Data Centers Should Be Designed to Support Communities and Nature

Part one of a three-part series

Data center in Newark, California / istockphoto.com, JasonDoly

By Keith Bowers

Artificial intelligence (AI) has quickly become part of landscape architecture practice. Many of us are already using AI to organize research, explore design alternatives, write code for GIS analyses, and create images for project presentations. Like CADD, GIS, LiDAR, and satellite imagery before it, AI is becoming another tool that can help landscape architects better understand and design the landscapes around us.

AI is already revolutionizing our work. The speed of its evolution is causing many of us to be concerned about the lack of transparency of its models, its potential to spread misinformation, and its ability to engender an over-reliance on it that can lead to skill loss among people. But there is now a growing focus on the physical infrastructure - data centers - that enables these technologies.

In a series of three articles, I'll explore an applied ecology framework that can be applied by landscape architects and communities to address data center siting and design. Over the series, I'll also delve into AI's inherent biases, considerations related to transparency and professional ethics, and the importance of maintaining our judgment in the design process.

Every AI prompt, online meeting, driving direction, streamed movie, medical record, and banking transaction depends on a growing network of data centers. Although we often think of "the cloud" as something virtual, it is grounded in physical infrastructure - large buildings, electrical substations, cooling systems, transmission corridors, fiber optic networks, and the living landscapes that support them.

Data Center infrastructure in the U.S., November 2025 / National Renewable Energy Laboratory (NREL)

This strikes me as an important topic for landscape architects, not simply because AI is changing how we work but because the infrastructure enabling it is changing how landscapes are planned, developed, and managed. It also invites us to consider what forms of intelligence should inform those decisions - from computational intelligence to the ecological understanding embedded in living systems and long reflected in Indigenous ways of knowing.

Every generation inherits a new infrastructure that reshapes the landscape. Railroads, interstate highways, subdivisions, and shopping malls all challenged planners, designers, engineers, policymakers, and communities to rethink how infrastructure should fit within the places where we live and work. Data infrastructure represents the next chapter in that story. The difference is that the technological needs to support this infrastructure are advancing much faster than the planning systems intended to guide it.

Data centers are unlike most commercial or industrial buildings. They are among the most resource-intensive buildings ever constructed, requiring substantial amounts of electricity, water-based cooling, and supporting infrastructure. They also generate waste heat, noise, and significant demand on regional electrical grids, not to mention their toll on the living world. At a time of accelerating climate change and biodiversity loss, we can no longer afford to treat these demands as impacts to be managed at the margins. These demands must radically reshape how and where data centers are planned, sited, designed, and operated.

At the same time, today's data center should not be viewed as the inevitable model for tomorrow's. Advances in computing, cooling technologies, renewable energy, adaptive reuse, and regenerative design have the potential to fundamentally change how these facilities perform and how they relate to the landscapes around them. That raises a question:

What should communities reasonably expect from responsible data center infrastructure?

Today, there is no consistent answer. Several European countries have begun asking that question in new ways. Rather than viewing data centers primarily as economic development projects, they are increasingly treating them as long-term infrastructure investments with measurable environmental responsibilities.

Public reporting of energy and water use, biodiversity performance, waste heat recovery, and broader sustainability standards are beginning to influence where facilities are built and how they perform over time. The conversation is gradually shifting from Can we build a data center here? to What should this project contribute to the surrounding community and landscape? That seems like a useful question for us to ask as well.

In the U.S., planning and regulatory frameworks have been slower to evolve. Most zoning ordinances and building codes were never written with hyperscale data centers in mind. They are frequently treated as another industrial or commercial use, despite excessive demands on electricity, water, land, and infrastructure that differ substantially from conventional development.

As a result, site selection often follows available electrical capacity, fiber connectivity, relatively inexpensive land, permissive (re)zoning, and in many cases, marginalized communities with little political capital. Increasingly, that combination is found in rural communities, agrarian landscapes, and rapidly growing exurban areas where planning frameworks were never developed for infrastructure of this scale.

Without clearer planning and performance standards, data centers naturally follow the path of least resistance. Large, single-story campuses spread across inexpensive land because that is often the most economical approach, not necessarily the most sustainable one. Multi-story and below-grade facilities are already operating successfully in Europe and Asia where land is more constrained, demonstrating that today's dominant American data center footprint - hyperscale data centers exceeding 200,000-plus square feet - reflects economics and public policy - or the lack thereof - as much as engineering.

Hyperscale data center campus in Huuto, Texas / istockphoto.com, BackYardProduction

The planning process itself often compounds these challenges. Site acquisition may occur through subsidiaries that obscure ownership and discourage transparency. Nondisclosure agreements can limit what local governments are able to discuss publicly while negotiations are underway. Community engagement frequently begins after fundamental decisions about location, scale, and infrastructure have already been made. By then, the conversation has often shifted from Should this project be here? to How do we mitigate its impacts?, which are fundamentally different questions.

