I spent the first half of my career comparing tower companies by the wrong number.

Honestly, it wasn't my fault. Every analyst report, every earnings call, every “American Tower vs Crown Castle” comparison starts with total site count. American Tower says hundreds of thousands of sites globally. Crown Castle has tens of thousands of towers plus a large small-cell network. The implication: more sites equals better coverage equals better investment. And for a while, I bought it.

Then I cost my company a lot of money. Enough that I now keep a checklist taped to my desk. And I'm writing this because if you're comparing tower REITs, you're probably making the same mistake I made in 2021.

The Surface Problem: Site Count Is the Blood Pressure Monitor

A blood pressure monitor is a useful tool. It tells you if your pressure is high. But it doesn't tell you if you have blocked arteries, an irregular heartbeat, or a tumor. You'd never accept a heart-health diagnosis from a BP reading alone. Yet that's exactly what happens when someone compares tower REITs by site count.

The question everyone asks is: “Which one has more sites?” The question they should ask is: “Which sites are actually generating reliable revenue with minimal hidden costs?”

Site count tells you scale, not quality.

Most buyers focus on site count and completely miss backhaul costs, power availability, ground lease escalations, and equipment age. These can add 30–50% to the lifetime cost of a tenant lease. That's not a rounding error.

What I Missed: Site Count Doesn't Measure Total Cost

In 2021, I was helping a regional carrier evaluate two options: a deal with a national REIT and a lease portfolio from a smaller operator. I ran the numbers like most people run the numbers—monthly rent per site, multiplied by term, plus a rough allowance for installation. The smaller operator looked 15% cheaper. We signed.

The problem was that I was comparing rent, not total cost of ownership. The smaller portfolio had older towers. Two of them needed structural modifications to support a 5G antenna array. One site had no backup generator. The ground lease on three sites was subject to a 10% escalation every five years. The “cheap” portfolio ended up costing us around $412,000 in retrofits, legal fees, and emergency generator rentals over two years. I say around—$412,000 is the number I remember, but I'd have to check the actual project ledger; it might be slightly higher.

The mistake wasn't that I chose a bad vendor. It was that I used the wrong framework. Site count and monthly rent are surface numbers. They're easy to compare, which is exactly why they get used. But they don't tell you:

  • Who owns the ground under the tower? If ground lease rents are reset to market value, your tenant rent can jump.
  • Does the site have fiber backhaul? If not, you're leasing connectivity separately, and that can double the cost.
  • What's the power arrangement? Commercial power vs. metered power vs. first-come backup power changes outage risk and cost.
  • How old is the structure? A tower built in 2005 with no structural analysis for new antennas is a different asset than one built last year.

I'm not saying site count is irrelevant. It matters. But it's an input, not a conclusion. It tells you scale, not quality. And scale without quality is how you end up with a 10-year lease on a site that can't support your actual network plan.

The Cost of Ignoring These Details

Let me give you a real scenario. In September 2022, we had a fiber cut at one of those “cheap” sites. The site was down for nine hours because there was no redundant path. The carrier's SLA penalty was $11,000. Plus we had to pay for extra construction crews. That one incident ate the first year of “savings.”

The most frustrating part? This was avoidable. The fiber route was public information. A decent due-diligence process would have caught it in an afternoon.

People think expensive vendors deliver better quality. Actually, vendors who can document structural modifications, power redundancy, and backhaul diversity—and can prove it in a lease agreement—charge more because the asset is genuinely better. The causation runs the other way: quality determines price, not the other way around. At least, that's been my experience with regional wireless deployments.

Why the Upgraded American Tower Corporation (AMT) Changes the Conversation

Now, I'm not writing this to say American Tower is always the right answer. Crown Castle has strengths, especially around small cells and dense urban fiber. But there is a reason the conversation around “American Tower Crown Castle” has shifted after the American Tower Corporation (AMT) upgraded its portfolio with the CoreSite acquisition and an expanded edge data center footprint.

CoreSite gave AMT data center space and interconnections. That matters because the boundary between macro tower, small cell, and edge computing is blurring. A site with an edge data center on the same property can host compute and storage closer to the user. That lowers latency and backhaul pressure. For a carrier, that can mean avoiding a separate lease and a separate power bill. In TCO terms, the “higher” AMT lease rate can be cheaper than a lower-priced site with no edge capacity.

The upgraded American Tower Corporation (AMT) balance sheet also gives it a longer runway for capital improvements. Again, not a guarantee. But when you're modeling a 15-year lease, you want a counterparty that can maintain the asset. You don't want to sign with an underfunded REIT and then wait two years for a leaking shelter to be repaired.

The Surface Question Hides the Real One

When someone searches “how to set voicemail on phone,” they want a simple answer: press the voicemail key, enter a password, record a greeting. Done. But if the tower carrying that phone's calls has no backup generator, or the backhaul is congested, the voicemail notification may never arrive. The surface question is easy. The underlying infrastructure is not.

That's the same problem I see in infrastructure negotiations. Buyers ask “which REIT has the lower per-site rent?” or “which stock has better momentum?”—but those are surface questions. The deep question is: “What does this portfolio cost to operate over 15 years, including risk?”

If you don't answer that question, you end up with what I had: a lease on paper that looked affordable and a network that wasn't. Honestly, the fact that our rollout survived at all was luck. I'd rather not rely on luck.

The Short Solution: TCO, Not Sticker Price

So what do I do now? I run every potential site through a checklist before the lease is signed. It keeps me honest, and it has caught 47 potential errors in the last 18 months.

  1. Pull the FCC structure registration (47 CFR Part 17) for every tower. It gives you owner, height, and modification history.
  2. Build a TCO model with at least these lines: base rent, ground lease escalation, power, backhaul, maintenance, insurance, structural modification, and exit/remediation costs.
  3. Stress-test the site for a 72-hour outage. Which assets have no generator backup? What is the likely SLA penalty per hour?
  4. Ask the landlord directly about backhaul diversity and power metering. If they won't put it in writing, that's a red flag.
  5. Compare the lease rate over the full term, not the first-year rate. Escalation clauses can turn a “cheap” site into the most expensive one.

This isn't glamorous. It's a spreadsheet and a lot of patience. But it works. The last time we did this, we walked away from a vendor whose bid was 20% lower on rent. Our TCO model showed that when we added generator rental and fiber leases, it was actually 8% more expensive over 10 years. We would've paid more to save money, if that makes sense. Actually, we would've paid more to feel like we were saving money—that's the trap.

Bottom line: the next time you're debating American Tower vs Crown Castle, or any infrastructure provider, ignore the headline site count. Ask about the assets behind it. The answer should always be “what's the total cost to deliver a service that actually works?” If you answer that first, you'll make fewer mistakes than I did.

Technical planning note: validate insertion loss dB, PIM dBc, grounding resistance, and relevant 3GPP TS 38.xxx requirements before final RAN acceptance.