Data center power considerations: Host community perspectives

The United States is engaged in a competition for AI dominance that will dramatically impact the economic, national security, and global soft and hard power leadership outcomes for decades.

The rapid growth of AI is transforming data center development and placing new demands on power, water, and local infrastructure. For host communities, understanding these shifts is critical to capturing economic benefits while managing long-term costs and risks.

Strategic planning helps communities evaluate opportunities with confidence. Learn more about the following below:

• Data centers: what changed?
• 4 data center scaling options
• Understanding host community readiness
• Power costs and data center impacts
• A host’s perspective

Understanding host community readiness

For host communities, data centers create both opportunities and challenges. Understanding the type of data center, the suitability of existing infrastructure, the project’s financial viability, and the project’s timeline are all crucial to screening offers and negotiating cost sharing, incentives, and long term commitments. Several key questions include:

1. Is the proposed project financially viable?
2. Is there a plan for power?
3. Is there a plan for water?
4. How do we accommodate the transient construction labor in our community?
5. How do we balance long term jobs and tax incentives?
6. Can data centers be attracted while maintaining energy affordability?

We address each of these in detail below.

Project financial and technical viability

Data center projects are fundamentally real estate projects with unique construction and operational characteristics. They can be unique in their resource intensity, specifically water and power, and their success is very dependent on supply chains for key internal and energy components that can demand extensive upfront payments.

Determining viability requires an evaluation of the developer’s history, the commitments they have for occupancy and offtake, the credit quality of the developer and the offtaker, and the supply chain reservations and financial investments they’ve made.

Host communities should be prepared to support a disciplined diligence process that includes infrastructure and construction issues as well as financial modeling that reflects their obligations for investment and bonding, the implications of tax abatements and other incentives, and the risk adjusted revenue impact of the proposed project.

The power plan

Power is the key constraint on data centers of all sizes. AI workloads may be specialized today, but AI is being built into every platform application from email to video editing and travel reservations—and the future is more intense.

These types of power loads are different than those for which the power grid was designed , and their impact on connected grids can be enormously expensive to accommodate and actually decrease overall grid reliability.

A large data center can exceed the load of an aluminum smelter or steel mill, require very clean power, meaning consistent voltage, and are not amenable to curtailment or other service interruption without extensive on-site storage or backup generation. Their power demand can also swing widely, with drops and spikes of 15% in milliseconds.

For these reasons, as well as multi-year delays and tremendous costs associated with getting large scale power, state and local regulations on CO2 and concerns around affordability and cost transfer to consumers many data center projects want to or are being forced to bring their own power. General reindustrialization is having the same struggle.

These solutions generally use natural gas combined with fast response turbines and batteries in standalone microgrids. They are expensive and complicated to build and operate, essentially becoming standalone power islands  with a single customer, but, when done correctly, can be built with no additional costs to local power ratepayers.

Comparing C&I to small data center electric load

For data center projects that assume they can access the local power grid for all or part of their needs, careful study is required of their power plan with a focus on their natural gas supply in both quantity and contractual security, their ability to be curtailed, the number of on-site hours of fuel for backup generators, and the power conditioning and stabilization system being used.

A detailed interconnection study should be conducted and reviewed by the host diligence team that reflects the totality of their impact including substations and switchgear, line rating under stress nodal congestion, circuit looping, and any N+1 power requirements needed to meet uptime expectations.

Finally, host communities need to carefully evaluate any assumptions for natural gas supply, including assumptions of municipal gas supply availability, city-gate expansion, new pipelines, and new or expanded rights of way. Projects should evaluate emissions from both primary power and secondary backup, including diesels, based on expected annual run hours.

Financial models should be evaluated for assumed power curtailment and related indemnifications that can be a burden on supporting utilities or on the facilities’ operating economics.

Host communities, working with local distribution utilities, may be able to negotiate access to some of the facility power under certain circumstances to improve local grid reliability and resilience and generally will want to provide design input into the use of the facility waste heat, whether to operate chillers or generate clean power and reduce the impact of creating large local heat islands.

The water plan

Data center water consumption varies extensively based on the project’s cooling design. Edge data centers are still generally built using an air cooled model, circulating chilled air through the facility. Next generation edge AI systems are adopting liquid cooling to reduce energy costs and increase energy density per square foot, which improves economic performance.

