My family first got online with Prodigy dial-up when I was 8. We connected to it via a local phone number, and so I thought that meant the offices of our local phone operator down the street was where “internet” lived. As I got older I began to understand it was far more complex than that.
One of my next-door neighbors growing up was a tall man named Craig. He spoke in a loud, but friendly, North Carolina accent that likely wouldn’t make you think he was the kind of guy who gets really into computers. With his characteristic strawberry-blond horseshoe mustache, he looked, talked and even dressed like a long-lost cousin of Dale Earnhardt.
But in the same way my dad just knew how to work on any car, Craig knew computers. Every so often he’d help my family out, upgrading our first desktop from a 2GB hard drive to 8GB and later helping us make the jump from Windows 95 to Windows 98.
I’ve been thinking about Craig recently because he was the first person I ever knew who worked in a data center. When I think about data centers, I still picture them exactly the way he described them: large, cold rooms filled with rows of whirring computers packed into racks and watched over day and night by people like him.
I can’t remember where Craig worked, and I may not have known at the time. For all I know, what he oversaw was a private network for a single company and not even part of the public internet. But the image sticks with me because it was likely the first time I understood, even vaguely, that rooms full of computers somewhere were doing something important enough that somebody had to watch over them day and night.
Where Internet Lives
As a kid, I understood that going online meant my computer was talking to other computers, but I mostly imagined they belonged to other people like me. And, to be fair, my first experiences with the internet in the 1990s came at a time when it was still personal and individual enough that sometimes that was true.
That changed pretty quickly with the dot-com boom and Web 2.0. By the time I was a sophomore in high school, I’d set up my first blog on Xanga.com. Another followed pretty quickly on Blogger, and then WordPress. By then I understood that these websites weren’t running on somebody’s personal computer somewhere, but I definitely had no idea where they were actually hosted.
Three decades after I first got online, I now make web-based software for a living and I still don’t have a better answer to that question. I could look, but I don’t have to. My work requires me to trust, but rarely verify, that a data center exists somewhere, anywhere, ready to run whatever I put there.
In my head, I’ve never assigned “cyberspace” a physical address. I still tend to picture it the way the “information superhighway” was depicted on the PBS show Ghostwriter, which apparently made a lasting impression on my childhood understanding of how the internet worked.
It’s so heavily abstracted now that I never really have to think about it. And for a long time, that’s worked pretty well for me. I mean, I’ll probably never see the buildings, hear the machines or meet the people keeping them running. If my site loads quickly, stays online and doesn’t cost too much, should I even care where those machines actually live?
Aside from this blog and some GitHub Pages sites, the majority of my projects are hosted by Cloudflare. Until I looked just now, I had no idea Cloudflare even had a data center in Oklahoma City. I didn’t know because the serverless workers I use don’t require me to choose a region. So I don’t. Cloudflare figures out where my software should run, and I get to pretend the physical infrastructure behind it is somebody else’s problem.
It’s a pretty convenient arrangement when you’re the person using the cloud. It’s a lot more complicated when your backyard is where somebody else wants to put it.
Which is why the sudden data center boom in my hometown has been something of a head-scratcher. Tulsa isn’t exactly a place I’d ever thought of as an obvious destination for a bunch of enormous data centers. Oklahoma may be close to the geographic center of the country, but it’s definitely not the center of the country’s population mass. I had always assumed you built data centers to solve for latency and geography. That’s not why they’re building these at all.
Boomtown
Last week, Google announced it was behind a new 506-acre hyper-scale data center being built in the greater Tulsa area. This marks Google’s second data center in Oklahoma, and they have plans for two more. Earlier this year, Meta also broke ground on a data center in east Tulsa, their first in the state.
It’d be easy to just say my hometown needs data centers like I need a hole in the head. But I can accept that the internet, and the ever-growing amount of computing and storage required to power it, ultimately does have to live somewhere. I mean, the internet is a core part of our infrastructure now, so why should a data center be any less necessary than a power plant, refinery or sewer system?
Modern life has always depended on grotesque behemoths most of us would rather not live next to. Data centers just spent a lot of their existence hiding in plain sight. Calling them “the cloud” only reinforced that illusion, making the whole thing sound simple, abstract and almost weightless when in reality data centers are every bit as physical and resource-hungry as the industrial infrastructure we’re already familiar with, if uncomfortable with.
But looking at a map of Google’s existing and planned data centers, Tulsa doesn’t appear to fill some obvious hole in the network. That’s because adding capacity to the region isn’t really the point. This facility isn’t here to connect Oklahomans or bring the internet closer to its users. It’s coming because Oklahoma has dirt cheap land, equally cheap electricity and even cheaper politicians.
