By Maksym Misichenko · ZeroHedge ·
What AI agents think about this news
The panel agrees that the recent market correction was primarily driven by a mechanical unwind of leveraged momentum trades and a spike in long-end rates, but they disagree on the extent to which fundamentals are at play. While some panelists see this as a rotation rather than a breakdown, others warn of systemic fragilities and potential tail risks.
Risk: A sustained capex slowdown, margin compression, or delayed AI payoff could keep high discount rates intact and force multiple contraction even if AWS/Azure beat.
Opportunity: Investors should treat current rallies as exit liquidity for over-leveraged tech positions, as the 're-test' of support levels is statistically probable.
This analysis is generated by the StockScreener pipeline — four leading LLMs (Claude, GPT, Gemini, Grok) receive identical prompts with built-in anti-hallucination guards. Read methodology →
Authored by Lance Roberts via RealInvestmentAdvice.com,
What a week that was...
Despite a hopeful bounce to end the month, it was a bloodbath for most assets. It was the Nasdaq’s worst July in 22 years, bonds’ biggest July yield spike since 2005, and oil’s biggest July jump in over 30 years.
Leaving investors with one big trillion-dollar question: is the momentum crash over?
As I discussed on Thursday on the ** Real Investment Show**, the average retail investor portfolio is likely faring far worse than the broad market index. The momentum crash we just lived through was the fastest on record. It ended last Thursday with a $45 billion hedge fund handing its entire public equity book to Citadel in a single block trade.
None of it should have been a surprise. On June 22, in ** The Technical Backdrop: When Flows Meet a Hawkish Fed**, I wrote that a market running on flows, leverage, and shrinking leadership could melt up into July. It could also reverse hard the moment those mechanical buyers turned into sellers.
The close of that piece was blunt,
and was published on the exact day momentum peaked.It is also named the mechanism.
“
“Lastly, watch the long end of the curve.If Warsh’s signal keeps the ten-year climbing, the most expensive, most crowded, most rate-sensitive corner of this market, the same one soaking up forty cents of every dollar, is the corner that pays for it first.
The most crowded corner of this market, the one soaking up forty cents of every S&P 500 dollar, would pay for rising yields first. That is precisely what happened. Two weeks ago, ** Momentum Meltdown Catches Traders By Surprise** flagged the same divergence in miniature. Last week,
walked through the $800 billion megacap air pocket. The only question left is whether the correction is finished or whether this was the first act.
The AI Capex Bill Comes DueMomentum Crashed. The Average Stock Did Not.
Start with the magnitude, because the numbers are without precedent. Morgan Stanley’s sector-neutral momentum index fell 17.4% over four sessions, the worst four-day stretch in the history of the series. The comparable declines were roughly 11% after the dot-com peak and again in the 2022 inflation bear, and 14% after the Covid crash. The technology and media slice of that basket dropped 36% in four days, against a prior record near 20% set in the 1999 to 2001 unwind.
You can see the same thing in instruments you can actually trade. The iShares Momentum ETF fell 18.0% from its June 22 peak to its July 29 low, and semiconductors, measured by SOXX, surrendered 29.0% over those same twenty-five sessions. Momentum broke. The equal-weight S&P 500 closed at a record high on July 28. Right in the middle of the wreckage.
While for many retail investors, it may “feel” like a market crash, it wasn’t. It was a rotation, and leveraged traders were liquidated.
None of that is new. In ** More Market Wisdom: Jesse Livermore, Part 2**, we walked through how leadership rotates across cycles. The Nifty Fifty became the laggards of the late 1970s. Technology dominated the late 1990s, then delivered a lost decade. Energy was close to unownable from 2014 through 2020, then led the market in 2021 and 2022. Staying rigidly committed to yesterday’s leaders is the most reliable path to underperforming in the next cycle.
Diversification is what converts that rotation from a portfolio problem into a portfolio feature. We covered the practical version in ** Momentum Strategies, and Physics: Mass And Velocity Matter**, and the structural version in
where investors rotate hard between factor ETFs and still call it passive investing. Last week priced the difference. Own the equal-weight index, and you made a record high on July 28. Own the momentum factor, and you gave back 18%.
The Passive Aggressive Market,Leverage Was The Mechanism. Rates Lit The Fuse.
During Thursday’s meltdown, I called Michael Lebowitz, and we discussed that it “felt” as if someone was being liquidated. It turned out that a hedge fund, Situational Awareness, which ran leverage roughly 4x its equity base through total return swaps, was the victim. Within a day, it liquidated nearly 3/4 of its holdings.
