AI Momentum Unwind Has Been Nasty – Seeing Early Signs of Capitulation…
Executive Summary:
AI Momentum Unwind & Sector Dispersion: Recent market volatility is being driven by a sharp, momentum-based selloff in AI and semiconductor equities, with the SOX index entering an official bear market after dropping 20% from its June peak. Meanwhile, the broader S&P 500 index has remained resilient, masking extreme underlying sector dispersion and a de-rating in the "Mag Seven" as its forward P/E multiple has compressed from 29x down to 24x.
Catalysts for the Tech Drawdown: The pullback was accelerated by the release of China's cost-effective Kimi-K3 open-source AI model from Moonshot, which threatens U.S. frontier model makers and forces an industry-wide shift toward operational efficiency. Concurrently, disappointments out of industry bellwethers like IBM—which plunged 25% following a moderate top-line revenue miss and software growth doubts—and Netflix, which fell sharply due to weak third-quarter revenue guidance – added to what was an already weak tape.
Geopolitical Inflation & The Fed Dilemma: Renewed military strikes and the end of the ceasefire in the Middle East have driven spot WTI oil prices up 15%, pushing the 10-year Treasury yield back up to 4.6%. This supply-side inflation pressure creates a complex backdrop for the Federal Reserve, where strong real economic activity contrasts sharply with depressed interest-rate-sensitive sectors, shifting futures market expectations toward an October rate hike.
Full Commentary:
The market is showing investors that momentum-driven parabolic upside moves in equities can be just as relentless on the downside. The operative word for last week's price action is volatility, along with a dramatic loss of upside momentum for the recent darlings of the AI trade. The SOX (semiconductor index) is now in an official bear market, down -20% from the nearby June peak. The index sank by -10% last week, its steepest weekly drop since April 2025. The S&P 500 slipped below its 50-day moving average – a line that has held twice since mid-June. If the Nasdaq 100 breaks the June 10th low of 28,508 (closed Friday at 28,592), the next stop on the charts is the April 28th level of 27,029.
There has been a slew of technical bear markets, with the Korean Kospi -27% from peak, US AI Tech Beneficiaries -25%, Global Memory -36%, EU Semis -23%, US MOMO -27%, and US TMT MOMO down -40%. It is a true headscratcher to see how resilient the cap-weighted and equal-weighted S&P 500 index has been in the face of such extreme turbulence under the hood. This is what market practitioners mean by low-correlation rotational moves, or dispersion, rather than a broad sell-off. The key point is that the S&P 500 has increasingly become a poor proxy for the average stock's performance. Friday's price action is a perfect example: the Roundhill Memory ETF (DRAM), which has become the go-to ETF for the subsector, traded in a 13% range only to finish flat, while the semiconductor space (SOX) traded in a 7% range and ended the day down 1%. Keep in mind we are talking about a sector where the average market cap of a company in the SOX index is almost $500 billion – massive companies exhibiting massive volatility.
The Morgan Stanley QDS team notes that we “are now 17 business days into the selloff, with MSZZMOMO -28% peak-to-trough. The move is already both faster and deeper than the median historical Momentum drawdown, which has been -22% only 33 business days since 1999. This is the worst Momentum drawdown since the Dec 2022 to Feb 2023 episode, which reached -29%. Outside of that period, it is the largest drawdown since Feb to May 2021, when Momentum fell -40% over 65 business days. The 2021 episode also featured a similar mix of long-side underperformance and short-side outperformance. The long leg was -29%, while the short leg was +18% vs the S&P 500’s +5% over the same period. In the current drawdown, the S&P 500 is up +1% since June 22.”
And look at some of the Tech bellwethers and what they have done from their nearby peaks (including two industrials who have become linked to the AI trade) – the Mag Seven, which had led the market by a wide margin for four years, has now lagged far behind so far in 2026. The group has been re-rated lower, with the forward P/E multiple compressed to 24x from 29x at the end of 2025:
Intel: -33%
SanDisk: -31%
Micron: -30%
Microsoft: -27%
Broadcom: -17%
Meta: -16%
Caterpillar: -15%
Honeywell: -14%
Nvidia: -14%
Cisco: -14%
Alphabet: -13%
Amazon: -9%
AMD: -8%
The persistence and magnitude of selling since early June point to a significant shift in investor positioning, with some signs of capitulation starting to emerge. Goldman's Vol Panic Index has surged in recent sessions. Unlike the VIX, it captures a broader set of volatility signals, offering a deeper read on market stress (see chart below). My advice: don’t get too eager, too soon, but the indicators I’m monitoring suggest it's time to start dipping your toes back into some areas that have been whacked and other areas that have become collateral damage.