A century after oilmen rushed to Tulsa in pursuit of untapped natural resources, tech companies are rushing to exploit a whole new set of them.
Construction on Google’s project was unanimously approved by the Tulsa County Board of County Commissioners about a year ago. Typically, local governments approve projects and offer incentives to companies because, despite the tradeoffs, they expect a project to deliver one or more of the following:
- provide something the community needs
- create a meaningful number of new, long-term jobs
- generate new tax revenue
- attract complementary industries (creating even more jobs and tax revenue)
Because of what I do, and how much I personally benefit from cheap and effective data centers, I started writing this expecting to feel pretty neutral about this. I thought that data centers were just infrastructure and we didn’t have to like it, but we should suck it up and deal with it. But once I started examining those points carefully, it became harder to argue there were ever enough ends to justify the means, and the more I looked, the more frustrated I became.
I’ll start with the easiest one: Tulsa does not need this particular data center. The internet needs data centers somewhere, but this facility isn’t being built because Oklahomans are underserved or because Google needs to bring its network closer to us. The project solves a problem for Google, not Tulsa.
Next are the jobs. Project Clydesdale will definitely employ a lot of people while it’s being built, but construction jobs are temporary by definition. And with no requirement that those workers be hired locally, it’s hard to know how much of that activity represents genuinely new employment for the Tulsa area rather than existing crews moving from one project to another.
Once construction ends, the headcount gets much smaller. The initial phase is expected to employ a little more than 50 people permanently, while the project’s developer has touted as many as 100 permanent jobs overall, though it’s unclear at what stage of the buildout that figure applies. Even taking the higher number, that’s a remarkably small workforce for a 506-acre development. I used to work at a company with 350 employees on a 15-acre campus with separate office and manufacturing buildings. That’s less than 0.05 acres per employee, still more land per person than the footprint of my house. If Clydesdale ultimately employs 100 people across the full site, it would use more than 100 times as much land per permanent job.
And the permanent jobs are likely to look less like a tech hub and more like the skilled trades and operations needed to keep a data center running: HVAC technicians, electricians, facilities staff, warehouse operations and security. These are real jobs, but it’s a very different economic proposition from the high-paying knowledge work people might imagine when they hear Google is coming to town.
If you think the permanent jobs figure is underwhelming, the tax arrangement is even worse. Project Clydesdale will receive a 100% exemption from ordinary ad valorem property taxes for 25 years. In exchange, for each phase the project will make a $1.5 million annual “payment in lieu of taxes,” or PILOT, increasing 1% each year. The first phase also includes a separate $500,000 annual payment to Tulsa County for community improvements.
That sounds like a lot of money until you compare it to what the project would otherwise pay. The first phase alone is expected to represent about $1 billion in investment. At today’s rates, that could otherwise generate roughly $11 million a year in property taxes before depreciation. A $1.5 million payment against a billion-dollar investment is a literal rounding error to someone as big as Google. That’s less than a penny of every dollar invested.
How this all intersects with Oklahoma’s school-funding formula is also interesting. Normally, adding hundreds of millions of dollars to a school district’s property-tax base would make the district look wealthier to the state, which reduces the amount of state aid they receive. Under this arrangement, schools get to have their cake and eat it too. Google’s property stays off the tax roll while they still receive a slice of the PILOT.
The largest beneficiary of the PILOT would be Owasso Public Schools, one of the largest and best-resourced districts in the state, with an annual budget over $100 million. Its share would amount to roughly another one percent of its budget each year. That’s not nothing, and no superintendent is going to turn it down, but it’s hardly transformative money for a district of its size.
But while they’re projected to come out ahead under the PILOT, other local services lose out. On the first $1 billion phase alone, ordinary taxation could have meant roughly $500,000 more for libraries, $600,000 more for community colleges and more than $1 million for technical education each year.
We are running out of potential wins, and the last best hope is that new data centers might attract complementary industry. There is little evidence of that around Google’s Mayes County facility. When Google moved into the industrial park nearly two decades ago, the site already hosted more than 75 companies and thousands of workers. More businesses have moved in since, but none that directly benefit from proximity to Google. In short, their data center did not create an economic hotspot. It moved into one.
The thing data centers are good at attracting is… more data centers. While Google remains the sole hyper-scale tenant in Pryor, building out substations, high-voltage transmission lines, and high-capacity fiber turns surrounding acreage into plug-and-play territory for the next operator looking to develop nearby.
