Earnings season is here! Citadel Securities: The next few weeks will be the real stress test for U.S. stock fundamentals

This test sets a high bar. The market consensus expects S&P 500 earnings per share to grow about 27% year-over-year in Q3, the third-highest forecast level before an earnings season since 2010, only surpassed by two post-pandemic recovery quarters. Meanwhile, valuations have contracted, investor demand is beginning to recover, and early shifts in market leadership are emerging.

The earnings calendar is highly concentrated. According to Citadel Securities, during the five trading days between October 28 and November 3, roughly 47% of S&P 500 index-weighted companies will report results, contributing about one-third of expected earnings growth. This means the market will test whether high earnings expectations can be realized in a very short period.

For investors, the key question is whether earnings growth can broaden from a handful of leaders to a wider set of companies. Retail buying has rebounded, equal-weighted indices have rallied, and single-stock volatility is rising—all signals that internal market dynamics are changing. However, index volatility remains low, and if rotation shifts to broad-based selling, correlations could rise quickly.

Citadel Securities notes that despite pressure on US equities in September, the earnings bar remains extremely high. Over the past 20 quarters, actual corporate earnings growth has beaten consensus expectations heading into the season in 19 of them. In the last two quarters, actual growth exceeded preseason expectations by about 10 percentage points. In Q2, the proportion of S&P 500 companies beating EPS estimates reached 88%, the highest level excluding the 2020–2021 recovery period.

Expectations for Q3 continue to be revised upward into earnings season. According to Citadel Securities data, from July 1 to just before earnings began, S&P 500 overall earnings estimates were raised by 1.6%—the second strongest pre-season upward revision in the past five years, only behind Q2 2026. This also marks the fifth consecutive quarter of pre-season upward revisions at the index level.

The divergence between earnings improvement and valuation compression is particularly prominent in semiconductors, memory, and AI infrastructure. Since the end of June, memory sector’s next-12-months EPS forecasts have risen 25% while their forward P/E has dropped 34%. AI infrastructure’s next-12-months EPS forecast increased by 25%, with forward P/E declining 22%. Meanwhile, for the S&P 500 overall, next-12-months EPS is up 10% and forward P/E down 6%.

Corporate guidance remains robust. In the past four quarters, S&P 500 companies have issued nearly three times as many positive as negative guidance revisions. Citadel Securities notes that, excluding the 2009–2010 and 2020–2021 recovery periods, this is one of the most supportive periods for corporate guidance in 20 years of data.

A core feature of Q3 earnings growth is its high concentration. Citadel Securities points out that Micron and Nvidia together account for 36% of S&P 500’s expected earnings growth. The top ten contributors together account for about two-thirds of expected growth. Excluding these ten companies, S&P 500 expected earnings growth falls to about 11%. The 300 smallest S&P 500 companies by market value are expected to grow earnings by about 5%. This shows that index-level high growth still relies heavily on a small number of sectors and companies.

Nevertheless, earnings breadth is not absent. About 74% of S&P 500 companies are expected to post year-over-year earnings growth, above the median of 66% since 2021. The coming weeks will be key to confirming whether this breadth is validated by actual results and guidance.

The buyback window may also reopen. Currently, with the earnings blackout period at its peak, only 24% of S&P 500 index-weighted companies are outside the buyback-restricted window. Citadel Securities estimates this will rise to about 55% by the end of the month, 75% by November 3, and 85% by November 9.

So far this year, US companies have announced a record $1.3 trillion in share buyback authorizations, 8% higher than last year's record. The tech sector contributed $573 billion, or 43% of the total. As earnings are released, corporate buying may again provide technical support for the market.

One of the most notable recent developments within US equities is the relative improvement of equal-weighted index performance versus mega-cap weighted stocks. In the previous quarter, index performance was largely driven by a handful of large caps. Citadel Securities notes that recently, more sectors have participated in gains, indicating early signs of market breadth recovery.

The ratio of the S&P 500 Equal-Weight Index to the Nasdaq 100 fell to its lowest level since April 2000 on October 7. Since mid-August, this ratio underperformed by nearly 9%. In the following three trading days, the equal-weight index outperformed the Nasdaq 100 by 2.8%, a change in the 95th percentile since 1990.