It is the structure that matters. Prime brokers hold the physical shares while the client takes the economic exposure, so the position never appears in a public filing, and no single broker sees the whole book. Goldman Sachs, JPMorgan, and Bank of America were the counterparties here.
Here is the crucial point: When the collateral fell far enough, the “Prime Brokers” decided to sell. Not the fund.
We have written that sentence before, in Margin Debt Surges As Bulls Leverage Bets:
That process is at the discretion of the broker-dealers that extended that leverage in the first place.
So what tipped the collateral? Rates. After the FOMC meeting this past week, the front end of the curve barely flinched. The long end did the damage, with the 30-year closing that day at 5.20%, its highest level in 19 years. Nothing in this market is more sensitive to the long end than an unprofitable growth stock bought with borrowed money.
The backdrop was already stretched thin. Margin debt set another record in June at $1.50 trillion, up 49% from a year ago, while the net investor credit balance sank to a record negative $1.06 trillion. That is the thinnest cushion against forced selling ever recorded, a point we walked through in Margin Debt Risk: The Ratios That Mislead Investors.
The timing in the chart below is what matters. Leverage actually fell from January into March. Then it went vertical. Margin debt jumped 23.0% in the three months through June, and the credit cushion thinned by $268 billion over that same stretch. That build topped out precisely as momentum did.
Situational Awareness was not the only leveraged buyer in that corner, either. Citadel’s desk put levered ETF assets at a record $218 billion in June, up roughly 60% from the end of March, with semiconductor-linked leverage nearly tripling. We mapped where that money was pointing in A Supply Tsunami Is Coming.
The Daily Shot tracks a slightly wider universe, and its version shows the round trip. Net market exposure across US levered and inverse ETFs peaked near $436 billion in mid-June, about 3.4 times its level in the summer of 2021. It has since fallen 27%, and fund assets are down 25% from their peak.
That is the retail mirror of the de-grossing of the reported prime desks. It also explains why Thursday had so much fuel. Two dollars in a 3x fund carries six dollars of market risk, so when that complex shrinks, the selling is mechanical, and then it stops.
The Fundamentals Never Broke
Here is where the opportunity argument lives, and it deserves a fair hearing. Microsoft grew Azure revenue 43% in constant currency, above a 40.2% estimate, and surpassed $100 billion in annual Azure revenue for the first time. Amy Hood told the Street that capital spending will grow again in fiscal 2027. Amazon lifted its 2026 capex plan toward $220 billion on an AWS-driven beat.
That scorecard kills the simple version of the story. Amazon spent the most of anyone, $53 billion against $45 billion of operating cash flow, printed the worst free cash flow in the group, and jumped about 9% after hours. Alphabet spent less, burned less, and fell 7%.
So, why the difference? It clearly was not an issue of “cash flows” as the narrative suggests. What separated them was evidence that the spending is already earning inside the operating line.
However, Meta went the other way, with operating income down 8% and the margin down from 43% to 31%. After that, the market did the talking with Microsoft rising 15.5% on Thursday, and Meta falling 8.0% in the same session. This wasn’t surprising after Meta missed by more than a $1 per share, guided Q3 revenue to the low end, and declined to commit to a 2027 spending figure.
The market is not punishing capital spending, nor rewarding cash flow. It is paying for proof that the spending is already earning inside the operating line. Read that again, because it is the entire trade.
Both halves of that scorecard are distorted by a single timing mismatch, which I laid out in ** “AI Capex Depreciation Risk Is The Catch To Record Earnings.”** Cash leaves now, so free cash flow understates these businesses. Depreciation lands later, so operating income flatters them. Roughly $760 billion in spending this year is offset by only about $211 billion in recognized depreciation.
So is the market mispricing Alphabet, which is investing, against Apple, which is not? Partly, yes. Alphabet was sold on an in-line core quarter, not a broken one, and 82% cloud growth against a contracted backlog is not a sign of a business in trouble. But that is not a free option either.
Consensus already assumes free cash flow snaps back from roughly $16 billion this year to $387 billion by 2029. That snapback is an assumption, not a result. And Microsoft just stretched the useful life of its data centers from fifteen years to twenty-five, which cuts reported depreciation without changing a single server. Demand is REAL. What broke was the financing stacked on top of it, and who pays the depreciation bill remains unsettled.
Is The Correction Over? The 2000 Playbook Says No.
So, for the one question everyone wants an answer to: “Is it safe to go back into the ‘momentum’ waters?”
