The Pricing Strategy
How do you price a product when each query costs fractions of a cent but you've spent billions on R&D? This is the central pricing challenge facing every AI company. Let's explore the economic principles behind their decisions.
The Freemium Model
ChatGPT uses a freemium pricing model — a free tier attracts users, and a paid tier captures revenue from power users. This is a form of price discrimination, where different customers pay different prices for essentially the same product.
Here's the typical AI pricing ladder:
| Tier | Price | What You Get |
|---|---|---|
| Free | $0/month | Limited access to older models, usage caps |
| Plus | $20/month | Faster responses, newer models, higher limits |
| Team | $25/user/month | Workspace features, admin controls |
| Enterprise | Custom | SLA guarantees, data privacy, dedicated support |
| API | Pay-per-token | Direct model access, full control |
Price Discrimination: A Feature, Not a Bug
In economics, price discrimination means charging different prices to different customer segments. There are three types:
- First-degree — Charge each customer their maximum willingness to pay (perfect price discrimination). Almost impossible in practice.
- Second-degree — Offer different product tiers and let customers self-select (freemium, Plus vs. Enterprise). This is what ChatGPT does.
- Third-degree — Charge different prices to identifiable groups (student discounts, regional pricing).
The genius of tiered pricing is that it's self-selecting — customers reveal how much they value the product by choosing their tier. A casual user happy with the free tier would never pay $20/month, while a power user gladly pays to remove limitations.
Try It: Spot the Economics
Paste any text about AI pricing or business models. The Concept Spotter will highlight economic concepts it detects.
Why Free Matters
Offering a free tier seems counterintuitive when each query costs money. But it serves critical economic functions:
- Demand generation — Free users create word-of-mouth marketing worth millions in advertising
- Data collection — User interactions improve the model (a form of indirect network effects)
- Conversion funnel — A percentage of free users convert to paid plans
- Market share — Grabbing users early creates switching costs
The typical conversion rate from free to paid in SaaS is 2–5%. If ChatGPT has 100 million free users and converts 3%, that's 3 million paying subscribers generating $720 million per year at $20/month.
The Subsidy Question
OpenAI's free tier is a cross-subsidy — paying users and investors subsidize free users. This is common in platform economics:
- Google Search is free, subsidized by advertisers
- Instagram is free, subsidized by advertisers
- Spotify Free is subsidized by Premium subscribers
The key economic question is: when do you reduce the subsidy? Too early and you lose users to competitors. Too late and you burn through cash before reaching profitability.
API Pricing: The B2B Play
The API pricing model is fundamentally different from subscriptions. It's usage-based pricing — you pay exactly for what you consume:
- Advantage: Low barrier to entry, costs scale with revenue
- Disadvantage: Unpredictable bills, no committed revenue for the provider
API pricing also enables price segmentation by model: offering cheaper, faster models (GPT-3.5) alongside expensive, capable ones (GPT-4) lets customers optimize their cost-performance tradeoff.
Bundling and Unbundling
OpenAI bundles many features into ChatGPT Plus: GPT-4 access, DALL-E image generation, web browsing, code interpreter, and custom GPTs. In economics, bundling works when:
- Products have different values to different customers
- The marginal cost of adding items to the bundle is low
- It prevents competitors from cherry-picking the most popular features
By bundling, OpenAI makes it harder for a competitor to compete on just one feature — you'd need to match the entire bundle's value.
Key Takeaways
- AI companies use freemium (second-degree price discrimination) to segment customers
- Free tiers serve as marketing, data collection, and conversion funnels
- API pricing provides usage-based access for developers
- Bundling multiple AI capabilities creates a defensible product offering
- Cross-subsidies fund free tiers until conversion economics work