What exactly has been marked down?
The multiple, not the earnings. Across more than 200 listed software and white collar services companies, the 60 sitting in categories judged exposed trade at a median 8.1 times earnings before interest, tax, depreciation and amortization, against a five year average of 14.1 times. That is a compression of roughly 40 percent.
The forecasts underneath those same companies have not moved with the price. Of the more than 200 tracked, only 10 are currently expected to record declines in both revenue and earnings over the next two years, and the median expected earnings decline across those 10 is 3.4 percent. A category has been repriced. Almost nobody in it is forecast to shrink.
The 10 have been marked down harder still. They trade at a median 4.1 times earnings against a five year average of 9.1 times, a compression closer to 50 percent. For a decline that consensus puts at 3.4 percent, the market has taken out roughly half the multiple.
The operating record over the same period runs the other way. Software vendors posted record combined profits in the first quarter of 2026 and median revenue growth of 14 percent. Fundamentals held. The price did not.
| Measure | Reading | Date or period |
|---|---|---|
| Tracked listed software and white collar services companies | more than 200 | June 2026 |
| Of those, in categories judged exposed | 60 | June 2026 |
| Of those, forecast to decline on both revenue and earnings | 10 | To 2027, consensus |
| Median forecast earnings decline across those 10 | -3.4% | To 2027, consensus |
| Median multiple, exposed categories, current | 8.1x | June 2026 |
| Median multiple, exposed categories, five year average | 14.1x | Five years to June 2026 |
| Median multiple, the 10 forecast decliners, current | 4.1x | June 2026 |
| Median multiple, the 10 forecast decliners, five year average | 9.1x | Five years to June 2026 |
| Median revenue growth, software vendors | 14% | Quarter to 31 March 2026 |
If the forecasts are unchanged, why is the multiple down?
Because the two numbers answer different questions on different clocks. A forecast extends what a business is doing now over a defined window. A multiple is what a buyer will pay today for a claim on everything after that window, which is where a durability question lands.
Apollo's own reading of the gap is that investors have begun pricing in the possibility of disruption without fully incorporating its potential effect on earnings, margins, and long term credit fundamentals. On that account the market is not disagreeing with the forecasts. It is discounting what it thinks the forecasts have not yet reached.
That leaves two honest resolutions and no way to pick between them yet. Either the market is early and the discount unwinds as the forecasts prove out, or the forecasts are late and the earnings follow the multiple down. Both are consistent with everything above.
What can be said without choosing is that the sequence has already happened. The repricing came first, it was applied at the level of the category, and it did not wait for a single set of results to confirm it.
“A category discount is not a verdict on a company, and owners are far too quick to accept it as one. What it tells you is that the buyer has reached a view about durability that your trailing numbers do not address, because trailing numbers never address durability. The work is not to argue the multiple down. It is to produce the evidence the multiple is standing in for, and that evidence takes years to accumulate and a fortnight to look self-serving.”
Does this stop at software?
Apollo says it does not, and names where it expects the pressure to travel. Software is described as the first sector where markets have begun to price the distinction, not the only one exposed to it. The categories flagged next are business services, including staffing, consulting, contact centres and business process outsourcing, and the administrative infrastructure of healthcare.
The stated logic is about what a business monetizes rather than what industry it reports in. Where revenue is tied to the labour, billable hours or seats required to deliver a service, the exposure is direct. Apollo's expectation is that value shifts away from businesses that monetize labour and toward those supplying the data, software and infrastructure that automate it.
Three channels are set out. Direct replacement, where the same task is performed at a lower cost per outcome. Labour displacement, where demand falls for the employees, contractors or users a revenue model depends on. And execution risk, where faster moving competitors take share.
Note that these are categories of exposure, not findings about any particular business. A staffing company with contracted, regulated or hard to substitute work sits in a flagged category and may carry none of the exposure the label implies. The label is applied first and tested later, which is the whole problem for the owner.
How should an owner read a discount applied to their category?
As a starting position taken by someone else, not as a measurement of the business. Apollo's universe is listed companies, and it says nothing about private middle market processes. The read-across is ours: a private company is valued against listed comparables, so when the comparable set is marked down by roughly 40 percent, the reference point a buyer opens with has moved even though nothing in the company has.
It changes what is worth proving. If the discount is a durability question wearing a valuation costume, then evidence about durability is the only thing that answers it. Contracted revenue with real switching costs, retention measured through a full renewal cycle rather than a good year, and revenue that is not a function of headcount are worth more than they were when the whole category was assumed durable.
It changes when that proof has to exist. Durability evidence is a record, and a record cannot be assembled inside a live process. Produced during diligence it reads as advocacy. Produced from three years of consistent reporting it reads as fact, and the difference is entirely in the timing.
It also changes what to ask a buyer. The useful question is not what multiple the market is paying. It is which comparable set the buyer is pricing against, and whether the business has been placed inside or outside the categories that set has been marked down for. That is a question with an answer, and it is rarely volunteered.
What would settle the disagreement?
Consensus forecasts moving. The cleanest single indicator is the count of tracked companies expected to decline on both revenue and earnings. At 10 out of more than 200 the forecasts are effectively saying the discount is wrong. A rising count would mean the market was early and the analysts have started to agree with it.
Failing that, dispersion inside the flagged categories rather than the level of the discount. A category discount is a statement that buyers cannot yet tell members apart. When they can, the strong and the weak separate and the median stops being the story.
Treat one publication as one publication. This is a single large manager's analysis of a universe it has defined itself, with the exposed categories chosen by its own analysts. The figures are worth reading because the framework is explicit and the underlying data is disclosed, not because a manager's categorization is authoritative.
The practical conclusion is about sequence again. The multiple moved before the earnings did, and it moved on a category rather than a company. The interval between those two events is the window in which a business that can document its own durability still gets valued on its record instead of its label.
As of August 2026
Sources: Apollo Global Management, 2026 Midyear Credit Outlook: Adoption, Financing, and Investing in the Age of AI, published August 2026, read from the whitepaper, for the statement that of more than 200 publicly traded software and white collar services companies tracked, only 10 are currently expected to experience both revenue and earnings declines over the next two years; and from Exhibit 17 of the same paper, sourced there to Capital IQ and Apollo analysts with data as of June 2026, for the count of 60 companies in high risk subsectors, for median valuations across those subsectors having compressed by approximately 40 percent against five year averages at 8.1 times current against a 14.1 times five year average, and for the companies with sell-side projected earnings declines at a median of 3.4 percent having compressed by approximately 50 percent at 4.1 times current against a 9.1 times five year average; and from the body of the same paper, attributed there to Jefferies, May 2026, for software vendors posting record combined profits and median revenue growth of 14 percent in the first quarter of 2026; and from the same paper for the statement that investors have begun pricing in the possibility of disruption without fully incorporating its potential impact on earnings, margins and long term credit fundamentals, for the framing of exposure through direct replacement, labour displacement and execution risk, and for the identification of business services including staffing, consulting, contact centres and business process outsourcing, together with the administrative infrastructure of healthcare, as the categories expected to follow software. The read-across from a listed comparable set to a private company being valued against it is our inference and not a statement by Apollo. The recommendations about what evidence to build and when are our own judgement and carry no figure. A companion article on this site reads the same sector divide as it appears in secondary loan prices rather than in equity multiples, and another covers how borrowers have been extending maturities rather than resetting terms.