AI infrastructure planning used to focus largely on cloud availability, compute capacity, data architecture, security, and cost. While those factors still matter, reliable power and realistic interconnection timelines are becoming increasingly important to where organizations can deploy AI workloads—and how confidently they can plan around that capacity.

Texas initially directed energy regulators to review pending data-center interconnection requests before projects could advance through ERCOT’s process. On September 21, Governor Greg Abbott expanded the review by directing the Texas Commission on Environmental Quality to halt data-center permits until state audits of power and water use are complete.

For CIOs and CTOs, this is less about following utility policy and more about understanding future capacity. A provider may announce a new site or quote a delivery date well before the facility has the approvals, grid connection, and infrastructure needed to serve customers.

What happened in Texas

On August 3, Governor Greg Abbott directed the Public Utility Commission of Texas and ERCOT to verify and audit data-center projects moving through the grid-interconnection process. On September 21, Abbott expanded the action by directing the Texas Commission on Environmental Quality to halt data-center permits until audits are finished.

ERCOT was considering more than 474 GW of new large-load connection requests when Abbott issued the directive. That’s roughly five times Texas’s record peak electricity demand, and the governor’s office said data centers accounted for about 90% of those requests.

The 474 GW queue includes projects at very different stages of development. Some developers have secured a site, financing, and customer commitments, while others are still exploring locations or seeking capacity from more than one utility. An interconnection request doesn’t guarantee that a facility will be built or that power will be available on a specific timeline, however.

The audit requires developers to provide more detail about their projects before they can progress. Regulators are reviewing project ownership, financing, taxpayer incentives, projected power use, onsite generation, water use and reuse, and potential community impacts. 

The review gives regulators a clearer picture of which projects are more likely to move toward construction. For companies buying future capacity, it’s a reminder to look past an announcement and ask what has been secured, what’s still pending, and what could affect the delivery date.

Planned capacity isn’t the same as usable capacity

One of the most important distinctions for technology leaders is where a provider's capacity sits in the development process. A planned data center may have a site and an announced capacity target, but that doesn’t necessarily mean the facility has secured its power, completed interconnection work, obtained the necessary permits, or started construction.

For planning purposes, teams can think about capacity as a progression: announced, site-secured, permitted, financially committed, grid-connected, under construction, and energized. Each stage reduces a different category of delivery risk. The closer a project is to being energized, the more evidence a buyer has that the capacity can support a specific workload on a defined timeline.

This is important to consider when comparing providers. Two vendors may both advertise hundreds of megawatts of future capacity while having very different levels of project maturity behind those numbers. CIOs should therefore evaluate the evidence supporting a capacity commitment rather than treating the headline capacity figure as immediately available infrastructure.

Why data center demand is under review

Texas is part of a larger trend. Reuters reported on September 1st that large power users had requested more than 700 GW of electricity across parts of the Midwest, Mid-Atlantic, and South, much of it tied to data-center development. That’s more than 10 times estimated current US data-center electricity use. 

The number has drawn attention because it includes projects at very different stages. Some requests come from developers with land, financing, customers, and a path to construction. Others reflect early site evaluations, competing locations, duplicate requests, or projects that may never advance.

Reuters describes some of these requests as “ghost demand”: proposed power demand that appears in utility planning queues but may never become an operating data center. The queues can include duplicate applications, early-stage proposals, and projects without the financing, customer commitments, or final site plan required to move forward.

Utilities are starting to use deposits, collateral, and study fees to separate more advanced projects from early-stage requests. Reuters reported that Exelon’s estimate of high-probability data-center demand fell about 40% after it tightened collateral requirements, while AEP Ohio’s pipeline dropped by more than half after the state introduced connection-study fees. 

The remaining demand can still be substantial. Data centers are growing alongside industrial development and electrification, while the transmission, substations, and generation needed to support major new loads can take years to build.

Pennsylvania’s project pipeline offers another view of the gap between proposed and active development. Reuters reported that more than 100 data centers had been proposed in the state, but only 20 had submitted required permits, and most hadn’t secured power or an anchor customer. 