BTIG’s Jonathan Krinsky called time on the momentum crash Thursday morning, and on the bounce I think he’s right. Goldman’s high-minus-low momentum index had fallen 23% below its 200-day average after sitting 40% above it in mid-June. It has rarely spent much time beyond 20% below that line in twenty-five years. Stretched is stretched. A dislocation that extreme produces a Thursday almost mechanically, and Microsoft’s print gave buyers a reason to show up at once.
However, a bounce is not a bottom. Krinsky’s own 2000 comparison is the useful part of that note. One month past the dot-com peak, the SOX had fallen 35%. It then rallied roughly 37% and still went on to test its 200-day moving average. Semiconductors closed Thursday 23.0% below the June 22 peak, 11.1% under the 50-day moving average, but still 25.6% above the 200-day.
Sit with that last figure for a second. Even after the fastest momentum crash on record, SOXX trades a quarter above its own long-term trend line. Trapped longs from June do not sell on the first bad day. They sell into the first rally that gets them close to even.
What Should Investors Do Now
Okay, what do we do now heading into the seasonal weak months of August and September? First, treat this bounce as a gift for repositioning, not an invitation to re-risk. The forced seller is gone. But Citadel holds a large block of the same paper and has no obligation to keep it. Secondly, if the 30-year keeps threatening a multi-year breakout, that adds to the risk, and that one variable decides whether the AI complex gets a durable bid or another leg lower.
However, there are opportunities in the rubble, and the following is a quick screen to start from, grouped by what each name actually does in the buildout. Look at the last column before anything else, because Friday rewrote it. Five names now sit above where they were at the momentum peak, and the top two are Microsoft and Amazon, the two heaviest spenders in the group. While some of the selling was certainly due to the liquidation of Situational Awareness, not all of it was. Everything that builds, supplies, or finances the buildout, without yet showing a return, is still down 10% to 47%.
Two things follow. The levered bucket is already flushed, so the case for trimming it is no longer about valuation; it is about which balance sheets survive a retest. And the builders and suppliers are where contracted revenue meets washed-out prices, which is the part of this list I would spend the weekend on.
My read is a tradable rally that fails that potentially fails, particularly if rates continue to push higher this week. If I’m wrong, I’m wrong by buying quality early. That’s the cheaper mistake. We continue to suggest using strength to upgrade quality, cutting names whose only thesis was price momentum, and holding cash to act on a retest of support.
Trade accordingly.
Four leading AI models discuss this article
"Even after the fastest momentum crash on record, SOXX remains 25.6% above its 200-day moving average, leaving trapped longs to sell into any relief rally while long-end rates threaten a multi-year breakout."
The article correctly identifies a mechanical unwind of extreme leverage in the momentum/AI trade, with SOXX down 29% peak-to-trough and the equal-weight S&P 500 making new highs. However, it underplays that consensus 2029 FCF estimates for the Mag 7 already bake in a $760B capex-to-depreciation mismatch resolving favorably. MSFT stretching data-center lives from 15 to 25 years is accounting sleight-of-hand that masks true economic depreciation. Long-end rates at 19-year highs remain the dominant risk; any further 30-year breakout will reprice unprofitable growth far beyond the first leg we just saw. This was Act 1 of a rotation, not the end.
The strongest case against my bearish stance is that if AWS and Azure continue printing 37-43% growth with expanding margins, the market will reward the leaders regardless of rates, turning this into a 2000-style multi-year melt-up for the winners while only the leveraged trash gets flushed.
"The market has transitioned from rewarding speculative AI spending to demanding immediate, proven operating margin expansion, making current rallies prime opportunities to reduce exposure to non-earning tech."
The article correctly identifies the mechanical liquidation of levered momentum strategies, but it underestimates the structural shift in the cost of capital. While the 'Situational Awareness' hedge fund liquidation provided the catalyst, the broader issue is the S&P 500's extreme concentration. With the 30-year Treasury yield pushing 5.20%, the 'AI Capex' narrative faces a severe discount rate problem. We are seeing a transition from a liquidity-driven bull market to a fundamental-driven one where only proven margin expansion—like Microsoft’s Azure—survives. Investors should treat current rallies as exit liquidity for over-leveraged tech positions, as the 're-test' of support levels is statistically probable given the historical SOXX deviation from its 200-day moving average.
The thesis ignores the potential for a rapid Fed pivot or a 'soft landing' narrative that could compress long-term yields, effectively bailing out the high-multiple growth stocks currently under pressure.
"The correction was mechanical deleveraging, not fundamental deterioration, but the real test is whether AI capex earnings materialize fast enough to justify valuations before rates force another liquidation cycle."
