This week I want to talk with you about an idea that has been developing in my mind about how the market has behaved ever since AI became a significant factor in the workplace. My central insight going forward is this: multiples, as a broad and general matter, will come down over the years, and short-term standard deviations will rise. You can see first evidence for this theory in the very sharp moves of large stocks in recent weeks: Alphabet lost about 15% in the days after its report at the end of July, Meta fell close to 10% in a single trading day, and HubSpot and Atlassian dropped close to 13% and 12% within two days, with no report and no change in their profit at all, simply because OpenAI presented a new tool that walks into their territory. A good example of multiple compression can be found at Nvidia and IBM, two stocks I addressed separately in earlier letters.
This week's letter is a bit more theoretical, and a bit "deeper" relative to everyday concepts, but I'm here to stop and explain the terms so that we're all aligned as we move further into the piece.
So let's start with a short explanation of the term multiples. If I own a business that earns $10 every month, and someone is willing to buy the business from me for $100, he is effectively paying today at a price of 10 years of profitability ($10 of profit a year, times 10 years forward, equals $100). This buyer expects that after holding the business for 10 years, he'll have made his investment back, and from there on it's all profit. That expectation rests on the assumption that this business will know how to preserve its profitability over many years.
In the same way, let's talk now about a case where the buyer is willing to pay $500 for my business. That is, 50 years of profit. Is he really building a business plan that lets him start earning 50 years from now? Probably not. What this buyer is actually saying (through the price he's willing to pay) is that he believes that after the purchase, he'll manage to bring the business to far higher profitability, and let's set aside for a moment the ways of doing that. So he is in fact looking not at the current profitability of the business ($10), but, for example, at profitability of $50 a year — after the moves he expects to pull off. And here we're back at a multiple of 10 ($50 of expected annual profit, times 10 years).
What am I really saying? That there are 2 types of multiple deals. The first type buys an existing state of affairs, meaning it assumes stable profitability over the years, and is therefore willing to price at relatively low multiples (around 10), and the second type sees growth and is willing to pay a "future multiple" for it, meaning to buy today at a much higher multiple that is expected to come down as the company grows, bringing it to a more reasonable multiple relative to the price that was paid.
The truth is there's one more type, a third one, and that's the multiple of a shrinking company — deals that assume profitability will decline over the years, and there the logic obviously works exactly in reverse: the multiple paid will be very low (single digit, sometimes even close to 1), out of an understanding that the profitability won't survive over time and so the buyer needs to earn the purchase price back as close as possible to the deal itself. But we'll talk about that less today, because the market by its nature houses fewer companies like that (they usually disappear from it over time).
How does the multiples discussion connect to AI? Glad you asked. Through the uncertainty component. The greater my certainty, as a buyer, about the future profitability of companies, the more I can "afford" to pay a higher multiple. Meaning, to build on future profits. In the opposite direction, if I'm not sure that what worked today will also work tomorrow, I'll be willing to pay a lower multiple, because I'm not "locked in" on how much this company will keep growing and generating profits for me.
If I look today at the reaction of entire sectors — mainly technology and software, but not only — that's exactly what I see. Companies that keep making money and sometimes even grow aggressively are falling tens of percent in their share price. Monday.com is an excellent example. Its stock lost a little more than half its value in the first half of 2026, while revenue in the first quarter grew 24% against the same quarter a year earlier, and its operating profit doubled. At the end of July it announced an efficiency plan and a cut of about 20% of its workforce, the stock fell another 8% that day, and in that very same announcement the company reaffirmed growth guidance of 19% to 20% for the full year. Nothing in the business itself broke. What broke is investors' willingness to pay for its future. This is essentially investors' way of saying: we have no idea whether this business won't be replaced by AI, we have no confidence in its level of profitability, and therefore we won't be willing to pay high earnings multiples, even though the profitability is already here, and it has grown (past tense) quarter after quarter until now.
