AI Data Centers: How to Spot Projects That Can Get Built11 min read2,198 words

AI Data Centers: How to Spot Projects That Can Get Built

Learn why many AI data center proposals stall, how to assess project readiness, and what developers must prove to lenders, utilities, and local communities.

AI data centersdata center developmentAI infrastructuredata center power demanddata center permitting
AI Data Centers: How to Spot Projects That Can Get Built

Why Many AI Data Center Projects Will Never Get Built

AI data center announcements can make the future of computing feel like a construction checklist: find a site, line up investment, build enough capacity to meet rising demand. But an announcement is usually only the first step. To actually get built, a project still has to clear harder hurdles - finding a credible tenant, securing financing, getting a realistic path to electricity, passing permitting, and earning enough local support to keep moving.

That gap matters because "proposed capacity" is not the same thing as capacity that reaches the grid. Developers, investors, utilities, and communities all make decisions based on projects that are often early-stage, speculative, or still resting on assumptions that haven't been proven. Brent Nelson of Ascend Analytics made this point in a TechCrunch Equity podcast: the data center pipeline can look much bigger than the set of projects that are truly likely to get built.

This article lays out a practical way to evaluate that gap: how to sort projects by maturity, what lenders and utilities want to see, why power and permitting stall, and how developers can reduce risk for both the infrastructure and the people nearby.

Why is the announced data center pipeline so inflated?

A pipeline is a list of potential projects, not a forecast of completed buildings. That difference gets muddied when companies announce huge capacity targets or describe a site as "planned" before the basic dependencies are settled. Something can be real - someone is actively pursuing it - while still having a low chance of reaching construction.

So the useful question isn't just how many megawatts a developer claims. It's what's actually secured, what's still conditional, and who pays the price if a key dependency fails. A proposed site with no committed customer and no believable power path is not the same thing as a project with a signed tenant, financed capital, and a defined route through utility and permitting.

Separate interest from commitment

A quick screening method is to sort projects into three broad stages:

  • Speculative: Based on an announced plan, a site option, or general expectations for AI demand. Major commitments are still open.
  • Developing: The developer is working through specific needs - utility discussions, planning approvals, or tenant negotiations. Outcomes aren't settled yet.
  • Construction-ready: Credible customers and financing are in place, there's a workable power plan, and the approvals and contracts required to start building are lined up.

These aren't a substitute for diligence. They're a way to stop treating very different projects as if they have the same odds. Even "construction-ready" projects can run into execution risk, but they've cleared more meaningful tests than a concept announced to attract attention or early investment.

Ask what would make the project stop

A realistic assessment focuses on failure conditions, not just upside. What happens if the tenant doesn't sign? If the utility can't deliver power on the schedule assumed by the business plan? If permitting drags longer than financing allows? Or if equipment costs and demand assumptions change?

Nelson's warning about an inflated pipeline matters for that reason: counts alone can make it look like infrastructure is coming when it may not be. For investors and communities, the practical move is to treat announced capacity as a set of possibilities - then figure out which projects have turned those possibilities into commitments.

What makes an AI data center project financeable?

A data center becomes financeable when the people providing money can see how it will earn revenue, secure power, and handle both construction and operating risks. The building itself is only part of the story. For AI infrastructure, expensive equipment only pays off if customers will actually use it - and if the facility can run at the scale required.

Lenders don't need every uncertainty erased. They do need clarity on what's been handled, what's still unclear, and who owns each risk. When a plan depends on a future customer, a future grid connection, and future permits, those dependencies are linked. If one slips, the others often get harder - or more expensive - to make work.

Start with the tenant, not the headline demand

A broad forecast for AI growth doesn't automatically translate into revenue for a specific site. Developers need to know who will occupy the capacity, what contractual commitments exist, and how the facility design matches that customer's needs. A tenant's credit quality and willingness to commit matter because a facility built for one customer or use case may not be easy to repurpose without breaking the economics.

That matters even more if the AI investment cycle cools. A developer with durable, long-term customer commitments is in a different position than one relying on short-term expectations or a small set of tenants that may never fully materialize. The real risk isn't only that demand falls. It's that the facility, the power arrangements, and the financing were sized around a demand that never actually arrives.

Match the financing plan to project maturity

A practical diligence review should ask:

  • Customer: Is there a signed agreement, a credible negotiation, or only an expectation of future demand?
  • Power: Is there a defined utility path and schedule, or just a nearby transmission line shown on a map?
  • Approvals: Which permits are already granted, which are pending, and which could change the design or timing in a meaningful way?
  • Capital: Does the financing plan account for delays and cost changes, or does it assume everything runs on the original timeline?
  • Fallbacks: If a tenant, power source, or approval slips, can the project be resized, delayed, or repurposed?

To be clear: you don't need every early-stage project to have every contract fully nailed down. Early development is inherently uncertain. The issue is that uncertainty has to be described and priced honestly. A project shouldn't be marketed as near-term capacity if the key agreements are still hypothetical.

Why do power and permits delay data centers?

A data center can be technically ready and still have no realistic way to connect at the scale it needs. AI facilities require a lot of electricity, and the generation and grid upgrades needed to serve them may involve separate investments, approvals, and build schedules. A developer's request for power doesn't automatically create the power delivery capacity.