A data center's location dictates community and ecological outcomes long before design even begins. Whether a facility consumes prime agricultural land instead of a brownfield, fragments wildlife corridors, impacts wetlands, or increases flood risk through massive impervious cover are foundational impacts that cannot be fixed by design. They can only be minimized. Consequently, these critical siting and zoning decisions must be addressed before design - led by planners, ecologists, landscape architects, and most importantly, community members.

Recognizing this gap, Biohabitats began asking what an applied ecology framework for evaluating AI infrastructure might look like. Drawing on more than four decades of ecological planning and design experience, we developed An Applied Ecology Framework for Data Center Siting and Design.

Rather than functioning as a rating system or regulatory checklist, the framework is intended to help planners, landscape architects, architects, engineers, developers, utilities, local governments, Tribal Nations, and communities ask better questions before decisions become irreversible. The framework can also be used to inform updates to zoning ordinances and building codes.

The framework complements emerging guidance, including Architecture 2030's Data Center Evaluation Checklist. It expands the conversation beyond building performance to consider the landscapes, watersheds, communities, and living systems in which these facilities are embedded.

Rather than beginning with landscape and building performance, the framework begins with a different question: Should this project be here?

From that starting point, it guides multidisciplinary teams through the broader questions, including:

• How does the project relate to communities?
• How does it affect living systems?
• How should it adapt over time?

The framework is organized around four interwoven themes that reflect the progression of responsible decision-making throughout the life of a data center project:

• Site selection
• Communities
• Living systems
• Adaptive futures

Together, these themes encourage multidisciplinary teams to consider how decisions made during site selection, planning, design, construction, operations, adaptive management, deconstruction, and future reuse influence communities and living systems. Instead of prescribing solutions, the framework provides:

• Guiding principles
• Framework elements
• Evaluation criteria
• Guiding questions

The framework supports informed decision-making throughout the life of a project.

At its core, the framework recognizes that data centers are not conventional buildings. They become long-term components of watersheds, landscapes, and ecological and human communities. Understanding those relationships through the lens of applied ecology creates opportunities to make better decisions before they become difficult - or impossible - to change.

The framework also encourages project teams to think beyond impact reduction. It asks:

• Does the site selection avoid unnecessary ecological and community impacts before design begins?
• Can water, biodiversity, healthy soils, and ecological processes be protected, conserved, restored, and even strengthened over time?
• Can projects be designed to adapt as technologies, communities, and landscapes continue to evolve?

These planning and design questions will shape every decision that follows.

These ideas are consistent with the direction many professions are already moving. Landscape Architecture 2040: ASLA Climate & Biodiversity Action Plan calls on landscape architects to achieve zero greenhouse gas emissions, conserve and restore biodiversity, and advance climate-positive design. Architecture 2030 has challenged the building industry to rethink the environmental impacts of data centers. Together, these efforts suggest that data center infrastructure can be viewed as a part of climate and biodiversity action. These facilities have the potential to help us achieve a more livable and just future.

Data center paired with wind energy system in Eemshaven, The Netherlands / istockphoto.com, Hugo Kurk

The opportunity before us is to rethink how AI infrastructure fits within the landscapes and communities that support it. By bringing responsible applied ecology into decisions about where these facilities are located, how they are designed, and how they evolve over time, landscape architects and our allied professions can help shape data center infrastructure that is more resilient, more adaptable, and more compatible with the living systems upon which all life depends.

Data center paired with solar photovoltaic energy system in The Netherlands / istockphoto.com, Hugo Kurk

In doing so, we may discover that some of the most valuable intelligence informing the future of AI has been quietly evolving within nature for billions of years.

In the next article, I'll explore the four themes of the framework in greater detail and discuss practical strategies for improving the planning, siting, design, and long-term management of data center infrastructure. I'll examine how an applied ecology perspective can inform decisions from the earliest stages of site selection through long-term stewardship and adaptive management.

Keith Bowers, FASLA, PLA, is the founder and practice leader of Biohabitats. He is a member of the ASLA Climate & Biodiversity Action Committee.

Transparency statement: This article was conceived, researched, directed, and substantially written by the author. AI was used to assist with organization, editing, and stylistic refinement. Source material was independently reviewed and cited information was verified. Analysis, opinions, and final editorial decisions are solely those of the author.

Based on published estimates, the AI-assisted development of this article likely consumed on the order of 0.1-0.3 kWh of electricity and approximately 0.2-1.5 liters of water, although the actual environmental footprint cannot currently be determined because AI providers do not disclose the operational data needed to calculate the electricity, water, carbon, and land impacts of individual user sessions. Estimates are informed by published literature, including the United Nations University Institute for Water, Environment, and Health's Environmental Cost of AI's Energy Use: Carbon, Water and Land Footprints.