Data Center Flexibility: Chapter 1 (Introduction)

A New Vision for Data Center Flexibility

1.1 The new reality: Interconnection bottlenecks and regional grid strain

The interconnection challenge facing the United States is increasingly complex. It’s not just the size of the queue; it’s the structure of the system itself. Data center load interconnection requests account for the majority of the large load interconnection queues.¹ CenterPoint Energy in Texas has seen new large-load applications surge from 1 GW to 8 GW in under two years.² In Virginia, Dominion Energy connects around 15 new data centers each year, adding 1 GW of new load, while another 50 GW of projects remain active in the queue, pushing wait time up to 7 years or longer.³ In Arizona, utilities report backlogs that could take the remainder of the decade to clear, reflecting both permitting complexity and the limits of existing transmission infrastructure.⁴

Figure 1: Estimated interconnection timeline in VA - Source: Dominion Energy

1.1.1 The speed mismatch

At the root of the problem lies a fundamental timing mismatch. Today, many high-voltage transmission projects can take over 5 years to permit, though each market and site is different.⁵ Interconnection studies, performed sequentially (or semi-sequentially in cluster studies) and under worst-case assumptions, compound the delay through potentially assigning later entrants the full cumulative cost of upstream upgrades.⁶ New on-site generation resources require 3-8 years depending on technology.⁷ Even natural gas turbine/engine plants, one of the fastest conventional generation option, can require 3-5 years and face increasing challenges from supply chain crunches and environmental and community opposition.⁸

These delays are more than an administrative issue; they carry significant economic and policy consequences. For developers, each year of delay translates to hundreds of millions of dollars in idle capital and lost digital capacity, which is increasingly becoming a driver of the overall economy and a national security concern.

1.1.2 Regional impacts vary but share common characteristics

The interconnection crisis is not uniformly distributed across the United States. Specific regions have become pressure points where digital infrastructure demand most acutely conflicts with grid constraints. Northern Virginia’s Loudoun County, home to around 70% of global internet traffic, faces fundamental capacity limits despite hosting more data center infrastructure than most countries.⁹ Texas (specifically Dallas, Houston, Austin, and San Antonio) has seen unprecedented load growth driven by crypto mining and hyperscale data centers, straining the ERCOT grid’s generation and transmission capacity.¹⁰¹¹

Figure 2: Data center forecasted demand growth - Source: DOE

Phoenix represents a particularly acute case study. The region’s combination of favorable tax policies, low-cost land, and abundant solar resources has attracted massive data center investment.¹² However, the local utility system was designed for residential and commercial loads with predictable seasonal patterns.¹³ Arizona Public Service (APS) has pioneered demand response programs specifically for data centers, but these remain pilot-scale relative to the magnitude of development pressure.¹⁴

Western markets face growing complexity as data center expansion collides with transmission bottlenecks and resource constraints. California’s restrictions on new data center water usage in certain counties have pushed development toward neighboring Western states, but many of these areas lack the grid infrastructure needed to support large-scale digital load additions.¹⁵ The result is a regional game of infrastructure arbitrage, developers chasing available capacity rather than sustainable long-term solutions, a pattern that’s increasingly difficult to maintain at scale.

The evidence is clear: interconnection timelines are lengthening, costs are escalating, and regional grid stress is spreading. Without new coordination models, particularly those that leverage data center flexibility, the United States risks letting physical bottlenecks determine the pace of its digital economy.

1.2 Data center growth projections and grid impact

1.2.1 The AI revolution in numbers

AI is transforming electricity demand faster than any industrial shift in recent memory. Bloomberg projects that U.S. data center capacity will grow from 35 GW in 2025 to 78 GW by 2035, more than doubling within a decade.¹⁶ It’s the steepest and most sustained increase in power use by a single sector since the post-world war electrification of American homes.

The new energy intensity of AI

AI data centers don’t just consume more energy and require more power; they change the very shape of demand. A single ChatGPT query consumes nearly 10 times more energy than a Google search.¹⁷ Standard-density server racks typically operate at 7–10 kW, AI training and inference racks can reach 30–100+ kW per rack, representing a step-change in both energy demand and cooling requirements.¹⁸ This intensity is already driving a shift in how data centers are designed and operated: bigger substations and transformers, high-performance cooling, and round-the-clock utilization.