Permitting adds its own layer of uncertainty. Energy projects often involve multiple agencies and steps, and delays in one area can ripple into others. A generation project may need approvals before it can supply a new load; a grid connection may depend on upgrades with their own planning and permitting requirements. When the timelines don't line up, projects can get stuck in a loop - each party waits for progress from the others before moving further.

Treat the power path as a critical dependency

A credible power plan should be more than "we'll use X source." It should lay out the route from generation and transmission to the facility, who is responsible for each step, the expected sequence of approvals, and the schedule assumptions behind it. If any portion depends on another project, that dependency needs to be part of the risk assessment.

This is also where an "announced megawatt" figure can mislead. It may reflect the facility size a developer wants to build, not the power the grid can deliver when the building is ready. That gap affects construction timing, financing costs, and whether customers can actually be served.

Permitting reform needs to address the whole process

Permitting delays aren't only caused by one slow office. The bigger issue is often the structure of the system: overlapping rules, sequential approvals, and uncertainty about what will be required. As Nelson noted in the Equity discussion, administrations from both major US political parties have contributed to a complicated environment for energy projects. That's why reform is about process, not partisan messaging.

The goal should be to make reviews more predictable and coordinated, without removing meaningful scrutiny. Faster decisions help if regulators also provide clear requirements and timelines. Simply removing checks without fixing the underlying causes of delay can move risk around rather than solve it - especially if projects later run into legal challenges or local opposition. The end result should be a process that's rigorous, understandable, and capable of reaching a decision.

How can developers earn local support and secure reliable power?

Local opposition isn't something you can safely handle after a project is already designed. Residents may have legitimate concerns about land use, noise, water, traffic, construction disruption, and what added electricity demand could mean locally. If developers treat those concerns like a public relations problem instead of project requirements, they can lose time and trust before construction even starts.

Engagement doesn't mean every objection will disappear, or that every project should automatically move forward. It means identifying impacts early enough to change the plan when changes actually matter. If a developer explains power needs, the construction schedule, and proposed mitigations in concrete terms, neighbors can evaluate the proposal. Vague promises about economic benefits usually don't answer the real questions people are asking about local costs.

Make the project legible to the people affected

Before asking for support, developers should be able to explain:

  • What's being built, at what scale, and on what timeline.
  • How the project plans to obtain electricity and what that could mean for the local system.
  • What impacts are likely during construction and operation, and what mitigation is planned.
  • What remains uncertain and when residents can expect updates or decisions.
  • How community feedback could lead to changes in design, operations, or construction plans.

The last point is often the difference between consultation and a presentation. If feedback can't change anything, residents are being informed - not meaningfully engaged. Early, specific conversations can surface constraints before they turn into expensive redesigns or formal disputes.

Do not assume natural gas is a simple reliability fix

Some developers treat natural gas like an obvious fallback when grid power is delayed or uncertain. But a fuel source isn't "reliable" just because it can generate electricity on demand under normal conditions. Nelson pointed out that natural gas can be vulnerable during extreme winter weather, when system stress affects availability. That makes it risky to treat gas as a guaranteed answer to every power problem.

One alternative worth evaluating is battery storage combined with a grid connection that can be temporarily curtailed during periods of high demand. The trade-off is straightforward: curtailment may help a project connect sooner or reduce stress on the grid, but the operator has to accept limits on when it can draw power. Batteries add flexibility, but they aren't a substitute for sustained supply. The right design depends on operating needs, local grid conditions, and the cost of downtime.

Key takeaways

  • Count commitments, not announcements, when estimating likely capacity.
  • Treat tenant demand, financing, permits, and power access as connected risks.
  • Test the project's fallback plan, including what happens during grid stress.
  • Engage local communities early enough for feedback to change the project.
  • Evaluate reliability options based on their limits as well as their benefits.

Conclusion

The data center pipeline isn't a construction forecast. To judge which AI infrastructure projects are likely to reach operation, look for evidence that demand, financing, power, approvals, and local realities are aligning - not just a large capacity announcement. Each dependency should have an owner, a credible timeline, and a fallback plan.

For developers, the standard should be making risk visible early. For investors and communities, it's distinguishing commitments from assumptions. That's a more useful way to assess the AI buildout than counting proposed projects and assuming they'll all arrive.

Frequently Asked Questions

The project may lack a committed tenant, financing, permits, or a viable path to enough electricity. Any one of those gaps can delay development; several at once can make the project uneconomic. An announcement signals intent, not proof that the dependencies needed for construction are secured.
Look for unresolved dependencies. If the customer is hypothetical, the power plan lacks a defined utility path, or key approvals and financing are still uncertain, the project is early-stage. Compare what has been committed with what is still described as planned, expected, or subject to approval.
Large facilities need substantial, dependable power, but generation, transmission, and grid connections take time to plan and approve. A site can be attractive and a customer ready while the power infrastructure is not. The delivery schedule and the project’s demand therefore need to be assessed together.
Not automatically. Natural gas can be vulnerable during extreme winter weather, so developers should test its availability under the conditions when the wider energy system is stressed. Battery storage and curtailment arrangements may offer other forms of flexibility, but each comes with limits and operating trade-offs.

Related articles