TL;DR
Peter Steinberger, creator of OpenClaw, used over $1.3 million in OpenAI API tokens in a single month. The spending was driven by extensive automation using GPT-5.5-based Codex agents. The event underscores the high costs of AI development at scale and raises questions about sustainability.
Peter Steinberger, the Austrian developer behind the open-source project OpenClaw and an employee at OpenAI, posted a screenshot on Friday showing his team spent over $1.3 million on OpenAI API tokens in a single month. This figure highlights the extreme costs associated with large-scale AI automation in software development, even when covered by his employer.
The bill, totaling $1,305,088.81, was accumulated over 30 days and covered 603 billion tokens across 7.6 million requests. Steinberger’s team operates roughly 100 Codex instances, which autonomously review pull requests, scan for vulnerabilities, deduplicate issues, and generate code fixes. The bulk of the expenditure was in ‘Fast Mode,’ a high-cost API setting, which alone could be disabled to reduce costs to approximately $300,000.
Steinberger clarified that the high costs are related to ‘Fast Mode,’ which consumes credits at a significantly higher rate than standard operations. He noted that without Fast Mode, the monthly cost would be around $300,000, still representing a substantial investment. The usage was driven by a focus on stress-testing AI-assisted development at scale, with the project remaining open source.
Why It Matters
This event underscores the potential financial scale of AI-powered development when used extensively, raising questions about the economic sustainability of such approaches outside of corporate or research environments. It also highlights the gap between consumer API pricing and the actual compute costs of running large AI models, which could influence future development strategies and pricing models.
For developers and companies, the high costs exemplify the need to carefully manage AI automation workflows and consider the implications of high-volume API usage. It also spotlights the role of AI tools as both research instruments and potential cost sinks.
AI for Solo Lawyers: A Practical Guide to AI Tools that Save You Time and Grow Your Practice (AI for Professionals)
As an affiliate, we earn on qualifying purchases.
As an affiliate, we earn on qualifying purchases.
Background
In recent months, AI coding tools like Codex and Claude Code have been competing for developer adoption, often subsidizing inference costs to attract users. OpenAI shifted Codex to token-based billing in April, increasing transparency but also exposing high costs for intensive users. Steinberger’s usage reflects an experimental approach, pushing the boundaries of AI automation without immediate cost concerns, since OpenAI covers the expenses.
OpenClaw has been in the public eye for turbulent reasons, including controversial interactions with Meta’s AI team and prompting Nvidia to develop competitors. Steinberger has described the project as a laboratory for stress-testing AI-assisted development, not as a commercial product.
“The $1.3 million figure reflects Codex’s ‘Fast Mode’ pricing, which consumes credits at a significantly higher rate than standard execution. Disabling Fast Mode alone would reduce the raw API cost to around $300,000.”
— 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
high-performance coding laptops for AI developers
As an affiliate, we earn on qualifying purchases.
As an affiliate, we earn on qualifying purchases.
What Remains Unclear
It is not yet clear how sustainable such high API costs are for other projects or whether OpenAI will modify pricing or usage policies in response to such high-volume utilization. The long-term implications for AI development economics remain uncertain.
API 560 500 Series Graphic Equalizer
Modular 10-band Graphic Equalizer
As an affiliate, we earn on qualifying purchases.
As an affiliate, we earn on qualifying purchases.
What’s Next
Further analysis is expected to examine whether OpenAI will adjust pricing for high-volume users or introduce new billing models. Additionally, other developers may explore similar high-cost experiments, influencing industry discussions on AI cost economics and automation scalability.
"Looks Good To Me": Constructive code reviews
As an affiliate, we earn on qualifying purchases.
As an affiliate, we earn on qualifying purchases.
Key Questions
Why did Steinberger’s team spend so much on API tokens?
The team used extensive automation with high-cost ‘Fast Mode’ settings to stress-test AI-assisted development workflows at large scale, not for typical production use.
Could this level of spending be sustainable for other projects?
Likely not, unless covered by large budgets or institutional support. The high costs highlight the economic challenges of scaling AI automation without cost controls.
Will OpenAI change its pricing policies because of this?
It remains unclear. OpenAI has not publicly indicated any policy changes, but the incident may prompt industry discussions on API pricing and high-volume usage limits.
