Oil crashed, the Dow rose, and the chips broke anyway
Brent fell 8.7% and the Dow gained 0.5% on Monday. The main US chip ETF still closed below a floor that had held for five sessions.
The risk read for this session
These are model estimates, not forecasts anyone can guarantee, and not investment advice. The dot before each probability is its band: 🟢 low · 🟡 moderate · 🟠 high · 🔴 very high, with thresholds set separately for each scenario. SOXX is the main US semiconductor ETF, the basket I track; from peak is the fall from the highest close of this cycle; VIX is the market’s fear gauge; the high-yield spread is the extra interest riskier companies pay to borrow, my early-warning line for funding stress. My internal alert is at Level 3, on a scale where 3 is the highest.
📌 The one-line takeaway
Monday, July 27, handed the market the best macro news it has had in a month — and the chip complex broke its floor anyway. That combination is the whole story.
Today I read 🟠 ALERT, and my internal alert stays at LEVEL 3, the highest, for an 8th day — a level only steps down after a full recovery checklist, and none of it is met.
📰 What happened
Over the weekend the United States and Iran paused their attacks on each other. Oil, which had been the market's biggest outside worry for two weeks, collapsed: Brent crude fell 8.7% to $88.36 a barrel, its lowest in over a week, after topping $100 only days earlier. Cheaper oil means less inflation pressure, which means less pressure on interest rates. The broad market took the gift: the Dow Jones Industrial Average rose 0.51% to 52,210.08, and the S&P 500 (the broad US market index) added 0.02% to 7,413.18.
Semiconductors went the other way. SOXX — the main US semiconductor ETF, a basket that holds the big chip stocks and the benchmark I track — fell 2.05% to 516.23, touching 500.00 at its worst point before recovering part of the drop. Nvidia fell 4.99% to 196.51. AMD fell 5.17%, ASML 5.80%, SanDisk 11.02%.
Across my full watchlist of 52 US-listed AI and chip names, the average stock fell 1.28% and only 15 of them rose.
Why this one matters more than the others
A 2% down day is not, by itself, remarkable. Where it closed is.
For five sessions SOXX had held a floor at 521.81 — the lowest close of this decline, set on July 17. It had also clawed its way back above the line 20% below its June 22 peak, which sits at 524.76, after closing just under it on July 17 and 20. On Monday it closed at 516.23, below both. That puts the drawdown (the fall from the highest point) at 21.30%, and it means the 20% decline I track as a scenario is realized again on a closing basis.
One honest detail cuts the other way, and I would rather give it to you than leave it out: the intraday low from July 17, at 498.54, was not broken. Monday's worst price was 500.00. So the closing floor gave way while the intraday floor held. That is a real distinction, and it is why I treat this as a confirmed deterioration rather than a capitulation.
Two more things changed underneath. Nvidia — the flagship, and the stock whose behavior tends to set the tone for everything else — lost its last bullish technical signal (its MACD, a momentum measure built from two moving averages, crossed downward), and now trades just 1.84% above its 200-day moving average, the long-term trend line that separates a correction from a broken trend. And Monday's selling came on volume 1.30 times the 50-day average. A floor breaking on heavy trading is a different event from a floor breaking on a quiet day.
What actually drove it — and neither reason is about the economy
One: who pays
Late Friday in San Francisco, Nvidia and South Korea's SK Group announced a multi-year agreement covering memory chips and more than 2 gigawatts of new data centers in Korea, valued by the parties at over $500 billion. Separately, Bloomberg reported that Nvidia is considering — considering, not deciding — a guarantee of up to $250 billion to help OpenAI lease a 10-gigawatt data center hub in Ohio being built by a SoftBank subsidiary. In total the financial press now counts more than $750 billion of new or potential Nvidia commitments.
The market did not read this as a demand story. It read it as vendor financing on an unprecedented scale: the supplier underwriting the spending power of the customer who buys its products. The clearest evidence is the counterparty. SK Hynix — the company on the receiving end of the $500 billion order — fell 7.47% in its US listing. When the firm collecting the money falls alongside the firm writing the check, the market is pricing the structure, not the sales.
Two: the moat
The Information reported — a single-source exclusive, and I label it as such — that a state-backed company in Shanghai has begun mass-producing homegrown immersion DUV lithography machines: five this year, twenty next, for Chinese chipmakers. DUV machines are the tools that print circuits onto silicon. China was cut off from the newest generation by export controls, which left older DUV tools as ASML's last big Chinese market. If China can now build its own, that market shrinks. ASML fell 5.80% and dragged the whole equipment group down 4.32%.
The rotation nobody is naming
Look at the three most recent sessions in sequence. On July 23 the market sold the companies that pay for AI — the cloud giants, punished for rising capital spending. On July 24 it sold the companies that build it — the chipmakers. On Monday it did the opposite of both: the cloud giants rose 1.11% on average (Oracle +4.27%, Alphabet +2.13%, Microsoft +1.94%) while memory (−5.81%) and equipment (−4.32%) were sold.
Three sessions, three different victims, same net result: money leaving the AI trade in an orderly way. That is the signature of a rotation, not of one bad headline.
