by Jurrien Timmer, Director of Global Macro, Fidelity Investments
The market gets rebalanced one way or another.
While we point to the sweeping midterms as the catalyst for the sharp drawdowns across the AI space last week, the die had already been cast in recent months as the insatiable demand for capital via both debt and equity markets got tested and reigned in. Markets self-regulate just like voters do, restoring balance when the pendulum swings too far in one direction. Mean reversion at work.
The S&P 500 index is down 6% from its October high while AI stocks are down 30%+ from their highs. The equal-weighted S&P 500 index is down only 1.3% from its August high, and the S&P 500 ex-AI index is up 0.5%. There has been fierce rotation for months now, masking what might have seemed like a quiet trading range. History is full of episodes like this, where growing earnings take a backseat to tightening financial conditions.
I suspect we will be in this mode for a while longer, not unlike 2018 or 2015. Friday’s modest recovery from Thursday’s 4% decline gave off a bit of a dead-cat vibe, which suggests to me that this rotation may not be over, even as many stocks remain in up-trends. Last week’s purge felt a bit like the tariff tantrum growth scare of 2025, but that one was easily reversed by a policy pivot. I’m not sure that there’s a quick pivot to the current correction, other than a successful ANT IPO and the potential for a re-rating in the Mag 7 if investors decide that the incumbents will win in a scaled-back capex cycle. They have been significantly derated since their cash flows started going to capex instead of shareholders.
For now, bonds are acting like bonds again, which is good to see. How long that anti-correlation lasts is unclear, but the history of secular correlation regimes suggests that it may not be long. That suggests to me that the next “trade” will be a rebalancing out of bonds back into equities. Given the still positive economic backdrop and a Fed which remains behind the curve in terms of policy restriction, I doubt that bond yields have much lower to go here.
Here’s a look at the drawdowns in the AI space. The GS AI basket is down 30% from its high, but there are plenty of names which are down much more than that. The ex-AI half of the market has been the relative winner during this rotation, but there’s a lot more cap on the AI side and that affects the overall index. When 10 stocks are 40% of the market, those 10 stocks will drive much of the beta.
The AI basket has returned to the “scene of the crime” (the most recent breakout earlier this year). The glass-half-full technical read is that the decline from the June high of 473 to 310 looks like a 3-wave ABC pattern, which would be considered corrective instead of impulsive. The AI vs ex-AI pair is down 34% from its high, while the 10-year yield has fallen 53 bps. Anti-correlation at work.
When markets fall hard and fast enough, correlations often converge, and we can see below that this happened last week as well. Earlier last week the ex-AI basket was only 10% correlated to the S&P 500 index, but by the end of the week it was 40%. The same applies to the S&P 500 equal-weighted index (SPW) and even REITs and Bitcoin.
How far will this correction go? For now, momentum is weak across the global regions and getting moderately oversold. As the momentum curves below suggest, markets tend to travel together on the way down.
Here’s a look at the S&P 500 index (SPX). We are back at support, and a deeper correction would find further support at the 7000 level. The GS risk appetite index is moderately oversold but has not reached those “close your eyes and buy” levels yet. Ideally, the eventual trading low will produce some bullish divergences which tend to help identify inflection points.
Another look at where the market stands technically is to consider where we are relative to the prevailing trendline. The S&P 500 is in both a cyclical and secular bull market, and while we remain well above the secular trendline, the index is now 5% below the cyclical trendline. That’s less than the previous two lows, set earlier this year at -9% and in 2025 at -14%, so again this suggests that the market has more consolidating/correcting to do.
With extensive chart damage evident across the AI space, we are left with the question of whether the entire AI trade is over. No one knows the answer, but one could argue that the S&P 500 semiconductor index has now confirmed a double top following a massive inflow into the space.
The late 1990’s internet analog would suggest that the AI boom is ending, but as I always caveat, price analogs are for “entertainment purposes” only. Then why show them at all, you might ask? Because looking at historical periods of similar price action can be an important conversation starter for how things might be the same or different this time. I use them as loose roadmaps, knowing that at some point the market may travel in a different direction.
I had failed to realize this before last week, but the inflection point in the 1998-2002 analog was always going to be the midterm election. That could have been a hint that a regime change was in store.
