The Data Center Boom Has Entered a New Phase: Power Is No Longer Just a Utility Issue It Is the Deal
The Data Center Boom Has Entered a New Phase: Power Is No Longer Just a Utility Issue It Is the Deal
By Data Center Resources | Data Center Insights Insights August 28, 2026
For much of the modern data center industry's history, real estate followed a recognizable formula.
Find land in the right market. Confirm fiber. Establish utility service. Navigate zoning and permitting. Build the facility. Lease capacity to customers.
That formula is being rewritten.
In 2026, the defining question surrounding many large-scale data center developments is no longer simply whether a site has enough acreage, favorable zoning, fiber connectivity or proximity to a major population center.
It is increasingly this:
Where will the power come from, how quickly can it be delivered, who will pay for the infrastructure required to deliver it, and will the surrounding community support the project once the consequences become clear?
Those questions are transforming data center development from a specialized real estate business into something much broader.
It is becoming an energy, infrastructure, regulatory, community-relations and capital-allocation business that happens to involve real estate.
That distinction matters.
According to the International Energy Agency, data centers consumed approximately 415 terawatt-hours of electricity globally in 2024, representing roughly 1.5% of global electricity consumption. The United States accounted for a significant share of that demand, and the IEA's base-case outlook projects worldwide data center electricity consumption could more than double by 2030.
The scale of individual facilities is changing just as dramatically.
The IEA has estimated that a typical AI-focused data center can consume as much electricity as roughly 100,000 households, while some of the largest campuses under development could consume many times that amount.
That is not incremental load in the traditional commercial real estate sense.
It is industrial-scale electricity demand.
When projects move into hundreds of megawatts — and proposed campuses contemplate gigawatt-scale requirements — the data center is no longer merely another customer connecting to an existing electrical system.
In some locations, the project becomes one of the factors around which the future electrical system itself must be planned.
That is the fundamental shift now taking place.
AI Didn't Create the Data Center Industry. It Changed Its Scale.
It is tempting to describe the current development cycle simply as an “AI data center boom.”
That description is incomplete.
Cloud computing, streaming, enterprise IT outsourcing, e-commerce, social media and digital services were already driving substantial data center demand before generative AI became a mainstream technology.
AI has added something different:
Extraordinary computational intensity and a new urgency around capacity.
The IEA has reported rapid growth in electricity consumption associated with AI-focused computing, alongside major increases in server power density.
That changes the physical infrastructure equation.
Higher-density computing affects electrical distribution, cooling, equipment requirements, redundancy strategies and ultimately the amount of power that must be concentrated at a particular location.
Meanwhile, capital continues pouring into the sector.
JLL has estimated that nearly 100 gigawatts of additional global data center capacity could be added between 2026 and 2030, requiring investment measured in the trillions of dollars.
Construction costs have also risen materially.
At that level of capital intensity, delays become enormously consequential.
A parcel that takes five years to energize can be fundamentally less valuable than another site where meaningful capacity can be delivered sooner.
And that is why the industry's traditional vocabulary is changing.
Developers once spoke primarily about location.
Today, they increasingly speak about speed-to-power.
The Most Valuable Commodity May Be Time
For decades, the old real estate maxim was “location, location, location.”
For the next generation of data center development, it may increasingly become:
Power, certainty and time.
CBRE's 2026 research has highlighted just how significantly power cost and delivery timelines are influencing site-selection decisions.
That is an extraordinary change for an industry whose location strategy historically placed tremendous weight on network infrastructure.
Fiber remains critical.
Latency remains critical for many workloads.
Redundancy remains critical.
But excellent fiber cannot energize servers.
The consequences are already visible in development schedules.
Projects that once could be contemplated on comparatively short construction timelines increasingly depend upon multi-year electrical infrastructure programs involving new substations, transmission upgrades, additional generation and long-lead equipment.
In some markets, power procurement timelines have become measured not in months, but in years.
Think about what that means for land valuation.
Two properties can sit twenty miles apart.
Both may have favorable topography.
Both may have access to fiber.
Both may have sufficient acreage.
Both may have political leadership interested in economic development.
Yet one may have a credible path to significant power within several years.
The other may require major transmission upgrades, new generation or an interconnection process whose outcome remains uncertain.
On a conventional land map, those parcels may look comparable.