In my estimation, a few years out from today, there won't be a business field that AI doesn't work its way into and change, and so the multiples event is in fact a market-wide event, not something unique to technology stocks. What we've seen there in recent months is likely to seep across the whole market. And this thing has very far-reaching implications:
The market is priced by investors through a number of parameters, and the central, longest-term one is the earnings multiple. Many times, when investment managers debate whether the market is cheap or expensive, you'll hear the line: "the long-term multiple of the S&P is 19, similar to the current market, and therefore stocks aren't expensive." When the current multiple rises significantly above the long-term multiple, that's a strong signal to investors to reduce exposure, and vice versa. I don't know many investment managers who would add to a market with an average multiple of 30, and I also don't know managers who wouldn't buy deep into a market at a multiple of 10.
That figure, in my view, is going to lose its credibility, and in fact the market's representative multiple, according to my thesis, is going to come down, for the reasons I listed earlier. If that's the case, an (average) stock at a multiple of 20 may trade in the future at a multiple of 12. So first of all, that pulls the ground out from under the benchmark for what a stock is worth — are stocks cheap or expensive? — but more than that, a stock that traded at a multiple of 20 and moves to trading at a multiple of 12 is in fact worth 40% less in price! Going back to that same example, of the business earning $10 a year, if instead of a multiple of 20 (that is, $200), investors are willing to pay a multiple of 12 only (that is, $120 in value terms), we're talking about a stock that falls from $200 to $120.
Which brings me to the second point I wanted to talk about — standard deviation. In the market, we use this term for the short-term movement of a stock. In simple language, how "hard" it rises or falls in a given period. Standard deviation is measured against the average move, meaning: if on an average day a stock (say Apple) moves 1% (up or down), then a move of 8% in a day is a standard deviation very far from the average. If the stock moves 8% on an average day, then a move like that is normal for it (which of course doesn't exist in this market).
My thesis says that investors are trying as hard as they can, and not necessarily successfully, to carry out the "multiple adjustment" as fast as possible, and therefore — given a piece of information released to the market (hello there, IBM stock) they simply "cut" the stock, sometimes over-aggressively, so as not to be stuck holding a stock that's about to go through a multiple adjustment (meaning, that its earnings multiple will come down as a result of uncertainty about future profits). We saw (and I wrote a letter about it) IBM falling 25% in a day, not because of bad reports (it kept making money and even grew its profits), but simply because of a fear about a change in the structure of its future sales, driven by customers shifting money to other spending, the kind suited to the AI world.
In that very drop, investors effectively "corrected" IBM's multiple from 22 to 16, and turned it into a "cheaper" stock, on the claim that you can build less on future profitability, and in such a case, as I explained, the multiple contracts. That strong drop was what we call in the market a high standard deviation. I've seen this phenomenon in recent weeks at Apple, Amazon, Nvidia and other stocks as well — not necessarily always downward, because the market is also trying to guess the "winners" and rewards them with pinpoint aggressiveness too, but in general, as long as this situation continues, in the short term I expect these swings to continue and even intensify, and in the long term — as astonishing as it sounds — the market may fall by significant percentages, if a multiple adjustment does indeed take place.
What can, or needs to, happen for the market not to fall despite the "multiple adjustment"? If on average a traded company grows its profits significantly, then the multiple adjustment gets offset, and the stock doesn't necessarily fall, or may even rise. An example — my business, the famous one from the start of the piece, earns $10 a year and trades at a multiple of 20 — that is, it costs $200. The multiple adjustment (say 40% down) brings it to a multiple of 12 and a value of $120. But if I manage, using AI, to double the annual profit and bring it to $20, then at a multiple of 12 it will already be worth $240. Meaning, despite the drop in the multiple, the stock rises from $200 to $240 (a gain of 20%), because of the improvement in profitability.
And if you ask me — investors in the market are building on exactly that today. Otherwise, they're facing significant losses as future multiples contract.
The truth is I have plenty more thoughts and extensions to this thesis, and I'll be glad to share them in future letters. Stay tuned, we'll get there yet.
We shall see.
See you in next week's letter, or during the week — if I can't hold myself back until then...