Stakeholder Perspectives and Communication Failures on Data Center Development

Independent Study by Prathyaj Mantha, Nomad Futurist Foundation Fellow & MBA Student at University of Virginia Darden Business School

Faculty Advisor: Prof. Steve Soltis 

Abstract

The United States data center buildout is causing major controversy due to the communication breakdown between different stakeholders. This study examines why. Drawing on 53 structured interviews and survey responses across ten stakeholder categories: host community residents, data center developers, investors, environmental organizations, university researchers, utilities, media, regulators, policy makers, and communications professionals. This study seeks to explore the different stakeholder perspectives, shortage of information, the mismatch in what each group believes the conversation is about and core misunderstandings. The study seeks to understand if it is a structural problem or a messaging error by the data center developers. This study will analyze the data and the potential solutions or at least first steps towards productive common dialogue between the different stakeholder groups.

1. Introduction

Data center development has become one of the most contested categories of American infrastructure in the recent years. The underlying demand is not in dispute: computing requirements associated with artificial intelligence and cloud services have grown sharply, and that growth has driven the demand for data center build up. What is in dispute is nearly everything else; whether the facilities belong near homes and farmland, NIMBYism, who absorbs the cost of the electricity and water they consume, whether the promised economic benefits in fact benefit the community they are housed in and if they are necessary to even be on American soil.

Industry participants typically frame these disputes as communication problems: opposition stems from misunderstanding, and better explanation would resolve it. The interviews conducted here underline the fundamental problem. There is a vastly different perspective of the issue surrounding AI and the economic growth it promises. It is entirely different for different stakeholders.

This paper therefore treats communication failure as the object of study rather than the explanation, asking what each group believes, what each believes about others, and where those beliefs are structurally incompatible.

1.1 Research questions

• How does each stakeholder group characterize the data center build-out, and how does each characterize the motives of the others?
• Which disagreements are empirical- resolvable with better data and which are disagreements about values?
• What role do regulators, policy makers, and media play in widening or narrowing these gaps?
• Which interventions plausibly change outcomes?

1.2 Scope and limitations

This is a qualitative study with a small, non-random sample. Respondents were recruited through community organizations, industry contacts, advocacy groups, and regulatory bodies, which introduces selection effects in every direction: community members who agreed to be interviewed are more likely to be engaged with the issue than the median resident, and developers who agreed are more likely to be reflective about engagement than the median operator. Percentages are reported for descriptive clarity, not as population estimates. Quotations are lightly edited for length and attributed by role rather than name under the confidentiality terms of the interviews.

1.3 Method and survey findings

Fifty-three respondents completed a seven-item instrument covering role, perceived communication quality, overall disposition toward data centers, primary concerns, and three binary policy questions. Each survey was paired with a semi-structured interview conducted by phone, which supplied the qualitative material in Section 3. Concerns were multi-select; all other items were single select.

2. Survey findings

2.1 Composition

Host community members form the largest single block, reflecting the study’s focus on the receiving end of siting decisions. Developers and investors together account for 13 respondents, roughly a quarter of the sample, providing the counterweight.

2.2 Perceived communication quality

Respondents rated the quality of communication between their own group and the data center industry on a five-point scale. The distribution is heavily weighted toward the negative end, and the absence of any “excellent” rating is notable: even respondents strongly favorable toward data centers did not describe the industry’s communication as good.

Figure 2. Rated communication quality between respondent group and the data center industry (N = 53).

2.3 Overall disposition toward data centers

Opinion is polarized rather than uniformly negative, and almost exactly evenly split: 21 respondents (40%) were favorable or very favorable, 21 (40%) were unfavorable or very unfavorable, and 11 (21%) were neutral. Critically, disposition sorts almost perfectly by stakeholder category, which is the central structural finding of this study.

Figure 3. Disposition disaggregated by stakeholder category. 

2.4 Primary concerns

Energy demand was the most frequently selected concern, followed closely by communication and transparency. The prominence of transparency is worth emphasizing; it is not a meta-concern layered on top of the substantive issues but is itself among the most cited substantive issues. Water use ranks third and while this study does analyze by region it can be assumed it is a bigger concern for homeowners in regions facing water shortages.

