Digital Infrastructure
Site Selection Used to Be a Real Estate Decision. It Is Now a Board-Level One.
In 2026, site readiness is a spectrum defined by power certainty, entitlement risk, delivery sequencing, and political exposure — all at once. This is now a board-level risk, not a site-selection footnote.
By Vaxa Strategy Team
3 min Read
Published on June 24, 2026
Sector Snapshot — Digital Infrastructure
01
~100 GW of new data center capacity expected globally between 2026 and 2030 — nearly doubling today's installed base
02
$3T in combined real estate, fit-out, and IT investment projected through 2030
03
140+ local community groups now mobilized against hyperscale data center projects across the U.S.
04
$30B project canceled by QTS in 2026 — a signal of how execution risk is repricing the sector
Market Consensus & Vaxa's Position
Site selection is still fundamentally a real estate decision — find the land, confirm power as a checklist item, negotiate incentives, build.
Most companies are underwriting growth plans against a "commercial operation date" that historically holds only 13% of the time. That's not a known risk being priced in — it's a mispricing hiding in plain sight, and it reaches far beyond real estate into any plan that depends on power-intensive capacity, anywhere in the chain.
The Shift
Site selection used to be a real estate decision. It is now a board-level one.
For most of the past two decades, data center site selection followed a familiar playbook: find land near fiber routes and population centers, negotiate tax incentives, confirm power availability as a checklist item, and build. Power access was assumed. The harder questions were about connectivity, climate, and cost.
That playbook no longer applies. In 2026, site readiness is not a binary condition — it is a spectrum defined by power certainty, entitlement risk, delivery sequencing, and political exposure, all at once. Developers are restructuring real estate strategy around energy strategy from the outset, not after a site is chosen. Power access is no longer a feasibility check. It is the foundation of the investment case.
The scale of capital now committed to this buildout means these are no longer operating decisions delegated to real estate and infrastructure teams. A five-year delay on a hyperscale project erodes competitive positioning, compresses IRR, and can cost an operator an enterprise contract to a faster-moving competitor. That is a board-level risk, not a site-selection footnote.
Two manufacturers each finalize an identical expansion plan the same month — same blueprint, same capital budget, same 90-megawatt power requirement. Both file for grid interconnection immediately.
The first is in a region with available substation capacity. Connection clears in a matter of months. Construction proceeds on schedule. Hiring for several hundred new positions opens on time, and the plant is running when the customer contract that justified the expansion needs it to be.
The second files an identical request in a congested queue region. The interconnection study alone takes years. The capital sits committed but idle. The hiring plan stalls, then quietly disappears from the next budget cycle. The customer contract that assumed the new capacity moves to a competitor who could actually deliver on time.
Same blueprint. Same budget. Same month. Two completely different outcomes — and the only variable that mattered was one neither company had ever treated as a planning input before now.
Power as the Bottleneck
Interconnection, not land, is now the pacing item.
Nearly two terawatts of generation capacity are currently stuck in U.S. interconnection queues — almost twice the country's installed base. PJM, the largest grid operator in the country, reopened its queue in 2026 after a multi-year pause, having accumulated a backlog exceeding 300 gigawatts of projects. ERCOT's Large Load interconnection process now has more than 225 gigawatts of large loads moving through it, prompting the grid operator to bring in outside help to redesign the process.
In response, developers are restructuring deployment plans around certainty rather than cost. Behind-the-meter generation, battery storage, and hybrid power strategies are increasingly built into project planning from day one, despite the higher upfront capital they require. A decade ago, developers optimized for the lowest delivered cost. In 2026, they are optimizing for the earliest energization date — because a five-year wait for grid access is now a bigger threat to a project than a modest cost premium.
This has cascaded into the labor market as well. Gas and combined-cycle generation expertise is resurging, battery storage and microgrid capability are in sustained demand, and utility relationship management — once a peripheral function — has become a core commercial capability for any operator building at scale.
The trend is accelerating, not stabilizing. Data center cancellations more than quadrupled from 6 in 2024 to 25 in 2025, and one industry tracker found that a quarter of the projects planned for 2026 haven't even disclosed a powering strategy — a direct sign that site selection is still happening before power access is confirmed, on exactly the sequence this piece argues has inverted. In early 2026, Oracle and OpenAI terminated a planned expansion of their flagship Abilene, Texas campus from 1.2 to 2.0 gigawatts, citing financing challenges and shifting energy infrastructure forecasts — a live, current example, not a historical one.