To a sophisticated data center buyer, they may not be remotely comparable.
That leads to one of the most important principles in the current market:
Data center land should not be valued solely by what physically exists on the property. It must increasingly be evaluated by what infrastructure can realistically be delivered to it — and when.
“Power Nearby” Is Not the Same as Power Available
This distinction is frequently misunderstood outside the industry.
A transmission line running near a property does not necessarily mean hundreds of megawatts can be delivered to that property.
A nearby substation does not necessarily mean that substation has available capacity.
A utility territory experiencing generation growth does not automatically mean a requested load can be connected on the developer's preferred schedule.
There are multiple layers between theoretical electricity supply and usable power at a data center campus.
Generation must exist.
Transmission must move electricity across the system.
Substations must transform voltage.
Infrastructure must reach the property.
Equipment must be procured.
Interconnection studies must be completed.
Reliability requirements must be satisfied.
And upgrades may need to be approved, financed and constructed.
Then comes another critical question:
Who pays for all of it?
The economics of those improvements may need to be allocated among utilities, developers, large-load customers and other stakeholders.
This is precisely why federal regulators have become involved.
In June 2026, the Federal Energy Regulatory Commission took action concerning the rules governing how large customers, including data centers and major industrial loads, connect to the grid.
FERC's involvement demonstrates how significantly data center development has expanded beyond the boundaries of conventional commercial property development.
This is no longer just a real estate issue.
It is federal energy policy.
America's Electricity Demand Curve Is Changing
For years, U.S. electricity planners operated in an environment of relatively flat demand.
That era appears to be ending.
The U.S. Energy Information Administration has identified data centers as an important contributor to renewed electricity-demand growth.
Its long-term modeling indicates computing-related demand could become one of the dominant forces shaping future commercial electricity consumption.
The International Energy Agency reaches a similarly consequential conclusion.
Its projections call for substantial increases in U.S. data center electricity use through the end of the decade.
But national totals only tell part of the story.
This growth is not distributed evenly.
Data centers cluster.
Northern Virginia became the world's most prominent data center market partly because of its dense fiber ecosystem, established operators, customers and infrastructure.
Other major markets including Dallas-Fort Worth, Phoenix, Atlanta and Chicago have expanded substantially.
That concentration matters.
Electricity demand that may appear manageable when viewed across an entire country becomes a very different engineering problem when enormous new loads concentrate within a handful of utility territories and counties.
That is why the national discussion about what percentage of electricity data centers consume can obscure the more important local question:
What percentage of incremental demand is arriving in a particular utility territory, transmission zone or community?
Infrastructure is ultimately built locally.
The Grid and the Data Center Are Moving at Different Speeds
This may be the central tension facing the industry.
Technology companies can deploy computing infrastructure extraordinarily quickly.
The electrical system does not operate on the same timetable.
The IEA has noted that data centers can often be developed more quickly than the large-scale energy infrastructure required to serve them.
Transformers, switchgear, turbines and other critical components can face procurement constraints.
Transmission projects may require rights-of-way, permitting, engineering, public review and years of construction.
New generation takes time.
Large nuclear projects take considerably longer under conventional development models, although interest in small modular reactors and other technologies continues to grow.
This mismatch creates a race for existing capacity.
And existing capacity has scarcity value.
It helps explain why sites with credible near-term power are attracting extraordinary attention and why developers are willing to investigate locations that might once have been viewed as secondary markets.
The market is not simply expanding geographically because developers want cheaper land.
It is following electrons.
The Rise of “Bring Your Own Power”
If the grid cannot deliver quickly enough, the next question is obvious:
Can the data center bring generation to the site?
That question is driving one of the most consequential trends in digital infrastructure.
CBRE has identified accelerating interest in behind-the-meter power strategies as developers and operators look for ways to reduce dependence on extended utility delivery schedules.
Natural gas generation, renewable energy, battery storage, nuclear technologies and co-location with existing power generation are all part of the broader conversation.
But “bring your own power” should not be confused with an easy workaround.
On-site generation introduces an entirely new set of questions.
What fuel supply is available?
How firm is that supply?
What air permits are required?
Can equipment operate continuously or only as backup?
What are the emissions implications?
How will redundancy be designed?
What grid connection remains necessary?
Can generation scale with later phases of the campus?
Who owns the generating assets?
