Data centers are at an inflection point. Once built on the fringes of urban areas, these massive energy consumers are increasingly moving into cities, driven by the need for low-latency artificial intelligence compute, cloud expansion, and edge processing. But this shift comes at a cost.
By 2030, data centers could account for 9% of U.S. electricity demand—more than double today’s levels. This surge is already straining grids and forcing industry leaders to rethink power procurement, infrastructure planning, and sustainability commitments.
The scale of data center expansion is unprecedented; Grid Strategies projects as much as 128 gigawatts of new load growth in the U.S. in the next five years, which is the equivalent of adding a new Texas grid. And the growth is colliding with an energy landscape constrained by slow grid expansion and volatile power markets.
Traditional power generation and transmission development can take up to 13 years. And a clogged load queue is making waits even longer.While renewable energy has long been the preferred decarbonization path, intermittency and land requirements present challenges for data centers’ high-density energy needs. For data center operators, these challenges translate into a new set of energy priorities: securing reliable power, mitigating long-term costs, and aligning with carbon commitments.
So data center companies have to be nimble. They need energy sources that meet their needs, buildings and infrastructure options that adapt as the landscape does, and systems they can deploy quickly and easily as demand changes. In response, the industry is reevaluating everything — from the sources of electricity to the form and function of the data centers themselves.
The promise of SMRs
So a growing number of tech giants are exploring a new solution: small modular nuclear reactors, or SMRs, which promise consistent, carbon-free power at a scale that aligns with data center growth. While SMRs have yet to reach full commercial deployment, the convergence of rising urban energy demand, data centers’ need for clean firm power, and the modular nature of these reactors is creating a rare window of opportunity to accelerate their adoption.
And major players are taking note. Google and Amazon have begun exploring nuclear-powered data center projects, and the U.S. Department of Energy is accelerating regulatory approvals for next-generation SMR deployments. Just last month, DOE reopened a $900-million funding opportunity for SMRs, which was initially issued by the Biden administration and funded by 2021’s Bipartisan Infrastructure Law
In data center applications, solar and wind generation require massive battery storage to meet reliability requirements. Meanwhile, land constraints and permitting bottlenecks make large-scale renewable deployment challenging near urban hubs.
But — especially compared to traditional nuclear plants — SMRs require significantly less land and water for cooling, making them viable for urban integration, SMRs require significantly less land and water for cooling, making them more viable for siting in or near dense urban areas where power-hungry data center clusters are emerging.
And with outputs typically ranging from 50 to 300 megawatts, SMRs offer a modular, right-sized solution that can scale alongside demand—without the geographic or permitting hurdles associated with large-scale energy infrastructure. For instance, the DOE-backed Carbon Free Power Project in Idaho, which includes plans for a NuScale SMR, has attracted interest from utility buyers serving urban markets and could serve as a model for colocated data center applications.
The appeals of urban adaptation and modularity
Meanwhile, the cities in which those clusters are located are changing. As urban data center footprints expand, real estate and infrastructure strategies must evolve to accommodate this growth. One emerging approach is adaptive reuse — repurposing vacant office buildings, warehouses, and retail spaces into high-density digital hubs.
Utilizing existing power and fiber connectivity reduces both development timelines and capital costs; converting underutilized buildings helps optimize land use in dense urban markets facing zoning and permitting challenges; and reducing construction waste aligns with corporate sustainability goals while accelerating deployment. At the same time, as energy and space constraints tighten, operators are rethinking how quickly — and how flexibly — they can bring capacity online.
And, with AI workloads demanding more computational power, data center operators are turning to modular designs for rapid scalability and efficiency. For instance, companies like Vertiv and Flexnode are offering prefabricated, modular data center solutions that integrate advanced cooling and power systems for rapid deployment. And the company I founded, Surfaice, is among emerging platforms that are helping operators optimize site selection, permitting, and infrastructure planning to accelerate delivery across dense urban markets.
Prefabricated offsite and assembled on location, modular data centers can be deployed in weeks instead of months. Advanced liquid cooling systems minimize water use while optimizing energy efficiency. And the flexible deployment models allow operators to scale infrastructure without committing to massive, centralized builds.
Microsoft, AWS, and Oracle have already integrated modular data centers into their growth strategies, highlighting their role in the next phase of digital infrastructure expansion.
The policy and planning outlook
With data center energy demand reshaping power markets, policymakers and utilities — and energy developers like those working on new SMR technologies — are grappling with how to allocate costs and integrate these facilities into the broader energy landscape.
For instance, Texas and Virginia have introduced bills to address the impact of hyperscale data center loads on local grids and rate structures. And California’s SB 57 proposes a new rate structure to ensure data center energy costs are fairly distributed.
Meanwhile, some data center operators are exploring direct partnerships with energy developers to build dedicated, colocated power infrastructure, bypassing traditional utility models.
The energy demands of AI and hyperscale computing are rewriting the rules of power generation, distribution, and infrastructure planning. The shift from traditional grid reliance to a diversified energy mix — one that blends SMRs, modular construction, and adaptive reuse — will define the next generation of data center development.
Companies that embrace these strategies early will position themselves at the forefront of an industry undergoing radical transformation. The winners won’t just be the ones with the fastest servers; they’ll be the ones with the most resilient, cost-effective, and sustainable energy solutions. The race to power AI is, at its core, a race to redefine energy itself.
Omar Zhan is the co-founder of Surfaice, an autonomous construction software company. The opinions represented in this article are solely those of the author and do not reflect the views of Latitude Media or any of its staff.