🚦 The signal board — my 9 gauges, in 6 families
| Family | Gauge | Reading |
|---|---|---|
| A · Price trend | Moving averages | AlertSOXX closed below its 50-day average for a 9th session, now 9.27% under it, on above-average volume. Nvidia is 5.79% under its own, and only 1.84% above its 200-day line |
| A · Price trend | Structure | CriticalThe floor broke. SOXX closed at 516.23, below the 521.81 pivot, with the pattern of lower highs since June still intact. This is the day's one escalation |
| A · Price trend | Momentum | AlertSOXX momentum deteriorating on both the daily and weekly clock; Nvidia's daily momentum crossed bearish for the first time in this leg |
| B · Participation | Breadth | AlertOnly 15 of 52 names rose (28.8%). But equal-weighted indexes beat cap-weighted ones for a third day: the broad market's participation is healthy — it is AI that is thinning out |
| B · Participation | Rotation and canaries | AlertMoney left AI on the best macro day of the month. My five "canary" stocks, though, do not confirm: as a basket they are 4.98 points ahead of SOXX over 20 sessions, and none is at a 52-week low |
| C · Volume | Distribution | AlertA new distribution day — a down session on rising volume, the fingerprint of institutional selling. That makes 12 active. Not red, because the break came on volume that was heavy but short of the capitulation this signal looks for |
| D · Volatility | Fear gauge | CalmSee my note below: I could not verify Monday's VIX close to my standard, so I am publishing no number. My API-based proxy nudged up slightly. Every candidate reading was below 20, so the gauge stays green either way |
| E · Credit | The debt tap | CalmThe high-yield spread — the extra interest riskier companies pay to borrow, my early-warning line for funding stress — sat at 268 basis points as of July 22. Credit has not moved |
| F · Froth | The over-inflated balloon | AlertSOXX is 29.47% above its 200-day average, down from 40.64% on July 22. The balloon keeps deflating. This gauge sits outside the count and only flags fragility |
Putting it together: Family A is now compromised (one red), Families B and C are under tension, D and E are calm. That is three families under tension and one compromised — enough to confirm ALERT. I do not read it as CRITICAL, which in my framework needs four families under tension, two families compromised, or the price-trend family breaking down at its deepest level: SOXX closing below its 200-day average, or a "death cross." SOXX is 29.47% above that average. This is the same reading I applied on July 21, the last time this gauge turned red, and I flag it as a judgment call rather than a mechanical one.
📊 Today's odds
I publish three probabilities every day, with the same definitions, so you can hold me to them over time. Each carries a band: 🟢 low · 🟡 moderate · 🟠 high · 🔴 very high, with the thresholds set separately for each scenario.
🔴 A 20% sector drop within 30 days — REALIZED. SOXX is 21.30% below its June peak. This is no longer a forecast; it happened again on Monday's close.
🔴 A 30% sector drop within 90 days — 32%, up one point from my previous read of 31%. What pushed it up: the floor breaking on above-average volume, Nvidia losing its last bullish signal, and the circular-financing question moving from background worry to price driver. What held it down: oil collapsing, the broad market refusing to follow, credit not moving, and a wide cushion — SOXX would need to fall another 11.05% to get there.
🟡 A 20% fall in the broad US market within 90 days, with AI at the center — 8%, unchanged. The S&P 500 is only 2.58% below its record. Falling oil and calm credit offset a deepening decline in AI.
These are model estimates. Short horizons are dominated by noise, and this sector has invalidated confident numbers in a single session before.
🛡️ Where this leaves me
My internal alert stays at LEVEL 3 for an 8th session, and the recovery checklist scores 0 out of 4: a new distribution day, nine sessions below the 50-day average, no valid follow-through rally, and breadth still far from the threshold. Nothing has begun to mend.
The level I now watch is 498.54 — the intraday low from July 17, the floor that held on Monday when the closing floor did not. And I watch Nvidia's 200-day line at 192.96, 1.84% below Monday's close.
Two events land immediately: the Federal Reserve's policy meeting on July 28-29, where futures markets put roughly 35% odds on a rate increase — a single-source reading I treat as an order of magnitude, not a precise number — and earnings from Microsoft, Meta and Amazon on July 29-30, where capital spending guidance is the number that has been moving these stocks for a week.
One postscript, from after Monday's US close and outside the numbers above: on Tuesday in Asia, SK Hynix fell 14.65% and Samsung Electronics more than 13%.
📚 Sources, and a note on one missing number
Prices, volumes, moving averages and momentum come from market data interfaces (Massive/Polygon), computed on Monday's close. Index levels come from Associated Press and Yahoo Finance wire copy, cross-checked against the SPY ETF, which moved +0.02% in line with the S&P 500. Oil prices are Reuters. The Nvidia-SK agreement is CNBC, July 25, reported as announced. The $250 billion guarantee is Bloomberg via Yahoo Finance, reported as under consideration — I keep that distinction because it matters. The lithography report is The Information, a single-source exclusive. Treasury yields (10-year 4.69%, 2-year 4.33%) are as of July 24, the latest published.
The missing number: I normally publish the VIX, the market's fear gauge. For Monday I could not verify it to my standard. One source reported 19.05, apparently an intraday reading; two others reported 18.58 in articles dated July 27 that actually describe Friday's session — a common trap, and 18.58 is exactly Friday's close. With no majority among independent sources, I publish nothing rather than a number I cannot stand behind. Both candidate readings were below 20, so the volatility gauge reads green either way.
My full method, including how every gauge and threshold is defined, is public in my methodology, and every previous daily read is in the archive.