Something Weird Is Happening Beneath the Stock Market’s Surface
Plus, big tariffs are back
By Spencer Jakab, The Wall Street Journal, July 21, 2026
…
” The Cboe Volatility Index, or VIX, uses options prices to get a rough measure of how much choppiness investors expect in the S&P 500 index in the coming 30 days. At 17.5, it’s below the index’s long-term average.
But look at how much traders are paying to hedge against big moves up or down in individual stocks and a totally different picture emerges. That can be seen from a related measure, VIXEQ, which is above 50.
It’s derived from a similar calculation for each stock in the index, weighted by market value. Individual stock volatility is always higher, but the gap between the two just hit a record… [end quote]
I never heard of VIXEQ before. Since both VIX and VIXEQ are calculated from the same stocks in the SPX that are market weighted, how are they different?
VIX is calculated from the SPX index as a whole. If some stocks go up when others go down they cancel each other out so VIX will be low. VIX only moves when the entire market moves together, such as during periods of uncertainty. The last time VIX spiked over 40 was on “Freedom Day” when President Trump announced high tariffs. Note that VIX did not reach 40 during the dot-com bubble collapse in 2000.
VIXEQ is more volatile than VIX because the movements of the stocks are separated. The ones that go up do NOT cancel the ones that go down. Over its multi-year history, VIXEQ has primarily traded in a range from the upper 40s to the low 50s. Traders often chart the VIXEQ minus VIX spread to gauge “dispersion” or market regime changes. A widening spread indicates mega-cap concentration or stress hiding beneath the index, while a narrowing spread points to balanced market health.
This seems like a valuable way to track the progress of the game of musical chairs. Right now, VIXEQ is rising while VIX is low and stable. VIXEQ minus VIX is trending upward, breaking out of its channel.
This indicates that volatility in individual market-weighted stocks is rising (VIXEQ) while the overall market (VIX) is stable. The money isn’t fleeing. It’s rotating from some stocks into other stocks. SPX has plateaued and is no longer rising.
The mega-caps are most likely to swing the VIXEQ and they are also most overpriced. During a market run-up, VIXEQ would be rising because the popular stocks are rising. But when SPX is not rising but VIXEQ is rising it shows that a significant number of stocks are changing at once. Some important stocks will be falling while others are rising.
When VIXEQ and VIX rise together – especially if SPX drops at the same time – watch out! That’s the sign of a rout. If it’s also combined with a spike in the Financial Stress Index it’s a sure sign of a financial crisis. This combination is rare and has only happened twice since 1994 – in 2008 and 2020. Not during the dot-com bubble burst in 2000.
I have added this to the Control Panel.
I asked Gemini what news items would signal a leading indicator that the AI bubble will collapse.
[from Gemini]
News that software vendors are spending $2.00 on cloud compute for every $1.00 of AI subscription revenue. When AI features degrade gross margins rather than expanding them, enterprise adoption halts.
Headlines that major tech giants or cloud vendors are renting out excess GPU server capacity at steep discounts. This signals that supply has officially outstripped real end-user demand.
Tech firms currently depreciate servers over 5 to 6 years. News that companies are forced to shorten AI server depreciation down to 3 years due to rapid obsolescence will instantly erase tens of billions in GAAP net income.
Spikes in yields for high-yield data center project bonds or private credit facilities backing power and compute infrastructure.
How the Unwind Unfolds
-
Phase 1 (Upstream Order Cuts): ASML/TSMC report order cancellations or shrinking backlogs.
-
Phase 2 (Capex Plateau): Hyperscalers announce lower-than-expected 12-month Capex guidance.
-
Phase 3 (Margin Compression): Software companies reveal shrinking gross margins due to inference cost overruns.
-
Phase 4 (Broad Market Contagion): $VIXEQ$ remains elevated while $VIX$ spikes above 35, compressing the spread as index-level forced selling begins. [end Gemini quote]
“Inference” means end-users actually using AI and “spending tokens.” That’s how AI makes money. It doesn’t count the AI internal ecosystem spending that boosts earnings reports. If end-users find that spending on AI doesn’t result in real-world improved profits they will stop using AI or cut back to only the most impactful operations.
And don’t forget “Distillation” where a super-expensive AI trains a much smaller AI student which can do 90% of the customer’s needs for a tiny fraction of the cost. And if it can be run on the customer’s own servers there will be ZERO token costs.
Why buy the cow when you can get the milk for free?
The AI hyperscalers are trying to prevent this. But Meta and the Chinese are releasing open-source AI to enable distillation.
That would cut the AI business case off at the ankles.
This is a call to all METARs to watch the leading indicators of the AI bubble collapse.
Wendy