Who operates them?
How does the energy structure affect financing?
And what happens when a project initially designed around temporary generation must ultimately integrate with the larger grid?
The industry is therefore not eliminating the energy problem.
It is internalizing more of it into the development project.
Then Comes the Equipment Problem
Even if capital, land and generation are available, the physical equipment required to move and manage electricity cannot simply be summoned into existence.
AI infrastructure is increasing power density rapidly.
That puts pressure not only on generation capacity but on transformers, power electronics, switchgear and other electrical equipment.
This creates another subtle shift in project risk.
Historically, a real estate developer might control land, entitlements, financing and construction while treating utility service as a critical external dependency.
For very large data center projects, developers increasingly require visibility much deeper into the electrical supply chain.
A project can have an interested customer and still face equipment constraints.
It can have utility support and still encounter long-lead components.
It can have capital and still lack certainty of delivery.
Therefore:
A credible megawatt is not merely a number discussed with a utility. It is a chain of infrastructure commitments that must ultimately become physical equipment in the ground.
The Market Is Still Exceptionally Tight
None of these constraints mean demand is collapsing.
Quite the opposite.
CBRE's 2026 North American data center research continues to show exceptionally tight conditions across major markets.
Northern Virginia remains enormous.
Atlanta continues to expand rapidly.
Large portions of capacity under construction are already committed before completion.
That creates an unusual paradox:
Demand for data centers is enormous at precisely the moment that building them is becoming more difficult.
That creates value for projects capable of solving the constraints.
It also creates danger for speculative projects that merely claim they can.
The Next Data Center Gold Rush Could Be a Due-Diligence Trap
Whenever a specialized asset class becomes highly valuable, landowners and intermediaries begin attaching the industry's label to properties.
We have seen versions of this in logistics, solar, battery storage and other infrastructure sectors.
Data centers are particularly susceptible because the potential difference between ordinary land value and perceived “data center land” value can be enormous.
A 200-acre property is not a data center site simply because someone says it is.
Neither is a parcel beside a transmission line.
Neither is industrial-zoned land with fiber somewhere nearby.
Neither is a property accompanied by an aspirational claim that “500 MW is available.”
The more valuable powered land becomes, the more important verification becomes.
A serious diligence process should distinguish among:
Power that exists somewhere in the region
Power a utility believes could eventually be served
Power included in a planning study
Power associated with an interconnection request
Power requiring substantial upgrades
Power with a defined delivery schedule
Power supported by binding agreements
Those are not the same asset.
The market will increasingly price those distinctions.
In many cases, it already does.
Community Acceptance Has Become Infrastructure
There is another scarce resource emerging in the data center market.
It is not land.
It is not electricity.
It is social license to operate.
CBRE has identified community opposition and zoning delays as increasingly important obstacles to development.
That deserves attention.
For years, data centers could be presented as unusually attractive economic-development projects.
They involve substantial capital investment.
They generally produce less vehicle traffic than large logistics operations.
They can significantly expand local tax bases.
They support digital infrastructure that increasingly underpins almost every sector of the economy.
Those benefits remain.
But communities are asking more sophisticated questions.
How much electricity will the project consume?
Will residents subsidize electrical infrastructure?
How much water will cooling require?
What will backup generators mean for emissions and noise?
How visible will transmission infrastructure be?
How many permanent jobs will be created relative to the scale of investment?
What happens to neighboring property?
Will the project change the character of the community?
Those questions should not simply be dismissed as obstacles to development.
They are becoming part of development.
Water Is Becoming Part of the Conversation
Power dominates data center discussions for good reason, but water is increasingly entering site-selection and entitlement debates.
The amount required depends heavily on facility design, climate, cooling technology and operating strategy.
It is therefore misleading to apply a single water-consumption figure to every data center.
Some cooling architectures can dramatically reduce direct water consumption.
Others use water to improve thermal performance and efficiency.
What matters for development is understanding the site's actual design and the local water context.
Georgia provides a useful example.
During 2026, Georgia's regional water-planning process has included direct discussion of data center development, cooling technologies, water withdrawals and infrastructure impacts.
That does not mean data centers and water resources are inherently incompatible.
It means developers can no longer assume water will remain a secondary technical detail invisible to the public.
Water strategy is becoming part of entitlement strategy.
