AI Infrastructure Fights Moving To Electricity; Developing A 'Contestibility Window'
I added three new signals articles this week. Two are related to how to meet the significant electricity demand from AI without making current rates skyrocket. While I think these two signals overall are positive in terms of figuring out how to provide electricity, they also open up new windows for viewing the concentration/dispersion possibilities.
The third signal is just an event I find intriguing at this point. Europe is actively exploring with public support creation of a public frontier model that might, in turn, become a core offering for smaller models to support small businesses, education, and governments. We are a LONG way from this becoming a reality, but there is nothing comparable in the US to this work. We are focused on public sector simply reacting/regulating whatever the private sector actors churn out. This just might become interesting later down the line, so I thought I'd flag it now.
Meanwhile, I've been working on developing the concept of a 'Window of Contestibility'. I know, it's a really catchy name . . . let me know if you what you recommend that might be a bit catchier. Anyway, the idea is that we need to develop ways to measure not only how concentration/dispersion is playing out, but how close are we to 'hardening' that would make the concentration/dispersion more difficult to change - not impossible, but more difficult. The Window of Contestibility is currently likely quite wide open, but we have no way to document that statement so it's only an opinion. This is the first concept essay of what I anticipate to be three developing the groundwork for the Window. When completed, I will be adding the Window of Contestibility to the Scoreboard as an additional way to track events in the Race.
Mark
New Signals
Can Frontier AI Become Civic Infrastructure? Posted June 25, 2 min. read
Europe’s EUROPA project is still early, and many of the most important details remain unresolved. But its basic premise is striking: frontier AI may not have to be only something we rent from private platforms — it might also become civic infrastructure that public institutions, researchers, businesses, and communities can inspect, adapt, and build on.
When the Data Center Brings Its Own Power Plant Posted June 22, 1 min. read
Reuters found that some AI data centers are trying to “bring their own power” through dedicated gas plants, sometimes moving through fast and opaque approval paths with limited public awareness. The issue is not that dedicated power is inherently bad; it is that state and local governments, regulators, and residents need to see and evaluate the full project before AI infrastructure quietly outruns public governance.
AI Is Now a Grid-Planning Problem Posted June 22, 2 min. read
FERC is pushing regional grid operators to update the rules for connecting AI-scale electricity users, so the grid can support major new data centers without unfairly shifting costs to existing ratepayers. The signal is that AI has become large enough to require serious public infrastructure governance: more power will be needed, but it should scale under clear grid rules while state and local siting authority remains intact.
New Concept Essay
The AI Time Horizon Problem: Why We May Overreact to the First Shocks of AI and Underreact to Its Deeper Transformation Posted June 21, 11 min. read
New technologies often look most dramatic at the beginning, when their first shocks are visible, emotional, and easy to debate. This paper argues that AI’s deeper democratic consequences may unfold more slowly — through defaults, contracts, institutions, infrastructure, and power — creating a window of contestability before temporary advantages harden into durable control.