This latest wrinkle in the AI revolution came on the back of news out of China where a tech start-up, Moonshot, released a new update for its open-source AI model (Kimi-K3) where early benchmark testing put it just behind leading US frontier models from Anthropic and OpenAI, but at a fraction of the cost. This is all eerily reminiscent of the DeepSeek selloff in the winter of 2025, the revenge of cheaper, customized models that now threaten the U.S. frontier model-makers. Looks like the complexion of the AI trade is in the process of changing yet again. No, this doesn’t mark an end, but rather one of many pivots that were always going to materialize as this life-altering technology evolves.
On another note, this is one manifestation of how inflation gets back on target. When part of the inflationary impulse is caused by a price surge in an industry that has committed a sizeable amount of capex in the pursuit of AI dominance and comes under competitive threat – prices will respond. This should force U.S.-based companies operating expensive models to focus on avenues of efficiency as well as force users to evaluate the cost/benefits of the respective models for their individual needs.
From an investment perspective, this structural evolution marks the maturation of the AI capex cycle and the onset of foundational "model deflation." As lower-cost alternatives erode the pricing power of generalized LLMs, the market is actively shifting its reward metrics away from a growth-at-all-costs infrastructure buildout and demanding a rigorous focus on unit economics, inference efficiency, and demonstrable ROI. Consequently, investors are no longer applying a broad valuation paintbrush across the entire tech sector; instead, they are aggressively discerning between generalized AI spenders and the targeted beneficiaries that supply indispensable technical bottlenecks—specifically enterprise security, specialized server architecture, and high-bandwidth memory products.
In addition to the news of a new competitive model out of China, the U.S. Tech sector got hit by a couple of announcements from corporate bellwethers that surprised investors. Let’s start with the bombshell out of IBM that propelled it to the worst single trading day in its history last week – where the cause ties directly back to the memory-chip theme that has been running through this missive. Keep in mind this was just a pre-announcement, not an actual earnings release (slated for July 22nd), which caused IBM shares to plummet by -25%. This was the company's worst day on record, exceeding the prior worst of October 19th, 1987, when shares fell by -23.7% (IBM has been NYSE-listed since 1916). In dollar terms, the decline wiped out roughly $67 billion in market value.
The miss itself was moderate; the reaction was not. IBM reported adjusted EPS of $2.93 per share on revenue of $17.2 billion, versus expectations of $3.01 EPS on $17.86 billion – about $660 million short on the top line. A ~3.5% revenue miss and an 8-cent EPS miss don't normally produce a -25% crash, so the scale tells you that the market read it as something more than one soft quarter – with doubts about the strength of IBM's software-led transformation, the quality of its growth, and the sustainability of its free cash flow. And it highlights the divide between AI winners and everyone else in the tech space; the winners being companies that focus on servers, storage, security software, and memory products used in AI computing. Investors are now thinking more about winners and losers in this AI trade rather than just using a broad paintbrush. Then there was Netflix – a textbook case of a stock falling hard on a report that was basically fine, which tells you the problem was expectations, guidance, and a self-inflicted disclosure decision rather than the quarter itself. The actual results were roughly in line.
So why the violent reaction to an OK quarter? It was all about the third-quarter guide lower: Netflix forecasted revenue of $12.86 billion in Q3, below the ~$13 billion Wall Street expected (the sales guidance points to a three-year low number), and guided EPS below the Street’s projection as well. Let’s see how things play out this week with roughly 20% of the S&P 500 set to report (notably Google, Tesla, Texas Instruments, Intel…), but early indications for the Tech space are that the bar is set pretty high going into results. Bank earnings were solid last week, and expectations weren’t nearly as lofty, which helps to explain why the sector performed well following results.
SpaceX rounded out the week’s disappointment as its stock price fell below its IPO level. From the peak in June, fully $1 trillion of market cap has been wiped out (the stock still trades with a forward P/E multiple of 182x, or a six-fold premium of the average S&P 500 member).
