A single engineer just burned through $1.3 million of AI in 30 days, and then posted the receipt on X for everyone to debate.
On Friday May 15, Peter Steinberger, creator of the viral AI agent OpenClaw, shared a screenshot from his own tool CodexBar showing the bill - with most of the spending tied to OpenClaw's development across roughly 100 Codex coding agents.
The Receipt
Steinberger no longer pays for that compute out of pocket, since he joined OpenAI in February and the company foots the token bill. He framed the spend as an experiment: what would software development look like if tokens were free?
In practice the fleet of agents autonomously reviews pull requests, scans commits for security issues, and writes fixes. The number may also be low, because when one user asked if the screenshot only reflected one account, Steinberger replied: "Yup. At least on this account."
The fastest read on stories like this lives in Market Briefs, every weekday morning in five minutes, with a free investing masterclass thrown in when you sign up.
Why Anyone Cares
This is the latest move in a Silicon Valley flex called "tokenmaxxing," where engineers show off how much AI compute they can pour into a workflow, with bigger token bills now signaling speed rather than waste. The frame is hard for normal companies to copy, which is why the replies under Steinberger's post turned hostile fast.
"Someone's burning through enough tokens to bankroll a small startup," one user wrote, while others questioned whether a project costing $1.3 million a month can really be called "lean," which is the word Steinberger used.
The deeper investing question is who pays once the OpenAI compute subsidy stops being free, since token prices have been falling but compute at this volume still adds up on a cloud bill.
Worth Noting
For OpenAI, having engineers run massive token bills internally doubles as a marketing message, since it tells outside developers the future workflow involves a lot more compute, not less. For everyone else, the math is harder to swallow.
If "tokenmaxxing" becomes the norm at startups paying market rates, AI infrastructure bills are about to get a lot bigger, which mostly benefits the companies selling the picks and shovels.
Want the daily breakdown of where AI dollars are flowing? Subscribe to Market Briefs here and you'll also get a 45-minute investing course as a sign-up bonus.
