Data Center Sites Powered by Gas Isn’t the Same As Ready to Build

Behind-the-meter-power can solve the timing problem. It doesn’t make the site buildable, permittable or connectable.

A data center site can have plenty of power lined up on paper and still fall apart before construction begins.
The problem may have nothing to do with megawatts. A pipeline cannot reach the parcel. Water access becomes a point of local opposition. A regulator changes how the power arrangement can be structured. Power may get a site into the conversation. It does not make the site executable or ready for capital.

Behind-the-meter generation, or bring your own generation (BYOG), has moved from workaround to default in the most constrained markets. Enverus Intelligence^® Research forecasts BYOG will serve 22.5 GW of new U.S. data center load between 2026 and 2030, roughly 36% of all data center capacity added over that period, concentrated in PJM and ERCOT. Gas accounts for 29.6 GW of the generation needed to serve it; committed solar and wind together account for 2.0 GW.

We estimate that BTM generation will serve 25.5 GW of demand from new facilities between 2026 and 2030, with data centers accounting for 22.5 GW, or 88% of the total. This implies that roughly 36% of U.S. data center capacity additions during the period will be served by BTM generation. 

Source: Enverus Intelligence® Research Behind-the-Meter Generation Forecast | Skipping the Queue

Three Ways a Powered Site Can Still Run Into Trouble

Each project appeared to have a viable power plan. Each still encountered a constraint that changed its path.

1. The regulatory structure changed: One proposed arrangement planned to supply a data center campus directly from a nuclear plant. When the original structure did not hold up under regulatory review, the project had to pursue a different grid-connected arrangement.
2. The pipeline route missed the mark: Another project identified natural gas supply, but the proposed pipeline route crossed land the developer did not control. Moving the route by less than a mile added significant time and cost.
3. Community concerns changed the plan: A third project faced local opposition tied in part to water availability and cooling. The project eventually broke ground with a revised cooling plan, but only after losing its original anchor tenant.

The pattern is not limited to individual sites. In its 2Q26 data center capacity work, Enverus Intelligence^® Research removed the Homer City campus in Pennsylvania and Fermi’s Matador project in Texas from its high-confidence project list, citing canceled contracts, missing offtakers or unidentified data center partners. Homer City is one of the largest planned gas-fired facilities in the country. Announced generation is not the same as a bankable project.

The common thread was not a lack of planned power. It was a gap between a powered site and an executable one. BYOG can shorten the path to power, but it also makes the rest of the site-validation work more important.

BYOG changes the risk map

Developers are not choosing this path because it is simpler. Interconnection study cycles in the most active markets now run in years rather than months, and load is materializing faster than transmission and new generation can be built to serve it. BYOG is a response to a timing problem, and it works on timing. What it does not do is settle everything downstream of that decision. More of the diligence shifts to the site, the infrastructure required to serve it and the fuel supply needed to keep it running. In a traditional grid-connected project, much of the early power diligence centers on queue position, study timelines, network upgrades and deliverability. BYOG keeps those questions on the table and adds several more:

• Can the parcel support the campus and generation footprint?
• Can pipelines and supporting infrastructure reach the site?
• Can the project clear permitting and local requirements?
• Is the grid and fuel supply reliable enough to support the operating plan?

The grid also remains part of the plan for many BYOG projects. Its role may be smaller or delayed, but it can still shape the power ramp, interconnection strategy and long-term economics.  Both sets of questions stay on the table.

Sites fail predictably, in a specific order

Many of the most expensive site setbacks can be traced to questions that could have been asked earlier. The challenge is knowing which questions to ask first. Start with the faster, lower-cost screens.
In days: parcel buildability, zoning and right-of-way constraints can eliminate weak candidates before a team invests heavily in engineering or power studies.

In weeks: pipeline routing, permitting and local requirements come next.

In months to years: grid deliverability, competing demand and long-term fuel supply. These issues can take far longer to resolve, which makes it even more important to clear the earlier hurdles first. The goal is not to eliminate every uncertainty. It is to expose the biggest disqualifiers before they become the most expensive problems.

What an investment-ready BYOG site must prove

Before capital is committed, a BYOG site should be able to pass five tests:

• Buildable: Can the campus and generation equipment fit on the site?
• Permittable: Is there a credible path through air, water, noise, zoning and local approvals?
• Connectable: Can the grid support the required load, ramp or backup strategy?
• Fuelable: Can the project secure enough gas, both physically and commercially?
• Defensible: Does the investment case account for competing projects and changing infrastructure conditions?

 
A site that passes all five is a site a capital committee can actually approve.

The next question: How much grid capacity is really available?

A site may clear its early land, permitting and community hurdles, but one of the most consequential questions is still unresolved: How much grid capacity can it actually count on?

