Article Archives
Article Categories
Articles
WATT is the Problem?
AI Data Centers, Grid Bottlenecks, and the New Rules of CRE Development
Commercial real estate teams are used to thinking in square feet, cap rates, and entitlement timelines. In the AI era, each of those variables is increasingly constrained by a fourth factor: available electrical capacity.
Anyone working in development, asset management, brokerage, facilities, or capital planning needs to understand this shift. AI data centers are the most visible new load. Still, they are also the leading indicator for a broader constraint that will affect industrial expansions, lab conversions, hospital projects, and building electrification.
The core issue is not only how much generation exists on a regional grid. It is the speed with which power can be delivered to a specific site through transmission and distribution infrastructure, with the reliability and redundancy a project requires. Transmission development can take years, and grid connection constraints are already delaying a share of planned data center projects.[5]
A helpful analogy is a modern port. The ocean may be full of ships, but if the channel is narrow and the cranes are booked, cargo still sits offshore. In the same way, electricity can exist in aggregate while projects stall in the interconnection queue, at the substation upgrade, or due to transformer delivery delays.[13][15]
That is why watts have become a site attribute. They shape where projects can be built, how fast they can be delivered, and how much unplanned capital may be required to make a location viable.
AI Compute Demand vs. Energy Capacity
As 2026 begins, AI growth is converting an abstract idea called “compute” into an uncomfortably tangible unit called “megawatts at the meter.” The International Energy Agency (IEA) estimates global data center electricity demand at about 415 terawatt hours in 2024, rising to roughly 945 terawatt hours by 2030 in its base case, with further growth beyond 2030.[1]
In the United States, Lawrence Berkeley National Laboratory (LBNL) estimates that data centers consumed 176 terawatt-hours in 2023, representing about 4.4% of U.S. electricity consumption. LBNL projects a range of 325 to 580 terawatt-hours by 2028, roughly 6.7% to 12% of U.S. electricity consumption, implying a peak load requirement of approximately 74 to 132 gigawatts.[2]
These projections are directionally consistent with other reputable estimates that project data centers will account for a fast-growing share of U.S. electricity demand through 2030.[3]
The demand curve is colliding with grid realities on three fronts.
- First, regional capacity margins are tightening as load forecasts rise and older generation retires. NERC projects summer peak demand growth of 132 gigawatts over the next 10 years across its footprint, and attributes a large share of that growth to large loads, including 58 gigawatts from data centers.[4]
- Second, even when a new generation is planned, it may not be ready when needed. NERC notes that a substantial portion of scheduled resources remains unbuilt, increasing timing risk.[4]
- Third, building supply and building delivery are not the same thing. A large volume of proposed generation and storage is waiting in interconnection queues, a well-documented bottleneck that slows the pace at which new supply can be connected and dispatched.[6]
A regional example illustrates how these factors stack. PJM, the largest U.S. grid operator by load served, has increased its load forecast in part due to expected data center growth and reports materially higher projected peaks than in its prior forecast cycle.[5] In parallel, market signals such as higher capacity prices indicate a grid being asked to do more, sooner, with less slack.[18]
The core mismatch is timing. The IEA flags that a meaningful share of planned data center projects may be delayed by grid connection constraints, and notes that transmission development can take 4 to 8 years in advanced economies.[1] Meanwhile, large transformers and related equipment often have multi-year lead times.[7]
How is it Impacting other CRE Development?
Power scarcity does not remain contained within the data center sector. It leaks into the broader commercial real estate development ecosystem in predictable ways.
Queue displacement is the first impact. When a single hyperscale campus requests hundreds of megawatts, it can consume the near-term upgrade headroom in a substation or corridor. Other projects, including industrial redevelopment, healthcare expansions, lab conversions, multifamily electrification, and conventional office repositioning, may face longer timelines, higher upgrade costs, or reduced service availability.
Site selection and land values are the second impact. Power availability has become a gating criterion, pushing developers toward emerging and tertiary markets and reshaping land pricing and entitlement strategies.[9] Market analytics also show rapid growth in delivered capacity and historically low vacancy in primary data center markets, a signal of demand pressure that can amplify local competition for infrastructure.[8]
Local governance is the third impact. Communities that were not previously “power constrained” are now writing zoning and design controls specifically for data centers, often to manage externalities such as noise, visual screening, and infrastructure impacts.[13]
Regulatory friction is the fourth impact. Ireland provides a clear example of regulators imposing conditions on new data center connections to protect system reliability and emissions objectives, including renewable electricity and operational flexibility requirements.[14]
In short, data centers function as a new class of anchor tenants for the grid itself. They compete for the same finite capacity required by other CRE projects for electrification, resilience upgrades, and growth.
