- Peter Steinberger claims to have spend $10 million in tokens in a single month at OpenAI
- Steinberger was chatting with Cloudflare CTO Dane Knecht at the Cloudflare Connect 2026 conference
- OpenClaw’s creator has been working with OpenAI on next-generation personal AI agents since February 2026
OpenAI’s Peter Steinberger – the creator of the OpenClaw AI agent platform – has admitted to spending an estimated $10 million in tokens to run AI workloads in a single month, a boast which has raised concerns over the way in which OpenAI is conducting operations, especially given than the tokens are paid for by the company’s investors.
The OpenClaw creator was chatting with Cloudflare CTO Dane Knecht at the Cloudflare Connect 2026 conference in San Francisco when the claim was made.
Over the coming months, OpenAI expects to improve its funding, with rival Anthropic believed to be planning an IPO. The comments from Steinberger indicate he is less concerned with the realities of business, and more focused on pushing AI agents.
Cloud vs. local
While discussing cloud AI at the Cloudflare Connect 2026 event, Knecht highlighted that Steinberger had admitted to using “a virtually unlimited token budget, which not all of us in the room have.” He added that “What people don’t always understand is not all of that is useful.”
In the competitive world of AI models and agents, where investment is vital, these words seem somewhat ill-considered.
Steinberger’s thoughts on how infrastructure supports AI is arguably far more useful than his profligate attitude to other people’s AI tokens. His view is that the cloud needs to be better, perhaps even providing seamless AI compute. “I feel like the main challenge a lot of people still have is that our infrastructure structure hasn’t really caught up yet,” he said. “Like why is this still using my computer? Shouldn’t this all be in the cloud?”
His opinions on how AI agent use will expand in the coming years are also worth consideration. “I think everyone will have some sort of intelligence that runs and thinks and helps them throughout the day or even predicts what could help you in the future.”
The implication of this is clear: “…instead of asking the question, maybe you already get the answer.”
Agent permissions
However, the agents you use may not reduce to just one. Steinberger feels that at least two may be required – one to do the work, and another to check it.
“You want to have another agent that can give it permissions or not and then this agent has to negotiate with another agent.”
He continued: “Many times the agents are like, ‘oh yeah, actually maybe I shouldn’t do this,’ or ‘there’s a simpler way that keeps it within the guardrails.’ So those those two systems combined greatly increase my trust that I just can let the agent rip.”