For true water cooled chilled water plant designs, initial fill can average approximately 6,000 gallons per MW, with annual replacement running <1% of fill volume, while on-chip cooling solutions can require roughly 3,000 gallons per MW and immersive cooling using synthetic oils can require none.

While not small, these are generally manageable resource burdens for communities. So where is the water challenge? Power.

100 MW generation at 100% capacity for a 24-hour period

Data center power can consume tremendous amounts of water depending design and size, with a 100 MW combined cycle gas turbine (CCGT) using as much water as a town of 8,000–10,000 people per day.

Even if the local utility is providing power, this can be a burden on regional water resources, especially in markets where water is sourced from aquifers that can take generations to recharge as opposed to surface water.

Some generation assets can run dry, however if a host community requires this, or has an automatic water shut off in drought conditions, this can decrease generation asset power output and efficiency, driving higher emissions per running hour and greater fuel consumption.

Understanding the asset mix proposed for on-site data center generation either stand alone or with grid power is critical to managing the impact on this local resource.

Employment impacts

A typical 100 MW data center can require between 1,200–1,500 construction jobs at peak including a massive number of skilled electricians and pipefitters to handle the complex energy and cooling platforms across a 12–18-month period, with an estimated 150 full time jobs during its operation.

Many regions need to import some or all the construction labor, creating a challenge in the spike of transient residents in the area requiring food, shelter, and medical care—and burdening water and wastewater, power, traffic, and public safety. Large hyperscale projects can also burden the local school system as construction employees relocate their families to the region for the project’s duration.

Host communities need to be proactive in working with data center developers to address these issues, including the identification of project specific burdens and the development of a cost recovery mechanism, for example hiring more police officers, reinforcement or replacement of roads and bridges carrying large construction loads, or creating camps with private buses to and from the job site to alleviate traffic congestion.

It’s critical host communities understand the limits of their infrastructure and creativity in their development to avoid over-building and abandonment. By working collaboratively, many host communities can use the construction period to reprioritize facility rebuild, modernization and extension plans, and even capture in kind funding from project developers to cover some costs.

Host communities need to be careful not to burden projects with shopping lists for deferred maintenance or other municipal agendas by relating each ask to a true impact, for example building a new fire station near a project site or replacing water mains to support the new facilities.

Developers have lots of options when it comes to locations, so host communities that have solid, rational business cases for their needs and fairly recognize the long-term tax benefits generally have a faster path to deal closure, lower transaction costs, and improved benefit realization timelines.

Balancing long-term jobs and tax incentives

Developers want tax incentives; communities want jobs and a reliable tax base. Data center developers are experts at pitting jurisdictions against each other to extract tax incentives that can significantly change their lifetime economics.

What do communities get in return? That of course depends on the jurisdiction. Is the primary tax basis property or income taxes? How do state revenues flow back down to host jurisdictions, or are data centers even eligible for tax incentives including sales/use, property, and income?

The scale of data centers can be so large—1.2 GW costs $3.9 billion for the shell and inside equipment—that normal incentives often get set aside and developers want to engage in custom crafting an incentives package that reflects the pace of capital spend, the timeline for benefits impact in taxes, jobs, and local spending, and the long term operating model that can include capital refresh every four to five years.

For most host communities, recognizing the jobs benefits, both long-term and short-term, is the toughest issue. Data center construction job ratios are well known based on the type, the scale, the reliability design, and the internal computing equipment.

A typical 100 MW data center may provide 2,400 job years of direct employment or $190 million in construction labor. During operations, the data center employment falls dramatically with an estimated 150–200 on-site full-time equivalent jobs, ranging from power management teams to HVAC techs, IT professionals, and security guards.

While long term jobs for the facility are limited, successful host communities build ecosystems around them to attract and retain a range of jobs in areas including tech startups that are building technology solutions, manufacturers who can benefit from local investments in power grid resilience and reliability, cloud or AI tenants who can benefit from being closer to the point of compute, and managed IT and cybersecurity firms serving other tenants.

For economic development teams, building these eco-system linkages including, for example, preferencing a local services provider or creating local vendor development programs can create a base of long-term employment and tax revenues. Ultimately, data centers at any scale can represent a long-term tax base contributor and employer for the host community.

Energy affordability

Most power grids in the US are simply not well suited to support data center loads, and in fact data centers can break normal utility infrastructures. Many US grids were already showing the strain of decades of underinvestment in aging equipment and increasing capital and labor costs while accelerated transitions to utility scale and low density renewables created a need for significant transmission to connect these facilities to areas of growing consumption.