A nearby substation is not proof of usable capacity. Competing loads, system conditions and network upgrades can change how much power is available and when it can be delivered. For a BYOG site, that answer helps determine how much generation must be built on-site, how quickly the campus can ramp and whether on-site generation serves as a temporary bridge or a long-term part of the power strategy.

In the next post, we’ll look at how developers can test grid capacity under different operating conditions and turn an initial headroom estimate into a more defensible range. If you would rather not wait: the on-demand session walks the grid-capacity screen and the gas-supply screen at the same candidate site.

Next in the series: How Much Grid Capacity Can Your Data Center Site Actually Count On?

Watch From Decision to Execution: How to Validate BYOG Sites Before You Commit to see how developers can test land, permitting, grid and fuel risks before committing capital.

The Hyperscaler PPA Is Coming: Can You Defend the Deal?

The next hyperscaler power agreement will be won by parties that can show, before a term sheet is signed, why the deal works for the plant, the grid, and the customers who may ultimately scrutinize it. Finding an available megawatt is only the beginning.

The Comanche Peak transaction illustrates the first half of that argument. Its owner signed a 20-year, 1,200 MW power purchase agreement with AWS in September 2025, then followed it with more than 2,600 MW across three other nuclear plants for Meta. Both deals closed in months. The next question is whether a similar agreement also works for the grid and the customers who will scrutinize it. That answer will shape how quickly the next deal can move.

Hyperscalers want firm, reliable power for rapidly expanding data center load. Plant owners want long-term revenue that can support continued operations, relicensing, maintenance capital, or uprates. Those objectives can align, but the path from a promising asset to a durable agreement depends on more than resource performance or buyer demand.

Merchant generators can often negotiate at industry speed, particularly in markets where wholesale transactions do not require a retail rate case. Other deals still move through existing federal market-based rate processes, so the speed advantage is real but not absolute. Vertically integrated utilities, municipal utilities, and cooperatives face additional questions, such as: Does the load change local reliability needs? How will replacement power  or additional reserve capacity be sourced? Who pays for new infrastructure? And can the cost allocation withstand review by a commission, board, city council, or ratepayer advocate?

For utilities, an indefensible agreement can create political backlash or fail regulatory review. For hyperscalers and developers, the same weakness can turn a promising site into a years-long delay. A defensible agreement gives both sides a clearer path to closing and helps protect the schedule after signing.

Three steps to a deal that closes and holds

1. Score the asset and grid exposure

Start with a screen that reflects how the deal will be evaluated across two dimensions. Asset exposure comes first and includes license runway, operating scale, multiunit configuration, current PPA commitments, uprate potential, buildable land, nearby data center development, hyperscaler presence, and zonal load growth. Ownership matters too, because the same plant can present a very different path to contract depending on who must defend the decision. Deliverability exposure comes next, and it is where most utility planners spend their time. That means assessing available headroom on the bulk and nodal systems, the upgrades a new interconnection would require, who funds them, and how the load curve behaves against existing commitments. A plant can score well on the first dimension and still fail on the second.

Enverus Intelligence® Research (EIR) screened operating U.S. nuclear plants for long-dated hyperscaler offtake using these kinds of criteria. Dresden, Comanche Peak, and South Texas Project emerged as Tier 1 candidates, while a broader group of plants offered more mixed or site-specific opportunities. Comanche Peak’s ranking already reflects its AWS deal. Half its capacity is contracted, almost exactly matching what that agreement accounts for. The remaining capacity represents the potential opportunity for a future hyperscaler agreement. The screen separates interesting from actionable. A plant with strong demand nearby may still have limited uncontracted capacity, constrained land, or an ownership structure that adds decision-makers and public process.

For a utility, scoring exposure early clarifies which assets are worth preparing before a buyer calls. For a hyperscaler or developer, it creates a realistic shortlist and surfaces execution risk before deal teams spend months on a non-starter.

The ownership chart matters because it shows why an attractive asset can still have a difficult path to contract. The more parties responsible for the decision, the more assumptions must be reconciled before a deal can move. No utility decision of this scale is made by a single stakeholder or team, and the more owners a plant has, the more boards, commissions, and city councils may need to sign off before an agreement can close.

Two plants in this screen show what that looks like. South Texas Project has three owners, and one of the municipal co-owners is currently asking its board to fund a bigger stake in the plant while studying new reactor technology, taking on more of this exposure, not less. Dominion’s North Anna and Surry carry the same exposure alone, with no partner to share the political weight of a PPA decision, exactly the kind of exhibit an intervenor reaches for in front of a state commission.

2. Bridge internal and external silos

A PPA can move quickly only when the people who will approve, model, regulate, and operate it are working from the same assumptions. No utility decision of this scale is unilateral.  Bring corporate strategy, regulatory affairs, resource planning, transmission, finance, and external development teams into the conversation before commercial terms harden. The teams should align on:

• The load ramp and delivery point
• Capacity treatment and reliability implications
• Replacement-power assumptions
• The questions a public reviewer is likely to ask

Misalignment shows up fast once a deal is in motion:

• A term sheet that looks complete to a developer may be an opening question to a utility board.
• A load forecast that supports a hyperscaler’s capacity target may not explain who bears the cost if the project ramps late or never reaches full utilization.
• A reliability study that satisfies engineering may still leave open who funds the upgrade it recommends.

