The Hidden Subsidy in Every Data Center Pro Forma

— by

I’ve been reading a lot of headlines lately about America’s data center boom — billions of dollars in investment, AI demand that just won’t quit, entire towns getting reshaped by a single building. What I haven’t seen covered nearly enough is the other half of the story: how much of that economics is actually being financed by what governments are choosing not to collect.

Take Ohio. The state forecast it would lose $136 million in 2025 to a sales tax exemption on data center equipment. The actual number came in at $1.6 billion — eleven times higher. That’s not a rounding error. That’s a state whose forecasting model was built for a different era of data center, and got run over by the scale of what’s actually being built today.

Three Layers of Incentive, Stacked on One Project

Here’s what’s actually happening underneath the demand story. Three layers of tax incentive stack on a single data center build. Federally, the One Big Beautiful Bill Act made 100% bonus depreciation permanent for equipment acquired after January 19, 2025 — and equipment (the chips, the servers) makes up roughly 60% of total project cost on a typical one-gigawatt facility. States layer on sales tax exemptions on that same equipment, often locked in for decades. Localities add property tax abatements on the land and the building.

If this pattern sounds familiar, it’s because finance teams have seen it before in renewable energy. Solar and wind developers have long relied on tax equity investors — capital that shows up specifically to monetize an investment tax credit, not because the raw energy economics alone clear the bar. Data centers are increasingly financed the same way: some of the capital in the stack isn’t underwriting AI demand at all. It’s underwriting the tax shield.

None of this is hidden exactly — it’s public record if you go dig for it. But it rarely makes it into the “AI demand is exploding” headline, even though it’s doing real work in the underlying math.

Why the Timing Matters As Much As the Size

The thing that actually changes a project’s financial profile isn’t just the size of the tax benefit — it’s when it lands. Bonus depreciation doesn’t spread the deduction over the life of the asset. It front-loads the entire benefit into year one. For a project with massive upfront capital spend and a long runway before revenue normalizes, that’s not a footnote on the tax return. It reshapes the entire discounted cash flow.

Separate the Demand Story from the Incentive Story

If you’re evaluating one of these projects — or just trying to make sense of why so many keep getting built even as public pushback grows — the useful exercise is running the model twice: once with the full incentive stack, once without it. The spread between those two numbers tells you how much of the return is actually coming from AI demand, and how much is coming from what the government isn’t collecting.

That spread also tells you something about risk. Incentives can be paused. Ohio froze new exemptions in May 2026. A Utah county passed a six-month data center moratorium that same year. If a project only clears its hurdle rate because of the incentive stack, a policy shift isn’t a footnote — it’s the difference between viable and underwater.

My honest read: some of what’s gotten labeled as a pure demand boom was probably an incentive-fueled boom wearing demand’s clothes. As those incentives come under more scrutiny — and they clearly are — we’re likely to get a cleaner read on what the real underlying demand actually looks like.

Key Takeaways

  • Stacked tax incentives — federal bonus depreciation, state sales tax exemptions, local property tax abatements — materially shape a data center’s after-tax returns, not just its tax bill.
  • Equipment is roughly 60% of total project cost, and it gets the most aggressive tax treatment — with the benefit landing in year one, not spread over time.
  • Incentive durability is a financial risk. Ohio and Utah both disrupted data center incentives in 2026 — proof this risk isn’t theoretical.
  • Run your models twice: with and without the incentive stack. The spread is the real measure of how dependent a project’s economics are on policy staying put.

Want the CPE credit? Take the full lesson on EverydayCPE and earn 0.2 CPE credits: https://everydaycpe.com/courses/technical-finance/the-hidden-subsidy-in-data-centers/

Today’s lesson


Leave a Reply

Discover more from EverydayCPE

Subscribe now to keep reading and get access to the full archive.

Continue reading