For utilities, this represents a fundamentally different kind of load, flatter, longer, and less weather-dependent than anything seen before. And even as processors become more efficient, total consumption keeps rising because computational demand is growing even faster.

A single 100 MW AI training cluster can draw as much electricity as a large manufacturing plant, but unlike a factory, it runs continuously.¹⁹ Cooling alone can account for roughly 40% of total energy use, while another 10–15% supports internal power conversion and distribution.²⁰ ²¹Together, these factors point to an urgent need for new power management mechanisms, solutions that can adjust power use dynamically without compromising uptime or performance.

1.2.2 Geographic concentration

Growth is also unevenly distributed. Research from EPRI finds that 15 states account for 80% of the national data center load, with amplifying grid stress in regions already hosting the densest clusters.²² Arizona’s Maricopa County has attracted multi-gigawatt commitments from major cloud providers, pushing local substations toward design limits.²³

Texas presents a different but equally challenging scenario. ERCOT’s load forecast documents substantial data center growth through 2030, with much of this growth concentrated in or near metropolitan areas.²⁴ However, over the last two decades, much of Texas’s transmission system evolved around dispersed wind generation in West Texas serving load centers in the eastern part of the state.²⁵ Adding large, concentrated loads in metropolitan areas requires new transmission routes that yet again will rework the system design.

The concentration pattern reflects several economic factors. Hyperscalers benefit from economies of scale and operational efficiency by co-locating facilities. Network latency requirements for certain applications favor proximity either to other data centers, where the latency is between data/compute services and each other, to population centers, where the latency is between data centers and the end-user, or both. Further still, skilled workforce availability concentrates in existing technology hubs. However, this geographic clustering creates systematic grid stress that utility systems were not designed to accommodate.

1.3 Why existing approaches are insufficient

1.3.1 The transmission-first paradigm

The traditional utility response to load growth has been infrastructure-centric: build new transmission lines, substations, and generation resources to meet projected peak demand. This approach served the industry well during the 20th century when demand growth was steady, capital was relatively inexpensive, and environmental and permitting constraints were minimal. However, applying this paradigm to rapid and massive data center growth creates a cascade of problems that render it extremely challenging in the best case.

New transmission development has become exponentially more difficult over the past two decades. Environmental review processes that once took months now require years.²⁶ Land acquisition costs and routing challenges have escalated dramatically, particularly in the suburban areas where data center development concentrates.²⁷ Community opposition has intensified as transmission lines increasingly traverse developed areas rather than rural corridors.²⁸ The result is that transmission projects can routinely require a decade or more from conception to energization.²⁹

Most data center developers have economics constructs that can’t afford to wait a decade for new transmission lines to come online. Financing constraints, underwriting, competitive pressure to reach the market quickly, and exposure to project risk all push them to look for faster options, different locations, or find creative connection strategies. Growth shifts to other regions without the same bottlenecks, or triggers pressure for expedited approvals that can erode safety standards and limit community input. Neither approach scales to meet the surge in demand expected through 2030.

The transmission challenge for data centers extends well beyond local distribution upgrades. While distribution utilities handle the last-mile connections, modern data center clusters often connect directly at transmission level, triggering upstream transmission upgrades managed by RTOs/ISOs. Large-scale AI/HPC loads, often hundreds of MWs, require complex, system wide planning, environmental review, and coordination across multiple utilities and jurisdictions.

There is still no clear playbook on cost allocation, regional benefit, and risk sharing at RTOs/ISOs level. While the hyperscalers may fund localized upgrades directly, the broader grid expansion costs are sometimes socialized across all customers through utility tariffs, sparking debate over fairness and regional economic impact. In 2024, utility customers across seven PJM states were billed $4.4B for data center related transmission upgrades.³⁰

Figure 3: Data center related transmission costs 2024, [$M]- source: The Union of Concerned Scientists

Transmission planning timelines, federal oversight, and financing mechanisms compound the problem. Those challenges must be solved at the grid scale, not just the utility scale, if the U.S. is to enable meaningful and sustainable data center integration. And this is not a regional anomaly. And this isn’t a one-off issue. Similar patterns are emerging in other regional markets across the U.S. and abroad, wherever data center growth clusters around limited grid headroom.