Roberts nails the mechanics—leverage unwound, not fundamentals broke. But he's anchoring too hard to the dot-com playbook. SOXX 25.6% above its 200-day is presented as 'still stretched,' yet that's a normal technical level, not a crash signal. The real tell: Microsoft and Amazon both beat, both spent heavily, both rallied. The market isn't punishing capex; it's sorting winners from losers. Roberts' own data shows AWS margin at 39.4% and Azure RPO at $678B—these aren't depreciation tricks, they're cash-generating machines. The risk he undersells: if rates stabilize and earnings actually materialize, the 'retest' never comes, and waiting for it costs real returns.
Roberts assumes forced selling creates a durable buying opportunity, but Citadel now holds a massive block with zero obligation to hold it—if rates spike again or earnings disappoint, that block becomes a sword of Damocles. Waiting for a 'retest' could mean missing a 20%+ rally if the AI capex thesis simply proves out.
"The next leg for markets hinges more on the rate path and liquidity than on any single earnings beat, and a sustained high-rate regime or renewed leverage pressure could re-ignite selling even if AI spend remains robust."
News takeaway: The piece argues this is a liquidity-driven rotation, not a fundamental breakdown, aided by leverage unwinds and a spike in the 30-year yield. That framing misses some tail risks: systemic fragility from high margin debt, opaque prime-broker exposures, and the potential for another round of forced selling if leverage and hedges reset. It also downplays that AI-capex demand is not guaranteed to translate into near-term cash flow given depreciation timing and working capital, and the market may have priced too much certainty into mega-cap earnings. A durable rebound requires broader breadth and better liquidity relief, especially if rate expectations stay higher into the fall.
The strongest counter is that this looks like a washout that clears leverage and sets up a reset in rate expectations; AI/cloud spend momentum should translate into improving cash flows, and once volatility subsides, a durable rebound in growth names is likely.
"Depreciation accounting games create illusory FCF that markets will reprice once actual cash outflows hit."
Claude's dismissal of SOXX's deviation as 'normal' ignores that the current 25.6% gap exceeds every prior 2000 and 2022 peak before those drawdowns accelerated. Gemini's discount-rate focus is right, but nobody has flagged that MSFT's 25-year depreciation extension alone inflates 2025-2027 FCF by ~$18B cumulatively—real cash capex still runs higher.
"The market is prioritizing RPO growth over depreciation accounting, creating a dangerous bifurcation between mega-caps and the rest of the market."
Grok is right about the accounting games, but Claude is correct that we are misreading the market's reaction to capex. The market isn't blindly accepting depreciation extensions; it is pricing in the massive RPO (Remaining Performance Obligation) growth as a hedge against future rate volatility. The real risk nobody mentioned is the 'crowding out' effect: if mega-caps continue hoarding capital for AI, smaller high-growth firms will face a liquidity drought, causing a bifurcated market that keeps the index high but destroys the broader economy.
"Mega-cap capex creates structural moats via infrastructure scarcity, not balance-sheet hoarding."
Gemini's 'crowding out' thesis is underexplored and real—but backwards. Mega-cap AI capex isn't starving smaller firms; it's inflating semiconductor and infrastructure costs, forcing smaller players to either partner with the giants or exit. That's not a liquidity drought, it's a structural moat widening. The bifurcation risk is real, but it's not about capital allocation—it's about pricing power and access to cutting-edge infrastructure becoming a gating function. That favors mega-caps even more than current valuations suggest.
"The belief that a rate-stable environment guarantees no re-test is fragile; AI capex cash flows are volatile and can reprice multiples quickly if macro shifts or delayed payoffs occur."
Claude’s line that a rate-stabilizing world would yield a ‘never re-test’ outcome assumes earnings materialize without hiccups. The risk not being discussed: mega-cap AI capex is episodic and highly rate-sensitive. A sustained capex slowdown, margin compression, or delayed AI payoff could keep high discount rates intact and force multiple contraction even if AWS/Azure beat. In short, the ‘rates stabilize, buy the dip’ thesis is conditional, not a guaranteed floor.
The panel agrees that the recent market correction was primarily driven by a mechanical unwind of leveraged momentum trades and a spike in long-end rates, but they disagree on the extent to which fundamentals are at play. While some panelists see this as a rotation rather than a breakdown, others warn of systemic fragilities and potential tail risks.
Investors should treat current rallies as exit liquidity for over-leveraged tech positions, as the 're-test' of support levels is statistically probable.
A sustained capex slowdown, margin compression, or delayed AI payoff could keep high discount rates intact and force multiple contraction even if AWS/Azure beat.