My dad only bought GM. What if we asked Meta for the product, too?

Growing up, my dad only bought GM.

It wasn't really about the cars. He ran a business, and a lot of his customers worked in the auto plants. When they did well, he did well. When the line slowed down, so did his week. Buying a Chevy was not loyalty to a logo. It was loyalty to the people across the counter, and to an industry that made something real close to home.

There was another part to that culture, and if you grew up around here you remember it. Domestic versus foreign. Better to drive a Ford in GM country than a Honda. Buying the wrong car did not feel like a consumer preference. It felt like picking a side.

The AI argument sounds familiar.

When Lt. Governor Micah Beckwith came to a Marion town hall on the last day of July, he said data centers are how "we win the technology war," and that they are a matter of national security. Build them here, or China builds them there.

I have heard that argument my whole life.

It used to be about cars.

I have been thinking about all of that while reading about the data center proposed for Marion.

A data center is not exactly a factory in the traditional sense. Nothing rolls out the door on a truck. But economically, I think it helps to think of it as one.

What it produces is computing power — the thing behind the AI models people are already using to write, research, design, code, analyze and automate work. That capacity is expensive. Demand for it is enormous. That is why a company would consider spending roughly $2 billion to build a facility here in the first place.

So here is the question I wish we had asked:

If somebody wants to build a factory in Grant County, why are we only negotiating for a check? Why not ask for some of what the factory makes?

I am asking late. The proposal may already be dead.

I am asking anyway, because I am not sure we ever got far enough to have the argument.

What was on the table

The summary sheet Mayor Ronald Morrell Jr. released in August listed $176.5 million in "community impact payments" over the life of the project.

That is real money.

Over twenty years, it averages about $8.8 million a year. Grant County has roughly 66,500 residents, so spread evenly across the population it is the equivalent of about $133 per person per year — around $11 a month.

That is not how community payments actually work. Nobody was proposing to hand every resident eleven bucks. Concentrated into roads, schools, public safety or infrastructure, $8.8 million a year could matter a great deal.

But the per-person number gives us a sense of scale.

And it raises a different question:

Could the same cost to the company create more value here if some of the bargain came in the thing the facility exists to produce?

Ask for what the factory makes

A technology company can produce computing capacity for less than a customer would pay to buy comparable computing or AI services on the open market. The company owns the chips, the buildings and the infrastructure.

The hard part is knowing how big that difference really is.

Meta does not publish a Grant County price list for its internal computing capacity. It is not Amazon Web Services. There is no clean retail number I can point to and say, this is exactly what one dollar of Meta's compute costs to produce and exactly what it is worth.

So what follows is a thought experiment.

Suppose a dollar of cost to the company could provide four dollars of comparable commercial computing value.

Under that assumption, instead of costing Meta $176.5 million to give Marion $176.5 million in cash, the same company cost could theoretically deliver something like $706 million worth of computing capacity over twenty years.

Four dollars of usable value here for every dollar of company cost.

Maybe the real ratio is two to one. Maybe it is six to one. Maybe the entire assumption falls apart.

But the principle is worth considering:

Ask for what the factory makes.

What a deal could look like

I can imagine four pieces.

One: Give every resident a real AI allowance.

Not a seven-day trial. Not a promotional account. A meaningful amount of access every year.

Using the hypothetical numbers above, if 60 percent of the compute pool went to residents, that would come to roughly $318 per person per year — about $26 a month in commercial value.

That is enough to cover the cost of many paid consumer AI plans today.

Every resident. Every year.

Two: Build a serious pool for local businesses.

Put the remaining 40 percent toward businesses.

Federal labor statistics counted 1,298 business establishments in Grant County in early 2026. That count only includes employers with payroll, so it misses some farmers, independent contractors and one-person businesses. The true denominator would be larger.

But using 1,298 gives you a sense of scale.

Our hypothetical pool would work out to roughly $10,900 a year per establishment in computing capacity.

For a small business that knew how to use it, that could buy an extraordinary amount of research, coding, design, analysis and automation.

Three: Whatever we don't use, sell it — and put the money into the schools.

Plenty of residents will never use their full allowance. That value should not simply disappear.

The agreement could require Meta to resell the unused capacity itself and send the proceeds back to Grant County, or buy the unused portion back at an agreed percentage of its market value.

Using the assumptions above, if 45 percent of the resident pool went unused and was converted back to cash at half of retail value, that would mean about $4.8 million a year — roughly $95 million over twenty years — for Grant County schools.

The point is simple: residents get first use of the product. What they do not use becomes cash for the schools.

Four: Give Grant County permanent access at the company's cost.

This may be the most valuable part.

Once a local business burns through its free allocation, let it buy more at some negotiated price tied to the provider's cost rather than normal market rates.

If our hypothetical four-to-one ratio were anywhere close to reality, a company here spending $100,000 could receive something approaching $400,000 in commercial computing value.

Now we are no longer talking about a community perk.

We are talking about an economic-development strategy.

Grant County has seen this before.

Natural gas helped turn Marion into an industrial city and turned a little village called Harrisburg into Gas City. Factories came because the fuel was here and it was cheap.

