The short version
The spark was core CPI. It came in at 0.3% against 0.2% expected, and the market priced a more aggressive Fed within minutes. Odds of an October 28 hike roughly doubled, from about 1 in 5 to about 1 in 3.
The damage was narrow and deep. The Nasdaq lost 1.9% and tech lost 3%, while the Dow finished nearly flat on a 2%+ bid in pharma and staples. That's a defensive rotation, and it kept the Dow standing without lifting it.
The tell was credit. CoreWeave dropped about 15% after a $3B convertible and a new share-sale program, with its unsecured bonds touching a 15% yield. When the market charges AI infrastructure junk-bond rates, the equity story has to pay the difference.
The tape was a midway game. I counted 13 valid bullish pivots in NQ, and all 13 were undercut. Overnight kept the streak alive. A late-Asia turn flipped every cross-asset driver risk-on for about an hour, then rolled over to new lows. By midnight the S&P and the Dow had broken their own overnight floors, and the hour-long cross-asset read was unanimous risk-off. Thursday's 8:30 data and TSMC's results now decide how deep this goes.
What actually happened Wednesday
The print.
September headline CPI rose 0.6% m/m, as expected.
The annual rate came in at 3.7% against 3.6% expected.
Core, the number the Fed watches, rose 0.3% against 0.2%. Shelter alone added roughly 0.15 point, with airfares and used cars chipping in.
That's the second straight 0.3% core month, so this wasn't a one-off.
The rates reaction.
The 10-year yield spiked to roughly 5.33% after the print, back near the highest levels since 2002, then reversed to finish around 5.20%.
The 2-year rose about 10 basis points intraday.
Fed funds sit at 3.75% to 4.00% after September's hike, and the market now leans toward another hike by December.
The scoreboard at the bell.
Nasdaq Composite: -1.88%
Russell 2000: -1.93%
S&P 500: -1.11%
Dow: -0.09%
VIX: +10% to 16.82
The rotation.
Technology fell 3%. Health care rose 2.4% and staples 2.2%.
The big pharma names (AstraZeneca, Merck, Bristol-Myers, AbbVie, Lilly, Pfizer) all rose about 4%.
Money was leaving the AI trade for shelter, not spreading across the Dow. Banks didn't help either: Bank of America fell 5% on a profit drop and a $1.2B+ anti-money-laundering fine.
The credit tell. This was the most important story of the day.
CoreWeave announced a $3.0B convertible note due 2033 plus an at-the-market program for up to 35 million new shares. The stock fell about 15%.
Its unsecured bonds touched a 15% yield, and a public argument broke out over whether AI buildout financing is turning into a private-credit bubble.
Nebius fell about 10% and Oracle about 9%. Dell fell 8% on a Barclays downgrade that questioned enterprise AI growth.
The AI demand story is intact. ASML said it can't fully meet customer demand in 2027 and 2028. What's being repriced is the financing and the valuation, and rising rates make both harder.
Thirteen Trapdoors
Every carnival has a booth where the prize looks easy and the floor drops out. Wednesday's NQ tape built thirteen of them. I counted 13 valid bullish pivots on the way down, and every one was undercut. I know better and still stepped on a couple.
The rule that would have saved me: after any short-covering rally, I want follow-on bull volume, and the reclaim levels have to hold with time. They have to act as support or build a durable shelf. Otherwise sellers press on any weakness.
The evening session showed every way that rule gets broken. The failed bounces came in four flavors:
No lift. Divergence prints at the low, nothing follows, and the low breaks within about five minutes. The strongest-looking divergence of the night failed this way, so follow-through matters more than how pretty the signal looks.
Capped bounce. The rally stalls under a level that had already rejected price, making a lower high. Undercut in five to fifteen minutes.
Rejected at the top. A strong pop, met at its high by heavy selling and a short-term divergence the other way. Undercut in about six minutes.
Reclaim, then rollover. Price reclaims a round number, builds a box that looks like a base, survives a 90-minute close, then rolls over half an hour later. This is the most expensive trapdoor because it looks like a shelf.
What every trapdoor had in common: each bounce made its lower high beneath a level that had already turned price away earlier in the session. The ladder of lower highs ran 30,152, then 30,105, 30,067, 30,021 and 29,973.
The one bounce that earned it.
The 19:53 low at 29,880 held for almost half an hour before anyone got paid. Every earlier low had died in minutes.
It built higher lows, and it launched at the 20:00 Asia desk open.
The trigger came from the S&P: ES swept Wednesday afternoon's NY PM low (about 7,741) carrying a weekly-scale divergence against the Dow, while NQ refused to make a new low.
The leader that had been dragging everyone down stopped falling, and the follower did the last flush.
That bounce ran about 125 points to 30,010. Even it couldn't clear the old lower-high ladder.
