Utilities have spent decades chasing efficiency across the grid. New transformers, smarter meters, automated switching, distributed energy programs. Every one of those investments exists for the same reason: deliver reliable power while wasting less of it.
Somewhere along the way, a lot of that same discipline stopped at the data center door.
Walk into most utility data centers today and you'll find applications built twenty or thirty years ago still running around the clock. Disaster recovery environments that haven't been touched in months, still drawing power just in case. Storage volumes ballooning as duplicate data piles up across business units. Cooling systems working overtime to support hardware that costs more to run than it delivers in value.
None of this shows up on an executive dashboard. It doesn't trip an outage alert or generate a customer complaint. It just quietly adds to the operating bill, month after month.
And the numbers back this up. According to the Uptime Institute's 2024 Global Data Center Survey, one in four data centers operates below 40% utilization. Other industry research puts average enterprise server utilization even lower, in the 10 to 15% range, while idle servers can still draw more than half their peak power doing essentially nothing. Some studies estimate that as much as a quarter of servers in enterprise environments qualify as "zombie servers," equipment that's powered on and cooled but performing no meaningful work at all.
Picture building three substations to serve a neighborhood that only needs one. No utility would sign off on that. Yet that's effectively what's happening inside a lot of data centers right now.
For years, sustainability strategy in this industry has centered on generation mix, renewable integration, EV programs, and grid modernization. Those investments still matter. But they're only half the picture.
Every digital system a utility depends on burns electricity too. Outage management platforms. GIS. Asset management. Customer information systems. Increasingly, AI layered across all of it to sharpen forecasting and maintenance planning.
That digital footprint is growing fast, and the electricity behind it is growing with it. The International Energy Agency reports that global electricity demand from data centers rose 17% in 2025 alone, right in line with its earlier projections. Demand from AI-focused data centers specifically jumped 50% in that same year. The IEA's base case now expects global data center electricity consumption to roughly double, from about 415 terawatt-hours in 2024 to close to 945 terawatt-hours by 2030.
That's not a far-off number for utilities. It's a planning problem today. Research validates that data centers already consumed 26% of Virginia's total electricity supply in 2023, along with double-digit shares in Nebraska, Iowa, Oregon, and North Dakota. By 2030, EPRI projects Virginia's share could climb as high as 59%, with several more states crossing the 20% threshold. Goldman Sachs Research is even more direct, forecasting a 165% increase in global data center power demand by 2030 compared with 2023 levels.
So the question for energy and utility leadership isn't whether digital infrastructure consumes energy. It's whether your own infrastructure is consuming more of it than it needs to.
Most utilities didn't set out to build an inefficient IT environment. They built a resilient one, one regulatory requirement, acquisition, and cybersecurity mandate at a time. Very few legacy systems ever get retired, because ripping out critical infrastructure feels riskier than just leaving it running.
The result is an environment full of overlap. Multiple applications doing similar jobs. Servers sized for demand that no longer exists. Nobody wants to be the one who unplugs something and causes a problem, so nothing gets unplugged.
McKinsey research puts a number on what that costs. Organizations carrying significant technical debt can see that debt eat up 20 to 40% of their total technology estate's value. That same research finds that infrastructure modernization, whether through cloud adoption or consolidation, can reduce total infrastructure costs by 30 to 50%.
Unused compute still pulls electricity from the grid. Idle storage still needs power. Underutilized servers still generate heat that cooling systems have to remove around the clock. None of that shows up as a single dramatic cost. It shows up as a slow, steady drain that compounds every single year.
Here's a misconception worth killing early: sustainability doesn't automatically improve just because workloads move to the cloud.
Lift a bloated application into a cloud environment without rethinking it, and you've usually just recreated the same inefficiency on someone else's hardware. Oversized virtual machines stay oversized. Duplicate data keeps duplicating.
The efficiency gain comes from what happens before migration, not after. Research from AWS, drawing on NRDC's data center efficiency assessment, found that cloud environments typically run at around 65% server utilization compared with 12 to 18% for on-premises infrastructure. Hyperscale facilities also operate at power usage effectiveness ratios of 1.1 to 1.2, compared with an industry average closer to 1.56, according to Uptime Institute's most recent survey. That gap represents real electricity, not a rounding error.
Getting there requires asking sharper questions before anything moves:
Answering those questions before migration is what separates real modernization from an expensive game of digital musical chairs.
Utilities can't approach this the way a retailer or a bank might, because reliability isn't negotiable. SCADA, energy management, outage restoration, and substation systems carry performance, security, and regulatory requirements that most business applications never have to meet.
That's why Green IT for utilities is almost never an "everything to the cloud" story. Most organizations are landing on hybrid models: business applications move to scalable cloud platforms, while operational technology stays closer to the grid until modernization can happen without introducing new risk. It's a deliberate, staged approach, not a wholesale leap.
Artificial intelligence has quickly become one of the biggest new sources of computing demand inside utilities, powering vegetation management, predictive maintenance, load forecasting, and customer experience tools. All of that runs on infrastructure, and infrastructure runs on electricity.
Organizations that clean up their infrastructure before scaling AI put themselves in a much stronger position than those layering AI on top of fragmented, aging systems. Efficient infrastructure makes AI efficient. Inefficient infrastructure just gives waste a faster engine.
The hyperscalers know this, and they're spending accordingly. Microsoft has invested heavily in direct-to-chip cooling to cut both energy and water use. Google continues expanding its carbon-free energy portfolio to power its growing AI footprint. Most utilities can't replicate that scale of investment on their own, but thoughtful cloud adoption lets them benefit from it anyway.
Environmental improvement rarely travels alone. Organizations that modernize their infrastructure tend to discover gains in resilience, cybersecurity, disaster recovery, and financial performance at the same time.
Fewer redundant systems means less to maintain. Consolidated applications are easier to govern. Automated resource allocation cuts down on manual admin work. Clearer visibility into how infrastructure is actually being used leads to smarter investment decisions down the line.
Maybe most importantly, a modern foundation makes everything that comes next easier, whether that's AI, advanced analytics, distributed energy resources, or business models that haven't been invented yet.
Before signing off on the next big infrastructure spend, it's worth stepping back with the leadership team and asking:
Utilities figured out a long time ago that squeezing efficiency out of thousands of physical assets adds up to real financial value. The same logic applies to the racks sitting in the data center down the hall.
Green IT isn't about making the technology function look greener on paper. It's about cutting unnecessary complexity, reducing wasted electricity, and building an environment that can actually handle what the next decade is going to throw at it.
For a lot of utilities, one of the biggest remaining opportunities to cut emissions isn't the next renewable generation project but making the technology that runs the business as efficient as the grid it was built to support.
Figuring out which of these questions actually apply to your environment, and what to do about them, is where most organizations get stuck. TSG works with utility leaders to assess infrastructure utilization, identify what can be consolidated or retired, and build modernization roadmaps that respect the reliability and compliance constraints unique to this industry while removing OT and IT silos.
Whether you're trying to right-size your data center footprint, plan a hybrid cloud strategy for business and operational systems, or build the infrastructure foundation your AI initiatives will actually need, TSG can help you turn these five questions into a concrete plan.