The future of datacenters depends on behind-the-meter energy generation

Posted

For the last decade, the datacenter industry has been a quiet engine of the digital economy. It was a background utility that just worked. However, datacenters are no longer passive consumers of electricity. They are becoming active participants in energy generation, storage, and grid management through behind-the-meter infrastructure. This is being driven by the explosive rise of AI and a manufacturing renaissance. Power demand is growing again, increasing an average of 2.1 percent annually over the past five years, with datacenters emerging as a dominant driver of long-term growth.

Electric Power Research Institute projects that datacenters could consume as much as 17 percent of U.S. electricity by 2030, more than triple their current 4-5 percent share. This massive spike in demand is clashing directly with a severely constrained power grid. Key regional utility transmission organizations are expected to face critical reserve shortfalls as early as 2028. At the end of 2025, more than 2,060 GW of new generation and storage capacity was represented by projects actively seeking interconnection to the U.S. grid. For the projects that successfully reached commercial operation, the process took a median of more than five years from request to completion.

To keep building at the speed of innovation, datacenter operators are fundamentally rethinking their relationship with power, shifting from being passive grid consumers to active, localized energy producers. Grid capacity is no longer the ceiling. Instead, the next chapter of digital infrastructure will be written behind the meter.

Behind-the-meter (BTM) refers to energy assets located on the customer side of the utility connection, including battery systems, microgrids, generators, fuel cells, and distributed energy resources that can operate independently of grid constraints.

The Role of Battery Energy Storage Systems (BESS)

While every datacenter developer faces grid constraints, the industry’s two dominant business models are adopting distinctly different strategies to bypass the power bottleneck.

Hyperscale Operators
Hyperscale operators (tech giants like Amazon, Google, and Microsoft) utilize their massive balance sheets to secure gigawatt-scale capacity directly. Rather than acting as passive grid customers, hyperscalers are taking direct, active control of their energy supply chains. They negotiate custom, state-level tariffs directly with utilities, co-locate facilities next to large nuclear or baseload assets, and sign massive off-site Power Purchase Agreements (PPAs). Increasingly, hyperscalers are also deploying large-scale, on-site Battery Energy Storage Systems (BESS) to bypass grid delays and accelerate their projects. Many are also turning to behind-the-meter natural gas generation to provide additional power on-site. With larger, hundreds-of-megawatts turbines facing long backlogs, some operators are instead deploying multiple smaller, roughly 50 MW turbines in parallel, either on-site or adjacent to the campus.

Colocation Providers
In contrast, colocation (colo) providers lease space, power, and cooling to third-party tenants. Colos operate on a highly competitive, faster-to-market model where speed-to-revenue is the primary metric. Colocation facilities are geographically constrained, as their facilities must be located close to business hubs to minimize latency rather than simply next to remote power plants.

Consequently, datacenter and colocation developers are increasingly integrating on-site, behind-the-meter energy solutions. In highly congested markets, BTM generation and battery storage can provide a path to power while developers wait for grid interconnection. In Virginia, for example, LS Power is pursuing a 300 MW datacenter project designed to receive power directly from its adjacent natural gas-fired Doswell power plant, creating a behind-the-meter supply arrangement that can help accelerate access to power.

At the same time, the rules for connecting large loads are beginning to change. In some U.S. markets, datacenters that can provide flexible load, on-site generation or storage are gaining new pathways to connect while giving grid operators greater control over demand during periods of constraint.

The New Coordinated Energy Stack: UPS + BESS + Generators

Historically, a datacenter’s power architecture was passive, consisting of an uninterruptible power supply (UPS) to bridge 5-to-15 minutes of downtime and backup diesel generators to handle prolonged outages.

AI has shattered this legacy design. Although many enterprise deployments still operate below 20 kW per rack, next-generation AI clusters are increasingly exceeding 100 kW per rack, with some designs targeting significantly higher densities.

Furthermore, AI training runs create highly volatile, synchronized compute loads, causing electricity draw to spike and drop in massive, sub-second waves.

To protect both sensitive computing equipment and the local grid from these fluctuations, operators are deploying a layered, grid-interactive energy stack:

  1. UPS (Instant Protection): Positioned in series with the servers, the traditional UPS remains essential for providing millisecond-fast power conditioning and preventing data loss during transient voltage sags.