1.3.2 The natural gas bridge mirage

Faced with transmission development timelines that cannot accommodate data center growth schedules, some regions are turning to natural gas-fired generation as a “bridge” solution. Simple-cycle combustion turbines (SCGT) can be deployed in 3-4 years, while combined-cycle plants (CCGT) require 4-5 years, both significantly faster than new transmission.³¹ Several utilities have announced plans for gas plants specifically to serve data center loads, particularly in constrained markets like Northern Virginia and Phoenix.³²³³ Leading gas turbine manufacturers, GE Vernova, Mitsubishi Heavy Industry, and Siemens Energy, which together supply 70+% of total markets³⁴, reported a total backlog of~ 130GW, of which ~ 45GW is for data center.

Figure 4: Gas turbine backlog from leading players - Source: Desk research, industry news, company’s news

Yet even these “quick-build” options face supply chain bottlenecks, equipment backlogs and labor shortages, stretching real-world delivery schedules well beyond early estimates. Developers are also exploring fuel cells, reciprocating engines, and hybrid systems as alternative pathways for reliable power, but these technologies face their own cost and operational challenges.

Power access and environmental trade-off

Natural gas as a bridge brings a parallel set of problems for data center growth. While gas-fired generation is widely used to provide reliable and dispatchable power at scale, its deployment risks locking in decades of CO2e emissions, upstream methane leakage, and stranded asset risk.³⁵ Data center operators are increasingly and voluntarily balancing carbon goals with practical reliability needs. Google signed a first-of-a-kind contract that will fund a carbon capture and storage system at a new 400 MW natural gas plant in Illinois, signaling a short-term compromise for reliability and project speed.³⁶ Yet across the industry, data center operators pair these near-term gas plant deals with ambitious carbon-neutral and net-zero commitments, creating an ongoing contradiction between short-term operational security and long-term climate goals. For most major operators, gas for immediacy remains a temporary fix, not a solution.

At the same time, new gas projects also face permitting challenges and local resistance. Local air quality impacts, NOₓ, VOCs, hazardous air pollutants (HAPs), particulate matter (PM₁₀ and PM_2.5), CO, and SOₓ, regularly trigger lengthy environmental reviews and community opposition, especially in high-growth regions where industrial expansion overlaps with residential areas. A clear example is the proposed gas fired data center campus in Pittsylvania County, VA. In late 2024, developer Balico LLC sought rezoning for a 3.5 GW gas plant and 84 data centers, but the plan met unified opposition from residents concerned about air pollution, water use, and the loss of farmland.³⁷³⁸ A health assessment by Dr. Francesca Dominici, Chair of the Harvard Data Science Initiative and Professor in the School of Public Health, estimated the project would emit more than 326 tons of PM₂.₅ each year, an air pollutant level that public health experts agree has no safe level of exposure.³⁹ After months of hearings, delays, and public pressure, the proposal was withdrawn.⁴⁰

Operational and economic constraints

Gas plants built to serve data center loads often face both operational and economic constraints, depending on the technology and how they’re used. Data centers require steady, around-the-clock electricity, which aligns better with high-efficiency CCGTs. Yet recent projects have seen sharp cost escalation and long-lead equipment (LLE) delays. S&P Global cites OEM quotes of up to 7 years for gas‑turbine deliveries depending on the model, with costs up as much as 2.5x versus just a few years ago.⁴¹ A recent report shows that installed costs for new CCGT builds are near $2,000/kW, nearly 90% higher than plants scheduled for completion in 2026–2027, between $1,116 -$1,424/kW.⁴²

