TL;DR
Thesis: The unit of account in data centers has shifted from square feet to secured megawatts, and the market is paying enormous premiums for contracted power because it is the one input that cannot be replicated quickly at any price.
The number that matters: Five-plus years, the time it now takes to replicate a grid connection in a constrained market (JLL; LBNL puts the median interconnection wait past four years).
Biggest risk: Phantom megawatts. ERCOT's large-load queue is 2.6 times the highest demand Texas has ever served, and utilities openly say most requests will never materialize. If speculative load washes out, regional power prices fall and queue-position "assets" mark down hard.
Bottom line: Underwrite the firmness of the power contract, not the headline megawatts. An executed, funded interconnection with a dated milestone holds a durable premium. A will-serve letter is paper.
In November 2025, Amazon closed on 189 acres in Prince William County, Virginia, for $700 million, roughly $3.7 million an acre, in what the Washington Business Journal reported as the priciest sale of undeveloped land in the state's history. The seller, homebuilder Stanley Martin, had assembled the full 270-acre site for about $51 million starting in 2021, then spent two years getting it rezoned for 3.5 million square feet of data centers and up to three substations.
Nobody pays a thirteenfold markup for dirt, and nobody pays it for buildings either. The price reflects one thing: entitlements plus a credible, dated path to electricity in the tightest power market in the country.
If you allocate to digital infrastructure, directly or through the funds that now dominate the space, this is the week-one fact to internalize. The unit of account in data centers has shifted from square feet to secured megawatts. A building can be replicated in 18 months. A signed grid interconnection in a constrained market can take five years or more to replicate, and that gap is now the single biggest driver of what these assets sell for.
The Comps Have Already Repriced
Let’s start with the largest transaction the sector has ever seen. In October 2025, a consortium of BlackRock's Global Infrastructure Partners, MGX, and the AI Infrastructure Partnership agreed to buy Aligned Data Centers from Macquarie Asset Management at an enterprise value of roughly $40 billion, the biggest digital infrastructure deal on record, according to Data Center Dynamics. Aligned brings more than 5 GW of capacity across 50-plus campuses, operational and planned. Run the numbers, and the consortium paid about $8 million per megawatt of platform capacity, a meaningful slice of which exists only as land, entitlements, permits, and power agreements.
The leasing market is no different. CBRE's H2 2025 North America report puts colocation vacancy in Northern Virginia, the largest market on earth, at 0.5%. The region ended the year with 21.5 megawatts of available supply against 4,040 megawatts of inventory, with most 2026 deliveries already committed and preleasing stretching into 2027. The national average asking rate for a 250 to 500 kW requirement rose 6.5% year over year to $195.94 per kW per month, the fourth straight annual increase, and CBRE tracked pricing on 10 MW-plus blocks rising by as much as 19% during 2025.
And underneath both sits the land itself. JLL figures cited by Bisnow show parcels that went for $10,000 to $30,000 per acre before the boom are now trading between $200,000 and $1 million, with established hubs like Northern Virginia ranging from $2 million to more than $6 million. The Amazon purchase sits comfortably inside that band.

Why a Megawatt Got Scarce
The four largest US hyperscalers (Alphabet, Amazon, Microsoft, Meta) have guided to roughly $630 billion of combined 2026 capital spending, up from $388 billion in 2025, with Amazon alone planning about $200 billion. In Virginia, Dominion Energy reported approximately 51 GW of contracted data center capacity as of March 2026, up 2.5 GW in a single quarter, against a request pipeline of about 70 GW. Notably, 10.4 GW of Dominion's contracted figure sits under electric service agreements that bill the customer whether or not they ever take the power. Even the utilities have learned to price firmness.
All of that demand now runs into the supply side, where every layer of the electrical stack is backed up at once.
The grid queue comes first. Just over 2,060 GW of generation and storage capacity sat in active US interconnection queues at the end of 2025, per LBNL's Queued Up dataset updated in 2026. That's more than 1.5x the country's installed generating base, waiting in line. And the line itself has slowed; LBNL finds the median time from interconnection request to commercial operation has roughly doubled since the 2000s and now runs past four years. Natural gas requests, the fastest-growing category, jumped 72% in a year to 136 GW while solar, wind, and storage requests all shrank.
Then there’s the equipment. Wood Mackenzie's Q2 2025 supply chain survey puts US power transformer lead times at an average of 128 weeks, with generator step-up units at 144 weeks and some large high-voltage orders stretching to four years. Power transformer prices are up 77% since 2019. Even if a utility wants to serve a new campus tomorrow, it can't get the equipment.
Generation hardware is no better. GE Vernova's gas turbine backlog and slot reservations grew from 83 GW to 100 GW in the first quarter of 2026 alone, with 21 GW of new agreements signed. The company now guides to at least 110 GW by year-end, per its Q1 filing. CEO Scott Strazik told investors that only about 10 GW of production capacity remains through 2030, with customers already fighting over 2030 slots. Wood Mackenzie projects the turbine crunch will push prices to roughly 195% above 2019 levels by the end of 2027.
Stack those layers, and the output is simple. JLL data cited by Bisnow shows the average grid connection lead time for a 50 MW facility in Atlanta has stretched to five years. A developer who locked in an interconnection in 2022 holds something a competitor starting today cannot buy at any price, only wait for. Scarcity with a five-year replication lag is what premiums are made of.
So the sharpest operators are bypassing the grid expansion process entirely.
Here's the part that should interest allocators who hold no data center exposure at all. The megawatt premium is cleared through wholesale power markets, which means everyone within the footprint pays a share of it.
PJM, the grid operator covering 67 million people across 13 states, cleared its 2026/27 capacity auction at $329.17 per megawatt-day, the FERC-approved price cap and a 22% jump from the prior year's record. Two years before that, the same product cleared at $28.92. The December 2025 auction for 2027/28 hit the updated cap of $333.44, the third consecutive record, and for the first time in PJM's history, the auction came up short of its reliability target by 6,625 MW. Monitoring Analytics, PJM's independent market monitor, attributes 40% of the latest auction's costs to data centers and calculates that data center forecasts drove 45% of the $47.2 billion in capacity costs across the last three auctions combined.

