Should you sell or refuse to buy a stock when it seems overvalued?

Published: 2025-09-24 02:45:52 ET ORCL

When I started writing this article, Oracle (that law firm disguised as a software publisher) soared 40% on the stockmarket after the publication of unexpected quarterly results, and notably a colossal order book. The stock was then trading at 50x earnings and 11.5x expected revenues in 2026… What an excellent introduction to this paper.

Etienne Monceau

Published on 09/24/2025 at 02:45 am EDT

In the markets, investors generally have high expectations of growth companies. These expectations are reflected in valuation levels. When you pay 50x next year's earnings for a stock, it's because you hope that earnings growth will be very strong in the years to come. Analysts, meanwhile, are often conservative. The accounting estimates they produce tend to cause upward surprises on publication day.

However, over the years, investors have come to understand this (that analysts are conservative). So much so that upside surprises are no longer really surprising. During the last earnings season, dozens of companies saw their share prices fall even though they reported figures above consensus on results day (Salesforce, SAP, Atoss Software, Amazon, Nvidia, Tesla, etc.). That said, the decline can sometimes be explained by information or concerns shared in the accompanying communication rather than by the quarterly figures themselves.

In short, in this context, where humanity seems to be on the verge of completing its latest invention (the next ones will be signed "your AI friend"), investors have become very demanding when it comes to growth stocks! And yet, as Oracle proves, some companies that are already highly valued are capable of gaining 40% in a single trading session.

But beyond a specific one-day movement, what surprises me most is the ability of these companies to reach ever higher highs...

Like you, the first finance books I read were by Graham, Buffett, Lynch... Intrinsic value, not God!

Then, five years later, I changed my religion and joined the data gogos. At data gogos, we like good stories, but above all we love data. My goal is no longer to build the typical portfolio of the rational investor, with a financial explanation behind each line in order to convince my friends that I'm doing things right (and that it's the fault of other investors if I underperform). My one and only goal is to understand how those who move the stockmarkets think and how the movements they initiate spread across the markets. This is with the sole aim of jumping on the bandwagon, making sure there are still a few stops left before it arrives.

And this conversion has not been without difficulty, as the truth can sometimes be very hard to hear.

Typically, in the quantitative strategies I have been working on in recent years, it is sometimes necessary to buy at the highest prices. No, but buying at the highest prices, can you believe it! What next? Not selling a stock when it is trading at 30x free cash flow for the current fiscal year? Well, believe it or not, yes...

At least over the last 15 years, since the markets haven't experienced a real major financial crisis, selling because something is expensive didn't seem like a good idea. Especially in Europe!

And it may well be that this is not just speculation in the sense of I'll find some idiot willing to buy at a higher price, but speculation on events that are likely to occur.

Understand that Oracle's order book has literally exploded. +360%, with $455bn in stock. To paraphrase my colleague Anthony Bondain, "The icing on the cake is that this growth, which analysts did not see coming, is notably based on cloud server rentals (whose revenues will increase from $10bn to over $140bn between 2025 and 2030, to give you an idea)."

From $10bn to $140bn in revenue for a business unit (I wouldn't say marginal until now, but almost) in the space of five years. Crazy.

The same goes for Palantir, which, after launching its Artificial Intelligence Platform (AIP) in the private sector, is likely to see its revenue triple in the space of four years, rising from $2.2bn in 2023 to $7.6bn in 2027 (Consensus based on estimates from 14 analysts as of September 10, 2025).

And let me stop you right there: it's not just American tech companies that are experiencing this kind of paradigm shift. It's the entire value chain. HVAC companies (Amphenol Corp, Schneider Electric, Comfort Systems, Vertiv Holdings, etc.), industrial companies, are trading at valuations never before seen in the sector.

And it's not over yet: it seems that the disregard for valuations is spreading to an even larger part of the market. And this is not a recent phenomenon, quite the contrary.

The four curves below represent the performance gap between low-valued and high-valued companies according to our Global Valuation in four major regions: North America, Europe, Japan, and Asia (excluding Japan). In the same way that E. Fama and K. French constructed their HML premium in their Asset Pricing model, we will retain this idea of annual rebalancing. In other words, each year, we will round things up and buy what is value and short what is expensive. We also control for size to properly isolate the value premium. However, unlike the researchers, we base our calculations on projected valuations rather than historical valuations.

Value premium in four major regions. Methodology: E. Fama and K. French (1993) with projected valuations. Source:

You're not crazy, what you see here are curves that are falling for all regions. However, there has been a plateau in recent years, particularly in Europe, with a recovery of value investing.

So, you might say that with annual rebalancing, we are not really embracing the logic of value investing. That investors who buy securities because they believe they are undervalued are sometimes willing to hold them for much longer than a year. And you are absolutely right.

But then how can we explain that it is this methodology that has led to the creation of asset pricing models that best explain the variance in stock market returns? In a way, by favoring this methodology—with annual rebalancing—over others, what researchers are telling us is that most investors seem to invest with a much shorter horizon than we would be inclined to believe.

And if you look closely, if we separate the value premium between big caps (large companies) and mid-small caps (small and medium-sized companies), the observation remains much the same: it has been a long time since we have seen significant and sustained outperformance of value companies over growth companies.

Value premium among small/mid caps and large caps in Europe and North America. Methodology: E. Fama and K. French (1993) with projected valuation. Source:

And here's a little bonus information. As you will have understood, these curves are the result of a long/short strategy, and what explains their decline is mainly the part where we sell companies that are expensive, rather than the part where we buy value stocks. So it would seem that what destroys performance is mainly selling when prices are too high or refusing to buy a stock because it is trading at too high a valuation ratio.

Now let's try something that may be more familiar to you: we're going to look at the practical consequences of the academic conclusions we've just drawn.

To do this, we will construct a strategy that involves sorting the US and European universe according to our Composite Valuation Rating and buying the top 50 stocks. Every day, we will check that the stocks still belong to this top 50. If this is not the case, the stock is immediately replaced by the outsider. There will be no rebalancing of any kind, no exit conditions other than those described above, no sector exposure control, and no take profit or stop loss.

Results of a backtest of a Quality strategy based on the Composite Fundamentals Rating. 2013-2025.

We can see that between 2013 and today, this strategy has outperformed the MSCI World, with a Sharpe Ratio of 0.75. The average holding period is 52 days, which is fairly realistic.

Now we are going to apply a new entry condition as well as an exit condition based on valuation. To enter the portfolio, a company must be among the 80% cheapest companies according to our Composite Valuation Rating. And if a company ends up in the top 20% of the world's most expensive companies, it is ejected.

Results of a backtest of a Quality at a Reasonable Price strategy based on the Composite Fundamentals Rating. 2013-2025.

So we now apply a filter on valuation. And what can we conclude from this? Quite simply that the strategy is much less effective. We end up with a Sharpe Ratio of 0.25.

It would therefore appear that over the last 15 years, selling because something is too expensive, or not buying because something is too expensive, was not the right thing to do.