is one. If you take it, size it as money you can lose entirely.

What you should not do: chase AAPL or SNDK because the dollar figures are enormous. AAPL's $124M is a financing trade with no direction, and SNDK's $70M "call buy" is one leg of a roll — the actual new money is $31M. If SanDisk's August 5 print tempts you, note the setup: forward P/E ≈6.8 versus ≈42 trailing is the classic peak-cycle-earnings signature, and the stock is already ≈46% below its June high.

📈 The Swing Trader

Two names deserve your attention for opposite reasons. MCHP has genuine tension: bookings have inflected (April was the largest booking month in nearly four years) and Texas Instruments corroborated the industrial recovery, yet the stock is ≈28% off its June high on sector beta. Earnings ≈Aug 6 resolves it.

SMH is the counterweight — someone just paid $51.7M for in-the-money downside protection on the whole sector out to June 2027. Holding a bullish chip view is fine; holding it without noticing that trade is not.

Watch AMZN's $240 level too. The desk that rolled there chose it deliberately.

💰 The Premium Collector

Today is a masterclass in what institutional premium harvesting actually pays. GXO captured $0.17 of time value on a $50.50 stock — ≈5.9% annualized — with zero directional exposure because it was hedged instantly. AAPL's conversion captured a one-cent spread on ≈$904M. That is the real world: tiny edges, enormous size, no risk taken.

If a retail "covered call" screen is showing you 40% annualized, you are being paid for risk, not for time. And note the trade you should not copy: QQQ's short $600 put and SMH's naked 2028 call both carry open-ended obligations. Cash-secured or not at all.

🌱 The Beginner

The single most useful idea on this page: the headline dollar number is usually wrong.

AAPL looked like a $124M bearish bet. It had no direction at all. QQQ looked like a $17M call buy — real money was $5.55M. AMZN looked like a $98M bet; it was a $49M roll of a position that already existed. SNDK looked like a $70M call purchase; it was one half of a roll costing $31M.

Three habits worth building from today:

1. Ask what the other legs were. Seven of ten trades here had more than one leg.
2. Ask whether stock traded with it. Seven of ten came with a stock block that cancelled the bet.
3. Ask whether the catalyst fits inside the expiration. CCJ's put expires 49 days before Cameco's next earnings. SBUX's January calls die before the holiday quarter is reported.

You do not need to trade any of this. Reading it correctly for a few weeks is worth more than one rushed position.

---

✅ RESOLVED — the August 3 OI Check Is In, and Three Reads Inverted

Updated August 3, 2026, pre-market. This section replaces the "provisional until tomorrow morning" note published July 31.

All 28 checkable legs across the 14 tickers resolved: 24 opened, 4 closed. Three of them inverted a published read — and in each case the trade turned out to be a roll rather than the structure we described. Full per-ticker detail is in each updated article.

🔄 The three inversions:

• MCHP — the biggest correction on the board. The Sep $85 call went 20,297 → 12,333 (−7,964). A sale that destroys open interest is closing a long, not opening a short. So this was never an income-financed diagonal — it is a bullish call roll out of OTM September calls into deep-ITM December calls. There is no short leg and no $85 cap on the upside. The Dec $65 leg confirmed open (50 → 11,225).
• SBUX — not a fresh ladder. The $110 call went 14,311 → 12,674 (−1,637): that leg retired contracts, most consistent with buying back a short call, not stacking a second long. The $105 leg did open, but only +2,472 of 7,500 (≈33%) was new. Both legs were still buys — that part held — but "all $12M at risk" was wrong; genuinely new capital is closer to ≈$6.83M.
• LLY — an inversion that strengthens the thesis. The Sep $1,040 call went 1,606 → 430 (−1,176), making it a buy-to-close on a short call, not a new long. You do not carry a short deep-ITM call unless you hold something underneath it — so this is independent evidence for the "hedge restructure on an existing stock position" read we could only infer on July 31. Both 2028 collar legs opened at exactly 1,600 each (100.0% of print).