By contrast, simple cycle peaker (SCGTs) turbines or reciprocating engines are designed for short, high-power bursts. Running them continuously drives up fuel use, emissions, and maintenance costs, with LCOE from $110-$251/MWh, and often exceeding $200/MWh at low-capacity factors-Figure 5.⁴³ Ongoing supply chain and labor pressures are pushing costs for all new gas projects higher, while the gap with clean-firm alternatives, such as advanced nuclear, long-duration storage, and hydrogen-ready turbines, should continue to close.⁴⁴ Together, these trends make it increasingly difficult to justify gas as a long-term solution for powering data centers, either economically or environmentally.⁴⁵

Figure 5: Estimated LCOE ranges of CCGT and gas peaking - Source: Lazard LCOE+ 2025, impactECI’s analysis, desk research

The “bridge” metaphor implies temporary operation until clean alternatives become available. However, once built, gas plants create economic pressure for continued operation to recover capital investments. The typical 20-30 year depreciation schedule for many generation assets extends well beyond the timeline when clean alternatives will be cost-competitive. What appears as a short-term solution risks becoming a long-term carbon commitment.

1.3.3 Traditional demand response limitations

Current demand response programs were not designed for large, always on data center loads. Most were built around commercial and industrial customers with flexible, low impact loads such as cooling or heating, that can be cycled briefly without major consequences. By contrast, data centers operate with near constant demand, strict uptime requirements, and limited tolerance for interruption. This difference in load profile and operational priority means traditional demand response models cannot effectively scale to meet the pace or magnitude of data center growth. New frameworks focused on predictable, high value flexibility rather than ad-hoc curtailment will be essential for integrating these loads into the grid.

Typical demand response programs allow utility customers to agree to reduce usage during peak periods in exchange for bill credits or payments. Events are called infrequently (typically fewer than 12 times per year) and last only a few hours.⁴⁶⁴⁷ Participation rates are modest, with most programs achieving 2-5% load reduction during peak events.⁴⁸

This model cannot scale to address data center growth for several reasons. The magnitude of load addition requires flexibility resources orders of magnitude larger than existing programs provide. Data centers have historically not been able to simply “power down” without major economic and service impacts; flexibility must be dynamic and controlled. The geographic concentration of data center loads means that traditional demand response, distributed across many customers, cannot provide sufficient localized relief.

Fundamentally, most traditional demand response programs treat reduction as an emergency measure rather than a normal escalating preventative measure and an integral part of system operations. Data center flexibility requires systematic integration into grid planning, market operations, and infrastructure investment decisions. The economic value of flexibility must be captured through escalating incentives and/or market mechanisms that reward reliability and performance rather than simple emergency curtailment.

1.3.4 The infrastructure financing challenge

The scale of data center growth creates infrastructure financing challenges that existing utility rate structures and regulatory frameworks cannot accommodate. Electric utilities are typically regulated monopolies that recover infrastructure investments through rate base mechanisms spread over decades, a system that works well for shared distribution assets. Under standard cost-allocation rules, large-load interconnection requests can trigger upgrades that must be fully funded by the requesting customer. For data center developers, that often means upfront commitments in the hundreds of millions of dollars or more, well before operations begin or revenue flows. These costs are usually secured through letters of credit or similar security during the interconnection process. If the project is delayed, scaled down, or canceled, utilities and other system participants remain financially protected, but the sunk development costs typically rest with the developer.

This financing structure creates risks that utility regulators are increasingly reluctant to accept. In many states, commissions now require data center developers to provide financial guarantees for transmission upgrade costs.⁴⁹ Others have mandated that customers requesting large interconnections must pay the majority of the upgrade costs upfront rather than amortizing them through ongoing rates.⁵⁰ While these rules protect utilities and ratepayers from financial exposure, they also raise significant barriers for developers, such as steep upfront capital requirements, higher financing costs, and more uncertainty on return on investment.⁵¹

The financing challenge extends beyond utilities to regional transmission organizations (RTOs) and independent system operators (ISOs). These entities coordinate multi-utility transmission planning but lack direct cost recovery mechanisms for projects driven by individual customers.⁵² When a single large load drives system-wide impacts that ripple across multiple utility territories, cost allocation and financing become exponentially more complex, slowing progress even as demand continues to surge.