So that’s it. There are still a few softer benefits around the edges, like promises of future phases and programs to train skilled tradespeople. The developer is also paying for infrastructure upgrades its own project requires. But after going through the usual case for economic development one point at a time, none of it feels substantial enough to justify the bargain county commissioners ultimately struck.
One final argument I haven’t really explored, and probably the strongest case I can steelman is that if we don’t accept projects like this now, something worse might come along and we’ll have less choice. Similar industrial land could end up hosting a fluoride-polluting aluminum smelter or a nuclear waste facility that also includes data centers. Those aren’t exactly appealing alternatives.
But “it could be worse” is a miserable standard for economic development. I’m not arguing we should welcome a fluoride-belching smelter or nuclear repository because the jobs numbers look better on paper. I’m arguing that “none of the above” should still be an option.
I refuse to accept that my corner of the country is in such dire straits that we have to choose between toxic heavy industry and server farms exempted from ordinary property taxes for 25 years. I consider myself a pragmatist and value compromise, but this isn’t pragmatic compromise. It’s subsidized extraction.
Talking Points
Tomorrow is Election Day for the Tulsa City Council—the leaders of my city, not the county officials who approved Google’s data center. None of the mailers piling up at my house from the candidates have said anything about their views on data centers. I’d considered reaching out to ask, but after considering it for a second I realized a “good” answer would be a mealy-mouthed diatribe about how their role is to seriously consider the merits of the applications put in front of them.
Normally, that’s exactly what I want a politician to say. I don’t want people running for office promising to reject things they haven’t seen yet. The problem is that local leaders have been saying some version of that for years, and the system they operate inside seems heavily biased toward saying yes.
I believe one of the biggest problems with new data centers is actually something quite old. For decades, state and local governments have competed in a kind of economic development sweepstakes, offering tax breaks, land and infrastructure to anything they can plausibly sell as a political win.
Data centers fit that bill almost too well: no polluting smokestacks, no dangerous chemicals to spill, and a headcount so low it won’t affect local traffic. But when a data center consumes hundreds of acres of viable ranch land, puts new pressure on the power grid and contributes relatively little to the tax base, residents can still end up seeing it as the Superfund site local leaders thought they were avoiding.
But demanding that leaders “just say no” isn’t necessarily viable policy either. Some data centers are moving forward simply because the rules already allow them to. If land is properly zoned and a project complies with the law, local government may have little basis to reject it just because it’s unpopular. Tulsa is now dealing with that, having paused new data centers while it rewrites the rules. Property rights don’t disappear just because we dislike the project.
Part of why I usually avoid writing about this stuff online is that the nuance tends to disappear pretty quickly. If you acknowledge the trade-offs, or point out that some of these projects are perfectly legal under the rules already on the books, somebody assumes you’re making excuses for them or that you’re some kind of “AI-pilled tech bro.” But I’ve spent this whole piece trying to take the strongest version of the case seriously. I can understand why these projects get built and still think we’re making a bad bargain.
But pitchforks and torches aren’t policy proposals, and “ban 100% of AI” isn’t a workable one either. I understand and respect the people who feel something genuinely unjust is being done to their communities, both physically and culturally. But some of the rhetoric I’ve seen online goes well beyond skepticism about AI or data centers. I’ve seen people argue, essentially, that it would be better to slow things like cancer research than allow AI to become more entrenched. If your anti-AI position leaves you siding with cancer, you may want to reconsider how absolute it’s become.
At some point, that stops being a disagreement over trade-offs and becomes a refusal to acknowledge that some trade-offs may be worth making. And that makes me wonder whether we’re actually ready to have serious conversations about what we want the future of this country to look like.
We’ve spent decades failing to have a serious national conversation about man-made climate change, severe drought and historically low waterways. But put a data center in the picture and suddenly everybody wants to talk about water usage and cooling systems.
Wages have stagnated, labor protections have eroded, and the federal minimum wage has been stuck at $7.25 since 2009, but we’re suddenly treating AI job displacement like it’s the biggest threat to working conditions and the thing driving the affordability crisis.
None of this means the effects of AI or data centers aren’t real, or that we shouldn’t take them seriously. We should. But if we’re going to have a serious conversation about them, we have to start with the world as it actually exists, not one where AI alone is blamed for drought, politics and aging infrastructure.
Communities have every right to decide what they’re willing to trade away in exchange for hosting data centers. But that starts with better zoning, an honest accounting of the costs, and making some of the richest companies in the world pay their fair share.
When the clouds do come home, they should at least be good neighbors.