Sunk Cost Fallacy
The sunk cost fallacy is a thinking error: people stick with something because they've already put a lot of money, time, or effort into it. The rational approach would be to decide based only on future benefit, since what has already been spent won't come back regardless.
You bought a movie ticket for 12 euros. After twenty minutes you realize: the film is bad. Still, you stay in your seat because you already paid for it. That’s exactly the sunk cost fallacy. The 12 euros are gone whether you stay or leave. They shouldn’t influence your decision at all anymore. The English expression literally means “the fallacy of sunk costs”: sunk means the money has been irretrievably spent.
Why billion-dollar projects keep running even when everyone knows better
On a large scale, this thinking error costs enormous amounts of money. Berlin Brandenburg Airport is the most well-known German example. Construction continued for years, even as costs rose from around 2 billion to over 6 billion euros. Cancellation was politically almost unthinkable because so much had already been invested. That very logic is the mistake.
In business, this pattern is also called escalation of commitment. This means a project team keeps pouring in more resources instead of pulling the plug. Behind it often lies not just poor calculation but shame. Whoever ends a project publicly admits that the money spent so far was wasted. Continuing pushes that admission into the future.
For investors, the effect is especially costly. Many hold on to a stock that has fallen sharply, out of principle. But the real question isn’t what you once paid. It is: would I buy this stock today at the current price? Whoever answers this question with no has effectively already made a decision to sell.
What goes wrong in our heads
Economics has a clear rule: only future costs and future benefits count. Everything already spent is irrelevant to the decision. But our brain doesn’t work by this rule. It prefers to count what has already been invested and doesn’t want to book that stake as a loss.
One important reason is so-called loss aversion. Behavioral research uses this term to describe how a loss feels about twice as intense as an equally sized gain. As long as the project continues, the loss isn’t final yet. One can convince oneself that things will still turn out fine in the end. Cancelling, on the other hand, makes the loss immediately real and tangible.
Added to this is the desire to appear consistent. Once someone has said yes, they don’t want to seem erratic. Economists Hal Arkes and Catherine Blumer demonstrated this in a famous 1985 experiment involving theater subscriptions. Those who had paid full price attended performances significantly more often than those who had randomly received a discount. The benefit of the performance was identical; only the amount already paid differed.
From phone contract terms to data centers
You encounter this effect constantly in everyday life. You finish playing a video game that’s no longer fun. You stay in a course that isn’t helpful because you’ve already paid the fee. You finish the plate even though you’re full. In all these cases, the stake is already lost.
In tech and finance news, the term comes up especially often in connection with AI. Companies pour billions into data centers and into training large models. When a model falls behind the competition, the question arises: improve it or start over? The compute costs already burned shouldn’t play a role here. What matters alone is which path yields more from today onward.
A common misconception, by the way, is to consider every cancellation reasonable and every persistence foolish. That’s not true. If a project will actually yield more in the future than it still costs, continuing is the right choice. The fallacy only applies when the past is the sole argument. A practical counter-check is: would I start this project today from scratch if I were starting at zero?