AInvest Option Flow Digest — 2026-08-28 · $492M Crossed the Tape. Six Dollars in Ten Were Plumbing.

Nine names, thirty-three legs, $491.61M of premium — and $301.59M of it, 61.3%, carried $5.68M of actual economics.

Four of today's nine trades are rolls, ladders and financing packages: deep-in-the-money structures where the premium is almost entirely intrinsic value, so there is nothing in them to be right or wrong about. SanDisk alone printed $202.13M and the whole net was $510,000 — a 396× overstatement, and Barrick printed $32.85M with about $470,000 of time value in it. The genuinely directional board is five names, $190.03M gross and $104.74M net.

⭐ And Barrick shows you exactly why these trades exist, because the deadline is visible. Its ex-dividend date is Monday. Under the settlement calendar, today is the last session on which exercising a call captures that dividend — and fourteen deep-in-the-money call prints duly crossed. Six of the fourteen priced below the $0.175 dividend, which is precisely the point at which exercising early stops costing you money. That is not a view on gold. It is a deadline.

That is the third session running where the headline dollar figure has been mostly noise, and it is why we lead with the split rather than the total.

⚠️ And today the feed was wrong on four of the nine names, once in an unusual direction. On Microsoft it under-reported: 7,250 contracts a leg shown against 17,000 on the tape — a $126.47M package, not $58M. It showed 10 of Barrick's 14 legs. And on SK hynix it dropped the leg that defines the trade — see below.

One more thing before the detail: every leg today printed as a negotiated block except two. Microsoft's $510 call buys went through lit, at the ask — the only place on the board where the aggressor is genuinely a buyer rather than a bookkeeping convention.

🔁 OI Review — Yesterday's Predictions, Scored

The 06:30 ET snapshot is in for the 2026-08-27 board. Sixty-seven legs across thirteen names, all resolved. The headline is not a miss — it is how little of yesterday's $257M created anything.

🔄 Forty-three legs traded. Zero contracts were created.

Ticker	Legs	Contracts traded	Net Δ open interest
NFLX	25	12,360	0 on every single line
META   ladder	10	2,250	0 on every single line
WMT	8 verifiable	13,310	−93 in total

We checked it was not a stale feed: the NFLX September $110 put has been pinned at 991 contracts since August 4 through 8,085 lots of volume, while the $110 call on the same expiry updates normally (53,642 → 53,370). The feed is live. Those put lines genuinely do not move. Yesterday's "$123.58M that means almost nothing" is now confirmed by open interest, not just by parity pricing.

🔄 Where the readings changed

• IGV — the protective leg never showed up. The $103 put went 5 → 30,051 (+30,046), a clean open. The $95 put went 34,965 → 35,001 (+36) — a transfer. So what entered the market was 30,046 new short puts and no new long puts. If you read that as a bounded, defined-risk spread, the tape does not support the "bounded" half.
• SMCI — the hedge expanded rather than renewed. January-2028 $10 put 49,099 → 38,612 (−10,487); December-2028 508 → 27,824 (+27,316), nearly double our forecast. ≈2.6 contracts added for every one retired, with ≈38,612 January contracts still standing. More defensive than we published, four days before the filing deadline.
• SOFI — both leans missed, in opposite directions. But the prediction that mattered held: resolving open-versus-close still did not tell us whether the package was a debit or a credit. We said it wouldn't. It didn't.

✅ And the confirmations

PCG's roll confirmed on both halves in one session — September $20 calls 316,090 → 128,633 (−187,457) while October went 63,024 → 309,968 (+246,944, 100.8%). EWY all three legs opened; PDD both at exactly 100.0%; HONA landed one contract from forecast; TOST at 84.0%, the top of its published band.

One honest failure worth naming: HRI's wide $170 leg opened but landed ≈398 contracts short of its own published arithmetic floor — offsetting same-day flow moved the net number. That is precisely the failure mode our four-branch rule exists for.

And a four-session backlog was cleared in the same run, producing two corrections that trace to one root cause: GLD and AA both had wrong prior open interest published (GLD's true figure was 136,136 against 7,400; AA's 12,387 against 4,600). A wrong baseline inverts everything built on it. From today, every quoted open-interest figure is verified against the open-interest tape before any arithmetic is computed from it.

