Tech Was Just De-Grossed Into The Biggest AI Capex Boom Ever
Upside pain
Tech has been aggressively de-grossed, NDX is bouncing from support and Nvidia has once again reinforced the structural AI capex story. With QQQ vol cheap, software breaking out and the calendar becoming increasingly supportive, the ingredients for renewed upside pain are falling into place.
Jackson Hole remains the immediate hurdle, but absent a genuine hawkish surprise, the path of least resistance may still be higher.
NDX
NDX futures are bouncing off the 100-day post-NVDA. We remain inside the broad range that has been in place for months, with the recent pullback essentially testing the lower end of it.
It wouldn't take much from here to squeeze back toward the upper part of the range and create some renewed upside pain. More here.
Source: LSEG Workspace
The tech sell
Info Tech was the most net-sold global sector last week, while also seeing its largest percentage de-grossing in two years.
Source: GS
Unstoppable
The most important line from Nvidia’s earnings call came from CFO Colette Kress:
“CapEx by the top five hyperscalers is expected to reach $800 billion in 2026 and $1.3 trillion in 2027.”
Consensus had $1.3 trillion penciled in for 2028. She just pulled it forward by a year. Full read here.
Source: Marlin
QQQ vol
QQQ implieds look rather cheap here.
Source: GS
Dead
SPX is barely moving. The 10-day average high/low band has collapsed to 0.52%, virtually matching the lowest level since 2021. Intraday movement is getting extremely compressed.
Source: GS
From here
Privorotsky’s modal view is that Jackson Hole is ultimately a push to nothing. Warsh could strike a hawkish tone, emphasizing the 2% target while giving little weight to softer inflation and labor data, but the broader setup remains constructive.
Several key risks are resolving better than feared: Nvidia demand remains strong, Chinese AI earnings point to accelerating inference usage, software is showing signs of monetizing AI, Hormuz is improving and the Fed is unlikely to deliberately crush growth. Add Monday’s UK bank holiday, the drift into Labor Day the following week and potentially supportive MSCI month-end flows, and in the absence of genuinely new negative information, the preference remains to grind higher.
The key tells now are how Nvidia trades, whether semis finally follow, and continued progress in oil/Hormuz.
Source: GS
King IGV
Software is breaking above the late-May mania highs and the upper end of the big range. The next resistance sits at $110. A close above that level would bring the ATHs into play.
This is getting close to the inverse setup we highlighted on April 14 (here), but we wouldn't dare fade this momentum just yet. Also worth noting is the 50/200-day golden cross.
Source: LSEG Workspace
Software vs SOX
The IGV/SOX ratio is breaking above the massive long-term trend line. This could easily squeeze further, especially with most still positioned the other way around.
Source: LSEG Workspace
Still depressed
Semis and software have long traded as photographic negatives of each other. That inverse relationship went into overdrive during the second half of last year as the market aggressively pressed the AI enablers-versus-disrupted narrative.
The fever has broken in recent months, partly due to positioning, including the July deleveraging, and partly because the fundamental debate around software has become more two-sided. Software nevertheless remains deeply depressed relative to semis. The wedge may retain some structural width, but with so much capital concentrated in the trade, expect plenty of ebb and flow along the way. More here.
Source: GS/Tony P
Midterm
Midterm years have historically chopped into the vote and rallied after it. Deutsche Bank's study of 20 postwar cycles shows the index has never been lower nine months after the election.
The shorter window tells a similar story.
Buying in November of a midterm year and holding through the following April has produced positive returns in every cycle since 1942, at an average of roughly 16%.
The dispersion around that average is enormous, so this is a tendency, not a guarantee.
But the historical pattern is difficult to ignore:
uncertainty has tended to peak before the vote and fade once the result is known. More on midterms here.