But the rebound is still in its early stages. Despite the upturn, this ratio remains in the 2nd percentile of the past five years. Citadel Securities points out that historically, less than half of similar bounces from extreme lows saw equal-weights continue to outperform in the following month.

Market breadth is also improving. Since the equal-weight to Nasdaq 100 ratio bottomed, 76% of S&P 500 components have outperformed the index. Defensive sectors rose 2.2%, rate-sensitive sectors up 2.0%, and cyclical sectors up 1.6%. Tech stocks lagged; only 42% of components rose.

At the same time, the proportion of S&P 500 components trading above their 50-day moving average has climbed from 25% to 37%. However, this still sits at the 16th percentile of the past year. For markets, earnings season will test whether this breadth recovery can be sustained.

Retail buying is rebounding even though overall trading activity remains below average. Citadel Securities platform data shows retail investors were net buyers of $9.1 billion in US equities last week, the largest weekly net inflow in 15 weeks and the ninth highest in the past year.

The increase in buying is mainly from stronger directional bias rather than a surge in trading volume. Last week, the retail buy bias exceeded sell bias by 2.7 percentage points, in the 90th percentile of weekly directional skew. Overall trading activity was only in the 42nd percentile over the past year. In the last three weeks, weekly net retail buying climbed from $4.8 billion to $6.4 billion, and then to $9.1 billion.

Flows are highly concentrated. Semiconductors saw $2.6 billion in net buying last week, the third-highest weekly total in the past year. At this pace, October could be the largest monthly net buy for semiconductors on Citadel Securities' platform, three times the average pace from the past year. AI infrastructure net buying was $2.4 billion, the fourth-highest in the past year. Neoclouds net buying was $700 million, the largest weekly accumulation observed for this category of stock.

Fixed income ETFs are also attracting retail demand at a record pace. In the first seven trading days of October, retail net buying of bond ETFs approached $3 billion, averaging $422 million per day—28% above the previous monthly record pace. Buying is focused on short-term Treasury bills, ultra-short duration, municipal bonds, and US Treasury ETFs. Demand for credit, loan, and CLO ETFs is only 0.4 times typical levels. If the current pace continues, October will be the largest month for retail fixed income ETF buying since 2017 in this dataset.

Citadel Securities highlights that one of the most prominent current market features is the ongoing divergence between single-stock volatility and index volatility. Rotation among winners and losers at the stock and sector level is raising individual and sector volatility, but offsetting moves within the index keep overall volatility near its past-year lows.

Since September 29, the average one-month at-the-money implied volatility for S&P 500 components has risen from 34.1% to 36.4%. Over the same period, S&P 500 index implied volatility fell from 13.2% to 11.7%. The spread widened by nearly 4 volatility points to 24.7, in the 90th percentile for the past year.

The ratio of single-stock to index volatility has risen to 3.1x, in the 98th percentile. One-month implied correlation has dropped to 7.6%, the lowest reading since 2006. Implied volatility for banks, biotech, AI, and software sectors has all climbed relative to S&P 500 volatility.

Realized index volatility is also subdued. Since June 5, the S&P 500 has not had a single trading day with a 2% or more move, marking 88 consecutive sessions without such a move. The index has also gone 52 trading days without a 1% decline. The 10-day realized volatility has dropped to 7.2%, in just the 5th percentile of the past year.

Broad-based selloffs have also been absent for an extended period. Citadel Securities notes the New York Stock Exchange hasn't had a day where declining volume comprised at least 80% of total volume in nearly a year. The last instance was November 17, 2025, about 225 trading days ago—the longest gap in nearly 30 years of data.

Citadel Securities concludes that the constructive view on US equities in Q4 remains intact. Earnings expectations are strong, valuations have compressed, retail buying is picking up, corporate buyback windows are about to reopen, and market participation is starting to improve.

But these supportive factors need confirmation from earnings. About half of S&P 500 index weight will report results over five trading days—a few companies still contribute the majority of earnings growth, and low index volatility masks rising single-stock risk.

Scott Rubner believes that the recent improvement in participation is notable but that the next few weeks will more effectively test whether this trend continues. In the current environment, Citadel Securities continues to favor selective single-stock exposure, which better aligns with the market’s structure of earnings dispersion and volatility divergence than simply adding index beta.