📊 Ainvest Option Flow Digest — 2026-08-03

Twenty-four names, ≈$218M net — and for once the big money was SELLING premium, not chasing it

✅ RESOLVED — August 4, 2026 pre-market. The next-day OPRA open-interest snapshot is in for every leg on this board: 46 legs across 24 names — 41 confirmed OPEN, 2 confirmed CLOSE, 2 partial, 1 inconclusive. Three published reads inverted, and one was materially revised:

• 🔄 GOOGL was not a new bear call spread. The $410 call's open interest fell 13,475 → 6,863: the buy was closing an existing short call. The real trade is a short call rolled DOWN from $410 to $390 — no capped $10.0M loss, and the ceiling moved ≈5% closer to the money.
• 🔄 TSM's $14M call sale was a long holder EXITING. Open interest fell 21,661 → 16,725. No new short call exists; the credit is exit proceeds, not compensation for risk.
• 🔄 VST was not a roll. Both legs' open interest rose by exactly 5,400. It is a brand-new deep-ITM $195 put time spread opened for a $3.32M credit, not a maturity extension of an old position.
• ⚠️ KKR revised: the $90 put's open interest fell 5,565, so this was a short put rolled UP from $90 to $95, not a cleanly-built spread — the $5.25M capped-loss figure is not supported.
• 📌 SPY anomaly: all three butterfly legs opened at exactly 2.00× the printed size (+300,023 / +600,147 / +300,057), in the same 1:2:1 ratio, while the tape shows only 1×. The program standing at those strikes is twice what printed.

Full per-leg tables are in each ticker's ✅ RESOLVED box.

---

🎯 Quick Read

Monday's board flipped last week's script. Fourteen of twenty-four names collected net premium rather than paying it. Institutions spent the session harvesting volatility, financing bullish positions with put sales, and rolling winners down in size — not buying upside.

• ≈$218M net across 24 names. Three separate desks sold more than $18M of options each and walked away with the cash.
• ⭐ The late standout is AZN. AstraZeneca fell ≈7% today on a Financial Times report it is in early talks to buy Bristol Myers Squibb for ≈$400B — an acquirer de-rating, with BMY rallying ≈8% as the tell. Into that drop, somebody bought $7.13M of October $170 calls at the offer, on a strike with just 270 contracts of prior open interest. That is fading deal speculation, not catching a falling knife on bad news.
• Two afternoon prints are textbook "looks like X, is actually Y" cases. RDDT sold a deep in-the-money put for $18.75M — that is a synthetic long, an agreement to buy ≈250,000 Reddit shares at an effective $145, not a bearish bet. And INFY bought a deep-ITM put with $0.17 of time value alongside stock — pure financing, no direction at all.
• FISV fenced ≈9.4 million shares and got paid for it. Sell 98,400 September $55 calls, buy 150,000 $50 puts, net $7.6M credit — a protective collar over roughly $510M of Fiserv stock, put on three days before the company reports on August 6, and days after an activist letter landed.
• ⭐ The trade of the day is NKE, and the equity tape proved it. Somebody sold 88,000 September $40 puts and bought 72,000 $45 calls — ≈$220M of long exposure on a stock down 33% this year, and they were PAID $3.52M to put it on. A 5,285,000-share block printed 24 seconds later, matching the structure's delta to within 1.4%. That block is what settles it: this was a bullish bet, not an income strangle.
• SPY's crash-hedge program added a third month. The same put-butterfly shape has now printed in Sep-30, Oct-16 and today Nov-20 — ≈$57M of crash convexity laddered across three expiries. Today's leg cost $25.35M and pays up to $1.8B if the index lands near $500. 📌 The next-day OI came in at exactly 2.00× the printed size on all three legs, so the position standing at those strikes is twice what the tape showed.
• AMZN is the most bullish name on the board — $52.2M of call buying, and part of it is follow-through we can verify. Friday's Nov-20 $240 roll built open interest 2,429 → 26,796 exactly as we predicted. Today the same strike got 7,250 more contracts.
• CRWV was the largest single premium sale: ≈$27.1M of calls sold 38–45% above spot on one of the market's most volatile AI names, with no stock hedge on the tape — and the stock rose ≈17% today. Earnings are confirmed for August 11, inside the expiry.
• MSFT's roll looks bullish and isn't. It banked $9.48M and cut the position's delta roughly in half. Read the delta, not the dollars. ❓ The next-day OI could not confirm the August leg's close — that strike traded 53,456 contracts against our 7,435-lot print, so our leg cannot be isolated. The roll read stands on structure, not on open interest.
• PDD put a ceiling on 1.35 million shares — and got paid $4.46M to do it. A negotiated block sold 13,500 June-2027 $125 calls and bought 13,500 $55 puts against them, both ≈39% away from the $90.18 price. That fences roughly $122M of PDD stock for the next ten and a half months: protected below $55, capped above $125. It is not a bullish trade and it is not a bearish one — it is somebody deciding they are content to exit at $125 and would rather not find out what happens below $55. PDD is down ≈22% over the past year, and the three most recent analyst targets ($80, $89, $89) all sit at or below today's price.
• Two late prints looked like ROLLS — the next-day OI says only one of them was. 🔄 VST crossed ≈$55M of gross premium in deep in-the-money $195 puts, and the package carried a net delta of just −53,244 shares. We called it a maturity extension. It was not: open interest rose by exactly 5,400 on both legs, so nothing was closed — this is a brand-new $195 put time spread (long the January 2027 leg, short the June 2027 leg) opened for a $3.32M credit, two days before Vistra reports on August 7. The tiny delta still holds; the "pushing an old obligation out" framing does not. PANW is smaller but sharper — the tape showed a second leg the screenshot missed. Somebody sold 920 August-7 $390 calls (worth $0.40, four days from expiry, effectively dead) and bought 920 September-4 $390 calls for $11.78. The August expiry dies before Palo Alto reports; the September one contains the print. They paid ≈$1.05M to move the same strike across the catalyst.
• QQQ printed $62M gross and means almost nothing — the same calls were bought at 10:31 and sold back eight minutes later.