1.4 How flexibility solves many grid problems

1.4.1 A paradigm shift beyond infrastructure

Easing the data center interconnection crunch will require more than just building new transmission lines and substations; it calls for a shift in vision. Data centers shouldn’t be seen as passive electricity consumers; they are now assets on the grid. With great power comes great responsibility: these facilities have both the capability and the obligation to help keep the system reliable, flexible, and efficient. The real question is now shifting from “How do we serve these loads?” to “How can these loads help serve the system?”

Grid flexibility from data centers offers a practical answer to some of the most pressing operational challenges facing today’s electricity systems, particularly as digital infrastructure growth accelerates.

Figure 6: Google compute load shifting - Source: Google.com

Addressing capacity breaching during peak hours

By actively managing their loads, data centers can shed, shift, or reschedule non-essential computing tasks during periods of grid stress-Figure 6. They can also dispatch energy assets or alter the operation of on-site infrastructure like cooling systems or smarter electrical equipment. Even a small adjustment, curtailing just 0.25% of annual operating hours, can unlock substantial additional grid capacity. For utilities, that flexibility translates into avoided or deferred investment in costly transmission reinforcements and peaker plants. Flexible interconnection agreements also allow utilities to approve large new loads far more quickly, with confidence that demand can be throttled in real time during rare but critical peak events.

Enabling low-voltage ride-through

Modern data centers are equipped with advanced energy management systems that can respond instantly to grid voltage dips or disturbances. By using onsite BESS, backup generators, and intelligent controls, they can maintain operations while supporting grid recovery-Figure 7. This ability to “ride through” low-voltage events, rather than tripping offline, helps stabilize grid frequency and voltage, which is becoming increasingly vital at regions with sensitive operations or where data centers represent large portions of load.

Figure 7: Peak shaving with BESS - Source: ABB

Providing fast-response services for reliability

Many data centers already operate as small-scale virtual power plants (VPPs). Their onsite BESS, UPS systems, and potentially thermal energy storage and cooling management assets can deliver grid services like frequency regulation, spinning reserve, and fast-ramping support. With proper orchestration, these facilities can shift load, inject or absorb power, and coordinate directly with utilities and RTOs to maintain system balance, often within seconds. This transforms them from being seen as burdens on the grid to being stabilizing partners that enhance reliability and resilience.

That shift isn’t only technical. It calls for coordination among utilities, grid operators, data center developers, technology vendors, and regulators, each with different goals and constraints. Utilities plan for reliability, operators for uptime, vendors for differentiation, and regulators for consumer protection. Acting alone, no one can unlock the full value of flexibility.

Real progress depends on shared incentives, open standards, and explicit rules that make flexibility scalable, adoptable, and fair. This coordination must occur across different sectors so that data centers and electric systems can grow together.

1.4.2 The critical insight: 350 days vs. 15 days

A 2024 study from the DOE shows that utilities can reliably serve data center demand for most of the year, around 350 days, and that its only the remaining 15 days, ~ 360 hours, that strain the grid.⁵³ This insight reframes the entire challenge. Flexibility isn’t about continuous curtailment or year-round operational sacrifice. It’s about targeted, high-value coordination during a handful of predictable, high-impact periods.⁵⁴

NV Energy case study

Using NV Energy as an example, Gridlab’s analysis demonstrated that data centers can deliver measurable grid value with minimal operational impact. Hourly heatmap modeling showed that curtailing 1 GW of load for only 0.5–1% of annual hours (roughly 500–880 hours), primarily during predictable summer evening peaks, can significantly reduce system costs and improve reliability-Figure 8. These results confirm that targeted, time-bound flexibility from data centers can serve as an effective non-wires alternative, accelerating interconnection and reducing the need for costly transmission upgrades.

Figure 8: NV Energy case study- Source: Gridlab.org

Figure 8: NV Energy case study - Source: Gridlab.org

ImpactECI-Dominion case study

The solution lies between activating flexibility only when it’s needed and supporting the grid during the toughest 15 days, while preserving uptime during the other 350 days.