The COVID-19 pandemic and the move to a more distributed work model for some created a need for higher reliability and resilience, which was challenged by weather events and under-investment in automation.

The trend of US national energy policy beginning in 2008 was to flatten and ultimately reduce national energy use. This created a planning assumption that didn’t invest in grid capacity over time, which would normally spread the economic impact over years, or even decades.

Total grid electricity consumption in 2020 was actually lower than in 2010, 3,717 TWh vs 3,754. During this decade, planners responded to mandates to reduce electricity demand and harden electrical infrastructure.

Increased energy efficiency largely held consumption flat while increasing renewable and related investments and aged plant retirements created an upward push on energy cost for consumers partially offset by federal subsidy and grant programs including the Inflation Reduction Act of 2022. Most of these have now disappeared.

This no growth trend is already reversing for several reasons, including:

• Reshoring of industry as the US rebuilds its critical industrial and manufacturing base.
• Heat pump mandates are moving winter heating from gas to an electric load including the use of highly inefficient electric heat strips in cold conditions.
• CO2 mandates are driving the replacement of industrial boilers and heating elements with electric facilities.
• Central air conditioning adoption increased from approximately 77% in 2001, to 88% in 2020. While most prevalent in the South and Midwest, regions like San Francisco have seen a nearly 40% increase in adoption over the last decade in response to change in the regional climate as well as concerns about respiratory health in the wake of large fires.

Add data center growth to this and the previously forecasted baseline around which utilities and regulators have made investment decisions is no longer valid. The US is shifting from a flat energy growth rate in the early 2000s, and data centers will make a definitive impact on communities over the next 25 years as they grow from an estimated 5.27% of total US grid load in 2025, to an estimated 17.19% by 2035.

While companies like Nvidia and AMD are working to improve the energy efficiency of their platforms, at the same time the use of these systems is expected to skyrocket, possibly slowing the rate of growth of power consumption less than forecast, but demand will grow and will create cost pressure for grid power as utilities compete with each other and well-funded behind-the-meter multi-GW data center and industrial developments for equipment and fuel, including natural gas and nuclear units.

The challenge of data center power for host communities is simple: do we support these facilities in our areas and what are the consequences for affordability on both residential and existing commercial rate payers?

Power costs and data center impacts

While there’s a lot of discussion about the economic impact of data centers, the reality is that today’s high electrical costs in parts of the United States haven’t been caused solely by utility investment in data center supporting infrastructure, but also reflect years of regional policy decisions.

Several regions, specifically Texas (ERCOT) and parts of the Midwest (MISO) have historically had energy costs below the national average, in large part due to their use natural gas and their continued use of coal even as wind and solar assets were introduced.

Within PJM, data centers and other loads coupled with policy decisions on generation and transmission are driving up prices as the capacity market surges to accommodate substantial increases in data center load predominantly in northern Virginia. Other regions have made different decisions and those are reflected in their costs.

A host’s perspective

So, for communities being approached by or looking to attract data centers, what should you be aware of and what should you do? Some emerging leading practices include:

1. Appreciate the developer has options. Most developers have a list of regional sites and generally are working several in parallel. The higher the friction with a community, the lower the probability that a deal gets done.
2. Be ready to move quickly. Communities should know if they’re interested in being facility hosts and understand their constraints, including public receptivity to avoid long, drawn out, and expensive negotiations that result in public costs and no benefits.
3. Know your infrastructure limits. Data center developers have teams of experts that understand the general capacity of a site and the possibilities of its expansion, so be prepared with your own data and costs. Make sure you have an equally skilled team on your side of the table including economic modeling, data center design, regulatory and permitting, and diverse power and water solutions skills to shape a solution with them.
4. Separate real costs from other public needs. Private developers may have the willingness to build some public infrastructure needed to support their project, so focus on those items and build a solid case for their direct relationship to the project.
5. Be careful with economic expectations and managing incentives. Have an independent economic modeler and negotiator on your team who understands this level of financial, tax, and operating complexity.
6. Review the transitional housing plan. While having several thousand guest workers is a short-term boom, they impose significant burdens on communities in traffic, housing costs, short-term school attendance spikes, service personnel costs, and infrastructure burden. Make sure the developer has a plan for dealing with these and related issues.
7. Build your team. Data center developers are sophisticated companies with teams of architects, professional engineers, energy professionals, and project management professionals. Make sure your team has the matching capabilities at scale.