The honest frame: premium sellers get paid up front and carry the obligation afterwards. Every credit on this page is compensation for a risk somebody accepted. Read what they'd owe, not what they collected.

---

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

Friday's next-day open-interest snapshot was the most consequential check we've run in weeks. Fourteen tickers carried ⏳ provisional flags across 31 legs. 28 were checkable and all 28 resolved: 24 opened, 4 closed. Three inverted a published read — and in every case the trade turned out to be a roll rather than the structure the intraday tape suggested.

🔄 Inversions

🐂 MCHP — the diagonal was a ROLL, and the upside was never capped. Sep $85 call OI 20,297 → 12,333 (−7,964) against an 11,000-lot sale. A sale that destroys open interest is closing a long, not opening a short. Corrected thesis: a bullish call roll down and out — liquidate OTM September calls, redeploy into deep-ITM December calls already $10.79 in the money. No short leg, no cap; delta per contract roughly doubled. This is a more bullish trade than we published.

🪜 SBUX — the "long call ladder" was half a close. Jan-2027 $110 call OI 14,311 → 12,674 (−1,637) against a 7,500-lot buy — most consistent with buying back a short call. The $105 leg opened, but only +2,472 of 7,500 (≈33%) was new. Genuinely new capital is closer to ≈$6.83M, not $11.96M. Worth stressing: both legs really were buys — that read held. Buy-vs-sell and open-vs-close are separate questions, and only open interest answers the second.

⚖️ LLY — an inversion that makes the thesis stronger. Sep $1,040 call OI 1,606 → 430 (−1,176) against a 1,500-lot buy. A buy that destroys open interest is buy-to-close: the desk was short those calls. You do not carry a short deep-ITM call ≈9% below spot unless you hold shares underneath it — which confirms the hedge-restructure reading we had explicitly marked inferred.