Georgia Is Becoming a National Case Study
Few markets illustrate the industry's transformation better than Georgia.
Atlanta has become one of North America's fastest-growing data center markets.
CBRE's 2026 research shows significant capacity under construction in the region, reflecting continued demand from hyperscale and digital infrastructure users.
But development interest is no longer confined to metropolitan Atlanta.
Communities farther across the state are beginning to confront the same questions around electricity, water, land use and economic development.
This week added another important development.
On August 26, Georgia Power announced regulatory approval of its contract to serve OpenAI's new project in Effingham County.
The significance extends beyond a single customer or project.
Georgia is also experimenting with the economic structure surrounding large-load growth.
Georgia Power has stated that rules approved through the Georgia Public Service Commission require new large-energy users, including data centers, to bear costs associated with serving them.
The utility has also projected customer savings associated with its growing large-load portfolio beginning later this decade.
Those projections will ultimately depend upon implementation and actual system economics.
But the broader policy question is more important.
The public discussion is evolving from:
“Should we allow data centers?”
toward:
“Under what economic structure should data centers connect, and who should bear the costs and benefits of the infrastructure they require?”
That is a much more mature question.
And it is likely to be asked across the country.
Ratepayer Protection Could Become a Competitive Advantage
Developers may initially view stricter large-load tariffs and financial requirements as additional costs.
There is another way to look at them.
If communities believe residential customers will subsidize data center infrastructure, opposition is predictable.
If regulators can demonstrate that large-load customers bear the incremental costs they create — and potentially produce broader system benefits — the political equation changes.
This suggests a counterintuitive possibility:
The markets with the clearest and most credible ratepayer protections may ultimately become easier places to develop.
Certainty has value.
Developers need to understand their costs.
Utilities need confidence that infrastructure investment will be recovered.
Regulators need to protect existing customers.
Communities need evidence that economic-development benefits are not being offset by higher household utility bills.
A transparent structure can serve all four constituencies better than ambiguity.
Reliability Cannot Be an Afterthought
The power challenge is not simply about generating enough annual megawatt-hours.
Electric systems must balance supply and demand continuously.
Data centers also have unusual reliability requirements.
AI workloads can introduce rapid changes in electrical demand.
Battery storage, on-site generation, sophisticated controls and demand-response mechanisms may therefore become increasingly important.
This summer provided a tangible example of the growing connection between data center growth and grid operations.
During periods of extreme heat, PJM confronted exceptionally high electricity demand and received emergency authority from the Department of Energy that included, as a last-resort measure, potential curtailment of certain large loads equipped with backup generation.
That deserves attention.
It demonstrates that the relationship between data centers and the grid is becoming operational — not merely theoretical.
The industry's future will therefore require not just more generation but more sophisticated coordination among data centers, utilities and regional grid operators.
A data center may eventually be evaluated not only by how much power it consumes, but by how intelligently it interacts with the electrical system around it.
The Grid Could Become a Two-Way Relationship
This is where the conversation becomes more interesting.
The simplistic narrative says data centers consume enormous quantities of electricity and therefore strain the grid.
There is truth in the first part.
But large, sophisticated energy users can potentially become resources as well as loads.
Consider what a major campus may possess:
Backup generation.
Battery storage.
Sophisticated energy-management software.
Predictable load characteristics.
Capital to finance generation.
Long-term electricity procurement agreements.
Potential flexibility for certain computational workloads.
These characteristics create opportunities.
Some computing tasks may eventually be shifted geographically or temporally depending on electricity availability.
Batteries can help manage power quality and peaks.
On-site generation can reduce dependence on constrained infrastructure.
Long-term energy contracts can support development of additional generation.
AI itself may help utilities and operators optimize energy systems.
The most successful next-generation projects may therefore stop thinking of electricity simply as a commodity purchased at the property line.
They will treat energy architecture as part of the data center product.
What This Means for Site Selection
The old data center site-selection checklist is no longer enough.
A credible evaluation should increasingly examine several layers simultaneously.
1. Land — Is there sufficient acreage for the entire campus, including substations, transmission corridors, generation, setbacks and cooling infrastructure?
2. Power — How much capacity can actually be delivered, at what voltage, through what infrastructure, at whose cost and by what date?
3. Generation — What resources exist locally, and can on-site generation, renewables or storage realistically supplement the grid?
