Understanding how data center growth is reshaping utility cost recovery, rate design, and grid operations — from interconnection to tariff structure.
Data center load growth is the most significant change in U.S. electricity demand in decades. After years of flat or declining load growth, utilities across the country are facing requests to serve individual customers that can rival the size of small cities.
U.S. data center electricity consumption rose from 58 TWh in 2014 to 176 TWh in 2023, and is estimated to reach 325–580 TWh by 2028 — translating to roughly 6.7% to 12% of total U.S. electricity consumption. U.S. data centers drove a 22% power demand increase in 2025 alone. Capacity has already reached approximately 75.8 GW as of early 2026, with industry tracking projecting 108 GW by 2028 and 134.4 GW by 2030. ERCOT alone forecasts 138 GW of large-load interconnection requests by 2030, up from 87 GW in 2025. However, not all of this will materialize — some analysts estimate up to 90% of proposed data centers in interconnection queues will never be built, driven by duplicative requests filed across multiple utility territories.
The geographic concentration is notable: the Mid-Atlantic corridor (particularly Northern Virginia), central Texas, and parts of California and the Southeast are seeing the heaviest development activity. Texas is actively positioned to overtake Virginia as the largest data center market globally. Individual projects from Microsoft, Meta, Google, and Amazon now routinely exceed 200–300 MW. The Stargate project (OpenAI/Oracle/SoftBank) has deployed over $100 billion in capital with $500 billion committed. Its Abilene, TX flagship site alone could reach nearly 1 GW by mid-2026, with five additional sites planned and total project targets of 10–15 GW. NVIDIA’s Vera Rubin architecture is expected to power the first gigawatt-scale facility by late 2026.
U.S. data center power capacity forecast (GW) — Baseline reflects ~76 GW actual in early 2026
Source: S&P Global Market Intelligence, 451 Research, Datacenter Services & Infrastructure Market Monitor & Forecast (October 2025).
These loads differ fundamentally from traditional demand growth. A single data center customer can double a utility’s load, require dedicated transmission infrastructure, demand specialized power supply arrangements, and introduce operational complexities around ramp-up schedules, load forecasting, and reliability requirements that existing rate structures were never designed to handle.
With the scale of these loads in mind, the next step is understanding what makes each type of large load unique — and why a one-size-fits-all approach to rate design won’t work.
Not all large loads are created equal. The characteristics of each load type — size, reliability requirements, infrastructure needs, scalability trajectory, and credit profile — determine the appropriate service offering and rate structure. Understanding these distinctions is the foundation for designing tariffs that are both fair and sustainable.
Select a load type to see its key characteristics and service considerations.
115 kV+ interconnection • Redundancy for power and delivery • BTM backup power • Reliability with power supply and delivery • Ability to shed load for power quality events
Knowing what these loads look like is one thing — understanding how they interact with the grid is another. Next, we examine the reliability and operational impacts these loads create.
Data centers are among the most power-sensitive facilities on the grid. They maintain their own backup power systems and will disconnect from the grid instantly when they detect voltage or frequency anomalies. While this protects the data center, it creates significant challenges for grid operators when hundreds or thousands of megawatts drop simultaneously.
On July 10, 2024, a 230 kV line fault occurred at 7:00 PM. Approximately 1,500 MW of data center load instantly disconnected from the system due to voltage fluctuations — a customer-initiated decision. The sudden loss of load caused subsequent voltage and frequency oscillations across the system, similar to the effect of a very large power plant suddenly coming online.
Source: NERC 2025 State of Reliability Report
When data centers drop from the grid, the impacts can range from 250 MW to 10,000+ MW — depending on the concentration of facilities in a region. This creates a reliability challenge that is fundamentally different from traditional load loss scenarios. In September 2025, the NERC Large Load Task Force issued a formal Level 2 Alert — a reliability warning highlighting grid disturbances, inadequate modeling, insufficient technical interconnection requirements, and a lack of operating protocols for the influx of large loads. Thirteen of 23 North American assessment areas now face elevated or high resource adequacy risks over the next five years, with MISO, PJM, ERCOT, WECC-Northwest, WECC-Basin, and SERC-Central all flagged. NERC’s January 2026 Long-Term Reliability Assessment projects summer peak demand could surge by 224 GW over the next decade — 69% higher than the prior year’s LTRA — with compound annual growth rates the highest since NERC tracking began in 1995.
Data centers with BTM backup disconnect in milliseconds, creating sudden imbalances equivalent to losing a major power plant.
The rapid loss triggers system-wide frequency and voltage oscillations that can cascade into broader stability issues.