The last headline I’ll hit on is the price action out of the popping of the Tech mania that cropped up over in Asia, where Korea’s $4 trillion AI-driven stock market is giving all of us a first-hand look at what happens when the leveraged tech trade heads south with no antidote from other sectors. The KOSPI has plunged by more than -25% from the June peak, and Taiwan is now also in official correction mode. Imagine what happens if moves like this begin to surface in the S&P 500.
We also have the renewed outbreak of tit-for-tat military strikes in the Middle East, where the ceasefire is dead. The Iranian regime is back to targeting US military bases in the area and energy infrastructure in the Middle East. The American strikes appear to be strategically surgical, aimed at Iran’s military targets, like command centers, missile sites, coastal surveillance facilities, and transportation networks. One can reasonably expect that the next phase will include bridges and power infrastructure. We could be on the path here for ground troops to overtake some of the key islands so that the U.S. forces can end the drone and missile threat once and for all. The effective closure of the Strait of Hormuz and the IRGC strikes against oil facilities in the Gulf region do risk taking crude prices up even further in the near-term, but the real problem for the global energy market resides in finished products and refining capacity.
Spot WTI oil prices surged 15% last week, which is putting upside pressure on bond yields, where the yield on the 10-year T-note is back to 4.6%, and the 30-year yield is up to 5.11% (closing in on the recent high of 5.18% reached back in the middle of May). One-year WTI futures pricing has moved off the recent lows, but at $72 per barrel, it would be manageable if oil can stay around this level. Whether the Fed sees this latest run-up in oil prices as being temporary remains to be seen. And whether the Fed sees the cracks in the tech trade and the sharp slide in the SOX as a sign that the boom in chip prices has run its course also has question marks in front of it. So far, many FOMC officials are unimpressed with the latest benign core CPI and PPI data – at least they are skeptical over their durability. The only one who does not appear hawkish is the Chairman himself, but he has a gang that wants to raise rates as the rhetoric suggests. Can he convince them otherwise?
As for Fed fund futures, not much has changed other than the timing of the first expected rate hike since July 2023, with futures markets shifting the timing for a hike to October while taking this month off the table. I will say, the Fed is not in an easy position with a rather complicated and muddled backdrop to discern whether a hike, a cut, or holding steady is the right call. Asset prices, consumer spending, and parts of the economy say to hike – did you see last week’s Philly Fed report where real economic activity reached a 5-year high? While housing, construction, and anything tied to interest rates are in some form of a recession or outright depression – indicating an interest rate cut is needed. My two cents is that holding where they are is the best plan of action. Call it the ‘Hippocratic oath’ approach – do no harm.
As for some closing thoughts, broadly speaking, I welcome the rotational action that is keeping the major averages in a sideways consolidation while at the same time cleansing the euphoria and excess that has built up in some areas. That doesn’t mean it's been easy for anyone who has capital allocated in some of these themes: AI, semis, memory, Mega-Cap Tech, nuclear, gold, miners, and emerging markets. But keep in mind, investing involves risk and corrections are part of the journey. Unless you’re a trader with impeccable timing, it’s virtually impossible to own an investment or company that doubles, triples, or 10x’s in value without having to endure through some 30-50% corrections. Sometimes the most appropriate action is to take your licks and endure through the correction – it will eventually run its course, and if you’re in a secular or multi-year cycle (which I think we are), then the longer-term trend will resume, and you’ll be in the right position to profit from it.
I don’t think the totality of this pullback has run its course, but as I stated earlier in this missive, I am starting to see some signs of the baby getting thrown out with the bathwater. Recently, the glue holding the overall market together includes Energy, Financials, and the REITs space (and of the Mag Seven, only Apple is behaving well), but I view this leadership as more tactical (short-term) than structural (long-term). In time, I expect past leadership to reassert itself (debasement, AI revolution, nuclear renaissance, reindustrialization / onshoring, deregulation). A key test lies ahead, as Tesla and Alphabet report this week, and next week we will see the numbers from Microsoft, Meta, Apple, and Amazon. These six giants together account for 25% of the S&P 500 market capitalization, so their guidance in particular will dominate the direction of the overall index from here.
The articles and opinions in "Capital Market Musings and Commentary" are for general information only, and not intended to provide specific investment advice. Performance, dividends and other figures have been obtained from sources believed reliable but have not been audited and cannot be guaranteed. Past performance does not ensure future results. Investing inherently contains risk including loss of principle. Advisory services offered through Casilio Leitch Investments, an SEC registered investment advisor. Copyright © 2026 Casilio Leitch Investments. All Rights Reserved.