📊 At a Glance

Ticker	Net	Expiration	Catalyst — and its date	The Option Play	What It Means
MSFT	$76.17M debit	Monthly (Oct-16 + Nov-20)	Earnings   ≈Oct 27–28, unconfirmed   — between the two expiries	17,000-lot four-leg roll, plus lit $510 call buys	Reads two ways, and they point opposite
SNDK	$0.51M credit	Weekly → Monthly (Sep-18)	No earnings before expiry	$900-strike roll, one leg expiring today	$202M printed; $510,000 was real
AMZN	$13.35M debit	Monthly (Nov-20)	Earnings   ≈Oct 29 , inside the window	Bull call spread $295/$350	Breakeven above the 52-week high
VIX	$9.61M debit	Monthly (Oct-21)	Sep 15–16 FOMC ; expires   before   the midterms	104,405 $28 calls	A hedge that stops short of the risk
ETHA	$3.32M debit	Monthly (Oct-16)	No catalyst — this is carry, not a view	Same-strike synthetic + 2M shares	A financing trade, textbook
IGV	$2.88M debit	Monthly (Nov-20)	PANW Sep 1, ORCL Sep 4, ADBE Sep 10   = 19.8% of the fund	Put spread $103/$97, unequal sizes	Buying what yesterday's desk sold
B	≈$0.47M of time value	Monthly (Sep-18) + one LEAP	Ex-dividend Monday — today is the last day to capture it	14-leg deep-ITM call ladder	A dividend deadline, not a bet
SKHY	$2.73M debit	Monthly (Sep-18)	No earnings inside the window ; union vote unresolved	Collar with the floor bounded at $135	A real hedge — and the only one today
GLW	$1.38M debit	LEAP (Jan-27 + Mar-27)	Ex-dividend Monday, Aug 31	Deep-ITM diagonal roll	A roll timed around a dividend

Gross across all thirty-three legs was $491.61M. Of that, $301.59M (61.3%) was mechanical — rolls, ladders and financing carrying just $5.68M. The directional board is five names, $190.03M gross / $104.74M net.

🔍 The Ones Worth Your Time

1. 🔁 MSFT — the same four legs read two ways, and they point in opposite directions

→ A 17,000-lot roll, and the day's only lit buyer

First, the size was wrong in the summary — and wrong in the rare direction. The reported figures showed 7,250 contracts a leg. The tape carries four prints per strike totalling 17,000, with consecutive sequence numbers matching each line's day volume. This is a $126.47M package.

Four legs: buy November $480 calls, sell November $580, sell October $545, sell October $595. And the open interest splits cleanly — both November legs sit below their existing open interest (21,709 and 18,648) while both October legs are near-empty. So something is being closed and something opened.

Here is where it gets interesting. Read at face value, it is a bull call spread with calls sold against it: +172,040 delta, mildly bullish.

But Microsoft gapped 15.51% on July 30 — $390.54 to $451.10, its biggest single day since 2008. A desk that was short a November $480/$580 call spread would have been run straight through by that. And closing a short spread means buying back the $480 and selling the $580 — exactly the sides that printed. Under that reading the only new risk is 34,000 short October calls: delta ≈−646,000 shares, neutral-to-bearish.

Same four legs. Opposite sign. It printed as an auction, so nobody outside the deal can prove which — and tomorrow's open interest separates opening from closing but never direction.

What we can say cleanly is where the strikes sit. Median target $550. Of twenty dated post-earnings targets, twelve (60%) sit at or above $545 — but only nine (45%) reach $580 or $595, and no target at all sits between those two. $580 and $595 are both above the 52-week high of $553.72; $545 is not — it is inside the year's range, essentially on the median.

⚠️ And the calendar carries a hidden risk. Earnings is expected ≈October 27–28, which would put it neatly between the expiries — October shorts dying before the print, November exposure spanning it. Microsoft has not confirmed the date. The pattern is uniformly the last week of the month, so it is likely, not certain. A print before October 16 would leave those short calls exposed.

2. 🔄 SNDK — $202 million printed, $510,000 was real

→ Why the biggest number on the board means the least

Two legs, 1,700 contracts each, same $900 strike — about 40% below the $1,494 share price, so deeply in the money — one expiring today, one in September.

The intrinsic value on both legs is $594.28. They printed at $593.00 and $596.00. The intrinsic parts cancel exactly, so the entire $510,000 net is the time-value difference. The $202.13M headline overstates it by 396×. The expiring leg actually printed below parity.

The question a reader should ask about any deep-in-the-money short call is whether it gets exercised early — and here the answer is a clean no. SanDisk has never paid a dividend, so there is no ex-date trigger; short interest is 5.24% with half a day to cover, so there is no borrow trigger either.