If computing power fuels the next generation of businesses as natural gas fueled the manufacturers of the gas boom, unusually cheap compute could give Grant County that kind of advantage again.

Why I think businesses could follow

If somebody handed me a generous frontier-AI account tomorrow and told me I could use it without worrying about the bill, I would build things.

Most would go nowhere. One or two might work.

Now multiply that possibility by every welder, teacher, farmer, nurse, programmer, student, manufacturer and shop owner in Grant County who has an idea and no budget to test it.

Then imagine a company deciding where to locate a new operation and discovering that one county has access to an important business input at a fraction of what competitors elsewhere pay.

That is a different economic-development pitch.

The jobs would not primarily come from the data center itself. The Marion summary sheet projected fifty full-time positions, which is in line with the relatively small permanent workforces large data centers typically employ.

The bigger opportunity, if there is one, would be jobs created through businesses using what the data center produces.

That feels familiar to me.

My dad's business did not work for GM. GM did not own it. But when thousands of people in this region made good money making cars and car parts, they hired people like my dad.

The factory's prosperity leaked out its doors.

What would it look like to intentionally design that effect for an AI economy?

Nobody has actually done this

I could not find a community anywhere that has successfully negotiated the kind of compute agreement I am describing.

Nobody has landed this.

But I did find evidence that the basic instinct — ask a data center for something besides ordinary tax revenue — is already showing up around the country.

The closest example came from Jordana Barton-García, who published a guide in March 2026 for communities negotiating with data centers. Among her suggestions: require developers to dedicate server capacity — "compute power" — to local innovation, research and startups. She also suggests using excess connectivity to improve local internet access.

That is a recommendation, not a deal anyone has signed.

New Jersey went a step further. Companies seeking state AI tax credits had to work with New Jersey universities or startups and document benefits including "price concessions, artificial intelligence support services, or other measures".

In plain English: part of the public benefit could be cheaper access to AI.

Then came the twist. On August 28, New Jersey passed legislation taking the remaining $250 million in that incentive program back. The state decided the credits were not worth continuing.

Two Brookings Institution researchers proposed another version in January: communities hosting data centers should receive at least half of one percent of the facility's gross revenue every year, through a legally binding and public agreement.

Their version is cash. Mine is partly product.

But the instinct is the same:

If a community is helping create the conditions that make an enormously valuable enterprise possible, share some of the value it creates.

Communities are already negotiating creatively.

Frederick County, Maryland announced a $110 million agreement with developer Catellus on September 1. It is broken out line by line: $40 million for a community center, $30 million for an elementary school renovation, $14.5 million for workforce training, $10.5 million for farmland preservation, $5 million for community solar and $1 million for a fire engine. The developer also agreed to reduce the project's size and dramatically reduce its drinking-water use.

Elsewhere, developers have offered direct payments to households, money directly to school districts and public broadband infrastructure.

The details are all over the map.

That is the point.

There is no rule that says the only thing a community can ask for is a check.

Where the idea breaks down

There are big problems with this idea.

First, the computing power might not actually be cheap for them.

The thought experiment depends on a gap between what computing costs the company and what comparable capacity is worth to us.

But cost is not the only thing that matters.

If every server is already running full tilt, then every unit of computing capacity given to Grant County is capacity the company could have used somewhere else. Its economic cost may be much closer to the market price than the electric bill and depreciation on the chips would suggest.

If that is true, the four-to-one premise disappears.

Then this is just a different-shaped check.

Second, you cannot fix a water main with computing power.

Cash buys fire trucks.

Cash repairs roads.

Cash can pay teachers.

A family that needs a roof does not need tokens for an AI model.

Anything we take in product could be something we did not take in money, and money remains the most flexible resource on earth.

Third, access is not distributed equally just because everybody gets the same amount.

The person most likely to turn $300 of AI capacity into something valuable is probably already comfortable with technology, already educated in how to use it, or already running a business that can take advantage of it.

Give everyone $300 in cash and everybody has $300.

Give everyone $300 in compute and some people will turn it into thousands of dollars of value while others will never log in.

That is not necessarily a reason not to do it. But the benefits would not be equal.

Fourth, twenty years is forever in technology.

A twenty-year promise of "AI access" sounds impressive until you remember what technology looked like twenty years ago.

There was no iPhone.

Facebook was two years old.

YouTube was one.

Nobody in Grant County should sign an agreement whose value depends on Meta — or any particular AI product — looking in 2046 anything like it looks today.

Whatever we negotiated would need hard definitions, minimum values, substitution provisions and a floor of real money underneath it.

And Grant County has seen the other side of this story, too.

If you grew up around Marion, you didn't need anyone to explain what Fisher Body was.

GM opened the plant in 1956, and for generations it was one of the anchors of the local economy. GM is still here. The plant we still call Fisher Body survived. It still stamps parts for GM vehicles, and GM put nearly half a billion dollars into it just a few years ago.

But Fisher Body was never the whole story.

There was Dana on Miller Avenue making universal joints and driveshafts. By 1960, Dana's Marion operation was making one out of every five driveshafts produced in the United States.