Figure 4. Concerns selected by respondents (multi-select, N = 53).

2.5 Binary policy questions

Three yes/no items probed the gap between opposition in principle and opposition in place. The results indicate that most respondents accept the necessity of data centers in the abstract while resisting them locally, and that support for restrictive local legislation is close to evenly divided.

Figure 5. Responses to three binary policy questions (N = 53).

3. Stakeholder perspectives

The sections below summarize each group’s account in its own terms. They are presented as reported positions, not as findings of fact; where groups make competing empirical claims, the disagreement itself is the object of interest.

3.1 Host communities

Host community respondents including homeowners, and elected community leaders held the most consistent and most negative position in the sample. Their objections clustered around four themes.

Land use. The most firmly held position was categorical rather than negotiable: data centers belong in industrial zones, and their appearance adjacent to residential subdivisions and working farmland represents a category error in land use planning. Respondents were not seeking better setbacks or improved landscaping; they were disputing the premise that the parcel was appropriate at all. 

1. Household cost. Respondents anticipated higher electricity and water bills and understood these as transfers rather than side effects especially capacity built to serve a facility, paid for through rates charged to everyone. Whether that expectation is borne out varies by tariff structure and jurisdiction; what matters here is that residents encountered no credible mechanism to verify claims to the contrary.
2. Precedent. A recurring concern was the “floodgates” effect: rezoning for a first facility establishes precedent making subsequent approvals procedurally easier. Respondents treated the first application as a decision about the county’s character over a decade, not about one building.
3. Absence of local benefit. Respondents reported neither meaningful employment nor tax revenue. Employment generated is low relative to footprint and skews toward specialized roles filled from outside the county.
4. Skepticism about the premise. Several respondents rejected the underlying growth narrative outright, characterizing AI as a speculative cycle rather than a durable transformation. This view was almost entirely unaddressed in developer communications, which typically assume the value of technology and argue only about local impacts.
5. Governance. Support for local moratoria was described by community leaders as effectively unanimous within their networks, paired with strong preference for decentralized authority each county setting its own rules. Notably, respondents did not attribute the problem to rogue actors. They described the opacity as structural, a feature of how the industry transacts rather than the behavior of a few firms.

3.2 Data center developers and operators

Developers described a fundamentally different situation: a broadly responsible industry whose reputation is being set by a small number of poor performers, operating in a climate of escalating and partly manufactured hostility.

1. Scale of impact. Developers considered water and energy concerns materially overstated relative to modern design practice, citing closed-loop cooling, efficiency gains, and contracted generation. They characterized community understanding of facility resource profiles as incomplete.
2. External interference. Several respondents raised the possibility that foreign actors are amplifying opposition, framing data sovereignty as a national interest and some organized resistance as adverse to it. This framing was not evidenced in the interviews and has a significant cost: it recodes ordinary civic objection as suspect, which forecloses the engagement developers say they want.
3. Capital mobility. A common argument held that a sufficiently hostile domestic environment will push development offshore, forfeiting revenue and opportunity. Respondents pointed to Loudoun County as evidence of achievable mutual benefit — a reference contested by environmental respondents, who cite the same jurisdiction as a cautionary case on cumulative load and water.
4. Self-critique. One developer diverged sharply from the group, arguing that the industry’s engagement model is failing on its own terms and advocating a trade body capable of setting and enforcing engagement standards. The same respondent doubted it would happen, noting that the incentives to defect are strong and the collective benefit is diffuse.
5. Regulation. Developers expected regulators to intervene and worried the intervention would be blunt. They favored decentralized regulation — the same structural preference host communities hold, arrived at from the opposite direction — and criticized policy makers for pursuing short-term positioning at the expense of long-term national interest.
6. Displacement. Several respondents attributed hostility to anxiety about AI and employment, arguing the industry is absorbing blame properly directed at large technology firms. They described themselves as willing to be transparent “in good faith,” and characterized non-transparent peers as outliers.