If the entire AI trade is “over,” the soldiers in the S&P 500 (ex-AI) will have a lot of heavy lifting to do to keep the bull market from stalling out. As we see below, the equal-weighted index has performed quite well lately but in a top-heavy market it may not be enough to provide positive beta to the headline index.
The fundamentals for the SPW remain attractive as well, with the index trading at 16x n12m EPS and the payout (PO) growing faster than earnings (leaving a payout ratio of 74%, which is higher than the cap-weighted index.
One potential upside to a slowdown in the capex cycle is that it may put the Mag7 back in favor. We can see below that the P/E ratio of the M7 has diverged considerably from the P/PO (payout) multiple, ever since the M7 started spending their cashflows on capex instead of buybacks. As the payout ratio (dividends plus buybacks as a percentage of earnings) started falling from 90% to 29%, the P/E-multiple declined accordingly. My math suggests that if this had not happened, the Mag 7 could be trading at twice their current multiple. Between strength in the broader tape and this possible outcome, I could see a scenario in which the bull market (both cyclical and secular) can keep going, even without help from the AI picks and shovels.
All in all, I can see a scenario in which the market continues to churn and rotate as growing but decelerating earnings growth gets offset by tightening financial conditions and widening credit spreads. That would put 2026 in the camp as 2018 or 2015 or even 2011 or 2022. Not the end of the world but not a melt-up either.
My investment clock suggest as much. We are out of the sweet spot (9-12 o’clock) and into a yellow zone (past midnight) where bullish fundamentals get offset by tighter liquidity conditions. Below I show the current cycle (black) against the 2020-2022 cycle (blue), 2017-2019 (orange), and 2014-2016 (green).
One question is whether we have finally reached peak earnings growth. It’s starting to look that way. The n12m earnings estimate is up 39% from a year ago but is starting to look toppy against an expected growth rate of 19% over the next 12 months. The base effect is kicking in.
A deceleration in earnings that are growing 30-40% is to be expected and history shows that if the level of earnings does not peak along side its rate of change, the market is more likely than not to keep advancing. So, all good there. The chart below shows all cycles since the 1920’s when the 5-year earnings CAGR peaked. Aside from the Great Depression, we have always peaked at the current 14-15% growth rate (5-year CAGR). What sticks out to me is that currently the estimates (from BBG) suggest a modest slowdown to 12% and then a re-acceleration back to 14%.
We can see that below a well. If we take the current Bloomberg estimates and project them from today’s level, we see that the market remains wildly optimistic that earnings will keep booming well into 2027. That would be quite different from past cycles and raises the possibility that we might transition from a period of upward revisions to downward revisions.
Unless it’s different this time, which are words that should always be chosen carefully. We are firmly on the right tail for both the 5-year EPS CAGR and the 5-year CAPE ratio, so history would suggest that earnings growth could decelerate from here.
On bonds, the 10-year yield peaked at 5.36% a few weeks ago and currently sits at 4.86%. Bonds are anti-correlated again, at least over the short term. Given their significant value at the recent yield highs (almost 3 % real), it’s not surprising to see them be a port in the AI storm. How long this will last is another question, and I suspect the answer is “not long.” The 10-year has re-entered the triangle apex zone, which to me suggests that further upside for bond prices is limited.
History shows that during secular regimes of positive stocks/bonds correlations, any near-term episodes of negative correlation such as happened in 1998 or 2025 will prove to be short-lived unless the secular regime is changing back to a deflationary one. With inflation above 3%, the economy strong amid ongoing fiscal expansion, and the Fed behind the tightening curve, such a regime shift seems unlikely. So, enjoy the bond gains while traditional 60/40 model offers a temporary reprieve to what is otherwise a secular 60/20/20 regime
This information is provided for educational purposes only and is not a recommendation or an offer or solicitation to buy or sell any security or for any investment advisory service. The views expressed are as of the date indicated, based on the information available at that time, and may change based on market or other conditions. Opinions discussed are those of the individual contributor, are subject to change, and do not necessarily represent the views of Fidelity. Fidelity does not assume any duty to update any of the information.
1290990.1.0
