Subscription economics

AI Subscription Margin by Utilization

A heatmap of provider gross margin as subscribers consume more of their API-equivalent value. The model assumes API list prices carry a 75% gross margin, so cost to serve equals 25% of API-equivalent usage.

Plans tracked 6 Anthropic Claude and OpenAI ChatGPT/Codex tiers with official pricing references.

Utilization points 10 From 1% casual use to 100% full plan extraction.

Fastest break-even 5.7% ChatGPT Pro 20x turns negative after very low average utilization.

Cost assumption 25% Cost to serve as a share of API-equivalent value consumed.

Heatmap

Gross Margin Falls as Utilization Rises

Green cells are positive gross margin, yellow is break-even, and red is negative gross margin. Highlight a utilization column to compare how each plan behaves at the same average usage level.

Highlight utilization

Plan	1%	5%	5.7%	10%	11.4%	15%	20%	25%	50%	100%
Anthropic - Claude
claude-pro 5x API-equivalent value, break-even at 20%	95%	75%	71%	50%	43%	25%	0%	-25%	-150%	-400%
claude-max-5x 5x API-equivalent value, break-even at 20%	95%	75%	71%	50%	43%	25%	0%	-25%	-150%	-400%
claude-max-20x 10x API-equivalent value, break-even at 10%	90%	50%	43%	0%	-14%	-50%	-100%	-150%	-400%	-900%
OpenAI - ChatGPT / Codex
chatgpt-plus 8.75x API-equivalent value, break-even at 11.4%	91%	56%	50%	12%	0%	-31%	-75%	-119%	-338%	-775%
chatgpt-pro-5x 8.75x API-equivalent value, break-even at 11.4%	91%	56%	50%	12%	0%	-31%	-75%	-119%	-338%	-775%
chatgpt-pro-20x 17.5x API-equivalent value, break-even at 5.7%	82%	12%	0%	-75%	-100%	-162%	-250%	-338%	-775%	-1650%

Selected utilization 20%

Positive plans 0

Average margin 0%

Generous plans rely on light users

At low utilization, subscription plans can look very profitable because subscribers leave most of the API-equivalent value unused.

Power users flip the economics

The same plans can move below break-even when average utilization rises, especially at the 20x tiers.

The table is gross margin only

It excludes sales, R&D, support, overhead, discounts, and any revenue share or cloud commitment structures.

Methodology

How to Read the Heatmap

Each cell estimates gross margin for a plan at a given average utilization level. The model assumes API list prices carry 75% gross margin, which makes serving cost equal to 25% of API-equivalent value consumed.

Gross margin is shown from the provider perspective. A 0% cell means subscription revenue and estimated serving cost are equal before other operating expenses.
Negative cells mean the average user consumes enough API-equivalent value that the plan would lose money on gross margin alone under the stated assumption.
Break-even utilization differs by plan because each subscription tier implies a different multiple of API-equivalent usage relative to its subscription price.
The modeled values are directional and sensitive to real model mix, routing, discounts, caching, inference optimization, and usage caps.

Sources

Official References

Only official OpenAI and Anthropic pages are linked here. The gross-margin heatmap is an Aistify model built from subscription and API-equivalent pricing context, not a provider-published margin statement.

OpenAI and Anthropic pricing pages

The tracker references official provider pricing and plan pages for subscription tiers and API-equivalent cost context.

Official OpenAI plan page for ChatGPT Plus, Pro, Business, Enterprise, and Codex access context.
ChatGPT pricing
Official OpenAI API pricing page used for API-equivalent cost and model pricing context.
OpenAI API pricing
Official Anthropic Claude plan page for Pro, Max, Team, and Enterprise subscription context.
Claude plans and pricing
Official Anthropic API pricing documentation used for API-equivalent cost and model pricing context.
Claude API pricing

Data note: values are modeled from public subscription and API pricing context. Update the tracker when OpenAI or Anthropic changes plan terms, API prices, model routing, or usage limits.