✅ Confirmations

• 🦋 SPY — all six butterfly legs opened, every one within 100 contracts of our published prediction; 1,050,000 contracts of new open interest.
• 🔄 AMZN — roll confirmed both ways: Nov $240 +24,367, Sep $220 −7,975.
• 🛡️ TSLA — all three legs opened, including the $400 call we could not prove (≈98% of print).
• 🪑 HNI both legs at exactly 5,434 · 🛡️ SMH both full size · 🧊 GXO 3,001 → 10,004 · ☢️ CCJ 0 → 15,109 · 📞 TIGO 96 → 5,288.
• 🔄 SNDK — Aug $800 call 30 → 2,053, which also proves the print was not busted. We flagged bust risk explicitly after SanDisk's July 21 cancellation; a cancelled trade never reaches open interest.
• 🧊 AAPL — exactly as promised, no snapshot exists and never will for the 0DTE legs. We said that at publication rather than quietly letting the check lapse.

⚠️ Two partial opens — direction settled, size overstated

QQQ's Jan-2028 $600 put built only +721 of 3,000 (≈24%) and SBUX's Jan-2027 $105 call only +2,472 of 7,500 (≈33%). Both opened — the positions are real — but most of each print was existing contracts changing hands, not new exposure. The lesson: a headline size is a ceiling on new risk, never a measurement of it.

---

📋 At a Glance

Net, not gross. Every figure above is net — premium paid minus premium collected. QQQ alone printed $62M gross and nets to a $6.6M credit; AMZN's two trades gross $52M and net the same because both were purchases. Gross numbers make small trades look enormous.

---

🔍 Three Worth Your Time

⭐ NKE — they got PAID $3.5M to take on $220M of Nike

Two legs crossed at 11:14:51 with the stock at $42.17: sold 88,000 September $40 puts for $11.62M and bought 72,000 September $45 calls for $8.10M. Net: a $3.52M credit.

The screenshot was ambiguous — it showed both a buy and a sell on the $45 call. The equity tape settled it. A 5,285,000-share block printed at $42.3549 twenty-four seconds later. The delta of sell-put-plus-buy-call computes to +5,212,240 shares — a match within 1.4%. Had the calls been sold instead, the structure's delta would have been ≈254,000, nowhere near the block. The calls were bought.

So: roughly $220M of bullish exposure on a stock down 33% year to date, established for a credit. Both legs are proven fresh opens.

What they owe: below $40, they are on the hook for 8.8 million shares — about $352M. The credit is payment for that obligation, not a gift. And note the timing: Nike's next earnings is expected ≈September 24–29, after the September 18 expiry. There is no earnings event inside this trade. It is a bet on drift and positioning, not on the print. Full breakdown →

🐂 AMZN — the rare case where we can check our own homework

On July 31 we published that a desk rolled out of September $220 calls into 25,000 November $240 calls, called it a proven open, and predicted open interest would build by at least ≈22,571 contracts.

It built by 24,367 — from 2,429 to 26,796. The prediction landed.

Today the same strike took another 7,250 contracts at $53.90, a $39.08M block cross. Add a separate lit purchase of 10,999 January 2027 $335 calls at the offer for $13.09M, and Amazon saw $52.17M of call buying in one session — the most bullish single name on the board.

The two trades are different animals. The January $335 call is 17% out of the money and lit — a real, aggressive directional bet where paying the offer tells you something. The November $240 call is deep in the money with only ≈$9.48 of time value — a high-delta, stock-like way to hold exposure. One is a wager; the other is position-building. Full breakdown →

⚡ POWL — $7.3M of calls sold hours before earnings

Powell Industries makes electrical switchgear for utilities, oil and gas, and data centres. The stock has roughly doubled this year (≈+83% or ≈+101% depending on how the April 3-for-1 split is applied to the year-end base — either way, doubled). At 10:50 somebody sold 1,400 February 2027 $220 calls at $52.40 — collecting $7.34M.