3.3 Environmental organizations and university researchers

Researchers and environmental advocates occupy a distinct position: they are the group most willing to be persuaded by evidence and the group least able to obtain it.

Respondents described repeated attempts at collaboration and offers to study consumption, climate effects, and grid effects were terminated due to data access. In the absence of operational figures, researchers apply a precautionary posture and model toward the upper bound of plausible impact. Several stated explicitly that they interpret the refusal itself as evidence, reasoning that favorable numbers would be disclosed.

This is the study’s clearest example of a self-reinforcing loop. Developers withhold data partly because they expect it to be used adversarial; researchers assume the worst because data is withheld; the resulting published estimates are then cited by opponents, confirming the developers’ expectation. The loop is breakable only from the disclosure side, because only one party holds the missing input.

3.4 Communications professionals

Communications consultants are a recent addition to data center project teams and they offered the most pragmatic diagnosis in the sample: the industry arrived at the conversation late, after positions had hardened, and is now attempting to solve at the messaging layer a problem created at the deal-structuring layer. This group validated community concerns as rational, read local political behavior as genuine constituent response rather than opportunism, and assessed developer leverage as weaker than developers believe. Approval, in their account, is a political outcome rather than an administrative one, and consent cannot be substituted with a permit.

3.5 Utilities

Utilities described substantial load growth and capital deployment opportunities while positioning themselves outside the controversy, explicit that they are service providers rather than developers and that they take direction from regulators rather than advocating independently. This positioning is coherent but consequential: utilities hold the interconnection and rate-design information most relevant to residents’ cost concerns, and their reticence leaves the most answerable question in the dispute who actually pays for incremental capacity.

3.6 Media

Media respondents characterized their role as presentation rather than persuasion, describing public opinion as forming naturally from reported facts. Some argued a degree of build-out is required for competitive reasons while rejecting unconstrained development, and several favored a temporary pause until regulatory frameworks mature. Notably, this group largely accepted the premise that the build-out is necessary precisely the premise host community respondents contest, which means it is rarely examined in coverage both groups read.

3.7 Investors

Investors were the most uniformly bullish group in the sample, describing the sector in terms of a generational reallocation of capital.

Respondents viewed community resistance as a misreading of opportunity and were comfortable deploying capital internationally if domestic conditions deteriorated. Their stated concern was not opposition but ambiguity: regulatory uncertainty, and an absence of clear direction from senior federal leadership, was described as the factor prompting strategic reassessment. Several indicated they could operate under strict rules or permissive ones, but not under unresolved ones.

3.8 Regulators 

Regulators described themselves as under-resourced and under-instructed: evaluating technically complex applications against guidance written for a different generation of industrial use, without clear direction from above. They favored case-by-case evaluation in practice while advocating a more centralized and standardized framework; the opposite of the preference expressed by both communities and data center developers.

3.9 Policy makers
Policy makers declined substantive participation, providing only referrals to published material and general statements of constituent commitment. That non-response is itself a finding. The one group with the authority to resolve the jurisdictional question is the group least willing to be recorded on it, which leaves regulators improvising, communities organizing, and investors waiting.

4. Analysis

4.1 The disagreements are mostly not empirical

It is tempting to read this dispute as factual, resolvable with better data. Some of it is: water withdrawal, consumptive use, load profiles, and rate impacts have determinate answers. But the core disagreements concern authority and value, and no measurement resolves them.

Host communities hold that land use should reflect the preferences of those who live with the consequences, and that a project generating no local benefit fails regardless of its efficiency. Developers hold that infrastructure serving national demand cannot be subject to parcel-level veto. Investors hold that capital allocation should follow market signal, with policy reducing variance rather than directing outcomes. Environmental organizations hold that cumulative climate impact is prior to local economics. These are positions about who decides and what counts: a respondent who accepted every empirical claim made by the other side could still hold their position without inconsistency.

