It is possible to be right that demand will grow and wrong about who will make money. That is the uncomfortable part of an infrastructure boom.
The equipment may be necessary. The buildings may fill. Customers may use the service. Shareholders can still earn disappointing returns if too much capacity is financed at the wrong price, if utilization arrives late or if competition transfers the savings to customers.
I would separate the operating achievement from the investment result. A company that completes a difficult build deserves credit for execution. Whether the shares were a good purchase depends on what that achievement cost, how much cash it produces and the expectations embedded in the entry price.
That is why I am less interested in asking management whether demand is strong than in asking what return the next dollar of spending is expected to earn. The answer should include downside assumptions, not just the project’s best utilization case. Friday’s weakness was not proof that every AI investment was misguided. It was a useful reminder that a popular demand forecast is only the first line of an investment case.