Why Do We Need So Many Data Centers?

A question has been bouncing around 𝕏 lately: “Why do we need so many data centers?”

It is a fair question. Most people do not interact with a data center directly. They do not walk into one, buy something off a shelf, or see a truckload of finished products roll out the back door. From the outside, a data center can look like a giant, windowless box that consumes electricity, uses land, and creates headaches for the local grid.

So the confusion is understandable.

But the common framing is badly incomplete.

The real answer is not “because AI companies want toys.” AI is certainly a major accelerator. It is changing the scale, power density, and economics of the buildout. But data centers are not new, and they are not merely an AI fad. They are the physical plant of the digital economy.

That sounds abstract, so let’s make it concrete.

The Internet Is Not “In the Cloud”

The phrase “the cloud” is useful shorthand, but it hides the reality.

Your photos are not floating in the mist. Your bank balance is not stored in the ether. Your email, tax records, medical portal, streaming library, maps, business documents, payroll system, logistics tracking, and video calls all live somewhere.

They live on servers. Those servers sit in racks. Those racks sit in buildings. Those buildings require electricity, cooling, fiber connections, physical security, backup power, fire suppression, monitoring, maintenance, and redundancy.

That building is a data center.

A data center is not just a warehouse full of computers. It is a purpose-built facility designed to keep digital services running continuously. If one piece of hardware fails, another takes over. If one network path fails, traffic reroutes. If power drops, backup systems keep the servers alive. If demand spikes, capacity scales.

That reliability is why you can tap your phone at 10:30 at night, check your bank account, send a file, watch a movie, order medicine, file paperwork, join a video meeting, or track a package, and expect the system to work.

The magic is not magic. It is infrastructure.

Data Centers Already Run Through Daily Life

Consider an ordinary day.

You wake up and check the weather. That forecast was generated, distributed, cached, and served through computing infrastructure.

You check email. Somewhere, servers authenticate your device, retrieve messages, filter spam, scan attachments, synchronize folders, and replicate data across regions.

You open a banking app. The transaction is authenticated, encrypted, fraud-checked, logged, routed, and reconciled. That is not happening inside the glass rectangle in your hand.

You drive somewhere using a map app. That service pulls traffic data, GPS information, routing algorithms, business listings, and live updates.

You buy something online. The system checks inventory, prices the item, processes payment, triggers fraud detection, updates fulfillment software, schedules logistics, and pushes shipping notifications.

Your doctor’s office uses electronic health records. Your pharmacy checks prescriptions electronically. Your insurance claim is processed through digital systems. Medical imaging is stored and retrieved from large data systems.

Your employer, bank, church, school, county government, utility provider, airline, grocery store, and phone carrier all depend on remote computing systems.

Even many “physical” businesses now depend on digital infrastructure. Modern farming uses sensors, satellite data, weather models, equipment telemetry, and logistics platforms. Manufacturing uses cloud-based design tools, supply-chain systems, inventory management, predictive maintenance, and quality analytics. Retail uses payments, point-of-sale systems, targeted inventory, security cameras, and customer data. Trucking and rail networks use routing, dispatch, maintenance, and freight visibility software.

This is why the “why do we need all these?” question often starts from the wrong baseline.

We are not choosing between a digital economy and some untouched 1995 alternative. We already live inside the digital economy. Data centers are the physical infrastructure that makes it function.

This Is Bigger Than AI

The current debate often treats data centers as if they suddenly appeared because ChatGPT became popular. That misses the history.

Data-center growth began long before generative AI. The commercial Internet, e-commerce, streaming video, smartphones, enterprise cloud computing, social media, cybersecurity, telemedicine, digital payments, remote work, and software-as-a-service all pushed more computing into centralized facilities.

The Bureau of Economic Analysis describes the digital economy in terms of digital infrastructure, e-commerce, priced digital services such as cloud services and internet/data services, and federal nondefense digital services [1]. That is a broad footprint. It is not just consumer apps, and it is not just AI.

Cloud services alone grew extraordinarily fast before the present AI boom. BEA found that cloud services grew 232.1 percent from 2017 to 2022, an average annual growth rate of 27.2 percent [1]. In other words, the demand curve was already steep.

AI made the slope steeper.

The International Energy Agency estimated that data centers consumed about 415 terawatt-hours globally in 2024, about 1.5 percent of world electricity use. The United States accounted for roughly 45 percent of that global data-center electricity consumption [2]. IEA projects global data-center electricity consumption to more than double to around 945 TWh by 2030, with AI as the most important driver, but not the only one [2].

That distinction matters.

AI is not replacing the traditional digital economy. It is being layered on top of it.

The Electricity Problem Is Real

None of this means critics are wrong to worry about electricity demand.

They are right to worry.

The Department of Energy reported that U.S. data centers consumed about 176 TWh in 2023, or about 4.4 percent of total U.S. electricity use. DOE’s Lawrence Berkeley National Laboratory projects that U.S. data-center electricity use could rise to 325 to 580 TWh by 2028, or roughly 6.7 to 12 percent of total U.S. electricity use [3].

