Price Discrimination

In The You On AI Field Guide

Economists distinguish three degrees of price discrimination. First-degree (perfect) price discrimination charges each customer exactly their maximum willingness to pay, which requires impossible levels of information. Second-degree price discrimination offers a menu of options (versions, quantities, bundles) and lets customers self-select — this is what versioning accomplishes. Third-degree price discrimination charges different prices to different identifiable customer groups (students, seniors, enterprise customers) based on their observable characteristics.

The AI market deploys all three forms simultaneously. First-degree approximations are attempted through dynamic pricing, usage-based billing, and negotiated enterprise contracts that price each customer near their specific willingness to pay. Second-degree versioning is visible in the free/pro/max tier structures that every major AI provider offers. Third-degree discrimination appears in academic pricing, startup credits, and enterprise versus consumer pricing.

Varian's analysis revealed that price discrimination in information markets is not exploitative in the way that uniform-pricing monopoly pricing is. It can actually expand access by making the product available to customers who could not afford a single uniform price set to maximize revenue from high-value customers. The student edition of software exists because the student could not pay the enterprise price, and the enterprise price could not be lowered to the student level without sacrificing the revenue needed to fund continued development. Price discrimination allows both the student and the enterprise to be served.

The same logic applies to AI subscription tiers. The hundred-dollar monthly price of Claude Max would be prohibitive for casual users; the free tier's existence allows those users to access genuine capability without eroding the revenue from serious users who can justify the higher tier. The versioning structure expands the market while extracting surplus from customers whose willingness to pay is high.

Origin

Price discrimination has a long history in economic theory, with foundational contributions from Arthur Pigou in 1920. Varian's contribution was to apply the framework systematically to information goods and to identify versioning as the specific implementation strategy most suited to zero-marginal-cost products.

Key Ideas

Zero marginal cost requires price discrimination. Uniform pricing cannot simultaneously extract high-willingness-to-pay value and maintain broad accessibility.

Versioning is self-selecting. Customers choose their tier based on their own valuation, eliminating the need for the seller to observe willingness to pay directly.

Access can expand through discrimination. Lower tiers exist because they would not otherwise be served; they do not cannibalize higher tiers if the versioning is designed correctly.

The AI market uses all three degrees simultaneously. Dynamic pricing, tier menus, and customer-segment pricing coexist in every major AI provider's structure.

Discrimination is not inherently exploitative. The welfare consequences depend on whether it expands access or concentrates surplus.