The AI boom is rewiring trade flows faster than any tariff ever could, and the West just realized it's been building on rented land.
The Summary
- AI infrastructure demand is forcing a hard reset on globalization, with compute, energy, and rare earth dependencies suddenly looking like national security liabilities
- Post-literacy concerns are emerging as AI handles more knowledge work, but the real question is what humans earn when AGI does the reasoning
- Solar deployment is accelerating exponentially, creating the energy abundance AI needs while reshaping geopolitical power centers away from oil states
The Signal
The AI race is doing what decades of political rhetoric couldn't: actually deglobalizing supply chains. Not because of ideology, but because of physics. Training frontier models requires chip fabs, data centers, and gigawatts of power in configurations that don't care about your comparative advantage theory.
Exponential View's latest analysis points to infrastructure demand as the forcing function. TSMC's Arizona fabs, Samsung's Texas expansion, and Intel's Ohio buildout aren't just political theater. They're insurance policies against the Taiwan risk premium that every AI lab is now pricing into their roadmaps. When your model training run costs $100M and takes six months, you can't afford a semiconductor chokepoint halfway around the world.
"The comparative advantage of the 20th century was cheap labor. The comparative advantage of the 2030s will be cheap, reliable electrons."
But here's where it gets interesting: the energy piece. AI's hunger for compute is colliding with solar's exponential cost curve. Solar installations are growing faster than analysts predicted even two years ago, and most of that new capacity is being built close to data centers, not exported. You're seeing energy nationalism in real time. Countries that can pair domestic chip production with abundant renewable power are building moats that look nothing like traditional trade barriers.
The post-literacy angle is subtler but might matter more. If AI agents handle synthesis, analysis, and most forms of written communication, what happens to the cognitive skills we've spent centuries optimizing for? Reading as we know it evolved for information scarcity. We're heading into information abundance managed by agents. The skill shifts from consumption to curation, from writing to prompting, from knowing to verifying.
Key shifts in the agent economy:
- Reasoning becomes a commodity service, not a human differentiator
- Value accrues to taste, judgment, and what to ask, not how to answer
- Post-AGI earning potential clusters around verification, emotional labor, and physical presence
This isn't about dumbing down. It's about specialization. Your great-grandfather probably couldn't solve a differential equation, but he could navigate by stars, preserve food, and build a fire in the rain. Different era, different essential skills. We're entering another transition, faster than most people realize.
The Implication
If you're building in the agent space, the infrastructure dependencies matter more than the algorithm. Focus on where compute happens, not just what compute does. Companies that control the full stack from chips to power to models will have regulatory and economic advantages that pure software plays won't.
For knowledge workers, start thinking about your work as agent supervision rather than agent replacement. The money won't be in doing the analysis. It will be in knowing which analysis to run, verifying outputs, and making judgment calls when the data is ambiguous. Develop taste. Build networks. Do things that require being physically present or emotionally intelligent. Everything else is getting automated, and faster than the optimists think.