Our analysis of Dominion Energy’s service territory in Virginia, one of the most data center dense regions in the U.S., shows just how much hidden headroom the grid already has. Looking at PJM’s Dominion Zone 2024 load data (acknowledging regional and circuit-level variations), the annual maximum demand occurs in only a handful of hours. When the data is viewed not as a single annual peak, but through quarterly, monthly, weekly, and hourly lenses, the picture changes dramatically.

At the annual scale, the grid appears “maxed out” (shown in deep red). But hour by hour, it reveals significant unused capacity, a gap between nameplate limits and real operational peaks-Figure 9. The real constraint, then, isn’t a lack of infrastructure; it’s a lack of flexibility.

We took the analysis a step further. By modeling what would happen if loads were briefly curtailed during those 50-350 peak hours, we found that the system could free up between 6 -17% of total capacity, equivalent to several GW of headroom, without building new transmission lines, power plants, or substations-Figure 10.

Figure 9: Dominion peak load heatmap with curtailment

Advances in forecasting, scheduling, and real-time control make this level of coordination entirely achievable. Grid operators can now anticipate stress events and give data centers time to pre-charge batteries, shift non-critical workloads, or migrate compute tasks to other regions. In doing so, data centers effectively become dispatchable reliability resources, partners that strengthen, rather than strain, the grid.

The next few years will be pivotal. Explosive digital demand, congested interconnection queues, and tightening climate goals are converging at once. To succeed, the industry must move beyond ad hoc pilots toward scalable, repeatable solutions: standardized contracts, shared data platforms, and aligned regulatory frameworks that work across utilities and markets.

1.5 The six-tier ecosystem model

The framework organizes stakeholders into six interconnected tiers, each with distinct roles and responsibilities in enabling data center flexibility at scale:

Figure 10: Six-tier ecosystems

Tier 1: Grid operators and utilities (the system coordinators)

These entities serve as the primary coordinators of system reliability and the source of flexibility signals. RTOs, such as PJM, ERCOT, and CAISO, provide market mechanisms and system-wide coordination. Distribution utilities manage local delivery and customer interfaces, often through substations that directly serve data center clusters. Instead of only providing passive consumption, both must adapt their operations to allow active participation by data center loads.

Tier 2: Aggregators and orchestration platforms (the translation and interface layer)

These services convert grid-level signals into site-specific actions across portfolios of data center assets. The software platforms that facilitate real-time coordination between grid requirements and data center operations are supplied by companies such as Emerald AI, Camus Energy, and Neuralwatt. This tier is important for scaling flexibility beyond individual pilot projects to systematic portfolio-level coordination.

Tier 3: Data center operators and developers (the base layer)

Data centers integrate flexibility capabilities into facility design, operations, and customer service delivery. This includes hyperscalers like Google, Microsoft, and Meta, as well as colocation providers and multi-tenant operators. Success requires aligning flexibility participation with customer service level agreements and business continuity requirements. Most importantly, they have the load and must make the decisions about optimization, risk, cost, etc. to protect their customers and reputations.

Tier 4: On-site flexibility assets (the physical enablers)

The physical basis of flexibility mainly consists of behind-the-meter assets. When integrated through automated control platforms, technologies like intelligent cooling systems, backup generators, and battery storage can react to grid events in seconds. The hardware and control systems that translate digital signals into quantifiable load adjustments are supplied by vendors such as Eaton, EnerSys, and FlexGen. This layer incorporates flexibility into physical infrastructure, allowing for both quick and long-term operational responses, in contrast to previous tiers that concentrated on coordination and policy.

Tier 5: Standards and community organizations (the scaling backbone)

Industry alliances and open-standards bodies ensure interoperability and prevent fragmentation. Initiatives (such as LF Energy, EPRI’s DCFlex, and the Open Compute Project) have shared protocols, certification frameworks, and performance benchmarks that enable competition and scalability. These standards convert isolated innovation into an ecosystem, ensuring that flexible solutions can interoperate across technologies, vendors, and jurisdictions.

Tier 6: Financial stakeholder (the enablers of capital)

Financial stakeholders provide the capital, assurance, and oversight that make data center flexibility financeable. This tier includes investors, lenders, insurers, auditors, and legal advisors who ensure projects are financially credible, insurable, and compliant with market and other standards. Their role turns technical flexibility into a bankable asset.