Late Asia added trapdoor fourteen.
Price swept the session low to about 29,866 at 21:47, then reclaimed it into the 22:00 close.
Within ten minutes every cross-asset driver flipped the same way: bonds sold off, yields rose, the yen weakened, and ES, the Dow and Nikkei all ticked up. It was the first unanimous risk-on read since the reopen.
It lasted about an hour. NQ undercut to 29,825.50 around 23:30, ES broke its overnight low (7,740.50) to about 7,732, and the Dow fell out of the bottom of its box.
The tell came from the slow read. Over a 60-minute window, every driver read risk-off: bonds bid, yields lower, the yen firmer, Nikkei down more than 3%, and ES and the Dow both lower. That held even while the 15-minute read bounced. A fast, unanimous bounce means little when the hour-long read disagrees.
By the rule above, it was a stall that never built its shelf.
One more lesson for the notebook: the Dow sat in a 62-point box most of the night while NQ and Nikkei bled. When one index decouples like that, its divergence against the others prints at every low and means nothing. The divergence that mattered came from the index moving with the leader: the S&P. And when the shelter itself finally gave way near midnight, the Dow breaking its box said more than any of its divergences had.
The world while you slept
Rates.
The long end got relief. The 10-year eased from about 5.33% at the post-CPI spike to about 5.14% at the Asia open, and near 5.09% by midnight. Bonds caught a steady bid all night.
The policy path got harder. Fed funds futures moved October hike odds from about 22% to about 32% overnight; intraday estimates ran as high as 46% to 50%. Most of the market now expects a December hike.
That's relief in yields, driven by people hiding in Treasuries. A dovish shift would look different. It only helps the bulls if yields keep easing while stocks firm up.
Oil.
Brent held $106 to $109 and WTI about $95 to $96, with WTI printing weekly highs in European trade.
The headlines run the wrong way for a relief rally. Houthi strikes hit an Aramco building in Dhahran, and a missile was fired at Riyadh.
Iran's Revolutionary Guards say a container ship sank after hitting a mine south of Hormuz.
Iranian officials told Reuters they're using new tactics at sea to keep prices high into the US midterms.
No ceasefire or peace headline showed up on Wednesday's wires.
The yen.
USDJPY slid from about 158.9 in the US afternoon to about 157.6 to 157.9 overnight.
A firmer yen alongside a falling Nasdaq is the carry trade flinching.
For context, Tokyo and Washington intervened together near 164 at the end of July. Today's levels are nowhere near that line.
Asia.
Korea's KOSPI fell 5.6%, its sixth straight decline, led by memory chips.
The Nikkei fell 1.3% early, with one report of a 2,500-point intraday drop, and Nikkei futures were down more than 3% from Wednesday evening by midnight.
Australia surprised with a 39,000-job loss.
The AI unwind is global. Tokyo and Seoul also sold ASML's strong guidance as a sell-the-news event.
Europe.
European shares edged up on ASML's outlook, which says AI demand runs into 2029 and 2030.
That split matters: the buildout story is intact, but the price of financing it went up this week.
Catalyst check: real or wishful?
The bulls have a wish list. Here's how each item looks as of publication.
Oil rolling over: not yet. Brent sits near $107, and the Gulf headlines are escalating, not easing. If this changes, it changes on a headline, and those cut both ways.
Rate relief, SOFR included: half real. Long-end yields eased overnight and front-end rate futures caught a small bid. But the Fed path got more hawkish, not less. Until yields fall while equities rally, the relief comes from fear.
Peace deals: nothing on the wire. No ceasefire or deal headline Wednesday, from either the Gulf or Ukraine. A surprise here would be the biggest upside shock on the board, which is exactly why nobody can plan on it.
Mid-month flows: plausible, unproven. Monthly options expire Friday, October 16, and mid-month money does arrive. My guess is the discretionary funds will be smart about deploying it and won't chase a falling knife at the open.
The catalysts that actually land in the next 36 hours:
TSMC earnings, Thursday before the US open. After KOSPI's 5.6% drop and six straight days of chip selling, this is the cleanest test of whether the AI unwind is about demand or about financing.
8:30 ET Thursday: September retail sales (consensus +0.3%), PPI (consensus +0.5%, core +0.3%), jobless claims and the Philly Fed survey. After a hot core CPI, a hot core PPI would push October hike odds higher still.
Friday: monthly opex and industrial production.
Next week: Netflix, Tesla, Meta, Lam Research, Texas Instruments and IBM.
October 27–28: the FOMC.
Levels and the scenario map
These are December futures levels as of publication. Judge them on closes, not wicks. A poke through a level that gets reclaimed by the next 30-minute close is a sweep, and Wednesday night had plenty.
