Chips, Compute and Counties 

Key Takeaways

Within a seven-day span, I visited two bookend projects driving our nation’s economy. Both are massive complexes transforming local communities while sparking intense yet necessary debates over energy, labor, land, noise, tax and water policy. 

First was a six-building, 4-million-square-foot data center campus for Google in Armstrong County, Texas, population 1,843, where 7,500 construction and craft workers show up each day. 

Then Onondaga County, New York, where Micron has broken ground on a semiconductor megafab, an investment of up to $100 billion over the next 20 years.  

One site will train and run the advanced AI models. The other will make the ultra-fast chips that underpin the endless progression of AI from generative to agentic to autonomous intelligence. Both projects sit atop a global transition as consequential as any county has ever faced and as unforgiving to those unprepared for it. 

We’re living through a second Gilded Age, the term historians used for the late 1800s’ extraordinary concentration of industrial capital in rail, steel and oil. This speed and scale of AI infrastructure investment is stunning: an estimated $800 billion this year alone across the country. 

At an estimated 3.6 percent of GDP, this dwarfs the startup eras of canals, highways, airports and the internet of below 2 percent. It is rivaled only by the railroad expansion of the 1860s through 1890s, estimated at 6 percent of GDP at its peak. 

Sit with the Armstrong County number again. More workers show up on this rural job site each morning than live in the entire county — by roughly four times. The project is part of Google’s $40 billion commitment across Armstrong and Haskell counties alone. That single commitment approaches two-thirds of the federal government’s entire highway and bridge budget. It’s 10 times the FAA’s entire airport improvement program. It’s larger than FEMA’s whole disaster relief appropriation for 700 to 900 counties dealing with major climate events each year. 

This is why NACo has been deeply engaged in tracking these massive investments, while also building peer-learning networks and hosting site visits in Oregon, South Carolina, Texas and Virginia, among other places. We are studying community benefit agreements negotiated by counties. We are collecting and sharing county plans and ordinances for air quality, noise pollution, building setbacks, water quality and much more. 

Here is what we’ve learned. Zoning decisions are intense and swift. Noise ordinances and decommission plans matter. Water agreements need teeth: tracking, public reporting and baseline studies before approval, not after. Setbacks, transmission corridors and workforce housing all need answers before the site plan arrives, not after the ribbon-cutting. 

Many of these decisions may sit outside the authorities of counties, yet residents still hold county officials accountable. 

So how should county leaders prepare? Study up. Go tour these facilities. Get ahead of it. Demand a seat at the table with federal, state and municipal officials. 

We’ve lived this before — the Gilded Age, the post-World War II manufacturing glory years, the boom-and-bust cycles of oil and gas, housing and the internet. Now it’s chips and compute. 

The fundamentals remain the same: demand a seat at the table, study the supply chain, understand the financing capital stacks, ask hard questions and fight for a deal that leaves your community better off — not used and abused. 

County government isn’t a passive observer on this journey. We’re on the ground where change is real, not in depots and switchyards this time, but in server farms and cleanrooms the size of NFL stadiums. 

The last Gilded Age caught plenty of communities flat-footed. With chips and compute, let’s make sure we’re the ones writing the terms, not just living with them.

Will we ever learn from history?

When steam-powered rail cars first rolled through the countryside in the late 1800s, people who’d spent their lives depending on horses and buggies didn’t see progress — they saw a threat. The automobile. The jet engine. The nuclear age. The internet. Each one triggered the same cycle: fear, resistance, adaptation, normalization. 

We’ve gone from mainframes to PCs to handheld tablets, from switchboard operators to the phone book to the internet in our pocket. Now AI is moving through its own version of that arc — from deep learning to generative to agentic to autonomous intelligence. 

Every one of these revolutions brought physical infrastructure, massive capital investment and workforce disruption that landed squarely on the doorstep of county officials. AI is no different — except this time, the speed, scale and thirst for resources are bigger and more concentrated than anything that came before.