How do Datacenter Renovation and Retrofit Cycles Compare to Traditional CRE?
Traditional CRE renovation cycles often revolve around leasing events and building system end-of-life. Data centers run on two overlapping clocks, one far faster than most building owners are used to.
- The first clock is the IT refresh. Industry survey research indicates that many operators refresh servers every three to five years, driven by performance gains, energy efficiency, and evolving workload requirements.[10]
- The second clock is electrical and mechanical infrastructure. Physical infrastructure, such as UPS systems, can operate for 15 to 20 years with proper maintenance, but component aging and risk management often drive earlier refreshes or modernizations.[12] Guidance on modernizing older facilities commonly targets sites around a decade old for significant upgrades to mechanical and electrical systems.[11]
This combination creates a capital profile that feels less like an office building and more like an airport runway. The surface can last decades, but the supporting systems are under continuous stress, and the operational tolerance for failure is low. That drives more frequent reinvestment, more rigorous commissioning and testing, and a stronger preference for modular, replaceable components.
Are there Solutions to the Demand=Capacity Collision?
There is no single fix. The collision is the product of physics, supply chains, regulation, and finance. The practical response is a portfolio of measures that reduce peak strain, accelerate delivery, and improve planning alignment.
Grid acceleration and buildout are the most direct paths, but they are slow. Transmission and substation upgrades have multi-year permitting, siting, and construction timelines, and critical component lead times remain a constraint.[1][7]
Load flexibility is an underused lever. If large training workloads can be scheduled, curtailed, or relocated across regions, data centers can function more like dispatchable industrial loads and less like immovable baseload. Regulators are starting to formalize expectations for flexibility as a condition of new connections in constrained systems.[14]
Behind-the-meter and firm supply contracting are expanding. The IEA expects dispatchable generation, including natural gas and nuclear, to play a material role in meeting incremental data center demand, and anticipates that small modular reactors will come online around 2030.[15] Governments are also funding pathways to deploy new nuclear technologies.[16]
Design and efficiency still matter, but they are not sufficient on their own. AI-driven densification can outpace incremental efficiency gains, meaning that efficiency reduces the slope, but does not eliminate the rise. Survey research notes that densified IT for AI is placing new demands on facility infrastructure.[10]
Market diversification is already happening. Market reports identify power availability as a key constraint in established hubs, with developers showing greater interest in emerging markets where power is less of a near-term constraint.[9]
What does this mean for Building Lifecycle Management?
Building Lifecycle Management becomes less abstract in a power-constrained world. It shifts from being a data integration ambition to being a practical survival tool for planning, risk, and capital allocation.
- First, lifecycle management connects the early design decisions to downstream grid feasibility. When interconnection timelines can exceed project schedules, requirements definition and utility coordination must begin earlier and be tracked as a critical-path scope.
- Second, lifecycle management forces a common language across disciplines. ISO 55000 frames asset management as a coordinated activity to realize value from assets across their life cycles.[17] ISO 19650 frames information management across the whole life cycle of built assets.[19] Together, they reinforce a discipline directly applicable to data center infrastructure, where power and cooling constraints make late changes costly.
- Third, lifecycle management improves governance and accountability for upgrades. The frequency of retrofit events in data centers means the asset is constantly moving through mini lifecycle loops. Without consistent asset data, configuration records, testing history, and change control, owners cannot reliably answer basic questions such as what capacity is truly available, what redundancy exists, what can be safely shut down, and what must be replaced next.
A forecast for likely industry responses follows.
- Data center and utility stakeholders will increasingly standardize “power requirement passports” for sites and projects, including timelines, ramp profiles, flexibility options, redundancy assumptions, and evidence of commissioning. These will become part of entitlement, financing, and underwriting packages.
- CRE owners will treat grid capacity as a strategic asset, not a utility afterthought. That will push more portfolio-level planning, earlier site screening for electrical headroom, and tighter coordination between capital plans and utility upgrade roadmaps.
- Building Lifecycle Management will serve as the operating system for this coordination. In a world where the constraint is watts, the organizations that manage asset information across the lifecycle will be the ones that can build, retrofit, and operate at the pace AI is demanding.