NQ above:
29,866 to 29,884: the old overnight floor, now overhead
29,940 to 29,973: the last rungs of the lower-high ladder
30,010: the best bounce of the night
30,020 to 30,037: the double top from 18:50 and 19:03, where sellers stepped in twice
30,067, then 30,105
30,152: the reopen rejection
About 30,183: Wednesday's settle
NQ below:
29,825.50: the new overnight low, set around 23:30
29,669 to 29,691: the next shelf
The September accumulation zone they already distributed into. Look at mid-September on your own daily chart; that's my next magnet if the floor gives.
Cash Nasdaq 100 references from last week's analysis: 28,900, then the June low near 28,300. Futures trade above cash, so adjust.
ES:
About 7,732: the new overnight low
7,740.5 to 7,741: the old Asia low and the swept NY PM low, now resistance
7,747: the first higher low of the failed turn
7,755 to 7,762: supply
7,760.25: last week's low, now overhead
7,769.5: the Asia high
YM: the 51,838 to 51,905 box held almost all night, then broke near midnight (about 51,802). The shelter gave way, and that's a signal.
Base case (50%): trapdoor fourteen stays open
All three indices have broken their overnight floors, and the hour-long cross-asset read is unanimous risk-off.
Rallies into 29,866 to 29,884 get sold unless they bring follow-on volume and hold a shelf across a 30-minute close.
A hot PPI, a TSMC disappointment, or an oil spike on Gulf headlines adds fuel.
Downside path: 29,825.50, then the 29,669 to 29,691 shelf, then the September accumulation zone.
Alt case (30%): flush and revert
NQ keeps refusing to follow ES lower. As of midnight NQ sat above its 23:30 low while ES sat on its own. That divergence is the one thing still arguing for a turn.
The trigger is a soft core PPI (0.2% or lower) or a reassuring TSMC report, with bonds and the yen turning risk-on on the slow read and not just the fast one.
NQ reclaims 29,884, then 29,973, on 30-minute closes with higher lows underneath.
Tail case (20%): the squeeze
Soft PPI, a TSMC beat, and a peace or oil-relief headline land together.
Shorts pressed into a three-index breakdown get caught. NQ runs through 30,010 to 30,067 and toward the 30,152 reopen rejection.
Wednesday's settle near 30,183 is the stretch target. Squeezes from a crowded short can run further than anyone expects.
How to tell which one you're in: watch which index makes the next low, and check the slow cross-asset read before trusting any bounce. If NQ breaks 29,825.50 together with ES, with no divergence and bonds and the yen still risk-off, step aside and let the trapdoor work. If NQ keeps refusing while ES and the Dow flush, and the hour-long read starts turning, take the revert seriously.
Rules of the midway
Proof before prizes. After a short-covering pop, wait for follow-on bull volume. The pop on its own is the barker talking.
Count the clock. A reclaimed level has to hold through time and build a shelf. Wednesday's failed lows died in minutes. The one that worked survived half an hour first.
Judge on closes. A wick through a level that gets reclaimed by the close is a sweep, not a break. Three sweeps-and-reclaims landed right into 30- and 90-minute closes Wednesday night.
Watch the follower, ignore the stray. The divergence that matters comes from the index moving with the leader. When one index decouples into a quiet box, as the Dow did, its divergence prints everywhere and means nothing, until the box itself breaks.
Check the plumbing, on two clocks. Bonds, the yen and front-end rates told the truth all evening. A bounce they don't confirm on the slow read is a trapdoor with fresh paint.
Stay nimble. Bears lose steam fast once they stall, and late Asia stalled for an hour before rolling over. I still prefer we go lower, and the slow read agrees. Respect both.
The calendar (ET)
Thursday, before the US open: TSMC Q3 earnings
Thursday, 8:30: retail sales, PPI, jobless claims, Philly Fed
Thursday: PNC, U.S. Bancorp and JB Hunt earnings
Friday, Oct 16: monthly options expiration, industrial production, Travelers and Truist earnings
Tue–Wed, Oct 20–21: Netflix, Tesla, Meta, Lam Research, Texas Instruments, IBM
Tue–Wed, Oct 27–28: FOMC
Sources
Reuters, Bloomberg, CNBC, TheStreet, Charles Schwab, Babypips, Fortune, Business Standard, Nikkei Asia, NewsOnJapan, BigGo Finance, 24/7 Wall St., MarketBeat, GuruFocus, StocksToTrade, Admiral Markets, Forex.com, Just Security. Market levels are live market info as of publication time.
Got a question about a level, a scenario, or why I'm ignoring the Dow's divergences? Drop it in the replies. I read them all.
This is market commentary and education, not financial advice. Futures carry real risk of loss. Size accordingly, and never trade money you'd miss.
I use AI for research and to get more done. The bad calls and the typos are all mine.
Buy my bots a bourbon. 🥃