OpenClaw creator’s $1.3 million monthly OpenAI bill reveals the real cost of autonomous AI coding at scale
May 18, 2026 – 7:01 pm
Summary
Peter Steinberger, the creator of OpenClaw and an engineer at OpenAI, spent $1.3 million in a single month running approximately 100 Codex instances on his open-source project. This bill, covered by OpenAI, includes 603 billion tokens across 7.6 million requests, providing the most concrete data point on the cost of autonomous AI coding at scale.
Peter Steinberger’s Experiment
Steinberger joined OpenAI in February 2026 and used their API to power his open-source project, racking up a substantial bill. The spending is part of a research investment aimed at understanding software development without token economics as a limiting factor.
What the AI Agents Do
The 100 Codex instances aren’t just generating code; they form an autonomous development pipeline where AI agents perform various tasks, including:
- Reviewing pull requests
- Scanning commits for security vulnerabilities
- Deduplicating GitHub issues
- Writing fixes and opening new pull requests based on project roadmap
- Monitoring performance benchmarks and flagging regressions
- Even attending meetings and generating pull requests based on conversation topics
Human Oversight
Despite the extensive AI operations, a three-person team oversees the entire process. They utilize additional tools like Clawpatch.ai, Vercel’s Deepsec, and Codex Security for enhanced bug and security analysis.
Cost Considerations
Steinberger was transparent about the economics:
- The $1.3 million figure includes "Fast Mode" pricing, which is significantly more expensive than standard execution. Disabling Fast Mode would reduce the monthly cost to approximately $300,000.
- At standard pricing, the operation would cost $3.6 million annually.
- The disparity between the headline figure and underlying economics highlights how pricing tiers can inflate reported costs.
When asked about ROI…
Steinberger stated that while they haven’t yet measured a direct return on investment, the insights gained from this experiment are invaluable for future development strategies.