4. Fiber — Are multiple carriers available, and can physically diverse routes be established?
5. Water & Cooling — What cooling architecture is contemplated, what resources does it require and what local limitations apply?
6. Entitlements — Is the proposed use permitted by right, or will zoning and other approvals create schedule risk?
7. Community — Has local leadership been engaged, and can the project's benefits and infrastructure impacts be credibly explained?
8. Schedule — What is actually on the critical path: power, equipment, transmission, entitlements, water or customer commitment?
9. Capital — Who funds infrastructure before occupancy, what deposits or guarantees are required, and who bears the risk if delivery slips?
10. Customer — Who actually needs the capacity, and when?
The final point may be the most important.
A technically viable project delivered after the customer's requirement can still be commercially unsuccessful.
The Definition of “Shovel Ready” Must Change
Economic-development organizations frequently use the phrase “shovel ready.”
For data centers, that phrase should be treated cautiously.
A cleared and entitled site without a credible power schedule may not be ready at all.
Likewise, a property with encouraging utility discussions but unresolved zoning may still carry substantial execution risk.
The next generation of truly marketable data center sites will require a much more rigorous definition.
A genuinely advanced site should be able to answer, with documentation rather than marketing language:
How much power?
From where?
Under what agreement?
By what date?
At what cost?
Through what infrastructure?
With what remaining approvals?
And with what expansion path?
The industry needs fewer claimed megawatts and more documented megawatts.
Not Every Announced Gigawatt Will Be Built
There is another reality worth acknowledging.
The current development pipeline contains enormous numbers.
Not all of it will become operational.
Projects compete for the same equipment, power, customers, capital and construction resources.
Some announced campuses are early-stage concepts.
Some generation proposals are speculative.
Some interconnection requests will never advance.
Some customers will change strategy.
Technology itself will continue to evolve.
PJM's forecasting work has already illustrated the need to distinguish between large-load requests and projects likely to materialize.
That is healthy.
Forecasting should distinguish between an inquiry and a project.
A request for 1 gigawatt is not the same thing as a contracted 1-gigawatt load.
A proposed 1-gigawatt campus is not the same thing as an operating campus.
And a generation pipeline is not the same thing as generation capacity.
For investors, utilities and communities alike, probability-weighted infrastructure analysis will become increasingly important.
The Winners Will Be Projects That Reduce Uncertainty
The current market understandably rewards scale.
But scale alone will not determine the winners.
The highest-value projects may be those that systematically remove uncertainty.
A site with documented land control, credible power, clear utility milestones, diverse fiber, an appropriate water strategy, community support, entitlements, equipment-procurement visibility, realistic construction scheduling and identifiable customer demand is fundamentally different from a site with a large acreage number and an ambitious presentation.
That difference will matter more as capital becomes increasingly sophisticated.
It will matter to lenders.
It will matter to institutional investors.
It will matter to hyperscalers.
It will matter to utilities.
And eventually it will matter to landowners deciding whether the headline price they were promised is actually achievable.
Data Centers Are Becoming Infrastructure Projects
This may be the most useful way to understand what is happening.
A modern AI campus should not be thought of simply as a large industrial building filled with servers.
At sufficient scale, it begins to resemble an infrastructure ecosystem.
There is the computing infrastructure.
The electrical infrastructure.
The telecommunications infrastructure.
The cooling infrastructure.
Potential generation infrastructure.
Potential natural-gas infrastructure.
Road improvements.
Water infrastructure.
Security infrastructure.
And the financial and regulatory architecture holding everything together.
Once viewed this way, many current industry developments become easier to understand.
Why are utilities central to real estate negotiations?
Because the energy system increasingly determines the development schedule.
Why are federal regulators involved?
Because large-load interconnection affects the broader electric system.
Why are governors and economic-development agencies paying attention?
Because these projects can involve billions of dollars of investment.
Why are communities organizing?
Because infrastructure impacts are local.
Why are hyperscalers exploring nuclear, natural gas, renewables and storage?
Because computing growth is becoming inseparable from energy strategy.
Why are developers searching outside traditional Tier I markets?
Because available power can create a market where one barely existed before.
The data center is becoming infrastructure.
The Next Great Data Center Markets May Be Created, Not Discovered
Northern Virginia did not become Northern Virginia by accident.
