I largely agree with Krugman's point. Power, water, rare earths, semiconductors, transmission capacity, and data centers are becoming the strategic choke points of the AI era.
What strikes me, however, is that this concentration is not merely a political choice. It is becoming an economic necessity.
As nations, businesses, and individuals become increasingly dependent on AI infrastructure, control over the physical inputs that make AI possible becomes more valuable than control over the algorithms themselves. Models can be copied, improved, and commoditized. Electricity, rare minerals, grid capacity, advanced chips, and secure supply chains cannot.
This is why the AI economy is beginning to resemble the utility industry. AI appears digital, but underneath it rests on a growing stack of physical infrastructure that exhibits the same characteristics as power grids, railroads, pipelines, and telecommunications networks: enormous capital requirements, economies of scale, and natural tendencies toward concentration.
The debate is no longer whether power lives upstream. The deeper question is who controls the upstream assets in a world where AI becomes as essential as electricity itself.
Excellent points, I think there is an even deeper layer beneath the infrastructure question.
Every major civilization eventually develops critical choke points. Water systems. Railroads. Electrical grids. Telecommunications. The internet.
AI is simply creating a new layer of dependency.
The question is not only who controls the upstream assets. The question is whether the signal flowing through those assets remains visible to the people who depend on them.
As infrastructure concentrates, information tends to concentrate as well.
This is what I call Signal Compression.
More and more of reality becomes interpreted by fewer and fewer organizations, algorithms, and decision makers.
The physical bottlenecks, power, chips, minerals, transmission capacity, may determine who owns the infrastructure.
But the informational bottlenecks may determine who shapes reality.
The greatest strategic risk may not be resource scarcity.
It may be losing visibility into the systems we have become dependent upon.
Great take! I write about similar themes around AI strategy, ethics, and human-AI co-evolution, feel free to check out my Substack if it's of interest: drmatthasan.substack.com.
I largely agree with Krugman's point. Power, water, rare earths, semiconductors, transmission capacity, and data centers are becoming the strategic choke points of the AI era.
What strikes me, however, is that this concentration is not merely a political choice. It is becoming an economic necessity.
As nations, businesses, and individuals become increasingly dependent on AI infrastructure, control over the physical inputs that make AI possible becomes more valuable than control over the algorithms themselves. Models can be copied, improved, and commoditized. Electricity, rare minerals, grid capacity, advanced chips, and secure supply chains cannot.
This is why the AI economy is beginning to resemble the utility industry. AI appears digital, but underneath it rests on a growing stack of physical infrastructure that exhibits the same characteristics as power grids, railroads, pipelines, and telecommunications networks: enormous capital requirements, economies of scale, and natural tendencies toward concentration.
The debate is no longer whether power lives upstream. The deeper question is who controls the upstream assets in a world where AI becomes as essential as electricity itself.
Excellent points, I think there is an even deeper layer beneath the infrastructure question.
Every major civilization eventually develops critical choke points. Water systems. Railroads. Electrical grids. Telecommunications. The internet.
AI is simply creating a new layer of dependency.
The question is not only who controls the upstream assets. The question is whether the signal flowing through those assets remains visible to the people who depend on them.
As infrastructure concentrates, information tends to concentrate as well.
This is what I call Signal Compression.
More and more of reality becomes interpreted by fewer and fewer organizations, algorithms, and decision makers.
The physical bottlenecks, power, chips, minerals, transmission capacity, may determine who owns the infrastructure.
But the informational bottlenecks may determine who shapes reality.
The greatest strategic risk may not be resource scarcity.
It may be losing visibility into the systems we have become dependent upon.
Great take! I write about similar themes around AI strategy, ethics, and human-AI co-evolution, feel free to check out my Substack if it's of interest: drmatthasan.substack.com.
Happy to connect and explore a collab as well!
The Tempered Signal publishes Tuesday, Thursday, and Sunday.
Tuesday covers ownership discipline.
Thursday covers AI and infrastructure systems.
Sunday synthesizes the week into one argument.
If the upstream control thesis interests you, Thursday's edition is where it goes deeper.
Then hit the subscribe button.