The announcement signals the end of the “electricity era” for OpenAI, where companies simply paid for the number of tokens or words processed. Instead, the company is positioning itself as a co-sharer in the innovation value chain.
What is “Outcome-Based Pricing”?
Unlike a monthly $20 fee or a per-token API cost, OpenAI’s new proposed model targets the actual commercial success of its users’ outputs:
- The Logic: If a pharmaceutical company uses OpenAI’s models to identify a new drug molecule, OpenAI believes it should share in the billions of dollars of resulting revenue.
- The Mechanisms: Friar noted that new economic models will emerge, including licensing deals, IP-based agreements, and royalties on products created through AI reasoning.
- The Internet Parallel: OpenAI compares this shift to the evolution of the internet—from a tool for communication to the foundational infrastructure of the global economy.
The 2026 Financial Pressure: $115B Burn
The pivot to revenue sharing is driven by the staggering costs of the “Agentic Era.”
- Projected Losses: Internal documents suggest OpenAI will clock $14 billion in losses for 2026, with cumulative losses potentially reaching $115 billion by 2029.
- Compute Constraints: With computing capacity tripled in 2025, OpenAI is looking for ways to capture the “massive upside” of the intelligence it provides to global industries.
2026 Revenue Stream Breakdown
| Stream | Target Audience | Model Type |
| ChatGPT Plus/Pro | Individual Professionals | Subscription |
| ChatGPT Go | General Consumers | Ad-Supported (Free) |
| APIs | Developers/SaaS | Usage-Based (Tokens) |
| Discoveries | Pharma, Energy, Finance | Outcome-Based (Royalties) |
Industry Reaction: A “Finder’s Fee” for Your Own Work?
The proposal has immediately drawn criticism from the enterprise and developer communities.
- The Ownership Dispute: Critics argue that if a user pays for a subscription and provides their own proprietary data to reach a discovery, OpenAI should not be entitled to the results. One tech founder labeled it a “finder’s fee for success you found yourself.”
- The “Local” Pivot: Analysts at Constellation Research suggest this could drive many enterprises toward local, open-source models (like Llama) to avoid “taxing” their own innovation output.
- Legal Gray Areas: IP-based agreements on AI outputs remain legally murky, as most jurisdictions currently do not allow AI to be listed as an inventor on patents.
Conclusion: A Selective Rollout
OpenAI has clarified that these “royalty-based” models are intended for large-scale, negotiated enterprise partnerships rather than individual creators or small businesses. However, the shift marks a fundamental change in the relationship between AI providers and their clients. In 2026, OpenAI is no longer just selling a software tool; it is attempting to buy a stake in the future discoveries of the human race.