Its ecosystem compounded over decades.
Fiber attracted customers.
Customers attracted operators.
Operators attracted suppliers.
Infrastructure attracted more infrastructure.
The next generation of markets may develop differently.
Instead of starting with connectivity and building power around demand, some markets may begin with energy abundance and infrastructure capacity and build digital ecosystems around them.
That creates opportunities for areas with existing generation, transmission capacity, industrial infrastructure, available land, supportive governments, reasonable construction costs, fiber expansion potential and communities seeking new investment.
But there is an important caveat.
Cheap land does not compensate for a five-year infrastructure problem.
The winning secondary and tertiary markets will not simply be places where acreage is inexpensive.
They will be places capable of assembling the entire infrastructure package.
The Conversation Must Mature
The data center debate is becoming polarized.
On one side, supporters emphasize investment, technological leadership, AI competitiveness and economic development.
On the other, critics emphasize electricity costs, water consumption, emissions, land use and limited permanent employment relative to project scale.
Both sides can produce legitimate evidence.
The industry will not build durable public support by pretending the tradeoffs do not exist.
Nor should policymakers treat every data center as identical.
A project that pays for required infrastructure, minimizes water consumption, provides grid support and generates meaningful local tax revenue is not economically equivalent to a project whose infrastructure costs are socialized and whose local impacts are poorly managed.
Project design matters.
Tariff design matters.
Cooling technology matters.
Location matters.
Generation strategy matters.
Community engagement matters.
The industry should welcome that distinction.
Responsible projects benefit when the market differentiates them from poorly structured ones.
The Real Data Center Race Is an Infrastructure Race
AI is often described as a race for chips, models, talent and capital.
Increasingly, it is also a race for physical infrastructure.
The most advanced GPU is of little value if it cannot be energized.
A billion-dollar data center shell produces no computing capacity without electricity.
And a gigawatt-scale development plan has limited value if the grid cannot serve it on a commercially relevant schedule.
The United States has enormous advantages: deep capital markets, leading technology companies, significant energy resources, engineering expertise, available land and an established data center industry.
But those advantages do not eliminate physical constraints.
The Department of Energy continues to identify the need for additional transmission infrastructure as data centers, manufacturing and other large loads increase electricity demand.
FERC is examining the rules for connecting those loads.
Utilities are redesigning tariffs.
Developers are exploring private generation.
Communities are rewriting zoning rules.
Water planners are asking new questions.
Capital is chasing powered land.
All of those developments point in the same direction.
The bottleneck has moved upstream.
The race is no longer simply to construct more data center buildings.
It is to construct the energy and infrastructure ecosystem capable of supporting them.
What Happens Next
The next several years will likely produce some spectacular successes — and some equally spectacular miscalculations.
More capital will enter the sector.
More communities will receive enormous development proposals.
More land will be marketed as data-center-ready.
More utilities will create specialized large-load structures.
More developers will investigate behind-the-meter generation.
More projects will move into markets that barely appeared on institutional data center maps several years ago.
And more proposed developments will fail because one critical piece of the infrastructure stack could not be secured.
That is not evidence that the data center boom is ending.
It is evidence that the market is maturing.
Early in a cycle, demand rewards almost everyone with exposure.
Later, execution separates the winners.
The next phase of data center development will reward those who understand that megawatts, timelines, infrastructure, community acceptance and contractual certainty are interconnected assets.
Acreage matters.
Fiber matters.
Capital matters.
Customers matter.
But increasingly, none of them matter enough without a credible path to power.
That may be the defining lesson of the AI infrastructure era.
The most valuable data center site is not necessarily the property with the most acreage, the closest fiber or even the largest theoretical power allocation.
It is the site where power, infrastructure, approvals, capital and customer demand can converge at the same time — with enough certainty to actually build.
The next generation of digital infrastructure will be created at that intersection.
And the competition to control it has already begun.
Data Center Resources | DCR Insights
Tracking the infrastructure, real estate, energy and development forces shaping the next generation of data centers.
Research referenced: International Energy Agency (IEA), U.S. Energy Information Administration (EIA), Federal Energy Regulatory Commission (FERC), U.S. Department of Energy, CBRE, JLL, PJM Interconnection, Georgia Power and Georgia's Regional Water Planning Councils.
