45% of S&P 500 Stocks Now Carry a Negative Beta. Here Is the Index Math.

Open your brokerage app on a green day and you may find two screens telling opposite stories. The broad index fund is up. The watchlist below it — the ten or fifteen individual names you actually follow, the ones you read about — is a column of red. Nothing is broken. You are looking at a cap-weighted average and a personal sample, and they have drifted further apart than at almost any point in recent memory.

How far apart is measurable. About 45% of S&P 500 stocks now carry a negative beta over a trailing three-month window, according to a Goldman Sachs note described by CNBC on September 28, 2026. CNBC’s own calculation, using weekly returns, put the figure near 40% over three months and at 17% over a full year. A caveat before anything else: these are figures from a news report summarizing sell-side and third-party research. They are not filings, not index provider documentation, and not something you can pull from a regulator. Treat them as the best available public read, not as audited fact. The S&P 500 negative beta story is real, but its precision is softer than the decimals suggest.

What follows is the mechanism underneath that number — why an index can climb while a large share of its own members walk the other way, and what that does and does not tell an index investor.

Step one: the index is an average, and the average is weighted 📊

The S&P 500 does not count votes. Under S&P Dow Jones Indices’ published construction for the benchmark, each member is weighted by float-adjusted market capitalization, so a company worth several trillion dollars contributes vastly more to the index’s daily move than a company worth twenty billion. Two stocks in the index are not two equal inputs; they are two inputs scaled by size. That is the index provider’s own rule, not an interpretation of market behavior.

That is the whole first step, and it is the one most often skipped. Once weighting is in the picture, “the index rose while most stocks fell” stops being a paradox and becomes ordinary arithmetic. Adam Turnquist, chief technical strategist at LPL Financial, framed it plainly in the same CNBC report: “It only takes a few of those mega caps names to work, and a lot of the smaller weighted stocks don’t need to work.” Concentration is the reason a minority can outvote a majority.

The practical translation for a household portfolio: when you buy a broad S&P 500 fund, you are not buying 500 roughly equal bets on American business. You are buying a size-proportional claim, and the top of that distribution now does most of the talking. Your fund’s return can diverge sharply from the experience of the median company inside it.

Step two: what beta is actually measuring

Beta is the sensitivity of a stock’s returns to the index’s returns over a chosen window. A beta of 1.0 means the stock has tended to move roughly in step with the index. A beta of 0.5 means it moved about half as much in the same direction. A negative beta means that, over the measured period, the stock’s returns tended to move opposite the index’s.

Two properties of that definition matter more than the concept itself.

• It is a backward-looking statistic, not a property of the company. A stock does not “have” a negative beta the way it has a share count. It has one over a specific window, on a specific return frequency, and the number changes as the window rolls.
• The window length changes the answer enormously. CNBC’s own numbers show this directly: roughly 40% of S&P 500 members registered a negative three-month beta, but only 17% did over one year. Same universe, same method, different lens — and the reading falls by more than half.

That gap is the single most useful thing in the data. A short window catches a temporary de-synchronization; a longer one filters most of it out. Anyone citing a breadth statistic without naming its window is not telling you enough to act on.

Step three: correlation and volatility, taken apart

Bradley Krom, director of investing strategy at WisdomTree, gave CNBC the decomposition that makes the rest coherent: “Beta is a function of correlation and volatility.” Beta rises when a stock both moves with the index and moves violently; it collapses toward zero when the stock’s moves stop lining up with the index’s, however large those moves are in isolation.

So a market full of stocks making big individual moves on unrelated news — an earnings miss here, a regulatory ruling there, a guidance cut somewhere else, each landing on a different day for a different reason — produces low pairwise correlation. Krom’s point is that those moves substantially cancel at the index level. Turnquist separately pointed to unusually low correlations among S&P 500 constituents.

This explains something index investors often misread as calm. A quiet index is not evidence of a quiet market. It can equally be evidence of a loud market whose noise is out of phase. Realized index volatility can sit low while the stock-level experience underneath is turbulent, and an investor who reads the index’s steadiness as a risk signal is reading an artifact of offsetting moves.

Where the tape confirms it ⏳

Breadth statistics get abstract fast, so the single-session evidence is worth holding onto. CNBC reported that on a recent Monday the S&P 500 rallied 1.5% while 30 of its members touched a 52-week low and only seven set a new high. An index near records, advancing strongly, with new lows outnumbering new highs more than four to one.

Jason Goepfert, founder of SentimenTrader, told CNBC that the last comparable occurrence — the S&P 500 gaining at least 1% while sitting within 1% of a 52-week high, with new lows outnumbering new highs — was in December 1999, immediately before the top of the dot-com boom. That is a genuinely rare configuration. It is also a single historical precedent, which is close to the weakest form of statistical evidence there is. One prior instance cannot establish a base rate, and the honest reading is that the condition is unusual, not that it is predictive.

The driver behind the split 🏭

The dispersion is not random. CNBC reported that in July 2026, AllianceBernstein — using one-year trailing returns — found an unprecedented share of U.S. stocks displaying negative beta as AI winners powered market gains. Semiconductor makers, hardware companies, and other beneficiaries of AI infrastructure capital spending have been on one side of that line.

Capital spending is the mechanism worth naming here, because it has two sides. Enormous outlays on AI infrastructure arrive as revenue for the companies selling into it and as cost for the companies buying. When a single spending cycle is large enough to dominate index returns, it mechanically sorts the market into a beneficiary bloc and everything else, and the two blocs stop moving together. The negative beta reading is a symptom of that sort, not an independent phenomenon.

Traps in this explanation ⚠️

Several places where the above stops holding, and where the statistic gets over-read:

• Negative beta is not a hedge. A stock that moved opposite the index last quarter carries no obligation to do so next quarter. Treating the label as portfolio insurance mistakes a description for a mechanism.
• The measurement disagrees with itself. Goldman’s ~45% and CNBC’s ~40% cover the same idea with different methods. Neither is wrong; the spread is the error bar, and it is wide enough to matter.
• Weekly-return betas are coarse. CNBC’s calculation uses weekly data. Fewer observations mean noisier estimates, and over three months the observation count is small.
• Breadth is not a timing tool. Divergences of this kind have historically persisted for extended periods. “Unusual” and “about to resolve” are different claims, and only the first is supported here.

The condition under which none of this changes your holdings 🔑

If your position is a broad index fund held on a multi-decade horizon, the three-month beta distribution is close to irrelevant to you. It is a short-window statistic; your holding period is not. The 17% one-year figure is the more relevant lens, and it describes a minority.

What the data does change is your interpretation of what you own. Concentration means the diversification you get from 500 names is smaller than 500 suggests — a point about the structure of your exposure, not a signal to alter it. And when your individual holdings fall while your index fund rises, the arithmetic above is the likely explanation. That is a sampling difference, not a sign that you are picking badly.

The reading that would genuinely change the picture is not a further widening of the divergence. It is a re-convergence on the downside — correlations rising while the mega-cap weights fall. Under today’s dispersion, weakness in most stocks has been absorbed by strength in a few. If correlation climbs back toward one at the same time the heaviest weights turn, the offsetting mechanism Krom described stops working, and the index would begin reflecting the breadth rather than masking it. Whether that happens is not something a beta table can tell you.