That price is the story: $52.40 on a $213 stock is 24.6% of the share price, for a strike barely 3% above spot. And Powell reports fiscal Q3 earnings tonight, after the close. Selling near-the-money calls hours before a binary event is exactly when option premiums are richest — and exactly when they are richest for a reason.

The seller keeps everything below $220 and only starts losing above $272.40 (the $220 strike plus the $52.40 collected). Worth noting where that sits: below Powell's June 25 closing high of $309.20 and below the S&P Global average analyst target of $316.25. The seller is fading both the recent tape and the sell-side — and three earnings prints fall inside this contract's life, starting tonight. Whether that is smart or reckless depends entirely on something the tape cannot show us: whether they own the shares. Covered, it caps a big winner. Naked, the loss is open-ended. We cannot tell, and we are not going to pretend otherwise. One more caution — Powell's option chain is thin, so a retail trader would give up a painful amount just crossing the spread. Full breakdown →

---

📅 Upcoming Catalysts — and Which Expiry Each One Lands In

⚠️ A catalyst only matters to a contract that is still alive when it happens. Several trades today expire before their own company reports.

---

👥 Four Ways to Read This Board

🎲 The YOLO Trader

Today offers you two things and you should be honest about which is which. TSM's $420 call expiring Friday cost $3.00 and needs a 4.6% move in four days — a genuine lottery ticket, priced like one. SPY's butterfly advertises 71:1, but that is the payoff at one exact index level on one exact day, with the body sitting 34% below spot. The realistic outcome for a deep out-of-the-money butterfly is a total loss.

What you should not do is read the credits on this page as free money. CRWV's sellers collected $27M and now carry unlimited upside risk on an AI name that can gap 20% overnight.

📈 The Swing Trader

NKE is the one to study. A desk took ≈$220M of long exposure on a name down 33% YTD, for a credit, with no earnings inside the expiry — that is a positioning bet, and the 5.3-million-share hedge tells you it was sized deliberately. Whether you agree is a separate question: consensus models Nike's earnings down next year and 25 of 39 analysts are on Hold.

SMH is the counterweight — a bullish semiconductor ladder built for ≈$1M of net premium, carrying ≈$249M of delta into NVIDIA's August 26 print. High conviction, thin cushion.

And note MSFT as a discipline lesson: the trade banked $9.5M and halved its delta. Institutions trim into strength. Retail usually does the opposite.

💰 The Premium Collector

This was your day — and also a warning. Three names sold serious volatility: CRWV ≈$27.1M, IONQ $7.73M (24.5% of spot), POWL $7.34M (24.6% of spot, hours before earnings). Those percentages are not generosity; they are the market's estimate of how violently these stocks can move.

The two structures actually worth studying are the defined-risk ones: KKR's bull put spread (collect $2.25M, risk $5.25M, needs KKR to hold above $95) and GOOGL's bear call spread (collect $5.49M, risk $10.0M, needs GOOGL below $390). Both risk more than they collect — that is normal for high-probability trades, and it is exactly why position sizing matters more than win rate. Note too that GOOGL's own analyst consensus sits above the short strike, so the seller is betting against the Street.

Never sell naked calls on a name like CoreWeave at retail size. Defined risk or nothing.

🌱 The Beginner

Three ideas from today, in order of usefulness:

1. A credit is not a profit. Nine names here collected premium. Every one of them accepted an obligation in exchange. NKE was paid $3.5M and could owe 8.8 million shares.
2. Gross premium lies. QQQ printed $62M — and it was the same calls bought at 10:31 and sold back at 10:39. Net effect: ≈$6.6M and almost no position. Always ask what the net was.
3. Check whether the catalyst is even inside the trade. Nike's earnings land after its options expire. Microsoft's next report lands after both of its expiries. TSM's Friday call dies three days before the company's monthly revenue release. This is the single most common mistake in reading options flow.

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

---

✅ Every Read Here Has Now Been Checked (resolved August 4, 2026, ≈06:30 ET)

This section was published as "provisional until tomorrow morning." Tomorrow arrived. The next-day OPRA open-interest snapshot has been pulled for all 46 legs, and here is how the flags landed.

