Fast Company's Francis Northwood has a bracing piece on Oracle's AI infrastructure debt, arguing that the old adage about selling shovels in a gold rush is more complicated when the shovels cost $50 billion and require continuous capital expenditure. This week also brought news that Peec, a Berlin startup that helps brands track their presence in AI search results, more than doubled its annualized revenue to $10 million in a matter of months. These two data points define the actual shape of AI economics right now: ruinous at the infrastructure layer, surprisingly healthy at the measurement and intelligence layer.

The AI Visibility Economy Is Real

Peec's growth is significant not just as a startup success story but as evidence that a new category of software is actually generating revenue: tools that help brands understand and optimize their presence in AI-generated answers rather than traditional search rankings. This is a legitimate new market, and it is growing fast. , analytics and observability tools for AI-mediated discovery, which has a clear buyer, a clear pain point, and margins that look nothing like compute infrastructure.

Infrastructure Debt vs. Intelligence Layer Returns

Oracle's problem is the inverse of Peec's opportunity. The company bet on being the foundational compute layer for AI workloads, taking on enormous debt to build out capacity. But as Northwood documents, the returns on that bet require AI adoption to continue accelerating at a rate that is far from guaranteed, especially as the AI brand backlash documented elsewhere this week begins to complicate enterprise purchasing decisions. The smart money, increasingly, is not on who owns the infrastructure but on who can tell you whether the infrastructure is working. Peec is the canary. Oracle is the mine.