For years, I treated American Tower like a simpler company than it actually is. Towers, long-term leases, steady rent. My mental model was a real estate investment trust with antennas. Then I spent 18 months working with a carrier evaluating edge data center capacity, and I discovered how wrong that model was. The mistake wasn't just embarrassing—it cost me the trust of a client and roughly $28,000 in rework and wasted analysis. This is what I learned.

Let me rephrase that: American Tower does own tens of thousands of towers, and that business is steady. But after the CoreSite acquisition, it also owns a data center platform that behaves very differently. If you assess the whole company with one valuation shortcut, you're going to make bad decisions. I made that exact mistake in 2024 when I told a client that the P/AFFO multiple for 2025 made the stock look overvalued. I was using the wrong comparative basis. (I had forgotten that the data center segment contributes a fast-growing, higher-multiple slice of the business.)

The First Mistake: Treating American Tower Like a Pure Tower REIT

The usual shorthand for valuing American Tower is the price-to-adjusted funds from operations—P/AFFO. For the legacy tower portfolio, that's a decent starting point. Towers are long-lived real estate assets with stable cash flows, and the market prices them on AFFO yield. I've spent years using that metric, and it works well when the portfolio is homogeneous.

The problem is that the portfolio isn't homogeneous anymore. American Tower's CoreSite business is a data center company, and data centers don't behave like towers. Their leases are shorter (typically 3 to 7 years vs. 10 years plus for towers), and their market value is more sensitive to capacity utilization and interconnection growth. For a data center, the total enterprise value—TEV—as a multiple of forward EBITDA often tells you more than P/AFFO does. When American Tower announced the CoreSite deal in 2021, the TEV was roughly $10 billion. I remember thinking, "why would a tower REIT spend that much on data centers?" The answer, which I missed at first, is that data centers are becoming the logical extension of the tower network.

What I Missed: The CoreSite Factor

Here's where I had to unlearn my habit. I kept looking at American Tower as a pure tower stock, so I applied a tower multiple to all of its cash flows. But CoreSite's edge data centers serve a different purpose. They interconnect networks, host cloud on-ramps, and reduce latency for content. The traffic isn't going to a tower; it's staying in a building.

This is also where the "networks vs Cisco" confusion comes in. I've seen internal teams compare networking hardware brands—Cisco, Juniper, Nokia—as if the device were the network. It's not. A network is defined by physical locations: where the towers are, where the fiber runs, where the data centers sit. The most expensive router in the world is useless if it's not in the right place. I once watched a team spend months negotiating a Cisco discount while ignoring the fact that their planned edge node had only one fiber route in and out. That's the equivalent of buying a high-performance car and parking it in a garage with no doors.

So when I think about CoreSite, I don't think about racks and servers first. I think about geography. Take De Soto, KS, for example. It's not a tech hub that shows up on most radar screens. But a site like that matters because it sits at the junction of regional fiber routes and can serve traffic that would otherwise have to backhaul through Kansas City. The latency difference is real. A carrier that learns to leverage edge points like De Soto can offer better performance than a competitor with fancier equipment but worse locations.

The Real Cost of Getting It Wrong

The practical consequences of this misjudgment were not academic. In March 2024, I delivered a report to a regional carrier recommending against expanding edge data center leases because I had valued the whole company's future cash flows using a tower P/AFFO model. The conclusion implied that American Tower's data center expansion was value-dilutive. That analysis missed the revenue growth from CoreSite's existing interconnect base and the new deployments near Kansas City and other secondary markets.

The client didn't follow my advice. They did their own diligence, signed a lease in De Soto, KS, and that decision worked out well for them. Our firm, meanwhile, had to redo the entire analysis. The wasted work was about $28,000 in consulting hours, plus a serious hit to our credibility. (I still have that report in a drawer as a reminder.)

That's when I started keeping a checklist. Since then, the checklist has caught 11 instances where someone was about to apply a standard REIT metric to a non-tower asset. It's not a perfect system, but it forces us to ask: "Which segment does this revenue come from?"

A Short, Honest Approach That Actually Works

Here's what I now recommend, and it's deliberately simple.

  1. Split the company. Separate the tower portfolio from the CoreSite data center segment. Use P/AFFO for the towers, but use TEV-to-EBITDA or similar multiples for the data center business.
  2. Watch the real estate, not just the charts. Look at where the towers and edge sites actually are. De Soto, KS, is a real reminder that an unglamorous location can have strategic leverage.
  3. Question device bias. When evaluating networks, the technology brand (Cisco or anything else) matters far less than the physical network design. The best devices won't save a poorly located site.
One number is never enough to value a company that contains two different businesses.

One caveat: this approach is for investors and analysts trying to value American Tower as a whole. If you're a small wireless tenant negotiating a lease on a single tower, none of this matters for you. You should focus on the commercial terms, not the REIT's data center strategy. Similarly, if you're evaluating CoreSite's data centers as an enterprise customer, the relevant metric isn't P/AFFO at all—it's the price per kilowatt and interconnection cost.

Honestly, I'm still not sure I fully understand all the accounting mechanics behind TEV and AFFO. But I understand that mixing up the two sides of this business is a mistake I won't repeat. If you're looking at American Tower for 2025, do yourself a favor and look under the hood of the P/AFFO number. The question isn't just "what's the multiple?" It's "what are you actually valuing this multiple on?"

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