Why Software Costs Nothing to Copy but Always Has a Price

The Simple Version

Once software is built, sending a copy to one more customer costs the company essentially nothing. But that has never stopped any software company from charging real money, and for reasons that are more defensible than most people realize.

What Economists Mean by Marginal Cost

Marginal cost is the cost to produce one additional unit of something. For a car manufacturer, each additional car requires steel, labor, paint, and assembly time. For a software company, each additional copy of their product requires a few milliseconds of server time and a database entry. The cost rounds to zero.

This is not an approximation. If you buy a seat of Microsoft Excel today, Microsoft’s cost to fulfill that transaction is genuinely close to nothing. The servers are already running. The software is already written. The bandwidth cost to download the installer is fractions of a cent.

Economists find this interesting because classical pricing theory says prices should settle near marginal cost in a competitive market. If it costs you nothing to make one more and nothing to make another, why would anyone pay more than zero?

The answer is that marginal cost and total cost are not the same thing.

The Fixed Cost That Changes Everything

Building software is expensive. Distributing it is nearly free. That split is what makes the software business structurally unusual.

Microsoft spent roughly $25 billion on research and development in its 2023 fiscal year. That money paid for engineers, product managers, designers, security researchers, and data centers running tests. Not one dollar of it maps cleanly to any single customer. It is a sunk cost, spread across everyone who ever buys anything Microsoft ships.

This is the economic reality that zero marginal cost conceals. The first copy of software might cost tens of millions of dollars to produce. The second copy costs almost nothing. The millionth copy costs almost nothing. But someone has to pay for that first copy, and that someone ends up being spread across all the copies that follow.

Software pricing, then, is fundamentally a cost-recovery exercise dressed up as a market transaction. The question companies are actually answering is not “what does it cost us to serve this customer” but “how much of our fixed costs can we recover from this customer given what they’re willing to pay.”

Why Willingness to Pay Varies So Dramatically

Because marginal cost is zero, software companies have enormous pricing flexibility. They are not constrained by material costs the way hardware companies are. This is why the same software can sell for $10 a month to a freelancer and $500,000 a year to an enterprise, and both prices can be rational at the same time.

Salesforce is the cleanest example. The company sells what is, at its core, a contact database with workflow automation. A small sales team might pay a few hundred dollars a month. A Fortune 500 deployment can run into eight figures annually. The underlying software is largely the same. The difference is support contracts, customization, integration work, and what the customer can afford to pay relative to the value they’re capturing.

This is not price gouging. It is a structural consequence of zero marginal cost combined with variable willingness to pay. When your cost to add a customer is near zero, you have every incentive to serve as many customers as possible at whatever price they’ll accept, as long as that price exceeds zero. The result is tiered pricing, enterprise contracts, and the byzantine packaging that software companies are known for. The engineers who design those pricing pages are doing something economically meaningful, not just marketing.

The Competitive Pressure That Keeps Prices Honest

Zero marginal cost should, in theory, lead to a race to the bottom. If your competitor can also produce software at near-zero marginal cost, why won’t they undercut you until prices collapse toward zero?

Sometimes they do. The history of consumer software is littered with categories that became free: email clients, basic office tools, navigation apps, photo editors. Google and Apple absorbed entire software markets into their operating systems, pricing them at zero because the distribution value outweighed the direct revenue.

But this doesn’t happen uniformly, for two reasons. First, switching costs are real. Enterprise software in particular builds deep integrations into a customer’s operations, making migration painful enough that vendors can sustain prices well above what pure competition would predict. Second, the software that doesn’t become a commodity is usually software that requires ongoing development to stay useful, which keeps the fixed cost base large and makes it hard for any player to sustainably price at zero.

The companies that get this wrong, that price too low thinking they’ll make it up in volume, often find themselves unable to fund the ongoing development that would make the product worth keeping. Pricing too low is its own kind of trap.

What Zero Marginal Cost Actually Means for the Market

The more interesting consequence of zero marginal cost is what it does to market structure. Industries with high marginal costs tend toward many competitors, because production constraints limit how big any one player can get. Industries with zero marginal cost tend toward concentration, because the player who recovers their fixed costs first can afford to price lower than anyone who hasn’t.

This is why software markets so often end up with one or two dominant players. The fixed costs are enormous and shared across the whole product. Once you’ve built the software and recovered those costs across a large customer base, you can price in ways that make it nearly impossible for a new entrant with a smaller base to compete on price.

It is also why software gross margins look so attractive relative to other industries. When your marginal cost is zero, almost every dollar above your fixed cost threshold is profit. Mature software companies routinely post gross margins above 70 or 80 percent. That is not a sign of monopoly exploitation in most cases. It is the mathematical result of having large fixed costs and near-zero variable costs.

The price of software is never zero because building software is never free. The distribution just hides that fact well enough that the economics look, at first glance, like magic.