✅ Confirmations — the other eleven tickers held:

• SPY — all six butterfly legs opened, every one within 100 contracts of our published prediction. 1,050,000 contracts of new open interest, no meaningful transfer anywhere. The cleanest resolution of the day at by far the largest size.
• AMZN — the roll confirmed in both directions: Nov $240 +24,367 (open), Sep $220 −7,975 (close). The fresh-short-call alternative is dead.
• TSLA — all three legs opened, including the $400 call we could not prove (6,041 → 8,484, ≈98% of print).
• HNI — both legs at exactly 5,434, a 100.0% match on strikes that did not exist before.
• SMH — both legs full size ($625 put 49 → 3,949; $730 call 3 → 2,004).
• GXO — fully fresh: 3,001 → 10,004, three contracts above the arithmetic ceiling of a clean open.
• CCJ — 0 → 15,109, inside our published 15,000–15,131 range.
• TIGO — 96 → 5,288, more than the block itself; ≈97% of the strike's whole day volume opened.
• SNDK — Aug $800 call 30 → 2,053, which also proves the print was not busted (a cancelled trade never reaches open interest).
• QQQ — both legs opened, but with an honest asterisk: the $600 put built only +721 of 3,000 (≈24%), so the market-wide short-put build is a quarter of the headline.
• AAPL — as promised, no snapshot exists and never will. Confirmed: OPRA published no August 3 open-interest record for either 0DTE leg. Same for SNDK's expiring $1,000 call. Their classification rests permanently on structure and economics.

⚠️ Two partial opens worth naming. SBUX's $105 call (≈33% new) and QQQ's $600 put (≈24% new) both opened, but most of each print was existing contracts changing hands. Direction settled; size overstated by the headline.

The standing lesson, sharpened by today. Intraday, MCHP's roll was indistinguishable from an income-financed diagonal — same two prints, same sizes, same net debit. Only the direction of the next morning's open-interest change revealed that the sold leg was destroyed rather than created, and that single fact flipped the structure, the tone, and the capped-upside claim. A big premium headline tells you capital moved. Only next-day open interest tells you what it did.

---

⚠️ Risk & Reality Check

Institutional flow tells you what large accounts did. It never tells you their reason, their holding period, what else is in the book, or whether they are right.

Today makes that unusually clear. ≈$131M of the headline carried no directional information at all, and two of the largest tickets were rolls of positions that already existed. Two more — SPY and QQQ — arrived in the feed with legs missing, which changed what they meant entirely. The genuinely new, genuinely directional money is a much smaller number than $339M.

And a specific caution on the SPY butterflies: a headline like "≈$3.0B of potential payoff" is a maximum at one exact index level on one exact day, not an expectation. The most likely outcome for a deep out-of-the-money butterfly is that it expires worthless. Someone buying insurance is not a forecast that the house will burn down.

What the tape cannot tell us: who traded, whether they own the underlying shares, whether an offsetting position exists elsewhere, or the sign of the open interest already there. One more caution: large prints can be cancelled by the exchange hours later — that is exactly what happened to yesterday's SMH ticket, and it took $4.4M off a number we had already published.

Practical discipline: size positions so a total loss is survivable, prefer defined-risk structures, never sell naked options against an index, and let the next-day open-interest check confirm a read before you build around it. Patience costs nothing. Chasing a headline premium number has cost a lot of people a lot of money.

This is market analysis and education, not investment advice. Options carry substantial risk of loss, including total loss of premium paid. Do your own research.

Last updated: 2026-08-03 — the next-day OPRA open-interest check resolved all 28 checkable legs from the July 31 session (24 open, 4 close). Three reads inverted: MCHP (diagonal → roll), SBUX (ladder → roll down), LLY (Sep call BTO → BTC). Glance table and the closing section updated accordingly.