TL;DR
OpenClaw creator Peter Steinberger spent approximately $1.3 million on OpenAI API tokens in a single month, using high-volume automation for AI-assisted development. The spending reflects the cost of large-scale AI operations and raises questions about AI tool economics.
Peter Steinberger, creator of the open-source project OpenClaw and an employee of OpenAI, posted a screenshot showing his team spent over $1.3 million on OpenAI API tokens in a 30-day period. This figure underscores the significant costs associated with large-scale AI automation for software development.
The expenditure covers 603 billion tokens across 7.6 million API requests, primarily generated by about 100 Codex instances operated by a team of three. The top model used was GPT-5.5, with the highest daily spend reaching nearly $20,000. Steinberger clarified that this total reflects ‘Fast Mode’ pricing, which increases token consumption. Disabling Fast Mode would reduce the cost to approximately $300,000, still a substantial sum.
Steinberger’s team uses these AI agents to review pull requests, scan for security issues, deduplicate GitHub issues, and even attend meetings to generate feature PRs. The project functions as a stress test for AI-assisted development without budget constraints, serving as a laboratory for future software engineering workflows.
Why It Matters
This development highlights the potential financial scale of AI automation in software development, especially for projects employing large models and high request volumes. It raises questions about the economic sustainability of AI-assisted coding tools and the true costs of AI infrastructure at enterprise levels. For developers and organizations, understanding such costs is critical as AI tools become more integrated into workflows.
AI developer API cost management tools
As an affiliate, we earn on qualifying purchases.
As an affiliate, we earn on qualifying purchases.
Background
OpenClaw, launched by Steinberger after joining OpenAI in February, has gained attention for its autonomous code review and development capabilities. The project has been publicly turbulent, with incidents involving AI communication and industry competition, including Nvidia developing its own AI coding tools. The high API costs reflect broader industry concerns over the economics of AI inference, especially as companies subsidize or subsidized AI usage to attract users.
OpenAI shifted Codex to token-based billing in April, making costs more transparent but also more variable for heavy users. Steinberger’s usage exemplifies the upper end of this cost spectrum, illustrating the financial implications of AI automation at scale.
“The $1.3 million figure reflects Codex’s ‘Fast Mode’ pricing, which consumes credits at a significantly higher rate than standard execution.”
— Peter Steinberger
“Everything we build remains open source. This spending is research into how software development would change if token costs weren’t a constraint.”
— Steinberger
As an affiliate, we earn on qualifying purchases.
What Remains Unclear
It is not yet clear how sustainable or typical such high expenditure is across other AI projects or industries. The exact breakdown of costs, the impact of potential future pricing changes, and the overall scalability of this approach remain uncertain.
large-scale AI development hardware
As an affiliate, we earn on qualifying purchases.
As an affiliate, we earn on qualifying purchases.
What’s Next
Further analysis of AI API costs at scale is expected as more developers and companies evaluate the economics of AI automation. OpenAI may adjust pricing or policies in response to such high-volume usage, and the community will watch for developments in cost management and efficiency improvements.
As an affiliate, we earn on qualifying purchases.
Key Questions
Why did Peter Steinberger spend so much on AI tokens?
He used high-volume, automated AI agents for open-source development tasks, leveraging ‘Fast Mode’ pricing that significantly increases token consumption.
Is this level of spending typical for AI development?
No, Steinberger’s usage is at the extreme end, representing a high-volume test case rather than standard practice for most developers or companies.
Could this cost be reduced?
Yes, disabling Fast Mode would lower the cost to an estimated $300,000, but it still reflects a substantial investment in AI automation at scale.
What does this mean for the future of AI-assisted coding?
This case highlights both the potential and the financial challenges of scaling AI tools for development, prompting industry discussions on sustainable economics and pricing models.