That is not a rounding error.

It is a major planning problem for utilities, regulators, grid operators, communities, and the data-center industry itself. It affects generation, transmission, transformers, substations, land use, water planning, air permitting, rate structures, and local politics.

This is where the debate should get more serious. Prosperity is proportional to available energy. Addition of capacity MUST be part of the discussion.

Saying “data centers use power” is true but incomplete. So do factories, hospitals, airports, military bases, refineries, steel mills, semiconductor fabs, and cities. The question is not whether a modern economy uses energy. It does. The question is whether the energy use supports something valuable enough to justify the infrastructure, and whether the costs are assigned honestly.

That is the right debate.

A bad debate says, “Data centers use electricity, therefore stop building them.”

A serious debate asks:

Where should they be built?
Who pays for grid upgrades?
Are residential customers protected from unfair cost shifting?
Are water and noise impacts being managed locally?
Are utilities allowed to phase in large loads responsibly?
Are data centers using efficient cooling and modern power-management designs?
Are we building enough generation and transmission to support the economy we are creating?

That is the grown-up conversation.

Capacity Is Already Tight

If these buildings were speculative boondoggles, we would expect to see high vacancy, falling rents, and weak leasing.

Instead, the market is tight.

CBRE reported that primary North American data-center supply reached a record 8,155 megawatts in the first half of 2025, up 43.4 percent year over year. Yet vacancy still fell to a record-low 1.6 percent. CBRE also reported that 74.3 percent of under-construction capacity was already preleased, and that power availability and infrastructure delivery timelines had become the decisive site-selection factors across major U.S. markets [4].

That tells us something important.

The buildout is not occurring because a few executives thought giant server farms would look good on a PowerPoint slide. Customers are reserving capacity years ahead of delivery because they expect to need it.

That does not prove every proposed project is wise. Some projects will be badly sited. Some local incentive deals may be too generous. Some utilities may overbuild. Some communities may get stuck with poor land-use decisions.

But the broad demand is not imaginary.

The Economic Footprint Is Not Small

Another misconception is that data centers are economically trivial because a single completed facility may not employ thousands of permanent workers.

There is a grain of truth there. A data center is not a 1950s auto plant. Once operational, a single building may employ dozens to hundreds of people, not thousands.

But stopping the analysis there is misleading.

The economic footprint includes land acquisition, site work, construction, electrical trades, mechanical systems, fiber, substations, switchgear, transformers, backup power systems, cooling systems, software, cybersecurity, operations, maintenance, professional services, tax revenue, and the companies that use the computing capacity.

A 2025 PwC study commissioned by the Data Center Coalition estimated that the U.S. data-center industry supported 4.7 million jobs in total in 2023, up from 2.9 million in 2017. It estimated the industry’s total annual contribution to U.S. GDP at $727 billion in 2023, up from $355 billion in 2017, and total fiscal support to federal, state, and local governments at $162.7 billion in 2023 [5].

Those figures come from an industry-commissioned study, so they should be read with the normal caution applied to economic impact studies. But they are directionally consistent with the obvious reality: the data-center sector is now a major part of construction, power infrastructure, telecom, cloud computing, and technology supply chains.

A more cautious way to state it is this:

Data centers may not look labor-intensive once the doors close, but they are capital-intensive, supply-chain-intensive, construction-intensive, grid-intensive, and economy-enabling.

That combination matters.

Be Careful With the “92% of GDP Growth” Claim

One striking claim circulating in this debate is that data centers accounted for roughly 92 percent of U.S. GDP growth in the first half of 2025.

That claim needs careful handling.

Harvard economist Jason Furman estimated that investment in information-processing equipment and software represented only about 4 percent of GDP but accounted for about 92 percent of GDP growth in the first half of 2025; excluding those categories, growth would have been close to flat [6].

That is important. It shows how much recent growth has been tied to the broader compute and software investment boom.

But it should not be simplified into “data centers alone created 92 percent of GDP growth.” The category is broader than data centers. It includes information-processing equipment and software. Also, GDP accounting can be tricky because imported chips and equipment may offset some of the domestic growth effect.

So the honest version is stronger than the sloppy version:

The compute buildout has become large enough to show up in national economic growth statistics. That does not make it magic. It does make it material.

Data Centers Are Also Strategic Infrastructure

This conversation cannot be reduced to consumer convenience.

Data centers support financial systems, healthcare systems, emergency services, logistics, defense contractors, research labs, universities, energy companies, transportation networks, and government operations.

They also matter in the strategic competition over AI, cyber, cloud platforms, semiconductor ecosystems, and military-relevant computing.

A country that cannot build enough compute capacity will not simply enjoy a quieter countryside and lower electric bills. It will import more digital capability from others, lose industrial leverage, slow innovation, and become more dependent on foreign-controlled infrastructure.

That matters for the United States.

It matters for AI, but also for cybersecurity, scientific computing, space systems, manufacturing, finance, logistics, missile defense, intelligence analysis, and critical infrastructure operations.