4.2 Trust is a function of structure

Host community distrust was not described as an impression to be corrected but as an inference from repeated experience: engagement occurring after decisions, commitments unenforceable at the point they matter, and information asymmetry preserved as a matter of course. The industry’s response; attributing opposition to misunderstanding, and in some cases to external influence supplies further evidence for the inference it is meant to rebut.

It follows that trust is unlikely to be restored through improved communication, because communication is not the variable generating distrust. Verified disclosure, enforceable commitments, and genuine decision authority are the variables. Each of these is costly to the party that must concede it, which is why they remain unconceded and why the problem persists.

4.3 The information vacuum manufactures its own conclusions

The absence of independently verified operational data does not produce neutrality; it produces upper-bound estimates. Researchers model conservatively, advocates cite those models, coverage reproduces them, and the resulting figures become the public baseline. An industry that believes its actual performance is better than the public baseline is, on this analysis, being harmed primarily by its own non-disclosure.

4.4 Jurisdictional ambiguity is the load-bearing failure

Communities want county-level control. Developers want decentralized rules for different reasons. Regulators want centralized standards. Investors want any rule that is stable. Policy makers have declined to specify. So the question of who decides is relitigated at every individual hearing, maximizing conflict, cost, and delay for all participants — including those who would prefer a rule they dislike to no rule at all.

5. Conclusions and recommendations

5.1 Summary of findings

5.2 Recommendations 

The interventions most likely to change outcomes share a common property: each operates on incentives or on verifiable obligation rather than on messaging. If the failure were informational, better explanation would resolve it. Because it is structural, only changes to what parties can verify and what they are bound to deliver will move it.

Data center developers hold the highest-leverage action available to any party: verified transparency. Third-party audited reporting of water withdrawal, consumptive use, energy draw, and generator runtime, published on a fixed schedule, removes the missing input that currently drives worst-case public estimates. Developers who believe their performance is better than the public assumption are harmed primarily by their own non-disclosure. The same logic applies to economic claims. Residents discounted projected employment and tax revenue not because they misjudged the arithmetic but because projections carry no remedy when they fail. Community benefit agreements with defined values, hiring commitments, and enforcement provisions change the claim from forecast to obligation. Transparency extended only to parties already judged sympathetic is not transparency and is read accurately as its absence.

Regulators should pursue decentralization rather than the centralized standardization they currently favor, and this study departs from their stated preference deliberately. Their frustration is legitimate, but additional procedural layers would lengthen review without improving it, and would move decisions about water systems, load profiles, and land use away from the authorities who understand them. Conditions vary enough between jurisdictions that a uniform national standard would be either too permissive to protect the tightest basins or too restrictive where capacity genuinely exists. What regulators need is not more bureaucracy above them but more capability within them: funded technical review, current guidance, and clear jurisdictional boundaries. Decentralization without that investment produces arbitrary outcomes; with it, capable decision-makers sit closest to the facts that matter.

Media organizations should balance weight over conflict. Coverage that frames each project as a confrontation between aggrieved residents and extractive corporations, or that answers substantive local objections by invoking competition with China, produces heat without informing either side. The whataboutism runs both directions and neither instance advances the dispute. The most useful contribution is the least dramatic one: reporting verifiable quantities and holding the premise of inevitable build-out open for examination rather than treating it as settled background, since that premise is precisely what host communities contest.

Policy makers must move past short-term positioning. Their non-response is itself a policy choice, and its cost is borne by regulators improvising without guidance, communities organizing without a forum, and investors deferring commitments they would otherwise make. The outstanding question is jurisdictional: what is decided locally and what is decided at state level. Every group prefers a clear allocation to the present ambiguity, including groups that would lose ground under it. Until that is specified, who decides will be relitigated at every hearing, at maximum cost to all participants. Rate allocation, meaning whether incremental generation and transmission costs fall on the facility or the ratepayer base, is equally determinate and equally unresolved.

Utilities should plan for exponential rather than incremental load growth and should abandon the neutrality that follows from defining themselves purely as service providers. That self-definition is coherent but consequential. Utilities hold the interconnection data and rate-design authority most relevant to household cost concerns, and their reticence leaves the most answerable question in the dispute without an authoritative voice. Planning calibrated to historical demand will under-build, and the resulting scarcity will be experienced by residents as exactly the cost transfer they anticipated.