There was Amcast in Gas City making aluminum wheels, most of them for GM.

Up the road in Anderson there was Delco and the enormous GM network around it.

There were plants making parts for plants that made parts for other plants.

That whole network put a lot of good paychecks into a relatively small part of Indiana.

Then it got smaller.

Delco is gone. Amcast is gone. Dana is gone. Other factories and suppliers closed, consolidated or learned to make the same amount of stuff with far fewer people.

GM survived in Marion.

Much of the ecosystem around it did not.

And now the old Dana property is the rumored site where Meta wanted to build its data center.

There is something almost too neat about that.

A place that once made driveshafts for the automotive economy could have become a place producing computing power for the AI economy.

The lesson from our history is not that we should fear the next industry because the last one changed.

It is that we should understand what makes the next industry valuable while we have the chance — and make sure some of that value stays here.

So ask for both

I would not trade the promised $176.5 million in cash for $176.5 million worth of Meta's internal costs.

That is too risky.

I would ask for both.

Real money for the things only money can do.

And then, on top of it, a negotiated share of the product the facility exists to create: guaranteed computing capacity for residents and local businesses, a way to convert unused capacity back into money for schools, and long-term preferred-cost access for businesses here that want more.

Maybe the economics make it impossible.

Maybe the company would refuse.

Then negotiate.

The important question is not whether my particular formula is right.

The question is why we were only pricing the deal in dollars.

We said no. I am not sure we ever heard the question.

The project is over, or close to it.

On August 26, Mayor Morrell said the proposal was "not the right fit for the city" and that he would not endorse it.

His explanation was straightforward: "Over the past several days, it has become clear that this is not the direction the citizens want."

On September 1, with residents holding signs in the council chamber, Council President Brian Cowgill said from the chair: "That's a done deal. It's not coming. It's over."

A petition went up on August 24 and collected more than a thousand signatures inside of a week. Signs appeared at the council meeting. A town hall was organized. The mayor explicitly cited public feedback when he stepped away.

That is democracy working the way democracy is supposed to work.

But after spending months reading county documents and listening to public recordings, I am not convinced Grant County ever got to the most useful version of the debate.

Much of the public conversation collapsed quickly into not here.

Sometimes that conclusion rested on serious concerns about water, electric rates, farmland, noise, secrecy and the wisdom of making a twenty-year bet on a massive technology company.

Those are real arguments.

Sometimes, I think, the conclusion came first and the reasons followed.

Either way, we moved very quickly from what is this? to we do not want it.

We rarely got to the next question:

If it did come, what could we demand in return?

But much of that problem started before the public ever got involved.

You cannot negotiate with a rumor

It is very hard to be open-minded about something nobody will describe.

For a year, the county's own economic-development study advised officials to soft-pedal the subject in public documents.

The summary sheet that finally surfaced did not say "data center." It did not name a company. It did not say where the project would be.

When residents began asking public bodies whether officials had signed non-disclosure agreements, the question had to be asked repeatedly — and it still remains largely unanswered. Our canvass of the officials involved shows who answered, who did not, and exactly what each person said.

Nobody handed Grant County a proposal and said:

Here is the company.

Here is the land.

Here is the water requirement.

Here is the electric load.

Here are the tax terms.

Here are the jobs.

Here is what we are offering the community.

Now tell us what you think.

Instead, residents got a rumor and eventually a fact sheet with the name cut out.

You cannot negotiate with a rumor. You can only object to it.

And a community that is only ever given something to object to will get good at objecting.

Before Grant County could ask what a $2 billion AI facility ought to give the community hosting it, residents first had to figure out whether the facility existed at all.

They did not tell us enough to negotiate.

We did not get far enough to ask what else we could negotiate for.

Together, those are how everybody ends up with nothing.

Where it actually stands

A mayor declining to endorse a project is not the same thing as the developer formally withdrawing it.

Meta — which Mayor Morrell named in an August 27 interview — has never publicly announced a Marion project. No application was ever filed. Morrell has said landowners had already agreed to sell, and those are private arrangements that his opposition does not automatically erase.

Council President Cowgill may be completely right that "it's over."

But "it's over" is his assessment, not a filing from the company.

So this is not a news story announcing that the project is coming back.

It is not an endorsement of the project, either.

It is an argument about the conversation I wish we had been able to have.

If another proposal comes — whether this one somehow reappears or another one arrives under a different company name — I hope we do better.

Demand the water numbers.

Demand the power numbers.

Demand the tax numbers.

Demand the location.

Demand the agreements in public.

Ask what can go wrong.

And then ask something else, too:

What do you make, and how much of it are you willing to leave here?

My dad bought GM because GM and the companies around it made things here, and his neighbors were paid to make them.

Their paychecks moved through grocery stores and restaurants and hardware stores and businesses like his.

That was the deal, even when nobody wrote it down.

If somebody wants to build the factory of the next economy in Grant County, I want us to understand what it makes.

And I want us to get some.

By the way, my dad now drives a Toyota Tundra.

He says it is the best truck he has ever owned.

Make of that what you will.