OpenClaw creator spent whopping $10M on AI tokens last month
What's the story
Peter Steinberger, the creator of the popular open-source AI agent OpenClaw and a member of the technical staff at OpenAI, has revealed that he spent an eye-watering $10 million on AI tokens last month. He shared this shocking number during a conversation with Cloudflare CTO Dane Knecht at Cloudflare Connect 2026 conference in San Francisco on Wednesday.
Token explained
Steinberger's massive AI token expense
Tokens are the basic units of data that AI models process.
Processing billions of these tokens requires immense computing power, which is why they come at such a high cost.
Despite the exorbitant expense, Steinberger's costs are being covered by investors, as he operates with a "virtually unlimited token budget," according to Knecht.
Cloud preference
The cloud conundrum for AI
Steinberger believes that AI should run on cloud servers instead of personal devices.
He said, "I feel like the main challenge a lot of people still have is that our infrastructure hasn't really caught up yet."
However, he also acknowledged that cloud-hosted AI agents are still very expensive, which is why many developers prefer running them locally to save money and ensure privacy.
Quality concerns
Addressing the low-quality code contributions issue
Knecht flagged a rising problem of open-source developers rejecting code contributions due to an influx of low-quality, AI-generated code.
Steinberger proposed that the submission process needs to change and developers should show how their AI reached its conclusion.
He believes this would increase trust in the output as compared to just throwing the finished work across the fence.
Streamlined future
The future of personal AI assistants
Steinberger doesn't see a future where everyday users will be running hundreds of AI agents at once. Instead, he envisions a more streamlined and proactive approach.
"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," he explained.
Security measures
AI agents as trusted negotiators
Steinberger warned about the risks of giving an AI agent permission to act on your behalf.
He suggested using a second AI agent as a security guard to negotiate permissions and catch mistakes.
"Those two systems combined greatly increase my trust that I just can let the agent rip," he added, emphasizing how this approach could make cloud-based AI assistants more reliable for managing daily tasks in the future.