Why does a $900 strike sit 40% in the money? Because SanDisk is up 525% this year and 2,824% over twelve months, on a fiscal fourth quarter that put revenue at $8.97B, +372%, with gross margin at 84.6% — against 22.5% five quarters ago.

⚠️ The counterweight, and we are not going to soften it. The $3.00 collected is essentially interest on the strike: the market prices $900 as unreachable. But this stock fell 46.6% in five weeks between June 30 and July 31, and has had six double-digit down days in two months. An identical repeat from here lands near $790 — through the strike. The market prices that tail as free. The stock's own record this summer says it is not.

3. 🎯 AMZN — a spread whose breakeven the stock has never seen

→ $13.35M that needs a new all-time high just to break even

25,000 November $295 calls bought, 25,000 $350 calls sold — a $55-wide bull call spread for $13.35M. Both legs wide-margin opens.

The number that frames it: breakeven is $300.34, and the 52-week high is $287.20. The entire post-earnings run only reached $284.02 on August 3. The stock has not been where this trade needs it to start paying.

The strikes sit either side of the Street. Of 26 enumerable targets, all 26 sit at or above the $295 long strike — the lowest is $310. But at or above the $350 they sold, only ten of roughly sixty analysts, with the median $25 below it. So they bought a level the Street already endorses and sold one it mostly does not.

Earnings around October 29 falls inside the window with ≈22 days to spare, carrying the holiday guide. The fundamentals genuinely cut both ways: AWS is accelerating — +24%, then +28%, then +37%, the fastest in eighteen quarters — while trailing free cash flow is −$7.6B against +$18.2B a year ago, and third-quarter guidance implies 9–12% growth against 20% delivered.

4. 📈 VIX — the strike is much closer than it looks, and the timing is the problem

→ Why "+93% away" is the wrong number

104,405 October $28 calls bought for $9.61M, on a line that has been building all week — 20,733 → 21,373 → 55,111 → 66,815, with today's purchase on top.

The obvious reading is wrong, and it is worth understanding why. VIX sits at 14.51, so $28 looks 93% away. But a VIX option prices off the matching VIX future, not the spot index — and the October contract settled at 18.53. The strike is +51.1% above what actually prices it, with breakeven at 28.92.

These are not equity options. They settle in cash against one Wednesday opening print on October 21. The path in between pays nothing — $9.6M turns on a single opening calculation.

On the odds: over the last 52 weeks VIX has closed at or above 28 on 3 of 257 sessions — 1.2%; over two years, 17 of 515.

⚠️ But the sharpest argument against it is a calendar point. The futures curve steps up 0.784 vol points from the October contract to the November one — and November is the contract that spans the midterms. This option expires before both that and the October 27–28 FOMC. The market is explicitly pricing event risk into a window this hedge does not reach.

5. 🔄 ETHA — the clearest financing trade we have shown you, and a correction we owe

→ A $3.32M package that isn't about Ethereum at all

20,000 October $17 calls bought and 20,000 $17 puts sold — same strike — with a 2,000,000-share block printing one minute later.

A call and a put at the same strike make a perfectly straight payoff line: delta 1.00 per contract, synthetic long stock. The maths says 20,000 contracts carry 1,985,200 shares of delta. The block was 2,000,000 — a 99.3% match. Held against stock, that is a conversion: a financing trade, priced here at roughly 4.4% annualised carry.

We are showing you this partly because we got the same call wrong two days ago. We described a package on another name as "financing" when its two strikes were $80 apart — which leaves a dead zone between them and a delta well under 1.00, meaning somebody genuinely had a view. We corrected it. The rule worth keeping: same strike means financing; split strikes mean an opinion.

⚠️ Honest limit: the usual motive for this trade is expensive stock borrow, and the borrow here looks to be getting cheaper — short interest is down 10.44% and the category just took its largest inflow in 203 sessions. The borrow rate itself we could not source, so we are not going to assert why they did it.

6. 🛡️ IGV — buying exactly what yesterday's desk was selling

→ Two sessions, two directions, one strike

Yesterday a floor block sold 30,000 December $103 puts for a $6.54M credit. This morning's open interest resolved it: the $103 line opened cleanly, the paired $95 line was only a transfer — so 30,046 new short puts entered the market and no new long puts did.

Today somebody bought 15,000 November $103 puts — same strike, different expiry, exactly half the size — against 12,500 sold at $97. Whether it is the same desk is not knowable from the tape.

