The OpenClaw Burn Rate is a Warning Sign for the AI Agent Economy

In the current gold rush of the creator and platform economy, there is a recurring pattern: the 'token-incentive loop.' This is a cycle where artificial liquidity, provided by deep-pocketed investors, masks the fact that the underlying product lacks sustainable unit economics. We are seeing this play out in real-time with OpenClaw.

As The Register first reported, Peter Steinberger—the creator of OpenClaw and a member of technical staff at OpenAI—disclosed the massive cost of his operations during the Cloudflare Connect 2026 conference in San Francisco. According to that reporting, Steinberger estimated that he spent approximately $10 million in tokens running AI workloads in a single month.

To be clear, this is my opinion: spending $10 million a month on tokens for a single project is not a sign of scaling success; it is a textbook example of venture-funded inefficiency. When Cloudflare CTO Dane Knecht noted that Steinberger possesses a "virtually unlimited token budget"—a luxury most developers do not share—he highlighted the disconnect between the current AI development bubble and the reality of the market.

What is most alarming is not just the sum, but the lack of utility derived from it. When the audience reacted with surprise to the $10 million figure, Steinberger admitted, "What people don't always understand is not all of that is useful."

From a monetization and platform perspective, this is a red flag. If a significant portion of a $10 million monthly spend is producing non-useful output, the unit economics of the project are fundamentally broken. In any other sector of the creator economy, a burn rate that yields negligible utility would be viewed as a failure of product-market fit. Here, however, it is framed as a necessary cost of innovation because, as The Register notes, the money is coming from investors' pockets.

This reliance on external funding is not an isolated incident. The Register reports that OpenAI is currently attempting to "accelerate funding" as it competes with the expected IPO of rival Anthropic. This suggests a broader industry trend where the goal is not to build a profitable, sustainable business model, but to secure enough capital to outlast the competition, regardless of the waste involved in the process.

Steinberger's vision for the future further underscores this detachment from economic reality. When discussing infrastructure, he questioned why AI workloads are still using local computers, asking, "Shouldn't this all be in the cloud?" While he views the cloud as the solution to infrastructure challenges, he ignores the financial gravity that Cloudflare's Dane Knecht pointed out: cost is a primary consideration for anyone not "setting venture funds alight."

Even the proposed technical solutions for trust and security within OpenClaw seem to compound the cost problem. To address the issue of permissions and guardrails, Steinberger suggested a system where one agent authorizes another. He described a process where an agent must "negotiate with another agent" to determine if a task is permissible or if there is a simpler way to operate. While this may increase trust, it effectively doubles the token consumption for a single action. In a world of sustainable margins, adding more expensive layers of AI to manage other AI is a non-starter. In the venture-backed world of OpenClaw, it is just another way to spend the unlimited budget.

We are seeing a dangerous precedent where the 'cost of doing business' is decoupled from the 'value created.' If the future of AI agents involves a world where everyone has intelligence that predicts their needs, as Steinberger suggests, that future cannot be built on a foundation of $10 million-a-month waste.

Ultimately, OpenClaw is not demonstrating the viability of AI agents; it is demonstrating the viability of venture capital. Until these platforms can prove they can operate without a constant infusion of investor cash to cover useless token spend, they aren't businesses—they are expensive experiments in artificial liquidity.