Highest P/E ratio stocks: near-zero EPS

Sorted by trailing P/E on 2026-09-07, the fifty highest P/E ratio stocks in the US market had a median share price of 24.21 dollars and, for the 42 that published a ratio, 30 implied trailing earnings under ten cents a share. A very large price to earnings ratio is almost never a statement about the price. It is a statement about a number near zero on the bottom of a fraction, and the arithmetic of dividing by something near zero does the rest.

Those fifty listings share a shape. Their share prices are unremarkable. The median price among the fifty was 24.21 dollars and 28 traded above 20 dollars, on prices stamped 2026-08-26. They are not small: the median market capitalisation was 2.05 billion dollars, 29 were above one billion and nine were above ten billion, the largest at 81.04 billion. What they have in common is that last year's earnings came to almost nothing.

The ten largest, with the denominator reconstructed

The implied trailing earnings per share below is the share price divided by the published ratio. It is arithmetic on the two figures the screener served, not a separate reading of the income statement, and it recovers the earnings figure the ratio was built from.

Every implied figure lands on a round number of cents, which is what you would expect if the ratio was built from an earnings per share figure reported to two decimal places. A company that earned two cents a share and trades at 38 dollars has a P/E of 1,922.50 whatever anybody thinks of it. Move the earnings to three cents and the ratio falls by a third, without the price moving at all.

Across the whole fifty, 42 published a ratio and 30 of those 42 imply trailing earnings below ten cents a share. Eleven imply less than two cents.

Eight of the fifty had the ratio withheld

The other eight are worth a paragraph, because they show the same thing from the other side.

Sorting is done on the stored figure. Publishing goes through an applicability check that can refuse a figure for a particular filer, and eight of the fifty were refused: three blank-cheque acquisition companies, a lodging trust and four others. Their stored ratio put them in the top fifty and the served row printed no ratio at all.

That is the correct behaviour, and it has a consequence for every count on this page and on any other. The counts here are over stored values. A listing whose displayed P/E is withheld still answers a numeric filter, so it still appears in a screener count even though no reader would see a number for it. The gap between what is counted and what is shown is small in the middle of a distribution and large in the tail, which is exactly where these fifty live.

Where the tail sits in the market

Of the 2,359 US listings that held a P/E and a sector on 2026-09-07, 169 carried a ratio of 100 or more and 68 carried 250 or more. That is 2.9 percent of the profitable sample above 250.

The other end of the same distribution is more instructive. The median listing sat between 20.75 and 20.94, and 766 sat below 15. The tail is thin and it is very long: the largest ratio in the sample is more than a thousand times the median. Any average that includes it is a number about one company.

That median is a count over listings. The published distribution puts the same day at 21.16 over deduplicated companies, and the gap between the two is the same market counted two different ways.

What a high multiple is and is not evidence of

It is evidence that the trailing year was a poor one for earnings. That is the arithmetic. Whether the poor year is the start of something, the end of something or an accounting artefact is a question the ratio cannot answer, and it is the only question that matters. Whether cash did the same thing that year is a separate reading.

It is not evidence that the shares are expensive. Take a share at 40 with two cents of earnings this year and fifty cents in a normal year: the headline ratio is 2,000 and the ratio against normal earnings is 80. Those are round illustrative numbers and not any company's, and the point survives whatever you put in them. The ratio you can compute is not the one that describes the business.

It is not comparable to anything. Two companies with ratios of 1,800 and 5,900 do not differ by a factor of three in any respect a reader cares about. They differ by two cents of earnings per share.

The measure built for this problem is the cyclically adjusted ratio, which divides by an average of ten years of earnings instead of one, and a normalised version that uses five. Both exist for far fewer listings than the trailing ratio does, which is its own kind of answer.

What this leaves out

Every figure is one reading of the public screener on 2026-09-07, on prices stamped 2026-08-26. Prices move, and a ratio built on a two-week-old price is a two-week-old ratio.

The stored P/E prefers a data vendor's published trailing twelve month ratio where one exists, and falls back to our own calculation from filed earnings where it does not. The implied earnings per share reconstructed above therefore reflects whichever of the two produced the ratio, and it is not independently read off a filing here.

The fifty rows are the fifty highest by stored value, which is not the same population as the fifty highest by displayed value. Eight of them display no ratio at all.

Naming a company here is a description of an arithmetic result on a date. It is not an assessment of the business, a comparison with its peers, or a view on its shares. Nothing on this page is advice.

Ratios are per listing. A company with two share classes appears twice, with the same earnings behind both.