Regulatory and Political Risk
Water, foreign ownership, and local opposition are now underwriting variables.
Power is no longer the only gatekeeper. Water availability has become a second, equally binding constraint — a data center project can clear every power and permitting hurdle and still be halted over water. Several jurisdictions have introduced consumptive-use permitting frameworks specifically for large-load data centers, and at least one major project has already been stopped over water permitting despite having secured local approvals.
State legislatures are also moving faster than the industry's planning cycles can absorb. New restrictions now require full cost-of-service accounting for large-load customers, and some states have gone further, barring utilities from serving foreign-controlled large-load customers altogether. New York's own one-year moratorium on new hyperscale permitting, driven by ratepayer electricity costs rather than power availability, is part of the same pattern — regulatory risk is compounding independently of grid capacity.
Local opposition has become organized enough to change outcomes, not just optics. More than 140 community groups have mobilized against hyperscale projects nationally. Ballot measures, council turnover, and federal environmental litigation are now realistic outcomes of a contested site — not tail risks. For developers underwriting a project, community and political risk now sit alongside power and water as line items, not afterthoughts.
What This Means for Strategy
Speed and certainty now outrank marginal cost.
The practical consequence for site selection teams is that the old sequence — find the site, then solve for power — has inverted. Power and water certainty now come first, because they determine whether a site is viable at all before real estate economics are even relevant. Regulatory and community risk assessment has to happen in parallel, not after entitlement, because a state legislative session or a local ballot measure can invalidate months of site work overnight.
For operators, this means site selection can no longer be owned solely by real estate and infrastructure teams — it requires the same rigor applied to political, regulatory, and utility relationship risk that used to be reserved for market entry decisions in regulated industries. For investors and boards, it means underwriting models built on delivered cost and land economics are missing the variable that is actually determining outcomes in 2026: schedule certainty.
The operators who treat this as a portfolio-level capability — tracking interconnection queues, water permitting frameworks, and state legislative activity the way they'd track any other market intelligence — will out-execute the ones still treating site selection as a real estate function with a power checklist attached.
Two manufacturers each finalize an identical expansion plan the same month — same blueprint, same capital budget, same 90-megawatt power requirement. Both file for grid interconnection immediately.
The first is in a region with available substation capacity. Connection clears in a matter of months. Construction proceeds on schedule. Hiring for several hundred new positions opens on time, and the plant is running when the customer contract that justified the expansion needs it to be.
The second files an identical request in a congested queue region. The interconnection study alone takes years. The capital sits committed but idle. The hiring plan stalls, then quietly disappears from the next budget cycle. The customer contract that assumed the new capacity moves to a competitor who could actually deliver on time.
Same blueprint. Same budget. Same month. Two completely different outcomes — and the only variable that mattered was one neither company had ever treated as a planning input before now.
DIAGNOSTICS CHECKLIST
What is the actual interconnection queue position — not the stated target date — for every facility, supplier, or landlord this plan depends on?
Has the powering strategy been independently verified, or is it being taken on the vendor's word?
Does the underlying contract address what happens if a partner's power access is delayed — and has anyone actually checked, or is it assumed to be covered?
Is delay-related insurance coverage actually available at a meaningful limit for this specific project, or has that been assumed rather than confirmed?
What is the realistic delay range for this specific region's queue — not the national average, which can differ from a specific utility's backlog by years?
Who inside the organization actually owns this risk — real estate, finance, procurement — or has it quietly fallen between all three?
Bringint it together
01
Only 13% of interconnection requests filed 2000-2019 had reached commercial operation by 2024 — the industry is underwriting billions in growth plans against a milestone that fails to hold roughly seven times out of eight.
02
This risk isn't confined to real estate or hyperscale data centers. Any company whose plans depend on new power-intensive capacity — their own, or a vendor's, landlord's, or supplier's — has inherited this exposure through the chain, whether or not they ever signed a permit application.
03
It can't simply be insured away. Delay-related coverage exists but is becoming harder and more expensive to obtain at meaningful limits — this is a risk that has to be actively tracked, not one that can be transferred and forgotten.
04
The exposure passes through contracts most companies never audit for it — liquidated damages clauses, financing covenants, and customer SLAs can all be triggered by someone else's power delay, several steps removed from the actual decision.
05