Can a free AI tool have an opportunity cost?

Yes. Learning and using it can consume time that has another valuable use.

Is opportunity cost the difference between two benefits?

No. It is the value forgone. The difference compares the chosen benefit with that forgone alternative.

Does previous spending determine what to do next?

Unrecoverable past spending is a sunk cost. The next choice should consider future consequences and still-available alternatives.

AI Glossary / Graphs and symbolic AI / Beginner

Opportunity cost

Definition

Opportunity cost is the value of the best available alternative that is given up when a choice uses a scarce resource. The resource may be money, time, computing capacity, attention, or something else that cannot serve every purpose simultaneously. In AI projects, opportunity cost helps compare a proposed use of resources with what those same resources could accomplish elsewhere.

What it means

A project can have a positive benefit and still be a poor choice if a better feasible alternative must be sacrificed. Conversely, a choice with no direct payment can have an opportunity cost because it consumes time or capacity with another valuable use.

The comparison is between relevant alternatives, not every imaginable possibility. An option that cannot actually be taken is not the next-best available use. Values also depend on the decision-maker's objective and time horizon. Personal enjoyment, service reliability, or learning can matter even when they are not expressed as revenue.

How it works

Identify the scarce resource and the choices competing for it. Estimate the value of each feasible alternative on a consistent basis, including relevant future costs. When one option is selected, the best alternative forgone provides the opportunity-cost comparison.

If several activities can fit together, the alternative may be a feasible combination rather than one project. A scheduling or allocation model may be needed to determine what is actually displaced.

Keep direct expenditure, forgone benefits, and sunk costs distinct. Money already spent and unrecoverable does not become recoverable by continuing a project. Future choices should examine what can still change. Care is also needed to avoid counting the same cost twice when alternatives are already expressed as net benefits.

Three practical examples

Model experimentation: allocating a fixed accelerator window to one training run means giving up the best other experiment that could use that window.
Staff priorities: asking an analyst to polish a dashboard may delay a more valuable data-quality investigation.
Personal learning: spending an evening testing an AI tool uses time that could have gone to another valued activity, even if the tool is free.

Think of it this way

Think of having one meeting room for one time slot. Reserving it for one event prevents another event from using it then. The relevant cost is the value of the best feasible use you displaced.

The analogy is limited because resources can sometimes be divided, rescheduled, or combined. The alternatives must be evaluated under the actual constraints rather than assumed to be mutually exclusive without checking.

Worked example

Suppose a team has one ten-hour computing window and three experiments that each require the entire window. After accounting for their separate direct costs, the team's hypothetical expected net benefit scores are A=120, B=95, and C=70. The scores use one common internal measure and the experiments cannot be combined or moved to another window.

Choosing A gives up B, the best remaining alternative, so the opportunity cost is 95 benefit units. A's advantage over that alternative is 120−95=25. The opportunity cost is not 25, and it is not 95+70=165, because B and C could not both have been performed in the single window.

If the team chooses C instead, it forgoes A's 120 units. C's positive score does not establish that it is the preferred use of the resource.

Now change the constraints: suppose B and C each require five hours and can both fit, with additive benefits and no interference. Their combined score is 165, making that combination a better alternative than A under the revised assumptions. This original illustration shows why feasible alternatives and resource constraints matter. The numbers are invented decision scores, not measured project returns or investment advice.

Illustrative case study

Fictional case: an AI team argues for continuing a demonstration because it has already spent several weeks building it. The remaining work would use the same specialists needed to fix an important data pipeline.

The manager separates unrecoverable past effort from the future effort still required. The team compares the expected value and urgency of the two remaining uses, including dependencies and deadlines. The case illustrates a forward-looking resource comparison; it does not imply that demonstrations are generally less valuable than infrastructure work.

Common misconceptions and mistakes

Opportunity cost is not the sum of every rejected option. It concerns the best feasible alternative, which may itself be a combination when the resource permits one.

It is not necessarily the amount paid. A prepaid resource can still have a valuable alternative use, while an apparently expensive choice may be justified by benefits that exceed the alternatives.

Do not mistake hindsight for the information available when a decision was made. Under uncertainty, comparisons should reflect the evidence and objectives available at that time. An unexpected outcome can motivate updated estimates without proving that the earlier choice ignored its opportunity cost.

Business use

For AI portfolios, opportunity-cost reasoning helps make tradeoffs visible when teams face limited budget, expertise, compute, or attention. It encourages comparison with meaningful baselines, including postponement or a simpler solution where those are feasible alternatives.

Teams should state which resources are genuinely scarce, what the proposed work displaces, and how benefits are estimated. Sensitivity analysis can show whether a choice changes under plausible assumptions. Not every value can be reduced credibly to money; transparent qualitative tradeoffs can be more useful than invented precision in a spreadsheet.

Check your understanding

Why does the best alternative change when experiments B and C each need only five hours?

Answer: They can now fit together in the same ten-hour window. The feasible alternative is their combined benefit of 165, rather than only B's 95.

Related terms

Expected utility
Decision analysis
Scarcity
tradeoff
sunk cost
marginal analysis
resource allocation
portfolio prioritization

Learn next

Choose one proposed AI activity and name the resource it uses, the best feasible alternative, and the evidence behind both values. Check whether alternatives can be combined or delayed, and distinguish direct costs from benefits already expressed on a net basis.

Sources

Federal Reserve Bank of St. Louis, Three Ways to Think Like an Economist: https://www.stlouisfed.org/open-vault/2020/march/three-ways-think-like-an-economist — opportunity cost, sunk costs, and marginal decisions. The computing-window example and project scenario are original educational illustrations with explicit feasibility assumptions.

Previous: Opportunistic reasoning Next: Opt-in