  2. BESS (On-site Shock Absorber & Optimizer): Connected in parallel at the facility level, BESS acts as a real-time shock absorber. It smooths out volatile AI compute spikes, stores cost-effective off-peak power for use during peak hours, and participates in grid-balancing services.

  3. On-site Generation (Long-Term Firm Power): Customer-sited generation, including natural gas microturbines, provides the ultimate firm baseload capacity, insulating the facility from grid failures for days or weeks.

By layering these technologies, operators can stop treating their power systems like a pricey “insurance policy” that sits idle, and start using them as a smart, money-saving asset. In many markets, demand charges (the fees utilities charge based on peak power draw) make up 30-to-70 percent of a datacenter’s monthly electricity bill. By discharging on-site batteries during peak demand periods — a strategy known as peak shaving — operators can reduce the peak loads used to calculate demand charges, potentially lowering electricity costs significantly.

Turning Backup Assets into Strategic Grid Partners

Historically, backup diesel generators sat untouched, accumulating maintenance costs unless a blackout occurred. Today, utilities and datacenter operators are establishing mutually beneficial partnerships. Through Virtual Power Plants (VPPs), utilities can aggregate a datacenter’s distributed backup generators and BESS capacity into a single, software-controlled resource. Under grid stress, the utility can dispatch these on-site assets, effectively turning the datacenter into a power supplier rather than a power drain.

A prime example is Google’s three-year partnership with Voltus, which will aggregate up to 100 MW of flexible distributed energy resources across the PJM footprint, the regional transmission organization that coordinates electricity supply and demand across 13 states and the District of Columbia. The effort is designed to support grid reliability and help unlock capacity for datacenter growth.

In some regions, regulators are increasingly requiring large energy users to build flexibility into how they operate. For example, in Texas, Senate Bill 6 establishes new requirements for large loads of 75 MW or more, including the ability to curtail grid consumption during energy emergencies. For datacenters, meeting those requirements will depend on flexible workloads, on-site generation, and battery energy storage that can help maintain critical operations while reducing grid demand.

Crucially, incorporating BESS into this stack minimizes the environmental impact of these curtailment events. Instead of firing up diesel generators for a brief 10-minute grid dip, the BESS handles the short-duration event, cutting backup generator runtimes and reducing local emissions.

The Rise of Energy-as-a-Service for Datacenters

For C-suite executives, the primary barriers to deploying behind-the-meter microgrids have historically been high upfront capital requirements, complex custom engineering, and a lack of on-site electrical expertise.

While Small Modular Reactors (SMRs) represent a promising future for carbon-free baseload power, large-scale commercial deployment is primarily anticipated in the early 2030s. In the years leading up to that transition, behind-the-meter strategies, such as natural gas microturbines, can serve as critical near-term power solutions.

To overcome these hurdles, the industry is standardizing microgrid architectures. Rather than custom-building every system from scratch, technology leaders are validating fully integrated electrical and control systems in factory environments before they ever reach the field. This standardized approach reduces engineering variability and can bring a complex microgrid online in as little as two weeks after site delivery.

Furthermore, the emergence of Energy-as-a-Service (EaaS) has rewritten the financial playbook. Under an EaaS model, third-party developers design, finance, construct, and operate the behind-the-meter energy assets. The datacenter operator avoids massive CapEx, paying only for the delivered power through predictable, long-term OpEx-based service contracts.

Major institutional capital is lining up to fund this transition.

New financing platforms are emerging to address capital constraints. Schneider Electric’s Accelerating Resilient Infrastructure (ARI) Initiative has assembled over $7.5 billion in private clean energy capital to deploy resilient microgrids at critical facilities. Similarly, the newly launched Faraday Energy platform — a joint venture backed by $500 million in equity capital from Seraya Partners and Schneider Electric — specifically targets the development and long-term ownership of distributed, customer-sited energy infrastructure for datacenters and industrial operators.

The New Energy Playbook

The days of relying solely on the grid are over; the relationship between datacenters and the energy network has fundamentally shifted. In this new era, power security is the ultimate competitive advantage. By moving from being passive consumers to proactive energy managers through microgrids and localized storage, the industry isn't only securing its own growth. It is actively architecting the resilient, grid-interactive infrastructure that our digital future demands.

Supplied by Schneider Electric

ai and ml