Queue requests that never materialize distort utility and grid operator demand forecasts, complicating resource planning.
On December 18, 2025, FERC issued a landmark order directing PJM to create entirely new tariff rules for co-locating data centers with power plants. FERC found PJM’s existing behind-the-meter generation (BTMG) rules “unjust and unreasonable” because they fail to account for large load impacts on resource adequacy. The order establishes three new transmission service types for co-located load, with PJM compliance filings due in January–February 2026. This fundamentally reshapes how co-located data centers can access power from adjacent generators across the largest U.S. grid operator.
Requires data centers and other large non-critical loads (≥ 75 MW installed after January 2025) to accept curtailment during firm load-shed events. Key provisions include mandatory ramp-down or switch to on-site generation upon utility request, two required DSM programs (mandatory and voluntary curtailment), interconnection disclosure and cost-sharing rules, and protocols for co-locating large loads with existing generators. Rulemaking on the most contentious provisions — particularly ratemaking practices for large loads within ERCOT — is ongoing and not expected to conclude until December 2026, leaving the final rules uncertain.
The state regulatory landscape has shifted dramatically. In just the first six weeks of 2026, over 300 data center bills were filed across 30+ states — a sharp acceleration from the 200+ bills introduced across 40 states in all of 2025. The policy conversation has pivoted from near-universal incentives to regulatory pushback and cost-allocation scrutiny. At least 18 states have introduced bills creating special rate classes for large energy users, with some requiring data centers to fund infrastructure improvements and demonstrate ratepayer benefits. Virginia, Georgia, and Oklahoma are proposing to reduce or eliminate previously generous tax credits, while several states (New York, South Dakota, Oklahoma) have introduced construction moratoriums. At the same time, at least 37 states still offer some form of incentive — with Kansas becoming the 37th in June 2025. The tension between economic development incentives and ratepayer protection is the defining state-level policy challenge.
At the federal level, PJM’s Critical Issue Fast Path process — launched in August 2025 — escalated to the FERC co-location order within months. Ohio and Indiana approved new large-load interconnection rules in 2025 with provisions for collateral, exit fees, and minimum billing demand requirements. NERC and EPRI continue to study ride-through requirements specifically for data center loads, and the DOE has invoked rare authority to direct FERC to expedite large-load interconnection rulemaking (discussed in Module 5).
Grid impacts are only half the equation. The next question is: how does a utility recover the costs of serving these massive new loads? That requires understanding the full cost-of-service framework.
Serving data centers doesn’t change the fundamental rate-making framework, but it stresses every element of it. The same five-step process applies — revenue requirements, functionalization, classification, allocation, and rate design — but each step requires new thinking when a single customer can represent a significant portion of system load.
Determine the total cost of serving all customers, including the incremental cost of infrastructure and power supply needed for data center loads.
Unbundle costs by function: production, transmission, distribution, and customer service. Data centers primarily impact production and transmission costs.
Classify costs as demand-related, energy-related, or customer-related. High load-factor data centers shift the balance toward demand costs.
Distribute costs among customer classes. The key question: should data centers pay embedded (average) costs or marginal (incremental) costs?
Structure the tariff components: customer charge, demand charge, energy charge, and riders. Data center tariffs typically require additional provisions for standby, balancing, and power cost adjustment.
This is the central tension in data center rate-making. The choice between embedded and marginal cost recovery determines who bears the cost of new infrastructure and generation — and whether existing customers are held harmless.
Consider a utility that procures 300 MW of new generation to serve a data center using embedded cost pricing. In year 5, new chip technology allows 30% more efficient power usage — the customer now needs only 175 MW. How does the utility recover 125 MW of stranded investment? Marginal cost pricing with exit fees and bonding provisions addresses this risk directly. Embedded cost pricing spreads it across all ratepayers.
The components of a revenue requirement vary by utility ownership type, which affects how data center cost recovery is structured.
| Component | IOU | Co-op | Municipal |
|---|---|---|---|
| O&M | ✓ | ✓ | ✓ |
| Depreciation | ✓ | ✓ | ✓ |
| Debt – Interest | ✓ | ✓ | ✓ |
| Debt – Principal | ✓ | ✓ | |
| State/Federal Taxes | ✓ | ||
| In-Lieu-of-Tax | ✓ | ||
| Return on Rate Base | ✓ | ||
| Margin over Debt Service | ✓ |
Allocating $100,000 in demand costs among three classes using coincident peak methods. Toggle methods to see how allocations shift.