The timing is worth noting: the puts were bought at 09:45 on the session the rally's leaders broke. Yesterday IGV rose 7.21% on CrowdStrike and Salesforce. Today CrowdStrike fell 4.52%, Fortinet 5.15% and Palo Alto 3.16%, while ServiceNow rose 4.00% and Salesforce 2.95%. The leadership inverted underneath a flat-looking index.

And there is a real asymmetry in what the buyer is covering. Palo Alto reports September 1, Oracle September 4, Adobe September 10 — 19.79% of the fund inside thirteen days. But CrowdStrike and Salesforce, 12.81% between them, reported on August 26 and do not report again before this option expires.

7. 💵 GLW — a roll timed around a dividend three days away

→ $60.73M gross, $224,000 of time value

A deep-in-the-money diagonal: buy January-2027 $55 calls, sell March-2027 $60 calls, 3,200 each. Gross $60.73M; net debit $1.38M; and of that, the time-value component is just −$224,000 — the rest buys $5 more intrinsic per share. Deltas of 0.988 and 0.979 leave the package nearly delta-flat, which is why no meaningful stock block appears at all.

The dividend is the point. Corning pays $0.28 a quarter, and the next ex-date is Monday, August 31. Three ex-dates fall before the March expiry — so the March leg picks up one the January leg never sees. The rate has been frozen since February 2023, and the February board meeting, historically when that gets revisited, sits inside the gap. The roll also buys an earnings print: Corning reports its full year in late January, which the January leg expires just ahead of.

8. 💵 B — a $32.85M ladder that is really a calendar deadline

→ Why fourteen deep-in-the-money call prints landed today and not tomorrow

Fourteen separate floor prints in Barrick calls, strikes from $10 to $33 against a $45.51 share price, across three batches. Gross $32.85M — and about $470,000 of time value in the whole thing.

Barrick's ex-dividend date is Monday. It declared $0.175 with a record date of August 31, and under the settlement calendar today is the last session on which exercising a call captures that payment.

Here is the mechanic, and it is worth two minutes because it explains a whole category of flow you will keep seeing. If you hold a deep in-the-money call, exercising early hands you the stock — and the dividend — but throws away whatever time value the option still has. So exercising is rational exactly when the remaining time value is less than the dividend. Six of these fourteen legs priced below $0.175: the $20 call at −$0.26, both $22s at −$0.36 and −$0.21, the $25 at −$0.01, the $27 at +$0.14, the $30 at +$0.04. Those 5,400 contracts cover 540,000 shares and $94,500 of dividend for −$27,400 of surrendered time value.

The market has priced these calls right at the indifference point, on the last day it matters. Nobody here is expressing a view on gold.

⚠️ One genuine open question we cannot close: Barrick is spinning out its North American gold assets via IPO by year-end. The legal form and any resulting contract adjustment are not sourced — which matters specifically for the January-2027 leg in this ladder.

9. 🛡️ SKHY — the only real hedge on the board, and the feed hid half of it

→ A collar whose floor stops right where the stock last went

SK hynix's American shares have existed for seven weeks — the ADR listed on July 10 in a $26.5 billion offering, the largest foreign listing on record, and it was a float sale backed by treasury stock rather than a capital raise, so nobody was diluted.

7,700 contracts on each of three legs, all expiring in three weeks: a $157.50 put bought, a $135 put sold, a $185 call sold. Net $2.73M. All three are clean new positions.

This is the one trade today with real content. It is a collar with a bounded floor — buy protection, sell a cheaper put to cap what that protection costs, sell a call to pay for it. And the put bought is only 2.5% below the share price. That is not a lottery ticket against a crash; that is somebody protecting stock they hold right now, over the next three weeks.

⚠️ The reported figures showed two legs. The tape carries three — and the missing one changes what the trade is. Without the $135 put, this looks like ordinary open-ended protection. With it, the downside cover stops at $135. Below that level the hedge simply stops working.

And that matters more than it sounds, because $135 is only about 8% above the $124.80 low this stock set on July 29 — seven weeks into its life. The floor runs out almost exactly where the shares last went. On the other side, the $185 call they sold sits below every individual analyst target on the Street.

⭐ One last thing worth noticing, because it spans three sessions. Two days ago a desk put on a bullish ratio spread in EWY — a fund that is roughly 48% Samsung and SK hynix — running out to June 2027. Today SanDisk rolls a deep-in-the-money call, and SK hynix itself buys near-the-money protection for three weeks. One desk reaching for upside in the memory complex; another paying to be covered inside it. Both can be right.

