Four nines, five nines, six nines – the language of uptime, and the standard the data center industry is chasing today.

Resiliency in uptime has become a critical consideration where data center investors and end users are concerned. For years, though, it was treated as more of an afterthought, tacked on via diesel backup systems designed to cover outages. Capital was deployed without resiliency designed in from the outset, forming a reactive model that was neither secure nor cost-effective enough.

And that assumes grid connection can even be achieved in time. The US data center market is expanding faster than the electric grid can keep pace, leaving data center developers increasingly facing multi-year delays in grid interconnection. These holdups directly affect speed to revenue, pushing back opening dates and increasing financial risk.

By designing for resiliency upfront – where assets operate as prime power from day one and later transition into backup once utility supply is established – operators can avoid stranded capital and realize greater long-term value.

Where is the balance?

The AI race is driving a surge in power demand. With highly fluctuating requirements for computing power, characterized by spikes and troughs over a matter of milliseconds, data centers need to become more powerful and more resilient to handle these high-density, energy-intensive workloads.

With larger sites, faster load swings, and tighter grids, maintaining that balance is becoming increasingly difficult. Andy Capps, executive vice president of engineering, procurement, and construction at PowerSecure, illustrates:

“Sites that used to be 50 megawatts are now entering the gigawatt scale, putting a huge strain on utilities. These GPUs can have 25-30 percent load swings in cycles of less than a second, and even if you mitigate ten percent of those swings, that’s still a 20-megawatt fluctuation every second. From a balancing perspective, that wreaks havoc on the system.”

A simple generator may not be able to handle these load swings, and utilities are feeling the strain as well.

This, coupled with a shift toward less congested, more rural facility locations, has sparked a growing requirement for fast power solutions. Utilities may require several years to deliver the necessary transmission, distribution, and substation infrastructure, meaning access to permanent power is often delayed.

Even where joint bridge-utility solutions are possible, lead times can stretch years, highlighting just how difficult it is for utilities to build at the pace required – they cannot overbuild, given their fiduciary duty to ratepayers.

“Where data centers were traditionally built close to existing fiber, network, and power infrastructure, developers are now trying to find land for massive campus-scale deployments. Without utility interconnection, they could be waiting three, four, even five years for distribution lines before service is available,” says Capps.

As powered land becomes scarcer and available capacity condenses, developers are increasingly turning to alternative solutions to meet demand.

Faith in connectivity

Against that backdrop, a broader question emerges: what does this say about confidence in near-term grid availability altogether?

For some developers facing constrained infrastructure, the response is to bypass the grid entirely.

PowerSecure, by contrast, has long positioned itself as a grid-interactive company, using conventional generation to capture demand response value. As microgrids, renewables, storage, and fuel cells have matured, the company has evolved – maintaining that grid interaction while expanding its role in delivering integrated resiliency solutions.

For Capps, the issue ultimately comes down to alignment between energy infrastructure and data center timelines:

“Data center customers do not want to own energy infrastructure. They do not want to be in the energy business. They do not want to compete with the local utilities. They want to partner with them. The challenge is getting the transmission, distribution, and substation builds completed on a timeline that meets operators’ expectations – which is where bridge power comes in.”

A structural shock absorber

Where a microgrid is focused on long-term resilience and control, a bridge solution quite literally bridges the gap between construction and grid connection.

The concept of bridge power – whether combustion turbines, fuel cells, or natural gas – is that it takes the place of the utility until it can arrive.

A bridge power solution is often considered rental equipment designed to be removed once it has served its purpose. What, then, is the difference between short-term, single-purpose temporary power and infrastructure that is designed to last and evolve with the site?

“One way to do that is to create a microgrid that can serve as a bridge power technology from day one, and when the utility does come in, it can use that asset as a demand response resource, like a peaking plant,” Capps explains, adding:

“More operators are understanding that they’re going to have to bring some skin in the game to move up in the interconnection queue, or to alleviate concerns around rate-based players having to pick up the tab for a data center entering the market.”

Consider a site with 200MW of AI load and a bridge power plant sized slightly larger – say 225MW to provide redundancy. Unlike a utility grid with gigawatts of capacity, this system is not ‘stiff’ enough on its own to tolerate large, rapid load swings. If AI load is connected directly to the bridge plant, sharp spikes and drops in demand can create frequency and voltage instability. As a result, generators struggle to respond, and reliability suffers.

