The AI Boom Is Reshaping the US Economy — What I'm Noticing

Look past the stock charts for a second. The AI boom isn't just Nvidia's market cap or a handful of hyperscalers throwing billions at data centers. What's actually striking me is how the transformation is showing up in places you wouldn't expect — small manufacturing shops, logistics yards, mid-sized law firms, county-level healthcare systems. The "boom" narrative is real, but the texture of it is way messier and more interesting than the headlines suggest.

Here's what I'm seeing on the ground.

1. Capex is the canary. The hyperscaler spending numbers are almost comically large at this point — hundreds of billions annually across the major cloud providers. But the more telling signal is the second wave: utility companies renegotiating power contracts, transformer manufacturers with 18-month backlogs, and construction firms specializing in cooling infrastructure suddenly having more work than they can handle. That's not speculation; that's physical evidence of deployment.

2. The labor picture is more nuanced than "AI takes jobs." The BLS data we've seen so far doesn't show mass displacement — it shows something weirder. Job postings for "prompt engineering" and "LLM agent" roles have exploded, sure, but so have postings for "AI workflow coordinator" and "automation ops" at companies that have never used the word "AI" in their history. The real shift is in job descriptions being rewritten around tooling. A lot of these aren't new headcount — they're existing roles being redefined to assume AI-assisted throughput.

3. Productivity gains are showing up where nobody's looking. The most compelling real-world evidence I've found is in verticals like insurance claims processing and accounts payable. Companies are reporting 20-40% cycle-time reductions on document-heavy workflows, not because they deployed some massive custom model, but because they finally wrapped off-the-shelf LLM APIs around their existing processes. The boring stuff — OCR, data extraction, routing — is where the quiet wins are. That's the practical tutorial nobody writes about: it's not the deep dive, it's the plumbing.

4. The wage dispersion is widening. This is the uncomfortable part. The premium for workers who can actually operate these tools — not just prompt, but integrate, evaluate, and maintain — is significant. Meanwhile, the premium for pure data-entry competence is collapsing. That's not a doom scenario, but it does mean the "from scratch" learning curve is now a real economic filter. The gap between "knows a bit about AI" and "can ship a production workflow" is where the money is.

5. The regional story is lopsided. The boom is hitting metros with strong power grids and fiber differently than everywhere else. You're seeing secondary cities — places like Columbus, Charlotte, Salt Lake City — grabbing data center and model-ops work because of land and electricity costs. That's redistributing economic activity in ways the coastal narrative completely misses.

None of this is a clean "everything's fine" story, and I'm not claiming the job-market anxiety is baseless. It's not. But the transformation is happening faster at the operational layer than at the headline layer. The companies getting real returns aren't the ones making grand announcements — they're the ones quietly rebuilding their internal tooling around LLM agents and measuring the delta.

If you're watching this from outside the US, the takeaway isn't "AI is the new gold rush." It's that the deployment curve has finally caught up with the hype curve. The next 18 months will tell us whether the productivity numbers actually stick or whether this is another cycle of overbuild and correction. My bet is on the former, but I've been wrong before.