Meeting AI Demand: Alternate Power, Design, and Site Strategy

Deloitte's 2025 AI Infrastructure Survey projects that U.S. AI data center power demand could grow more than thirtyfold by 2035, reaching up to 123 GW, with grid stress as the leading obstacle. Developers are adopting alternative energy sources, including small modular reactors and behind-the-meter generation, with tech companies signing contracts for over 10 GW of new nuclear capacity in the past year, according to Goldman Sachs. Chris Hastings, Vanderweil Engineers' Power Group Managing Principal, emphasized that planning beyond traditional utility service can avoid grid bottlenecks as AI adoption accelerates.

Insight and analysis on the data center space from industry thought leaders. Meeting AI Demand: Alternate Power, Design, and Site Strategy AI’s explosive growth is forcing a fundamental reimagining of data center infrastructure. As artificial intelligence accelerates the transformation of industries, it is also reshaping the very infrastructure that supports it. Developers, architects, and engineers are taking bold steps toward the future by designing data centers specifically to meet the immense computational and energy needs of AI-driven workloads. Driven by the immense power demands of AI, traditional approaches to power, site design, and infrastructure are no longer sufficient to meet AI’s needs. Developers must deliver forward-thinking engineering and architectural solutions to redefine how data centers are planned, powered, and positioned for long-term resilience and growth. The following article showcases strategies that integrate cutting-edge energy solutions, AI-optimized floor plans, and site-selection methodologies that anticipate future demand. Strategies to Meet AI’s Energy Demand By 2035, power demand from AI data centers in the United States could grow more than thirtyfold, according to Deloitte’s 2025 AI Infrastructure Survey https://www.deloitte.com/us/en/insights/industry/power-and-utilities/data-center-infrastructure-artificial-intelligence.html , reaching up to 123 GW. The largest data centers in construction or planning for leading hyperscale clients like Amazon, Meta, Microsoft, and Google are expected to double or quadruple the capacities of their completed projects. Despite this upward momentum, grid stress /uptime/from-capacity-to-chaos-how-ai-data-centers-challenge-the-grid is the leading obstacle to the development of data center infrastructure. Training sophisticated AI models and running inference for systems like ChatGPT require power densities far exceeding typical data center norms. Recognizing the limitations of existing utility infrastructure, developers must consider alternative energy sources to ensure reliable, 24/7 service. On-site generation systems /energy-power-supply/why-data-centers-produce-their-own-power can be engineered to scale with AI workloads while providing redundancy and sustainability. To curb the reliance on gas plants and coal-fired plants, many tech companies are considering low-carbon energy sources such as nuclear power. The implementation of small modular reactors SMRs can reduce emissions and the demand for water as a coolant. Tech companies have signed new contracts for more than 10 GW of possible new nuclear capacity in the last year, according to Goldman Sachs https://www.goldmansachs.com/insights/articles/is-nuclear-energy-the-answer-to-ai-data-centers-power-consumption . Data centers can also operate behind the meter by co-locating load and clean energy generation, reducing the need for grid upgrades and fossil-fuel pollution. “By planning beyond the boundaries of traditional utility service, facilities can meet customers’ needs without being bottlenecked by local grid capacity or reliability issues – a critical differentiator as AI adoption accelerates,” said Chris Hastings, Vanderweil Engineers’ Power Group Managing Principal. Design Solutions for Supporting Alternative Power Sources High energy demand, power density, and complex AI workloads are rapidly changing how we design data centers. Optimizing a data center for AI isn’t simply a matter of adding more server capacity. The design, site, and floor plan must handle AI’s high-density computational requirements, with spaces that support the necessary cooling, power distribution, and flexibility for future upgrades. The history of designing and constructing programmed spaces and site layouts for data centers has changed over the years, and significantly in the past year. Twenty years ago, the ratio of data halls to support space was 50/50. Today, we’re seeing the building size reduced significantly, but the site area remains the same or increases drastically to accommodate cooling equipment and on-site power generation. These changes are driven by increasing rack densities /ai-data-centers/ai-rack-density-s-real-limits-power-cooling-failure-risk – from 25-30 kW per rack to as much as 600 kW per rack – as well as moving support space, such as utility power supply and power rooms, outside. As building size changes with higher densities, cooling requirements are rising sharply. Having enough cooling equipment on roofs to prevent the facility from overheating becomes infeasible. “The challenges of site and building planning are tied to the mechanical and electrical system design now more than ever,” said Dutch Wickes, Ci Design, Inc., Principal and Global Mission Critical Practice Leader. “All the variables make a Rubik’s Cube look easy Especially when you throw in on-site power generation, whether it is bridging power or permanent power. The typical rules of thumb are moot.” Once a potential site is identified, informed decision-making requires a conceptual plan developed with the client, input from mechanical and electrical engineers, and architectural planning. An integrated, on-site power generation and distribution system that harmonizes with the facility’s architecture could ensure not only operational efficiency but also scalability, enabling the end user to expand capacity without redesigning core systems. These innovations align with broader goals of resilience and sustainability, ensuring the facility is future-proofed against both technological and regulatory change. Strategic Site Selection Beyond the traditional criteria of connectivity, proximity to users, and climate risk /build-design/how-data-centers-are-adapting-to-extreme-weather , data center locations must be evaluated for their ability to accommodate off-grid power infrastructure, access to alternative fuels, and regulatory environments supportive of innovative energy solutions. “Utilities significantly impact data center site selection decisions. We’re seeing data centers emerge in more remote locations that have access to less strained grids, but the lack of adequate power interconnection and infrastructure can pose issues,” said Mason McPike, Provident Data Centers’ Executive Director of Vertical Development & Delivery. “Finding a successful data center site starts with building strong relationships between the end user, government entities, utility companies, and the local community.” Smart site selection ensures that each facility is located where it can deliver maximum value – not only today, but as AI demands grow exponentially in the coming years. Facilities built close to power plants can overcome these issues, but data centers that generate their own off-grid power using technologies such as fuel cells, batteries, renewable energy, and SMRs may be a more sustainable solution for the environment and surrounding communities. Positioning for the Future Early collaboration among the end user, developer, architect, and engineer can position a data center project at the forefront of a rapidly evolving industry. By addressing energy constraints head-on, designing for AI’s unique operational needs, and selecting sites that support a bold, off-grid vision, the industry can set a new standard for how data centers can and should evolve. “The changing market demand allows us to have an adaptive perspective in how we approach data center development,” McPike said. “Everything has to work in tandem, from finding the right site to sourcing power supply and creating an efficient design. We have to rapidly evolve as the online world continues to grow in response to new technologies.” For data center professionals, these lessons underscore the importance of rethinking traditional assumptions and embracing integrated solutions that bridge power, design, and location strategies. As AI continues to redefine what’s possible, these forward-looking strategies will define the next generation of mission-critical infrastructure.