A sale looks like the simplest thing in commerce. Someone hands over an apple and someone hands over a coin, and when it is done the apple has changed hands. The seller no longer has it. The buyer owns it, completely, to eat or resell or throw away, and the transaction closes behind them both. That is what the word meant for most of the history of trade. A transfer.
Most of what the modern economy calls a sale is nothing like this. When a company sells a data feed, or a piece of software, or access to a model, it still has the thing the moment the deal closes. It can sell the identical feed to a hundred other buyers that same afternoon, and it will. Nothing was handed over, because nothing left the seller’s hands. What moved was not the good. What moved was permission to use it.
Permission bounded by terms is a license, and the difference from a sale is not a matter of legal wording. It is a matter of what the good is. Some things cannot be transferred, because using them does not use them up. Economists call such goods non-rival: one person’s use takes nothing away from another’s. An apple is rival. A theorem is not. A person can give an apple away, but no one can give away a theorem, because after giving it they still have it. Nearly everything the technology economy runs on lives on that side of the line.
This puts the usual picture backwards. Business treats the sale as the normal case and the license as the exotic corner, the thing the lawyers handle for software. It is the reverse. Once a good is non-rival, the license is the only instrument available, because a sale in the true sense has become impossible: no one can hand over what they still hold. The outright sale, the apple, is the special case. It is the rare transaction where the permission granted happens to be total and permanent, so total that no one bothers to write the terms down. Licensing is the general form of commerce. The sale is the version of a license whose terms are so complete they turn invisible.
The sale is retreating even where it still looks solid. A tractor is sold, and the farmer owns the steel and can paint it any color. But the software that makes the steel run is licensed, and it can be gated, throttled, or switched off, which is how someone who owns a tractor outright can find they are not allowed to repair it. A car can leave the lot with features already built into the metal that stay locked until a separate payment unlocks them. The atoms are sold. The function is rented. What looks like ownership is more and more a shell around a license, and the part that makes the object do anything never transferred at all.
Once this is seen, the strangeness of selling data and software stops being strange. The transaction never really closes, because a bounded permission has an inside, an edge, and an expiry, so it has to be scoped and watched and renewed, and the relationship persists long after an ordinary sale would have ended. The price floats free of cost, because making one more copy costs almost nothing, so there is nothing for the price to anchor to, and it becomes instead a reading of what the buyer will bear. This is why the enterprise deal carries no sticker and the whole of it is negotiated. None of that is a quirk of the industry. It is what selling a non-rival good has to look like.
The word sale survives out of habit, carried over from an economy of apples into an economy of permissions. It is a comfortable word, and it lets everyone pretend the transaction ends when the signature dries. It rarely does. What most companies now do for a living is not sell. It is grant, and bind, and renew, over and over, on a thing they never actually let go of. The word has not caught up to the work.
Touchstone: the economics of non-rival goods, and Paul Romer’s point that ideas, unlike things, can be used by everyone at once without being used up.