To address this, some microgrids incorporate battery energy storage in unique ways – referred to by PowerSecure as an uninterruptible battery energy storage system, or UBESS. This approach aligns battery storage with an uninterruptible power supply (UPS).

“With the UBESS, the DC bus operates up to 1500 volts, allowing us to achieve two, three, even four megawatts of capacity per unit. That reduces the need for traditional UPS systems for shorter ride-through times, while providing AI load smoothing,” says Capps.

In practice, the UBESS buffers load swings so the utility – or bridge generation – sees a more constant, predictable profile rather than the volatility typical of AI-driven demand.

Positioned between the bridge power plant and the data hall, it follows the spiky AI load on one side while smoothing that profile on the other. It absorbs sharp demand spikes and fills in dips, acting as a shock absorber for the system.

For AI-heavy sites, where bridge power is intended to function as a primary supply during the interim period, this kind of buffering is effectively non-negotiable.

The difference lies in intent and integration

PowerSecure follows a “bridge plus” approach, moving from a bridge solution toward a fully integrated microgrid that can be used for resilience or as grid support post-utility interconnection, removing the risk of stranded capital – a strategy that aligns closely with the continued prominence of natural gas in large-scale data center deployments.

Natural gas remains the top choice for many operators in the US, due to its wide availability, transportability, and cost-effectiveness. It can be compressed, liquefied, and deployed at scale, making it well-suited to bridge power applications. At the same time, combustion turbine lead times are stretching.

“People are out scouring for used or canceled equipment so they can secure a place in the queue, as lead times now exceed even those for permanent power. As a result, we’re seeing more and more developers leaning toward more available natural gas,” says Capps.

Within this context, PowerSecure’s plug-and-play architecture enables modular scalability. A site might begin with natural gas generation as a prime power bridging asset, but the system is designed to accommodate additional technologies, such as solar or hydrogen fuel cells, over time, and to allow for other use cases when the utility arrives, without requiring a complete redesign.

“You’re using the asset either as a resiliency resource after the bridge period, or as an economic asset – a broader benefit to the overall grid system,” he adds.

With a modular, centralized system, capacity can also be added as demand grows. Over time, it can evolve into an asset that supports the utility, enhances site resiliency, or delivers both, as Capps illustrates:

“You don’t want to treat it like an all-you-can-eat buffet – your eyes will be bigger than your stomach. You need to be realistic about ramp rates and ensure your design has the modularity to meet that demand at the lowest cost possible.”

DC to DNA

Building on that shift toward modular, integrated infrastructure, the next evolution may lie in how power is delivered within the data center itself.

In the near future, data centers are expected to move toward high-voltage DC delivered directly into the data hall. In this model, solid-state transformers sit between the grid or on-site generation and the IT load, flexibly converting between AC and DC in both directions.

The result is a more efficient and modular power architecture, with significant implications for how on-site generation is integrated. With microgrids, batteries, and other distributed resources easier to incorporate, highly dynamic AI loads can be managed more effectively through precise control of power electronics and system topology.

“That’s going to change the dynamic of the data center as well, representing a realistic next step in how data center and microgrid power systems are physically designed,” explains Capps.

While practical 800VDC and AC/DC-flexible transformers represent a clear, engineering-led evolution in response to AI-driven demand, looking further ahead, Capps points to a more ambitious “what if”:

“Imagine we could encode data in DNA so that a teaspoon could hold everything now stored in an entire hyperscale data center. That fear of overbuilding conventional infrastructure is another reason bridge power and modular microgrids are so attractive.”

The point is not that this shift is imminent, but that the long-term trajectory of computing and storage is radically uncertain. That uncertainty strengthens the case for modular, bridge-style approaches – rather than committing to large, irreversible infrastructure bets that could become obsolete within ten to 20 years.

Taking control of your energy strategy

Within this context, taking control of the energy strategy becomes essential. Engaging experts early – alongside decisions on land and IT infrastructure – helps operators define a clear, adaptable path through uncertainty by balancing speed to market with long-term resilience and value.

As power constraints, evolving technologies, and shifting demand profiles continue to reshape the landscape, those that treat energy as a core strategic pillar – rather than a downstream consideration – will be best positioned to scale efficiently and remain competitive.

For more information, please visit powersecure.com/bridge-power.