The transformation ahead

The next five years, but especially the next 18 months, are pivotal. Trends such as explosive digital demand, tight interconnection queues, and decarbonization targets are converging to make data center flexibility necessary. Another trend worth mentioning is the emerging concerns with regard to national security and economic security coming from the domestic AI industry. This will continue to drive demand growth requiring swift action.

For success, the industry must move towards systematic, scalable solutions. This can include standardized contracts, shared platforms, and consistent regulatory frameworks that work across utilities and markets. The leaders in this transition will shape how America powers its next generation of computing infrastructure.

The stakes are high. With successful flexibility deployment, the U.S. can expand its digital economy without building redundant infrastructure, keeping energy reliable and aligned with climate goals. The other side of the coin involves an infrastructure-first approach, which risks slowing innovation just as AI and advanced computing are becoming essential to national competitiveness.

Schedule of Future Chapter Releases

1. https://gridlab.org/wp-content/uploads/2025/03/GridLab-Report-Large-Loads-Interim-Report.pdf

2. https://www.renewableenergyworld.com/energy-business/new-project-development/a-fundamental-shift-centerpoint-sees-700-increase-in-data-center-interconnection-request-queue/

3. https://www.renewableenergyworld.com/power-grid/transmission/dominion-energy-serves-data-center-alley-heres-how-they-feel-about-the-surging-demand/

4. https://www.kjzz.org/business/2025-09-19/arizona-corporation-commission-is-considering-utility-rates-just-for-data-centers-it-could-take-more-than-a-year-to-implement

5. https://cleanpower.org/wp-content/uploads/gateway/2024/04/ACP-Pass-Permitting-Reform_Fact-Sheet.pdf

6. https://www.wrightlaw.com/wp-content/uploads/2024/01/Order-No-2023-Improvements-to-Generator-Interconnection-Procedures-and-Agreements.pdf

7. https://onlocationinc.com/news/2025/05/data-centers-and-the-next-wave-of-distributed-generation/

8. https://www.reuters.com/business/energy/rush-us-gas-plants-drives-up-costs-lead-times-2025-07-21/

9. https://www.datacenterdynamics.com/en/news/loudoun-county-data-center-market-share-drops-as-new-virginia-jurisdictions-rise/

10. https://www.utilitydive.com/news/data-center-activity-has-exploded-in-ercot-spiking-grid-reliability-risk/752780/

11. https://www.trgdatacenters.com/resource/texas-data-center-markets-are-booming/

12. https://www.datacenterfrontier.com/special-reports/article/11427209/tax-incentives-and-connectivity-drive-phoenix-data-center-market-growth

13. https://www.utilitydive.com/news/data-center-grid-reliability-residential-cost-aps-load-growth/732480/

14. https://www.latitudemedia.com/news/nvidia-and-oracle-tapped-this-startup-to-flex-a-phoenix-data-center/

15. https://www.latimes.com/environment/story/2025-09-23/data-centers-water-use-bill

16. https://about.bnef.com/insights/commodities/power-for-ai-easier-said-than-built/

17. https://www.sciencedirect.com/science/article/pii/S1364032125008329

18. https://www.nlyte.com/blog/data-center-rack-power-costs-a-condensed-analysis/

19. https://epoch.ai/blog/power-demands-of-frontier-ai-training

20. https://www.boydcorp.com/blog/energy-consumption-in-data-centers-air-versus-liquid-cooling.html

21. https://solartechonline.com/blog/how-much-electricity-data-center-use-guide/

22. https://www.wpr.org/wp-content/uploads/2024/06/3002028905_Powering-Intelligence_-Analyzing-Artificial-Intelligence-and-Data-Center-Energy-Consumption.pdf

23. https://www.kjzz.org/business/2025-07-03/phoenix-sets-new-rules-for-data-centers-including-where-they-can-go-and-how-noisy-they-can-be

24. https://www.powwr.com/blog/how-data-centers-are-driving-demand-growth-in-ercot

