The first mistake I see in data center land deals is also the most common one: somebody quotes a price per acre and thinks they have said something useful. They have not. A 100 acre site that can deliver 20 megawatts and a 100 acre site that can deliver 200 megawatts are not two versions of the same asset. They are different assets that happen to share a fence line measurement.

That is why this market stopped pricing dirt by the acre and started pricing it by the megawatt. In my earlier piece on powered land, I laid out the headline figure: powered land in primary U.S. markets is averaging $584,000 per megawatt in 2026, up 51% from last year and 35% above its five-year average. This post is the method behind that number. It is how I actually value data center land when a landowner, a developer, or a lender asks me what a site is worth.

Step 1: Start with deliverable megawatts, not acres

Every valuation starts with one question: how many megawatts can this site actually deliver, confirmed by the utility, and by when? Not the broker's megawatts. Not the megawatts on the offering memorandum. The megawatts in an interconnection study, a utility letter, or a queue position with a date attached.

The reason is scarcity, and the scarcity is measurable. In one major regional footprint there are 60 GW of potential large-load power projects in planning and only 13 GW in advanced stages, with a 6 GW generation shortfall looming if everything planned gets built. Most of the demand chasing land cannot be served on any near-term timeline. So the value does not sit in the soil. It sits in the site's place in line for power.

My working rule: a megawatt that is confirmed and deliverable in 2027 is a fact. A megawatt that is hoped for in 2031 is an option. I value them very differently, and so should you.

Step 2: Convert every comp to $/MW

Once I know the deliverable megawatts, I convert every comparable sale to dollars per megawatt before I compare anything. The math is simple: price divided by deliverable megawatts.

Here is an illustrative example, using round numbers to show the method rather than to quote a market price. Say a site sells for $40 million, it covers 100 acres, and it can deliver 200 MW. That is $400,000 per acre, which sounds extreme, and $200,000 per megawatt, which is well below the $584,000 per MW primary-market average. Now take a second site: $10 million, 50 acres, but only 20 MW of deliverable power. That is $200,000 per acre, which sounds cheaper, and $500,000 per megawatt, which is nearly two and a half times the cost of the first site for the thing the buyer actually needs.

Same market. Same week. The per-acre ranking and the per-megawatt ranking point in opposite directions. If you only run one of those numbers, run the megawatt one.

Step 3: Adjust for what makes megawatts unequal

No two megawatts are identical, and the adjustments are where the appraisal work actually happens. These are the factors I adjust for, in roughly the order they move price:

Power certainty and timing. A signed interconnection agreement beats a completed study, which beats a queue position, which beats a conversation with the utility. Each step down that ladder is a discount, and a delivery date three years later than a comp's date is a discount on top of that.

Entitlement status. Zoning that already allows a data center by right is worth more than zoning that needs a special use permit, which is worth more than a site in a county that is actively debating a moratorium. Entitlement risk is not a footnote. It can be the whole value difference between two otherwise identical sites.

Cost to connect. The substation distance, the transmission upgrades, and who pays for them matter. In appraisal language these are costs to cure, and a site that needs eight figures of off-site work to reach its stated megawatts does not deliver the same megawatts as the site next to the substation.

Water, fiber, and site quality. Cooling water access, diverse fiber routes, topography, flood plain, and shape all feed the same calculation: what will it cost, in money and in time, to turn this parcel into an operating campus?

Market. The $584,000 per MW figure is a primary-market average. A powered site outside Dallas and a powered site in Loudoun County live in different markets, and a comp from one is not a comp for the other without serious adjustment.

Step 4: Cross-check with $/acre and with the total project

The $/MW method is the primary approach, but I never run it alone. Two cross-checks keep it honest.

First, the per-acre check. Farmland and ordinary commercial land in the same county set a floor and a sanity range. If a $/MW conclusion implies a per-acre price that no buyer in that county has ever approached, I want to know exactly which megawatt assumption is doing that work.

Second, the land share check. All-in greenfield development costs in the U.S. and Canada average $17.6 million per megawatt for modern facilities, excluding chips and GPUs, and power infrastructure alone is 21% of that total. Against that figure, powered land at $584,000 per MW is roughly 3% of the all-in cost per megawatt. That ratio explains a lot of buyer behavior: land is a small share of a very large project, which is why buyers stretch on the right site. It also gives you a tripwire. If a land price pushes the land share far above the level the total project economics can carry, the price is a speculation, not an underwriting conclusion. JLL puts the scale of the buildout at up to $3 trillion to support 100 GW of new capacity. At that scale, small errors in the land basis per megawatt multiply fast.

Step 5: Read the deal structure before you trust the price

A surprising share of data center land trades are not clean fee simple sales. They are options, phased takedowns, and long due diligence periods while the buyer confirms power and entitlement. The headline price in an option deal is the price if everything works. The effective price, after option payments, extension fees, and the time value of a two or three year runway, is a different and lower number.

When I use an option deal as a comp, I want the whole structure: the option consideration, the extension terms, what the buyer has spent on studies, and whether closing is conditioned on the interconnection. A comp without its structure is half a comp.

The honest gaps

Two things I will flag the way I flag them in a report. First, there is no public database of verified data center land sales priced per megawatt. The $584,000 per MW average comes from a brokerage cost guide, not from recorded deeds, and in nondisclosure states like Texas most sale prices never become public at all. If you need a verified price, you call the broker who did the deal. There is no shortcut, and anyone who quotes you a precise local $/MW figure without telling you where it came from is guessing with confidence.

Second, the demand side of this market is real but concentrated. Global data center occupancy ended 2025 at 97%, and 25% of sites now run on-site generation, up from 19% a year ago, which tells you buyers are paying to solve the power problem themselves when the grid cannot. Those are sector numbers. They support the method. They do not set the value of your parcel.

The bottom line

Value data center land the way buyers buy it: in dollars per deliverable megawatt, adjusted for certainty, timing, entitlement, and connection cost, then cross-checked against dollars per acre and against the land's share of the total project cost. Acres tell you how big the site is. Megawatts tell you what it is worth.

If someone brings you a data center land assignment, start where I start: how many megawatts, confirmed by whom, and when? Get that answer in writing, and the valuation has a foundation. Skip it, and you are appraising a field. And when the assignment grows from the dirt to the building on it, I run the same discipline through my appraiser's checklist for underwriting a data center.

Sources: Cushman and Wakefield 2026 Data Center Development Cost Guide (via Business Wire, September 2026); JLL 2026 Global Data Center Outlook; CRS R48762.3, Data Center Energy Infrastructure: Federal Permit Requirements (2025); AFCOM State of the Data Center 2026.