Resolving those gaps early gives both sides a faster path to an agreement that can survive diligence.

A partner can move quickly when its reviewers are aligned before the clock starts. That preparation gives the hyperscaler a clearer path through diligence and gives the utility fewer reasons to reopen the deal.

3. Defend the rate and cost structure

The commercial question is not simply whether a hyperscaler can pay for power. It is whether the agreement clearly assigns the costs and risks created by the new load. A defensible structure should address incremental generation and transmission investment, customer protections, credit and collateral, load-ramp commitments, minimum contract terms, outage and replacement-power treatment, and the consequences if expected load does not materialize.

That discipline is increasingly visible in large-load tariffs. Utilities are using ramp requirements, collateral, and long-term commitments not only to recover costs, but also to distinguish funded projects from speculative queue positions. The same logic can strengthen a nuclear PPA. Clear obligations make the project more credible to regulators and more dependable to the buyer.

The scrutiny around large-load agreements is also increasing. The White House’s Ratepayer Protection Pledge began in March 2026 with seven hyperscalers committing to cover their own power costs. By July, it had grown to 187 organizations across four stakeholder groups, including utilities and cooperatives. The pledge is not a contract or guarantee, but it signals a growing expectation that hyperscalers should account for the power costs and grid impacts associated with their growth.

A cost allocation that can be explained plainly to a commission, board, city council, or ratepayer advocate is a competitive advantage. It reduces unresolved questions that can reopen the transaction and gives the hyperscaler a better chance of receiving power on schedule.

The deal test: what makes an asset defensible

A screening report can identify the plants most attractive for hyperscaler offtake. It cannot, by itself, determine whether a specific agreement is right for the owner or its customers. That second question depends on ownership, existing obligations, market structure, local reliability, and the proposed allocation of costs and benefits.

Can both sides explain the agreement using the same load model, reliability assumptions, and answer to “who pays if the plan changes?” That answer determines whether the deal is ready, regardless of how compelling the asset looks on paper.

Why Enverus

Enverus Intelligence® Research provides the shared data foundation utilities, hyperscalers, and developers need to evaluate these decisions together. Utilities can use the screening framework to understand fleet exposure before an offer arrives. Developers and hyperscalers can use it to identify assets with meaningful capacity, site potential, demand, and a realistic path to contract.

The broader planning challenge is covered in Planning You Can Defend, an Enverus e-book on how utilities can defend, optimize, and grow the grid as load forecasts, interconnection queues, market behavior, and rate pressure become harder to reconcile. Download the e-book to build a planning process that stands up when the next major load arrives.

MISO hits the Price CAP $10,000

Power Market Insights

MISO prints $10K cap as evening load, solar collapse, imports weaken — spurring an EEA2. The full rundown below & our Mosaic timeseries highlighting the convergence

MISO’s real-time market printed at the $10,000/MWh cap for multiple intervals during HE19-20 (6-8pm ET) on September 2nd, driven by a textbook late-summer scarcity setup rather than a discrete generation failure. Load came in 4,300-5,700 MW above the day-ahead forecast through the back half of the day as temperatures ran hotter than expected, while solar output collapsed (neon green shaded area) from a midday peak of ~18,350 MW to just 137 MW by HE20 — removing nearly 18 GW of supply exactly as load peaked. The combination, layered on reduced import availability (orange area) and transfer capability, pushed MISO from an EEA1 at 6:48pm to EEA2 by 7:11pm — a 23-minute escalation that signals how little headroom the system had left.

Batteries discharged into the event but were drawn down before the worst of it hit. Real-time at Indy Hub averaged $1,630.70 (HE19) and $4,118.32 (HE20) against DA prints of $347 and $257, respectively. We flagged but cannot confirm a generation dip around Fermi and other units during the event window; until telemetry vs. trip is resolved, we’re treating solar ramp-down, weaker imports and load beat as the primary, confirmed drivers.

Why it matters: No unit trip, no wind miss — just ordinary load forecast error (~5 GW) stacking on the routine evening solar ramp-down with weaker imports. That combination alone was enough to trigger EEA2 and hit the cap, which means the reliability margin has shrunk further than the market has priced in.

Critical risks going forward:

• Imports can’t be assumed as a reliable hedge during the evening ramp — if neighboring-system tightness or transfer constraints reduce import availability at the same hours MISO’s own margin is thinnest, that’s a correlated risk across the footprint, not a diversified one.
• Forward curves and RT vol should reprice — if somewhat routine variance can trigger caps, cap-hitting probability across the summer strip is higher than a single-event read would suggest.