NQ / Nasdaq 100: The Money Didn't Leave, It Left Tech, August 6, 2026

The September contract traded 30,074 on Wednesday, the highest print in five sessions, and settled 459 points underneath it at 29,615. On the same afternoon the Dow closed at a record. Money didn't leave the market. It left technology.

Closing in the bottom 15.5 per cent of the daily range tells you whoever seized control after lunch never gave it back.

What broke it

Semiconductors, first. A leading American chip designer dropped about 7 per cent after third-quarter sales guidance came in under what buyers of the AI capital-expenditure story had already paid for, and the weakness spread through the complex. Another large-cap name adjacent to the AI trade lost better than 13 per cent, on higher projected spending of its own.

Then the governance news, which arrived all at once. One mega-cap search platform lost the chief scientist it had kept for 27 years, off to launch a venture, and changed the chief executive of its AI research unit. A hyperscaler disclosed that roughly 70 per cent of its AI revenue traces to a single model developer, that customer contributing 24.1 billion dollars in fiscal 2026.

Any one of those is digestible on its own. Arriving together, the AI trade stops being an unquestioned growth engine and becomes one carrying visible customer concentration and key-person exposure. That reframing is exactly what turns a one-month high into a reversal.

Fifty-five points short of the wall

Wednesday's high stopped just under the most important number on the board.

At 30,127 to 30,130 in futures terms sits the equivalent of the 30,000 cash strike. That strike doubles as the call-side dealer wall and as the primary options concentration in the current book. When two separate positioning measures settle on one round number is roughly the strongest resistance signature this market produces, and price ran out of buyers roughly 55 points beneath it.

Underneath, the strike ladder runs 29,627, then 29,127, then 28,677 in futures terms. The first of those deserves attention: 29,627 lands inside twelve points of a 29,615 settle and a 50-day of 29,639. Three independent measures packed into twenty-five points, and Thursday opens straight into them.

One average overhead, and it is the one that counts

Price closed above the 5-day, the 20-day, the 100-day and the 200-day. It closed 24 points under a 50-day of 29,639.

That single exception is the pivot Thursday resolves first. Trade sustained above it puts the burden back on sellers, and 29,740, the computed pivot, comes into play. Get rejected there and Wednesday's reversal is confirmed, and the 50-day becomes resistance, which it has not been at any point in this July recovery.

Below, the 40-day crossing at 29,435 is the final short-horizon average sitting under price. Lose that and the next equivalent reference is the 18-day zone down at 28,831, which is a lot of empty space.

The trend exists on one timeframe only

Nine-day directional strength reads 35.59, positive direction 26.03, negative 17.72. Genuinely trending.

Widen the window and it dissolves. Fourteen days gives 25.83 with the two lines nearly touching, 22.00 against 20.26. Twenty days gives 19.66 with negative direction already above positive, 21.23 to 20.44. Fifty days has negative on top again.

So the accurate description of the past week is a five-session surge, 2,273 points and 8.31 per cent off the 29 July low, sitting inside a three-week downtrend that hasn't actually turned.

Relative strength didn't confirm the extension either. Fourteen-day sits at 54.95, having shed 2.43 across the session. Here is the number worth carrying: the price that puts that reading at a neutral 50 is 29,076. Another 539 points of decline from the settle still would not make this technically oversold. That is a great deal of room, and it argues against expecting any quick bounce should 29,400 give way.

Where the selling actually showed up

This is the most directionally useful item in the whole review, and it requires holding two numbers side by side.

Broad-index hedging flow finished Wednesday at roughly positive 2 billion dollars of customer delta, made up of about 5 billion in put selling against 3 billion in call selling, overwhelmingly same-day expiry. That is volatility sellers harvesting the fat premium, and it mechanically damps intraday movement.

Technology-specific flow finished at roughly negative 3 billion, driven by longer-dated call selling.

Those aren't the same thing wearing different labels. Short-dated premium selling stabilises a session. Longer-dated call supply is a directional statement about upside conviction over weeks, not hours. When one index is being held steady by same-day flow while the other is being sold further out the curve, the second one is where the exposure sits.

Skew in the Nasdaq proxy prints right at the 100th percentile of where it has traded. Extended, rejected at a positioning wall, with the call side crowded is not a mix that resolves upward unless something fresh turns up.

Options are underpricing the movement

Average true range over fourteen days is 736.71 points. The options-implied one-day move works out near 435.

The proxy carries an implied move of 10.52 dollars against a 716.86 close, about 1.468 per cent. One-month realised volatility in that proxy reads 26.14 per cent against implied of 23.04. Both comparisons say the same thing from opposite ends: options cost less than this contract has been delivering.

Note also that nine-day range readings sit above the fourteen and twenty-day on both measures. Realised volatility expanded through the recovery instead of compressing into it, unusual behaviour for a rally, and consistent with how much of this move was short covering. When short-window volatility rises and price gets turned away at the upper edge of its range, overnight gaps tend to widen.

How to trade it

Fade rallies into the 29,740 through 29,950 band, scaling rather than committing at one price. That band runs from the computed pivot up to the first computed resistance, and the 50-day underneath it serving as the failed-reclaim marker.

Don't chase a short under 29,700. Should the open come in weak and never climb back into that zone, sit out and take the next session instead.

Stop 30,090, above Wednesday's high and beneath the one-month high at 30,094, so invalidation lands beyond the swing high instead of at some arbitrary distance. From a 29,850 average that is 240 points. Take a third at 29,406 where the first computed support meets the 40-day zone, a third at 29,197 where the 50 per cent retracement at 29,151 and a 29,000 cash strike equivalent of 29,127 thicken the area into a band, and run the last third toward 28,895 only if the morning trended. Roughly 1.85, 2.7 and 4.0 to one.

A half-hour close above 30,094 ends the structure. Treat it as an instruction to go flat and reassess, not to reverse. A reclaim of 29,740 that holds an hour with semiconductors leading higher weakens the setup materially even without touching the stop, and the long side then lives between 30,094 and 30,130, or wherever a retest of 30,060 manages to hold, stop 29,930, targeting 30,283 then 30,335 then 30,492.

Why Thursday is a positioning day

Nothing on the calendar's first-order for this index. Claims at 08:30 matter more than usual after Wednesday's payroll miss at 44,000 and a services employment sub-index in contraction at 47.4.

The line to actually watch is unit labour costs, forecast 2.1 per cent against 1.8 prior. Services prices paid on Wednesday accelerated to 70.3 where 65 was the forecast, and an upside surprise on labour costs extends precisely the rate repricing that pushed money out of long-duration technology to begin with. For an index carrying this much duration, that is the shortest path from a data point to the price.

Otherwise this is positioning into Friday's payrolls, and the 21 August expiry is the structural anchor capable of pinning everything if no catalyst turns up.

The five-session surge is real. So is the three-week downtrend it is sitting inside. Wednesday was the first session where the second one mattered more.

The broad index gave back a record on the same rotation, and it has the cushion this one lacks: ES / S&P 500: a record at 10am, gone by the bell.The complete data picture

Every number behind Friday’s plan, charted first; the full numeric reference follows underneath.