Compute is not the whole of national power. But in the 21st century, it is becoming one of the load-bearing beams.

The Local Objections Are Not Fake

At the same time, local objections should not be dismissed as ignorant NIMBYism.

Some communities have legitimate concerns.

Virginia’s Joint Legislative Audit and Review Commission, looking at the largest data-center market in the world, found that data centers provide positive economic benefits, mostly during construction, but also that the industry is forecast to drive immense increases in energy demand. JLARC warned that meeting unconstrained data-center demand would be very difficult and that rising demand could increase costs for non-data-center customers if not handled properly [7].

That is the kind of finding both sides should take seriously.

A community has every right to ask whether a proposed facility is properly sited, whether it is too close to homes, whether noise standards are adequate, whether water use is appropriate, whether backup generators are properly regulated, whether tax incentives are excessive, and whether the utility has protected ordinary ratepayers. Proper zoning is a key part of communities coexisting with data centers.

Those are not anti-technology questions.

They are governance questions.

If the industry wants public acceptance, it should welcome those questions and answer them with data, transparent cost allocation, better designs, and enforceable rules.

The Answer Is Not “Build Nothing”

The wrong answer is to pretend the digital economy can keep expanding while the physical infrastructure stays frozen.

That is fantasy.

If construction is artificially slowed across the board, the consequences are predictable:

Cloud capacity gets tighter. Costs rise. Smaller companies pay more. AI tools become more expensive. Streaming, storage, cybersecurity, analytics, and enterprise software face constraints. New services roll out more slowly. Utilities face more chaotic interconnection queues. Communities lose the chance to shape projects before demand becomes urgent. Strategic competitors keep building.

Bottlenecks do not eliminate demand. They ration it.

The better answer is disciplined expansion.

Build data centers where the grid can support them. Make large-load customers pay appropriately for the infrastructure they require. Require realistic interconnection studies. Protect residential ratepayers. Encourage efficient cooling and water practices. Update zoning rules. Build more generation. Build more transmission. Use nuclear, natural gas, geothermal, renewables, storage, demand response, and whatever else can reliably meet the load without pretending one favored technology solves everything.

The data-center debate should become an energy-abundance debate, a grid-modernization debate, a permitting-reform debate, and a local-governance debate.

It should not become a morality play where every data center is either a heroic temple of innovation or a villainous box of wasted electrons.

Reality is more complicated and more interesting.

The Simpler Answer

So, why do we need so many data centers?

Because almost the entire economy already runs through them.

Because “the cloud” is made of buildings, servers, power systems, cooling systems, fiber, and people.

Because banking, logistics, healthcare, retail, manufacturing, government, entertainment, education, research, cybersecurity, and national defense already depend on computing infrastructure.

Because AI is not the only driver, but it is adding a new layer of demand on top of an already digitized economy.

Because the alternative to building enough compute is not a return to a calmer, simpler world. It is higher costs, slower services, weaker innovation, and greater dependence on places that did build.

But “we need data centers” is not the same as “approve every project anywhere under any terms.”

The right question is not whether America needs data centers.

It does.

The right question is whether we are serious enough to build the power, grid, permitting, zoning, cooling, and cost-allocation systems required to support them responsibly.

If we want a modern economy, we need modern infrastructure.

Data centers are part of that infrastructure now.

Ignoring that will not make the demand go away. It will only make the bottlenecks worse.

[Disclaimer: Article was proof-read by AI, which did an admirable job of cleaning up my clumsy draft & double checked the references for me.]

References

[1] U.S. Bureau of Economic Analysis, “U.S. Digital Economy: New and Revised Estimates, 2017–2022,” December 2023.

https://apps.bea.gov/scb/issues/2023/12-december/1223-digital-economy.htm

[2] International Energy Agency, “Energy and AI,” Executive Summary, 2025.

https://www.iea.org/reports/energy-and-ai/executive-summary

[3] U.S. Department of Energy, “DOE Releases New Report Evaluating Increase in Electricity Demand from Data Centers,” December 20, 2024; citing Lawrence Berkeley National Laboratory’s 2024 U.S. data-center energy report.

https://www.energy.gov/articles/doe-releases-new-report-evaluating-increase-electricity-demand-data-centers

[4] CBRE, “North America Data Center Trends H1 2025,” September 2025.

https://www.cbre.com/insights/reports/north-america-data-center-trends-h1-2025

[5] Data Center Coalition / Center of Your Digital World, “2025 PwC Report: Economic Contributions of U.S. Data Centers, 2017–2023.”

https://www.centerofyourdigitalworld.org/2025-impact-study

[6] Fortune, “Without data centers, GDP growth was 0.1% in the first half of 2025, Harvard economist says,” October 7, 2025.

https://fortune.com/2025/10/07/data-centers-gdp-growth-zero-first-half-2025-jason-furman-harvard-economist/

[7] Virginia Joint Legislative Audit and Review Commission, “Data Centers in Virginia,” 2024.

https://jlarc.virginia.gov/landing-2024-data-centers-in-virginia.asp