⏳ Tomorrow Morning Is the Scorecard

Four branches on every leg, and no arithmetic floor where size sits below open interest.

Test	Full capture	Partial	Flat = transfer	Close
MSFT   Nov $480C / $580C	—	40–90%	listed first — size ≤ OI	a fall confirms the Reading-B close
MSFT   Oct $545C / $595C	≈19,091 / ≈17,234	≥+14,909 / ≥+16,766	—	—
MSFT   Oct $510C	—	40–90%	listed first — 9,366 vs 33,339	close branch live
AMZN   $295C / $350C	≈28,235 / ≈33,018	≥+21,765 / ≥+16,982	—	—
VIX   Oct $28C	≈171,220	≥+37,590	—	(a cash-settled index — no borrow or float dynamic)
ETHA   $17C / $17P	≈30,609 / ≈21,850	≥+9,391 / ≥+18,150	—	—
IGV   Nov $103P	≈15,982	≥+14,018	—	—
IGV   Nov $97P	—	40–90%	listed first — 12,500 vs 12,550	a fall means protection rolled UP, which is more bearish
GLW   Mar $60C	≈3,200	—	—	(prior OI was zero)
GLW   Jan $55C	—	40–90%	listed first — 3,200 vs 6,679	close branch live
SNDK   Sep $900C	≈1,960	≥+1,440	—	—
SNDK   Aug $900C	—	—	—	no check is possible — the contract expires today
SKHY   $157.5P / $135P / $185C	≈8,517 / ≈9,574 / ≈8,649	≥+6,883 / ≥+5,826 / ≥+6,751	—	—
B   — the whole 14-leg ladder	—	—	—	weak evidence here: exercise over the ex-date scrambles the open-interest print

👥 How to Read Today, By Who You Are

🎲 The YOLO trader. VIX is the one built like your trade — 104,405 calls for $9.6M on a strike that pays only if volatility roughly doubles. Before you copy it, understand two things. The strike is +51% above the October future, not +93% above spot — better odds than it looks. But it expires before both the October Fed meeting and the midterms, and the futures curve is visibly pricing more risk into the next contract. The trade stops short of the event.

📈 The swing trader. Your calendar this month is unusually crowded at the front. Palo Alto September 1, Oracle September 4, Adobe September 10 — nearly a fifth of IGV inside thirteen days — then the September 15–16 Fed meeting with a dot plot. Further out, Microsoft and Amazon both report in the last week of October, and in Microsoft's case the date is not yet confirmed, which is itself a risk to anything structured around it.

💵 The premium collector. Four of today's seven packages are rolls, and they are worth studying as craft rather than signal. GLW is the cleanest lesson: a deep-in-the-money call rolled up and out, timed so the new leg captures an extra dividend, for $0.70 a contract of time value. SNDK is the cautionary one. The $3.00 collected on that roll is essentially interest — the market says $900 cannot be reached — but the stock fell 46.6% in five weeks this July. Collecting carry against a tail the market calls impossible is exactly how premium sellers get hurt.

🌱 The beginner. Today's lesson is the most useful one we publish, and it takes thirty seconds. Before you react to a big options headline, check whether the premium is intrinsic or extrinsic. SanDisk printed $202 million and $510,000 of it was real. Corning printed $60.73 million with $224,000 of time value in it. If the strike is deep in the money and the price is close to what the option is already worth, the dollar figure is measuring the stock, not anybody's conviction. Then read ETHA, which shows the same idea from the other side: two options at the same strike are financing; two at different strikes mean somebody has a view.

⚠️ Risk, Plainly

A large trade is not a recommendation, and it is not a prediction. Every desk here has information, hedges and mandates you cannot see.

Check whether the premium is intrinsic before you believe the headline. Six dollars in ten today were plumbing — and Barrick shows the cleanest version: a ladder that exists because a dividend deadline falls tomorrow, not because anyone has a view.

On a negotiated block, nobody outside the deal can prove who initiated. Microsoft's two lit call buys were the only exception on the whole board.

The feed can be wrong in both directions. Today it under-reported Microsoft by more than half.

Same strike means financing; different strikes mean a view. That one distinction separated a real financing trade from one we misread earlier this week.

A hedge that expires before the event is not a hedge against that event. The VIX position stops short of both the October Fed meeting and the midterms, and the futures curve says so out loud.

Nothing here is investment advice. Options carry the risk of total loss of premium, and short option positions carry risk far beyond the premium collected.