4-CP Method: Allocates demand costs based on class contributions to the four highest monthly system peaks (typically summer). This is the most commonly used method and reflects seasonal demand patterns.
With the cost framework established, the next critical question is: how does the physical infrastructure get built, paid for, and managed? That’s the interconnection challenge.
Connecting a data center to the grid is a multi-year process involving interconnection studies, infrastructure construction, and power supply procurement — often running in parallel with overlapping risks and dependencies.
The customer typically requests information on tariff structure, system siting and availability, and whether production costs will be marginal or embedded. The utility evaluates customer type, credit rating, and preliminary load requirements. ISOs are adopting large-load-specific study processes and clustering approaches to streamline the queue.
System impact studies and construction typically span 6 to 18+ months, often overlapping with power procurement efforts. Key risks include schedule delays, cost overruns, changing data center requirements (size may change mid-process), and determining who pays for interconnection infrastructure versus adjusted transmission rates.
Typically, the data center customer pays for interconnection infrastructure — these are large, customer-centric investments. Options include limiting interconnection to specific grid locations, requiring contributed capital (CIAC), credit review and construction bonding, and clear ownership/operation agreements for substations and interconnection facilities. Some utilities are proactively identifying specific grid locations that can accommodate large loads to speed the process.
Two major federal actions in late 2025 have reshaped the large-load interconnection landscape.
On October 23, 2025, the DOE invoked rare authority under the Federal Power Act to direct FERC to initiate rulemaking to “rapidly accelerate the interconnection of large loads.” The directive applies to loads greater than 20 MW connecting to FERC-jurisdictional transmission. It proposes generator-like standardization for large load interconnections, potentially expediting projects that agree to curtail or dispatch. FERC issued an Advanced Notice of Proposed Rulemaking (ANOPR) with the DOE setting an April 30, 2026 deadline for final action. This directive has sparked significant federal-state jurisdictional concerns, since states traditionally have jurisdiction over retail load interconnections.
FERC’s December 18, 2025 order directs PJM to create three new transmission service types specifically for co-located load — loads physically adjacent to generating facilities that seek to consume some or all of that generator’s output. The order found existing behind-the-meter generation (BTMG) rules unjust and unreasonable because they don’t fully account for large load impacts on resource adequacy. PJM compliance filings were due in January–February 2026. This order is particularly significant because co-location has been one of the fastest pathways data centers have pursued to bypass congested interconnection queues.
FERC Order 1920 (issued May 2024, modified in November 2024 and April 2025) requires transmission providers to produce 20-year regional transmission plans at least every five years, incorporating multiple demand growth scenarios. This creates a framework for incorporating data center demand projections into regional transmission planning at scale for the first time. PJM’s cost allocation compliance filing is due June 2026.
With the physical infrastructure in place, the most complex and consequential challenge remains: securing and managing the power supply to serve these enormous loads.
Power supply is the single most complex and consequential element of data center service. A utility may need to procure hundreds of megawatts of new generation, manage complex renewable energy arrangements, and operate a power supply program tailored to a single customer — all while protecting existing ratepayers from cost exposure.
20+ year commitment with construction risk. Requires bonding, surety, and PUC cost recovery approval. Some IOUs creating new wholesale “GenCo” entities (e.g., NIPSCO, I&M in Indiana).
5–10 year market purchases. More flexible but exposes utility to market price risk. Customer may bring their own renewable PPA that utility must “firm up” with scheduling, balancing, and backup.
Direct pass-through of market prices (real-time or day-ahead). Shifts price risk to customer but requires sophisticated metering, scheduling, and settlement infrastructure.
Dedicated nuclear power supply at or adjacent to the data center site. Provides 24/7 carbon-free baseload matching the high load factor of data centers. Emerging rapidly with first SMR-powered facilities expected by 2027–2030.
Nuclear co-location has emerged as one of the most significant power supply trends since late 2025. Tech companies are signing unprecedented deals for dedicated nuclear capacity, driven by the convergence of 24/7 clean energy commitments and the limitations of intermittent renewables for baseload data center operations.
1.2 GW campus in Ohio using Oklo’s Aurora SMR technology. Oklo’s total customer pipeline exceeds 14 GW, including a 12 GW master agreement with Switch and 500 MW deal with Equinix.
17-year, 1.92 GW PPA from the Susquehanna nuclear plant. AWS has committed $20 billion to Pennsylvania data center development powered by this arrangement.
The SMR market reached $6.9 billion in 2025, projected to reach $13.8 billion by 2032. Tech giants have collectively committed over $10 billion to nuclear partnerships. First SMR-powered data centers expected by 2027–2030.