25. https://www.aep.com/news/stories/view/1338/ETT-Energizes-Last-of-Seven-CREZ-Transmission-Lines-in-West-Texas/

26. https://www.rff.org/publications/reports/how-long-does-it-take-national-environmental-policy-act-timelines-and-outcomes-for-clean-energy-projects/

27. https://nocapx2020.info/wp-content/uploads/2019/07/Transmission-Cost-Estimation-Guide-for-MTEP-2019337433.pdf

28. https://bendbulletin.com/2025/04/09/se-bend-residents-oppose-proposed-power-lines-through-neighborhood/

29. https://www.publicadvocates.cpuc.ca.gov/-/media/cal-advocates-website/files/press-room/reports-and-analyses/230612-caladvocates-transmission-development-timeline.pdf

30. https://www.utilitydive.com/news/pjm-data-center-transmission-costs-ratepayers/761579/

31. https://netl.doe.gov/sites/default/files/gas-turbine-handbook/1-1.pdf

32. https://www.datacenterfrontier.com/energy/article/55317213/utilities-race-to-meet-surging-data-center-demand-with-new-power-models

33. https://www.utilitydive.com/news/fossil-fuel-gas-coal-climate-data-centers/753565/

34. https://www.bloomberg.com/features/2025-bottlenecks-gas-turbines/

35. https://energyforgrowth.org/article/untangling-stranded-assets-and-carbon-lock-in/

36. https://trellis.net/article/google-funding-new-natural-gas-plant-outfitted-carbon-capture-storage/

37. https://www.datacenterdynamics.com/en/news/revised-gas-powered-300mw-data-center-campus-in-the-works-for-pittsylvania-county-virginia/

38. https://appvoices.org/2025/07/09/community-defeats-gas-plant-and-data-center-proposal/

39. https://www.selc.org/wp-content/uploads/2025/04/2025.04.12-Public-Health-Impacts-Analysis-Balico-Gas-Plant-FINAL-REPORT.pdf

40. https://www.selc.org/news/a-rural-virginia-county-is-a-case-study-in-community/

41. https://www.spglobal.com/commodity-insights/en/news-research/latest-news/electric-power/052025-us-gas-fired-turbine-wait-times-as-much-as-seven-years-costs-up-sharply

42. https://www.publicpower.org/periodical/article/new-report-finds-rising-cost-new-gas-plants-outpacing-planning-assumptions

43. Lazard LCOE+ 2025

44. https://www.reuters.com/business/energy/rush-us-gas-plants-drives-up-costs-lead-times-2025-07-21/

45. Plant size: reciprocating engine 5MW, CCGT 200MW, Heat rate: CCGT 6,150-6,900 btu/kWh, Reciprocating engine 7,400-8,800btu/kWh, gas price $3.27/mmbtu, average installed cost CCGT $2,000/kW, Recips $1,050-1,800/kW, O&M: CCGT $12-23/kW-year, Recips $25-28/kW-year, cost of debt 8%, cost of equity 12%, debt/equity ratio 60/40

46. https://www.utilitydive.com/news/demand-response-dr-utility-programs-resideo/754205/

47. https://www.energysage.com/electricity/demand-response-programs-explained/

48. https://www.aceee.org/blog/2017/02/demand-response-programs-can-reduce

49. https://www.jdsupra.com/legalnews/state-legislative-and-regulatory-8564473/

50. https://www.bracewell.com/resources/texas-senate-bill-6-ushers-in-major-overhaul-of-large-load-interconnection-and-grid-access-rules/

51. https://www.utilitydive.com/news/aep-ohio-data-center-crypto-rates-puc/716150/

52. https://efifoundation.org/wp-content/uploads/sites/3/2024/04/EF3-Regional-Planning-Cost-Allocation-Presentation-4.11.24.pdf

53. https://www.energy.gov/sites/default/files/2024-08/Powering%20AI%20and%20Data%20Center%20Infrastructure%20Recommendations%20July%202024.pdf

54. https://www.canarymedia.com/articles/utilities/one-way-data-centers-can-help-the-grid-by-being-flexible