AI companies and investors should orient toward long-horizon strategy rather than quarterly positioning. Capital entering this sector is priced on assumptions requiring sustained community consent, stable regulatory treatment, and durable access to power and water over asset lives measured in decades. Decisions optimized for near-term share price and speed to energization degrade precisely those conditions, and the resulting opposition raises the cost of the next site. Participating in building a stable framework serves the sector better than deferring to whatever framework emerges from a process it declined to join.

Host community members should move past reflexive opposition, but the route there is engagement rather than exhortation. Telling residents their objection is parochial has failed everywhere it has been tried and deserves to. Concerns about rates, precedent, and absent local benefit are real, and no one is obligated to abandon them on request. What changes the position is participation in a process where terms are genuinely open, where residents help set water conditions, load thresholds, and buffer requirements rather than approving or rejecting a finished proposal. Communities that enter that process discover which objections generalize and which attach to the specific site, and that distinction determines what is winnable. Insisting that a facility not be sited on farmland adjacent to housing survives contact with a negotiating table. Opposing all development anywhere forfeits the ability to shape development that proceeds regardless. The obligation runs both ways: communities cannot be asked to engage in good faith with counterparties who have not earned it, which makes the disclosure and enforceability asks directed at developer’s prerequisites rather than parallel commitments. Where developers meet that bar, durable outcomes come from converting opposition into specification, since standards survive changes in council composition while project-by-project resistance does not, and from coordinating terms regionally, which reduces venue-shopping and guards against the competitive lowering of standards that decentralized authority might otherwise invite.

Research institutions should be more vocal and should work with media to raise the visibility of what they find. Researchers described dead-ended collaboration and retreated to precautionary modeling but largely did not publicize either the findings or the refusals that produced them. Both are newsworthy. Publishing methodology and assumptions openly, documenting where access was denied, and partnering with journalists to translate technical findings would raise the cost of non-disclosure and give communities and regulators an independent evidentiary basis they currently lack.

Communications professionals emerge as the most grounded participants and are underused. This group alone held an accurate map of the whole system, validating community concerns as rational, reading local political behavior as genuine constituent response rather than opportunism, and assessing developer leverage as weaker than developers believe. Their diagnosis, that the industry arrived late and is solving at the messaging layer a problem created at the deal-structuring layer, is the correct one. Communications expertise belongs where sites are selected and benefits are structured, not appended after an application is filed. Retained early and given standing to shape terms rather than only explain them, this group can do what no other stakeholder currently can: translate between parties who have stopped being able to hear each other.

5.3 Further research

Three extensions would strengthen this work: a longitudinal study of counties five to ten years post-approval, measuring realized employment, tax receipts, and rates against approval-stage projections; a comparative analysis of jurisdictions that adopted mandatory disclosure, testing whether verified data narrows the perception gap; and an assessment of whether modular and edge architectures alter the siting calculus enough to shift the political economy described here.

References and data note

Findings draw on 53 structured interviews and paired survey responses with stakeholders in U.S. jurisdictions with active or proposed data center development. Respondents were recruited through community organizations, industry contacts, environmental advocacy groups, academic departments, and regulatory bodies, and are identified by role only under the confidentiality terms of the interviews. The survey instrument, coding scheme, and anonymized response data are retained by the author and available to faculty advisors on request. All charts are generated from that dataset; percentages are computed on N = 53 unless otherwise stated, and multi-select items sum to more than 100%.
This study involves anonymous surveys and interviews of adult participants and does not collect personally identifiable information or sensitive information that could reasonably place participants at risk. The research is therefore expected to qualify for an exemption under 45 CFR §46.104(d)(2), which covers certain research involving survey and interview procedures where participant identities cannot readily be ascertained and the research presents minimal risk. No personally identifiable information was collected or retained, and participation was voluntary. The study did not involve interventions, deception, or procedures presenting more than minimal risk to participants.