Regulators have noticed. Virginia approved Dominion's new GS-5 rate class in November 2025, which forces large data center customers to fund the infrastructure they require. Under SB 6, Texas now requires large-load applicants to disclose duplicate requests and pay for upgrades. The political economy of who pays for the buildout is becoming its own risk factor, and it lands hardest on operators whose power strategy was "show up and ask."
Dirt With Electrons Is Becoming an Asset Class
The market's response to all this is the rise of powered land: sites bought, entitled, studied, and equipped with a contracted path to energization before any concrete is poured. Hines Research estimates that roughly 20,000 acres of powered land sit under operational data centers globally and another 40,000 acres will be needed within five years, at a density of three to four megawatts per acre. JLL pegs the broader prize at nearly 100 GW of new data centers between 2026 and 2030, about $1.2 trillion in real estate value creation.
The work is slow and unglamorous. Taking a parcel from acquisition to shovel-ready runs three to seven years per JLL, covering load studies, transmission upgrades, substation siting, environmental review, and the local politics of entitlement. That's precisely why it commands a premium. JLL has roughly 100 staffers on powered land. CBRE has assigned part of its 6,000-person data center group to it. Hines' CIO David Steinbach frames the shift plainly: thirty years ago, the constraint was the building; today, it's megawatts and entitlements.
One caution from the law firms doing these deals. Ropes & Gray's 2026 outlook stresses that a utility "will-serve" letter is not powered land. What matters is a contract for transmission capacity by a date certain, or for on-site generation, permits and fuel supply actually in hand. The premium attaches to firmness, not paper.
The Bear Case is Phantom Megawatts
Now the honest part. A meaningful fraction of the demand signal driving this premium does not exist.
Developers shopping for sites routinely file interconnection requests with several utilities for the same project, then build in one place, leaving the other requests sitting in queues as live megawatts. ERCOT's large-load queue nearly quadrupled in a year, from 63 GW at the end of 2024 to 226 GW by November 2025, with data centers making up roughly 73% of it. Texas' all-time record peak demand is 85,508 MW. The queue is now 2.6 times the highest load the state has ever served, and ERCOT's own planning chief told the board the grid operator has "outgrown the process" built to review these requests.

Utilities are now admitting this on the record. Exelon's Q1 2026 investor materials show roughly 19 GW of committed data center projects in its territories, and only about 45% of even that committed pipeline is secured with transmission security agreements. The company has told investors that most of its long-run request pipeline is unlikely to materialize. Industry estimates cited by Latitude Media put speculative requests at five to ten times the number of real projects. PJM cut its 2028 peak forecast by 4.4 GW in January 2026 specifically because of stricter vetting of planned data centers. And the developers themselves are hedging. In a Foley & Lardner survey released in January 2026, 63% of industry respondents said they expect a "strategic correction" in data centers by 2030. Heatmap counted 25 US data center projects canceled by local opposition in 2025, four times the prior year.
So read the megawatt premium with both eyes open. Where interconnection is contracted and funded against a dated milestone, the scarcity is real, and the premium is durable. Where it's a queue position or a will-serve letter, the asset may be worth a fraction of its mark. The same desperate competition for power that created the premium also created the incentive to manufacture paper megawatts, and the gap between the two is where this cycle's losses will sit.
What an Allocator Should Ask
Sophisticated capital is already positioned along this divide. The Aligned buyers, BlackRock's GIP, MGX, and an AIP roster that includes Microsoft, NVIDIA, and the Kuwait Investment Authority, paid platform multiples for secured power positions. Macquarie, on the other hand, took the full cycle and exited. Brokerages have built powered-land desks. Utilities are rewriting tariffs to shift buildout risk onto developers. Each of these actors has decided the megawatt, not the building, is the scarce input, and is pricing accordingly.
For anyone evaluating data center exposure, in a fund, a co-invest, a REIT, or a credit sleeve, the diligence questions write themselves.
Does the project hold an executed interconnection agreement or a letter of intent?
What is the dated energization milestone, and what does the underwriting look like if it slips by 18 months?
Who funds the substation and transmission upgrades, the developer or the ratepayer, and does the local regulator agree?
How much of the surrounding utility's queue is duplicated demand that will vanish, and does the valuation survive if regional power prices fall when it does?
The buildings will keep going up. The premium belongs to whoever already has the electrons.
Bedrock Capital Research publishes independent research for informational and educational purposes only. Nothing in this memo is investment, legal, or tax advice. Authors may hold positions in assets discussed.