Nuclear co-location introduces new dimensions to the power supply decision: NRC licensing timelines (3–7 years for SMRs), spent fuel management, decommissioning liability, and federal regulatory requirements that most utilities have never navigated. It also raises cost allocation questions — if a data center customer co-locates with a nuclear facility, who bears the decommissioning and insurance costs? Existing ratepayers or the data center customer? The embedded vs. marginal cost framework from Module 4 applies directly.
Most hyperscaler customers have 100% clean energy commitments. This adds layers of complexity to the power supply arrangement:
Customer-provided PPA: The utility may need to “sleeve” the customer’s renewable PPA — purchasing all output and selling all power, with financial reconciliation. The utility provides capacity firming, backup, and scheduling services.
Hourly vs. Aggregate: How is 100% “clean energy” measured? Meta, for example, has transitioned to hourly reconciliation of RECs and carbon emissions, moving beyond annual aggregate matching.
Renewable Portfolio Standards: State RPS requirements may interact with customer clean energy goals, creating either synergies or conflicts in procurement planning.
Operating power supply for a single 500 MW customer requires a level of coordination and staffing that most utilities have never needed for individual accounts:
| Operational Element | Requirements |
|---|---|
| Ramping Schedule | Customer provides load ramp-up plan; technology changes (HVAC, computing) may alter timeline |
| Day-Ahead Profile | Customer provides daily energy forecast; frequency of updates varies by contract |
| Demand Balancing | Define impacts when operations vary from contracted amounts (over/under energy and demand) |
| Revenue Sharing | If customer is over/under contracted amounts: buy through at market, sell excess, share revenues (50/50?) |
| Key Account Staff | Dedicated customer representative with daily coordination across customer service, power supply, and planning |
| Demand Response | Power quality events, ability to shed load, backup power coordination |
With the cost and supply framework established, it’s time to assemble the pieces into an actual tariff structure — the binding document that governs the entire relationship.
The tariff is where all the analysis comes together — cost recovery, risk mitigation, operational requirements, and competitive positioning. A well-designed data center tariff must balance the utility’s need for cost recovery with the customer’s need for predictable, competitive pricing.
The comprehensive tariff framework for data center service covers ten major areas. Each requires specific provisions tailored to the load type and utility’s risk profile.
| Element | Key Considerations |
|---|---|
| Applicability | High load factor customers with demand ≥ XX MW |
| Service Characteristics | Firm power, full requirements at transmission voltage ≥ X kV |
| Land Rights / Easements | Utility maintains access to all necessary easements and land for infrastructure |
| Interconnection | Transmission-level; studies, facilities, ownership, maintenance responsibilities |
| Power Supply | Capacity/energy, contracted amounts, new resource vs. existing, market access |
| Contracting | Term, direct assignment or embedded, utility build options |
| Operations / Balancing | Hourly planning, over/under provisions, backup power, demand response |
| Transmission / Delivery | ISO market, vertically integrated, OATT, CP allocation management |
| Customer Service | Dedicated key account staff; direct assign or COS-allocated costs |
| Credit Requirements | Surety bonds, performance bonds, cash flow, default provisions |
A typical data center tariff includes several rate components, each serving a specific cost recovery function.
How the same cost-of-service translates to different rate designs
Data centers with behind-the-meter generation require standby service provisions. These rates cover the utility’s cost of maintaining capacity and infrastructure available for when the customer’s on-site generation is unavailable.
| Component | Description |
|---|---|
| Service Charge | Fixed monthly charge for administrative costs |
| Reservation Charge | Charge for contracted standby capacity — paid monthly regardless of usage |
| Backup Charge | Capacity and energy provided during unplanned outages |
| Maintenance Charge | Capacity and energy during scheduled maintenance of customer generation |
| Supplemental Power | Demand/energy needs beyond on-site generator capacity under normal conditions |
| Excess Power | Penalty for non-performance or non-compliance with standby contract terms |
Any data center tariff should satisfy the core Bonbright criteria: practical implementation, uncontroversial interpretation, meeting revenue requirements, revenue and rate stability, fairness among customer classes, avoidance of undue discrimination, and economic efficiency. The tension between “fairness among classes” and competitive positioning for economic development is the central policy challenge.
Theory meets practice when utilities actually implement these frameworks. Let’s examine real-world tariff examples and how one city navigated the challenges.
The data center tariff landscape is evolving rapidly. Across the country, utilities are developing new rate classes, contract structures, and interconnection rules. Examining existing approaches reveals both common patterns and innovative solutions.