• ✅ Confirmed OPEN (41 legs): every leg we called a proven open held — MRVL's Aug-14 $117/$118 pair (0 → 500 each), SMH's four legs (94–99% of print), MSFT's Oct $510 call (253 → 52,671), MRVL's Oct $300 put (57 → 1,053), all three SPY butterfly legs, IONQ's $35 put (1,191 → 9,879), both PDD collar legs (≈99% each), VST's Jun-2027 $195 put (5 → 5,405), PANW's Sep-04 $390 call (10 → 932), both NKE legs (≈100%), both CRWV strikes, AZN, HOOD, INFY, PG, POWL, RDDT, FISV, both AMZN legs — including AMZN's November $240 call, which was on the "cannot be proven" list and came in at ≈100% fresh (26,796 → 34,053). SKHY's two unprovable legs both resolved opening, within ≈2% of our predicted full-open levels.
• 🔄 INVERTED (3 legs, 3 tickers): GOOGL's $410 call (13,475 → 6,863 — a close, so the trade was a short call rolled down, not a spread), TSM's September $400 call (21,661 → 16,725 — a long holder exiting, not a new short), and VST's Jan-2027 $195 put (7,298 → 12,698 — an opening long, so the package was a fresh time spread, not a roll).
• ⚠️ Revised or partial (3 legs): KKR's $90 put fell 5,565 — partly closing, so this was a short put rolled up rather than a clean new spread. MO's September $70 call rose only 1,548 against a 7,500-lot sale (≈21% genuinely new). SMH's $500 put came in at ≈85% of print — comfortably above the ≈317-contract minimum we set, so the ladder read holds.
• ✅ Confirmed CLOSE (as published): PANW's Aug-07 $390 call fell 2,349 → 1,388, proving the roll exactly as the structure suggested.
• ❓ Inconclusive (1 leg): MSFT's August $460 call. Strike-level open interest rose 22,191, but that strike traded 53,456 contracts on the day against our 7,435-lot print — our leg cannot be isolated, and no future snapshot will change that. The roll read now rests on structure (same-instant paired legs, deep-ITM sale at the bid), not on open interest. We stated that test too strongly when we published it.
• ⚠️ The one we said mattered most — TSM's $400 call — resolved to the closing case. Open interest fell ≈4,936, almost exactly the ≈16,661 we forecast for a close. Somebody closed a long. It was never a new bearish short.
• 📌 One anomaly worth naming: all three SPY butterfly legs opened at exactly 2.00× their printed size, in the correct 1:2:1 ratio, while consolidated volume at those strikes matched only 1×. Seven other strikes in the same expiry moved normally, so this is not a data artifact. The most plausible explanation — inferred, not proven — is a second identical package cleared away from the public tape.

---

⚠️ Risk & Reality Check

Institutional flow tells you what large accounts did. It never tells you why, for how long, what else is in the book, or whether they were right.

Today's board makes one thing unusually clear: premium selling dominated, and premium selling is a business of small, frequent wins and rare large losses. Nine names collected credits. Those credits are compensation for obligations that only show up if the market moves against them — which is precisely when they hurt most.

Friday's open-interest check should also stay in mind. Three of our published reads inverted once the data arrived, and in each case the trade turned out to be a roll rather than the structure the intraday tape implied. We publish the corrections because you are making decisions with real money, and a confident wrong answer is worse than an honest uncertain one.

Practical discipline: size so a total loss is survivable, prefer defined-risk structures over naked ones, never sell uncovered calls on a volatile single name, and wait for the next-day open-interest confirmation before building a thesis around any print.

This is market analysis and education, not investment advice. Options carry substantial risk of loss, including total loss of premium paid — and for sellers, losses that can exceed the premium collected. Do your own research.

---

Last updated: August 4, 2026 — next-day OPRA open-interest resolution added for all 46 legs (see the ✅ RESOLVED block at the top and each ticker's ✅ RESOLVED box). Originally published August 3, 2026.