Governor Josh Shapiro’s executive order, signed on August 18th, adds disclosure and permitting requirements for large data-center projects, including commitments tied to grid costs, energy and water use, environmental performance, and community engagement.

What this means for technology leaders

CIOs don’t need to follow every utility filing or state executive order. They do, however, need to understand whether a provider’s future capacity is supported by completed utility agreements, active permits, financing, and a realistic construction plan. 

For teams weighing future capacity, the work starts with understanding how close a project is to serving customers.

Ask what “available capacity” means

“Available capacity” can mean very different things depending on the provider. It may refer to a planned site, a permitted facility, a project with an executed utility agreement, a location under construction, or a facility that’s already energized and ready to serve customers.

Ask where a specific project sits in that process. A facility with energized power has a different risk profile than one that still depends on interconnection approval, local permits, grid upgrades, equipment delivery, or new generation resources.

Review what could affect delivery

A provider’s delivery timeline can depend on milestones outside its direct control. Interconnection studies, transmission and distribution upgrades, substation construction, local approvals, generation availability, and long-lead electrical equipment can all affect when a facility is ready for production workloads.

Before tying a business-critical initiative to future capacity, ask which milestones are complete, which are still pending, and how the provider would handle a delay. The answers can inform vendor selection, contract terms, launch planning, and fallback options.

Include power in procurement

For large infrastructure commitments, power should be part of the procurement conversation early on. Infrastructure, procurement, legal, finance, and sustainability teams may need to work together to understand the dependencies behind a provider’s delivery plan.

Contracts can distinguish among planned, contracted, and energized capacity. They can also spell out notice requirements, remedies, or alternative-capacity options if an external dependency pushes back the delivery date.

It’s also worth asking whether the provider has completed the financial commitments required by the relevant utility or regulator. Deposits, collateral, study fees, minimum-bill commitments, and executed service agreements can provide a clearer picture of project maturity than an announcement alone.

Factor in regional energy costs

Energy prices and grid conditions can affect the economics of power-intensive workloads, including private AI platforms, colocated GPU deployments, and high-utilization training or inference environments. The impact varies by location, utility tariff, contract structure, and usage profile, so teams shouldn’t apply one national energy-cost assumption to every deployment.

PJM offers one example of why regional conditions belong in long-term infrastructure planning. Its independent market monitor attributed $29.4 billion in capacity charges across the last four base-capacity auctions to existing and forecast data-center demand. The figure won’t determine any one company’s energy bill, and it doesn’t apply directly to Texas’s ERCOT market, but it shows how fast-growing large loads can influence costs across a region.

Keep options open

Companies with time-sensitive AI programs may want to avoid tying a critical launch to one future facility or one provider’s projected energization date. The right level of redundancy will depend on the workload, budget, compliance needs, and tolerance for delay.

Teams may be able to qualify a second provider, use multiple cloud regions, deploy in phases, retain a hybrid architecture, or make workloads portable across environments. Those options can help the organization adapt if a site’s power, permitting, or construction timeline changes.

Our takeaway

Texas’s review is a reminder that a data-center announcement and production-ready capacity can be separated by a long list of dependencies. Utility agreements, interconnection work, grid upgrades, permits, construction, and site readiness all affect when a facility can serve customer workloads.

For CIOs, the practical move is to ask for the facts behind a projected delivery date. Knowing what has been completed, what remains pending, and how a provider plans for delays can help teams build more realistic AI roadmaps and avoid preventable surprises.

^{Need a stronger data foundation for your AI roadmap? Talk to FullStack about our AI, data, and analytics services today.}

Heading 1

Heading 2

Heading 3

Heading 4

Heading 5

Heading 6

Lorem ipsum dolor sit amet, consectetur adipiscing elit, sed do eiusmod tempor incididunt ut labore et dolore magna aliqua. Ut enim ad minim veniam, quis nostrud exercitation ullamco laboris nisi ut aliquip ex ea commodo consequat. Duis aute irure dolor in reprehenderit in voluptate velit esse cillum dolore eu fugiat nulla pariatur.

Block quote

Ordered list

1. Item 1
2. Item 2
3. Item 3

Unordered list

• Item A
• Item B
• Item C

Text link

Bold text

Emphasis

^Superscript

_Subscript