Why Only One-Third of Data Center Projects Are Active
Only a third of data center projects are active. Discover the hidden challenges and what this means for the industry.
The pipeline looks impressive on paper. Hundreds of planned data centers, billions in projected investment, gigawatts of future capacity — the numbers suggest an industry sprinting toward an electrified, AI-powered future. But strip away the announcements and the press releases, and a more complicated picture emerges: only about one-third of data center projects in development are actually under active development. The rest are stalled, shelved, or quietly headed nowhere.
That's not a rounding error. That's a structural reality of how the industry operates — and understanding it matters whether you're a developer, an investor, a grid planner, or a municipality that just got promised hundreds of jobs.
The Pipeline Illusion
"Pipeline" is one of those words that sounds more concrete than it is. In practice, a data center project can enter the pipeline at the moment someone files a zoning inquiry or signs a letter of intent on a piece of land. That's a long way from a shovel in the ground.
The gap between announced capacity and built capacity is one of the most consistently underappreciated dynamics in infrastructure development. For data centers specifically, the chasm is wide. Projects get announced when land is optioned, when permits are applied for, or when a developer wants to signal market presence to investors or customers. That's not deception, exactly — it's how capital-intensive industries build momentum. But it creates a distorted view of actual supply coming online.
When industry observers note that only a third of pipeline projects are under active development, they're describing a filtration process that's ruthless and largely invisible from the outside. The projects that survive it aren't just luckier — they're structurally different from the ones that don't.
Why Projects Stall Before They Start
Power Is the New Permitting
A decade ago, the primary bottleneck for data center development was land and permits. Today, it's electricity — specifically, the ability to secure a utility interconnection agreement in a timeline that makes economic sense.
Interconnection queues at major utilities have ballooned. In some markets, developers are waiting three to five years just to get a formal study completed. A hyperscaler can absorb that wait. A mid-tier developer with a speculative project and a debt-service clock ticking cannot. Many of the inactive projects in the pipeline aren't dead because the developer gave up — they're suspended because the power simply isn't available on the timeline the business model requires.
This dynamic is reshaping where data centers get built more than any zoning map or incentive program ever could.
Capital Doesn't Flow to Uncertainty
Funding is the second major filter. Data center development is expensive — utility-scale campus projects routinely run $1 billion or more before a single server is racked. That kind of capital requires committed tenants, creditworthy offtake agreements, or a hyperscaler pre-lease. Without one of those anchors, institutional financing is nearly impossible to close.
Speculative development — building capacity before a customer is signed — happens, but it's the exception and it's reserved for operators with deep balance sheets and established track records. For everyone else, the capital stack requires proof of demand before it assembles. Projects that can't demonstrate that demand don't die dramatically. They just quietly stop moving forward.
Regulatory and Community Friction
Zoning approvals, environmental reviews, water rights (cooling systems are thirsty), and increasingly, community opposition all add friction that kills projects on the margin. Data centers have a mixed reputation with local governments — they promise jobs but often deliver fewer than expected relative to their footprint, power consumption, and infrastructure demands. Some municipalities have enacted moratoriums. Others have introduced new utility cost-sharing requirements.
None of these individually is fatal to a well-capitalized project with real customer demand. But for a project already borderline on economics, each hurdle is another reason to redirect capital somewhere easier.
What This Means for Infrastructure Planning
The knock-on effects of a bloated, largely illusory pipeline are underappreciated at the infrastructure level.
Utilities build transmission and substation capacity based on interconnection requests — many of which will never materialize into actual load. Grid planners are making decade-long infrastructure decisions based on demand projections that assume most of the pipeline converts, when history suggests it won't. That's a coordination problem with real costs: either utilities over-build and ratepayers absorb stranded assets, or they under-build and the projects that do materialize can't get power.
The communities that site data centers based on pipeline projections face a parallel problem. Economic development plans get built around job creation numbers and tax revenue estimates that assume construction proceeds. When two-thirds of announced projects stall, the roads, water lines, and zoning accommodations made in anticipation of them don't unwind cleanly.
There's also a less obvious competitive effect. When the pipeline overstates supply, it can actually slow investment into markets where real demand exists — because investors and tenants think capacity is coming that isn't.
The Projects That Do Get Built
Look at the data center projects that successfully navigate from announcement to operation, and patterns emerge. They almost always have one of three things: a hyperscaler or large enterprise tenant with a signed lease, a developer with an established platform and repeat access to capital, or a market with a genuine supply-demand imbalance severe enough to justify speculative development.
Location still matters, but not for the reasons it used to. Cheap land is largely irrelevant when power costs dominate the economics. Fiber connectivity is table stakes, not a differentiator. What actually determines site viability now is proximity to transmission infrastructure, water availability, and — increasingly — access to renewable energy at scale. Markets like the Midwest, parts of the Southeast, and select Western regions are attracting serious capital partly because they check those boxes in ways the saturated Northern Virginia and Silicon Valley markets can no longer reliably offer.
Emerging technologies are beginning to reshape the calculus too. Liquid cooling systems reduce water consumption and enable higher rack densities, making previously marginal sites viable. AI workloads, with their extreme compute intensity, are pushing developers toward purpose-built facilities rather than retrofitted general-purpose space — which means the technical requirements of a "successful" project are themselves shifting underneath developers trying to plan years in advance.
What Developers and Investors Should Take From This
The one-third figure isn't a failure rate to be embarrassed about — it's a description of how high-stakes infrastructure development actually works. Capital flows to the highest-conviction opportunities and filters out the rest. The question worth asking isn't why so many projects stall, but whether the projects moving forward are actually the right ones.
A few things that separate serious projects from pipeline filler:
Power commitments before site control. The developers making real progress are often working utility relationships before they've finalized land deals — because they've learned that land without power is worthless.
Offtake before financing. The capital markets have effectively imposed a pre-leasing discipline that the industry needed. Projects with signed customers get built. Projects without them wait.
Genuine site differentiation. Not just "shovel-ready" — actually advantaged in terms of power cost, renewable access, fiber diversity, or proximity to demand centers.
The forward-looking reality is that the gap between pipeline and active development is unlikely to narrow dramatically. If anything, AI-driven demand will intensify competition for power, capital, and viable sites — which means the filtration process gets more aggressive, not less. The projects that move forward will be better positioned than the average of what's announced. That's actually how it should work.
What's less acceptable is when planners, policymakers, and communities make long-term infrastructure decisions based on pipeline projections rather than active project fundamentals. The one-third that's real deserves serious, coordinated support. The rest is noise — and treating it otherwise is how you end up with stranded infrastructure investments and broken economic development promises.
Explore the InfraSale Marketplace for more insights and opportunities.