Software Economics

Quick definition: The distinct economics of software businesses—characterized by high fixed costs, near-zero marginal costs per user, long-term data accumulation, and powerful network effects—that create competitive dynamics fundamentally different from traditional industries and incompatible with classical valuation metrics.

Key Takeaways

• Software has zero marginal cost. Once built, distributing software to one more user costs nearly nothing, unlike physical products or services. This enables extreme profitability scaling as revenue grows.
• Scale creates winner-take-most dynamics. The firm with the largest user base accrues the best data, the most valuable network, and the strongest competitive moat, making competition binary: win or lose relevance.
• The first profitable user subsidizes the unprofitable user. Because marginal cost is zero, a company can afford to subsidize unprofitable customer acquisition or product lines as long as the customer base overall is profitable.
• Traditional financial metrics become unmoored. Price-to-sales, price-to-earnings, and return on capital employed mean something completely different for software than for a traditional business.
• The flywheel compounds over decades. Early market share leads to data advantages, which improve products, which attract more users, which generate more data—a self-reinforcing cycle that favors early winners.

The Economics of Building vs. Distribution

A traditional manufacturing business has a binary cost structure: fixed costs (factory, design, management) and variable costs (materials, labor per unit). A car company invests $5 billion to design and build a factory that produces 500,000 units per year. The fixed cost is $10,000 per unit. Material and labor might be another $6,000 per unit. Total cost is $16,000 per car. If the company sells the car for $25,000, the gross margin per unit is about 36%.

A software company invests $100 million and 100 engineers to build a search engine, productivity suite, or social network. The fixed cost is sunk before the first user. The marginal cost to add the millionth user is effectively zero. There's no second unit; there are just users.

This changes everything. For software, the unit economics look brutal in year one. The company has burned $100 million with few users. Cost per user is thousands of dollars. But as the user base grows, cost per user plummets. If the company reaches 100 million users, the cost per user is $1. If it reaches a billion, cost per user is $0.10.

Meanwhile, because marginal cost is near-zero, the company can increase prices, add premium features, or introduce advertising with minimal impact on profitability. The gross margin for an additional user approaches 100%.

This is not an extreme case; it's the norm for software. Spotify, Netflix, Slack, Microsoft, Amazon Web Services—all have virtually zero marginal cost per additional user. The profitability of the marginal user is nearly 100% of the price charged.

The Scale Inverts Traditional Economics

In a traditional business, scale helps but faces limits. A larger auto manufacturer has better purchasing power and factory efficiency, but diseconomies of scale emerge: more management layers, complexity, wage pressure. Margin improvement from scale is real but bounded, perhaps a few percentage points.

For software, scale creates a discontinuity. When a company reaches critical mass, the following happen simultaneously:

• Marginal cost stays at zero while revenue scales. Each new user brings revenue with almost no additional cost.
• Data accumulates and improves the product. A search engine's algorithm improves with more searches. A social network's recommendations improve with more users and interactions. A language model improves with more training data. The larger the user base, the better the product, regardless of how much additional R&D the company invests.
• Network effects amplify value. A communication platform is worth more to each user the more users it has. A payment network is more valuable the more merchants and customers are on it.
• Lock-in increases switching costs. A company with a billion users has data, integrations, and habitual usage patterns that make switching to a competitor costly. The switching cost compounds with time.

The result: profitable software companies at scale experience declining cost per user and increasing value to users. This is a virtuous cycle that doesn't exist in traditional industries. A larger auto manufacturer isn't obviously better for customers. A larger oil refiner doesn't produce better oil. But a larger social network is unambiguously more useful, and a larger search engine is unambiguously more powerful.

Data as a Competitive Moat

Software's unique advantage is that the user base itself becomes an asset. The more users interact with a software product, the more data is generated. That data, fed into algorithms and machine learning models, improves the product further, which attracts more users and generates more data.

This creates a competitive moat that's difficult to penetrate. A new social network launching today might have better code, better design, and lower latency than Facebook or TikTok. But it starts with zero users and zero data. It cannot possibly compete with an incumbent that has billions of users generating petabytes of behavioral data per day.

The data advantage is long-term and self-reinforcing. Facebook's data advantage in 2010 helped it build a better product by 2011, which attracted more users by 2012, which generated more data by 2013, which enabled better algorithms by 2014. Each year, the gap widened. A competitor can't catch up by simply copying the code; the data advantage is unstoppable once achieved.

The Winner-Take-Most Outcome

These dynamics—zero marginal cost, data-driven improvement, network effects, and switching costs—combine to produce winner-take-most or winner-take-most outcomes in software categories.

Consider search. Google's dominance isn't primarily due to better indexing (which is hard but replicable). It's due to data: billions of searches per day, trillions of clicks from users telling Google which results are relevant. This data fed into machine learning created an unbeatable search algorithm. Meanwhile, Bing, despite Microsoft's resources, could never close the gap because it had far fewer searches and clicks to learn from.

The same pattern repeated in social networks (Facebook, then Instagram, then TikTok), e-commerce (Amazon), cloud infrastructure (AWS), and video streaming (Netflix until very recently).

For a value investor, this poses a problem. If a software category is destined for a winner-take-most outcome, valuing the incumbent based on current profitability is insufficient. The incumbent's value derives from the permanence of its market position. A 40% market share today might seem sustainable, but in winner-take-most dynamics, 40% is a transient state—the market will likely contract to one or two winners, and the question is whether the incumbent is one of them.

Unit Economics and Free Products

Software's marginal cost advantage enables a business model unknown in traditional industries: profitability despite giving the product away. Google, Facebook, YouTube, and Slack (in its early days) all became massive with free user bases. They make money through advertising, data monetization, or conversion to paying customers.

This inversion confuses traditional value investors. A company with a free product and no revenue per user appears to have broken unit economics. But the actual unit economics—the cost to acquire and serve the user plus the lifetime value of that user across all monetization paths—might be highly attractive.

A search engine company that offers search for free but monetizes through ads is not less profitable than one charging per search; it's more profitable, because the marginal user acquisition cost drops to zero (the product spreads virally) and the lifetime monetization per user can be optimized through ad targeting and pricing.

Freemium models—where the product is free for basic users and paid for advanced features—exploit the same dynamic. Once users are on the platform using zero-marginal-cost supply, converting a fraction to paid customers is nearly pure profit. A freemium company with a small percentage of users on paid plans can be more profitable than a traditional software company selling to 100% of users at a lower price, because it operates at much larger scale.

Implication for Valuation and Competition

For classical value investors, software economics pose a fundamental challenge. The discipline's core assumptions—that scale improves profitability within limits, that competition will erode margins, that bigger doesn't mean indefinitely better—don't hold.

Software leaders at scale achieve near-infinite margins on incremental users. Competition doesn't automatically erode margins; the winner takes an ever-larger share. Bigger means indefinitely better because of data and network effects.

This is why value investors struggled with technology stocks. Valuing Facebook based on its profitability relative to cost of capital didn't account for the data moat. Valuing Amazon based on current P/E ratios didn't capture the flywheel advantage of scale. Valuing Spotify based on current subscriber margin didn't reflect the inevitable cost per subscriber would decline as the company scaled.

Understanding software economics isn't necessary to become a great investor, but misunderstanding them is nearly sufficient to guarantee failure in the technology sector.

Next

Explore how network effects and winner-take-all dynamics shape competitive strategy and returns: Network Effects and Winner-Take-All.

---

See also: Pricing Power in Inflation — How dominant software businesses exercise pricing power in inflationary environments.