The biggest constraint facing AI data center operators is increasingly not energy generation, but the ability to deliver sufficient power quickly enough to keep new projects on schedule.
In this DCD>Talks episode, Joaquin Aguirre, director of strategic portfolio development at PowerSecure, explains why operators are turning to bridge power, microgrids and energy storage to overcome utility delays while preparing for increasingly dynamic AI workloads.
Bridging the utility gap
For Aguirre, the industry’s biggest challenge is no longer how much energy is available, but how quickly power can be delivered where and when it is needed.
“We are not having an energy problem today,” he says. “We have more of a power and timeline problem.”
As AI campuses grow larger, developers are increasingly constrained by the time required to secure utility interconnection agreements, upgrade transmission infrastructure and add new grid capacity. Those processes can take several years, while operators are under pressure to bring new facilities online much sooner.
“The bigger the data center gets, the bigger the problem gets in terms of how quickly they can get utility power delivered to them,” Aguirre explains.
The result is a growing mismatch between utility timelines and AI deployment schedules. Rather than waiting three, four or even five years for grid connections, operators are looking for ways to accelerate speed to power while maintaining the flexibility to integrate utility supply when it becomes available.
More broadly, Aguirre argues that the ability to secure power quickly is becoming a competitive differentiator. Increasingly, developers are competing not only on location or design, but on how rapidly they can bring capacity online.
“Speed to power is becoming a competitive advantage between these developers today,” he says.
Building before the grid arrives
One approach is what Aguirre describes as “bridge power” – on-site generation and microgrids that allow facilities to begin operating while grid connections are still under construction.
“Data centers can’t wait,” he says. “They need power today.”
These systems may combine natural gas generators, batteries, fuel cells and other distributed energy resources to create self-sufficient microgrids capable of supporting operations from day one. Instead of building campuses in phases as additional utility capacity becomes available, operators can complete larger developments on their preferred timetable before transitioning to grid power when it eventually arrives.
Crucially, Aguirre argues that this infrastructure does not become redundant once utility power is available. Instead, on-site generation can continue providing resilience and backup capacity throughout the life of the facility, allowing operators to maximize the value of their investment.
Energy storage takes on a new role
AI is not simply increasing power demand; it is changing the character of electrical demand itself. Unlike conventional enterprise workloads, AI training and inference create rapid, unpredictable swings in consumption that expose the limitations of traditional data center power architectures.
Conventional UPS systems were designed to provide ride-through during outages and keep critical IT systems running until backup generation came online. Rapid changes in workload create sharp fluctuations in power demand, requiring energy storage to play a more active role in day-to-day operations by smoothing sudden spikes before they reach the wider electrical infrastructure.
“We’re not really just backing up the load,” says Aguirre. “We are shaping it.”
Batteries are no longer expected to sit idle until an outage occurs. Instead, they are increasingly being used to smooth these spikes, helping generators operate more efficiently while reducing stress on utility connections. They can also support ride-through requirements and help operators meet emerging grid expectations for large electrical loads.
For Aguirre, that distinction is important. Energy storage is no longer simply a reserve power source. It is becoming an active part of the power architecture, continuously helping balance AI workloads while improving the efficiency of both on-site generation and utility connections.
Aguirre argues that this broader role is making energy storage an increasingly important part of AI-ready power infrastructure. Alongside improving resilience, batteries can help filter rapid load changes, optimize on-site generation and provide greater operational flexibility as workloads continue to evolve.
PowerSecure’s response is its Uninterruptible Battery Energy Storage System (UBIS), which combines UPS and battery energy storage functions into a single platform. Aguirre presents it as part of a broader microgrid architecture designed to support both conventional IT workloads and increasingly demanding AI applications.
Energy storage becomes integral
Looking ahead, Aguirre believes energy storage will become a fundamental part of AI-ready power infrastructure rather than an optional layer of resilience.
“It’s going to be a native component in the system,” he says.
Batteries will increasingly be used to shape AI workloads, support on-site generation and help operators meet evolving utility requirements. As AI deployments continue to grow, Aguirre expects energy storage to play a more active role in balancing increasingly dynamic loads while improving the overall flexibility of data center power systems.
For Aguirre, the defining challenge of AI infrastructure is no longer generating enough energy, but delivering enough power, quickly enough, to keep pace with deployment. As operators look beyond traditional grid connections, flexible power architectures combining on-site generation, microgrids and energy storage become essential tools for bringing new AI capacity online on schedule.
Watch the full DCD>Talks episode with PowerSecure’s Joaquin Aguirre here.
More from PowerSecure