Select a utility to see their approach to data center tariff design.
Serves historical data centers <100 MW with a conventional rate class for customers >5 MW. Includes a load development fee for customer payment of substation and connection costs ($/kVA). Demand charge of approximately $29.50/kW (adjusted 4% in January 2026 to address rising infrastructure costs) with tiered declining block energy rates. SVP also offers a Data Center Program Rebate of up to $1.5 million per customer (program year July 2025–June 2026) — a demand-side incentive alongside rate design. A straightforward approach reflecting SVP’s established data center market.
Large Load Substation service for customers >20 MW with minimum billing demand at 80%. Terms unique to each customer including load forecasting requirements. Seasonal demand charges: $9.99/kW winter, $16.77/kW summer, $29.00/kW summer peak generation & transmission. Published seasonal energy rates with cost adjustments.
Two large load classes approved by the Florida PSC on November 20, 2025, effective January 1, 2026, as part of FP&L’s four-year rate plan (2026–2029). LLCS-1: serves up to 3 GW of combined load in three zones (Martin, St. Lucie, and Palm Beach Counties) near existing 500 kV transmission. LLCS-2: optional rate for loads outside those three zones, not capped at 3 GW, with an Incremental Generation Charge (IGC) based on a formula rather than a stated rate. Customer pays for all infrastructure. 20-year contract term. Represents the trend toward marginal cost pricing for new loads — and is now the most significant approved IOU data center tariff in the U.S.
Dedicated Data Center Tariff for customers >25 MW with minimum billed demand at 85%. Utility can suspend service if demand exceeds contract by 1 MW. 12-year contract term with exit fee provisions. Collateral/credit rating requirements include proof of cash balance for 50% of minimum bill. Reflects the more prescriptive IOU approach with strong protections.
The City of Georgetown, Texas provides one of the most detailed public examples of a municipal utility developing a comprehensive data center tariff. Operating in the ERCOT market, Georgetown faced three categories of risk when large loads (4–25+ MW) began requesting service.
Significant capital outlay for substation and transmission upgrades. Mitigated by Line Extension Policy, Application Fee ($2,500/MW), and Plant Investment Fee.
Over/under-hedging exposure from inaccurate load forecasts. Binding semi-annual forecasts with ±3% tolerance band. Deviations trigger ERCOT real-time market cost pass-through.
Revenue concentration and 45-day billing lag. Mitigated by 3-month prepayment, letter of credit, and bills due on receipt.
Rate components for a 12 MW customer at 85% load factor
| Rate Component | Amount |
|---|---|
| Setup Fee | $25,645 (one-time — ERCOT metering, fees, personnel) |
| Customer Charge | $5,500 / month |
| Demand Charge | $19.25 / kW per month |
| Energy Charge | $0.05317 / kWh |
| Power Cost Adjustment | $0.0137 / kWh |
| Energy Rate Rider | $0.0082 / kWh (new — hedging cost recovery) |
Georgetown’s most innovative contribution is the VLC&I Energy Rate Rider — a mechanism specifically designed to recover incremental power supply hedging costs from data center customers without impacting other rate classes.
The rider is calculated and evaluated quarterly using two months of actual energy supply costs and six months of forecasted costs. It divides the total net hedging cost by total energy served over the 8-month period. The rider can be a charge or credit depending on the balance of the Over/Under Account. Key advantages: aligns with the utility’s risk management program, provides price stability for customers, ensures full power supply cost recovery, and reduces administrative burden compared to binding forecast methods.
Typically 10–20 years for built generation; shorter for PPAs. Exit fee and stranded cost provisions standard.
Minimum billing demand typically 80–85% of contracted capacity with demand ratchets.
Surety bonds, proof of cash balances (50%–200% of minimum bill), corporate guarantee requirements.
Industry moving toward marginal/incremental generation charges to hold existing customers harmless.
Data center service is the most complex rate-making challenge utilities face today. We help you navigate the interconnection, cost recovery, and tariff design decisions with clarity and confidence.
We’ve designed tariffs for data center customers across every utility type — IOUs, co-ops, and municipals. From interconnection cost recovery to power supply structuring, we bring real-world experience to the table.
We help utilities identify and mitigate the unique risks of data center service — stranded costs, construction overruns, credit exposure, and load forecasting uncertainty. Our frameworks protect existing ratepayers while enabling growth.
The regulatory landscape for data center service is changing rapidly. We track developments across every major jurisdiction and help utilities develop tariffs that meet current requirements while adapting to evolving rules.
Ready to start the conversation?
Connect With NewGen →