How to handle energy cost management for IT leaders

Why energy cost management is becoming an IT priority

The following structural forces drive up energy demand across enterprise IT.

AI and high-performance computing. GPU-intensive workloads require energy for computation and for the cooling, power distribution and backup infrastructure that surrounds them.

"On-premises or enterprise data center AI workloads … don't just require energy to run," said Todd Jacobs, founding member of the Theia Institute Think Tank. "They also impact the need for airflow and cooling. [They] require additional energy to power server racks, enterprise-grade uninterruptible power supplies and backup power, and other infrastructure and operational components of data centers that most people don't think about."

Data center power consumption. Electricity is already the single largest ongoing expense for data center operators. Data center electricity demand grew by 16% in 2025 alone, according to Gartner, and the electricity use of AI-optimized servers is expected to rise nearly fivefold by 2030.

Cloud and hybrid infrastructure expansion. As hybrid and multi-cloud deployments expand, so does the underlying energy consumption they generate, even when those costs are invisible inside provider invoices.

Electricity price volatility. Energy prices are exposed to geopolitical instability, extreme weather events and regional grid constraints that aren't related to how efficiently a data center runs. These factors make cost exposure increasingly difficult to forecast and budget.

Additionally, the sharp rise in cost per kilowatt-hour since 2022 caught most IT leaders off guard. Based on current projections, the trend is not reversing, said Pavels Gurskis, fractional CIO at IThesion.

Corporate sustainability goals. Regulatory and contractual requirements push organizations to account for and reduce energy-related emissions, adding compliance pressure on top of the operational cost problem.

The hidden cost of digital infrastructure

Metered electricity is only part of the problem. Energy costs also accumulate in cloud service fees, facilities budgets and capacity planning models built before today's AI workloads existed.

AI workloads and training costs. Organizations that deploy AI tools typically track token usage against a foundational model. They often miss the cost of updating the vector database that feeds it and the compute overhead of AI training workloads more broadly. Most organizations lack a full component inventory, so costs accumulate without attribution.

"If you have rapidly changing data and are regularly updating your vector database in your RAG [retrieval-augmented generation] solution with this new information, the cost of executing the embedding model can become a hidden day-to-day cost that is not even usage dependent but data dependent," said Jim Olsen, CTO at ModelOp.

Cloud pricing opacity. Energy costs absorbed into service fees do not appear as discrete line items in most capacity planning models, which makes them nearly impossible to manage actively.

"Energy costs are embedded in pricing by cloud providers and usually aren't broken out separately," Jacobs said.

Inefficient server utilization and data storage growth. Underutilized servers consume power around the clock, even when idle. Rapidly scaling data storage compounds this, adding continuous energy overhead that rarely appears as a visible line item until it is already embedded in operational costs.

Supply planning mismatches. Facilities are sized on five- to 10-year growth projections. A merger, acquisition or rapid GPU deployment can outpace those projections years ahead of schedule, pushing power density well beyond what the facility was built to support.

"You pick a facility based on five- to 10-year growth projections, and at year five, you need double what's planned due to M&A," Gurskis said.

Legacy cooling and power infrastructure. There are also hidden costs associated with aging cooling systems and power distribution infrastructure.

"As we continue to rebalance cloud and on-prem, the inefficient, legacy cooling and UPS [uninterruptible power supply] power distribution systems are often ignored, resulting in inefficiency and excess capacity," said Greg Sanker, principal advisory director at Info-Tech Research Group.

Risks of poor energy cost management

Poor energy cost management exposes the organization across six distinct areas.

Rising operational expenses. Energy costs that were not budgeted for and electricity prices that shift without warning put direct pressure on operating expenses.

"A sustained increase in the price or demand for energy would be a cross-cutting business concern," Jacobs said.

Infrastructure scalability limits. Facilities running at or near power capacity have no headroom to support new workloads. Organizations that have not actively managed energy density find themselves constrained precisely when they need to scale.

Reduced margins on digital services. When IT leaders don't understand the full compute cost of an AI or digital service, they build pricing and ROI calculations on incomplete data. Without visibility into total costs across all components, organizations cannot accurately assess whether a given initiative generates or erodes value, Olsen said.

Sustainability compliance challenges. Contracts and compliance frameworks that require data availability and system integrity impose energy obligations that grow alongside costs. Also, compliance requirements can catch organizations off guard during rapid expansion into new territories, where local energy regulations and grid conditions may differ significantly from existing operations, Gurskis said.

Increased exposure to energy price volatility. Organizations without deliberate energy procurement strategies are fully exposed to market fluctuations caused by geopolitical instability, extreme weather and grid constraints.

Lack of business continuity. Rising energy costs increase operating expenses and the full cost of continuity coverage that businesses depend on.

"Each fully redundant hot site can often consume as much energy as the primary site unless it's been deliberately scaled down," Jacobs said.

Strategic energy cost management approaches for CIOs

While most CIOs can't influence the actual price of energy, they can manage the effects of rising costs.

IT leaders have the most direct influence across the following four areas.

Infrastructure modernization

• Upgrade to energy-efficient hardware. Modern servers offer more compute per watt than equipment five or more years old, thereby reducing energy costs per unit of work.
• Consolidate servers and virtualize workloads. Reducing the physical footprint to power and cool is one of the fastest paths to measurable savings. "Targeted modernization and consolidation remain strong opportunities for improvement while rarely impacting front-edge innovation and digital growth," Sanker said.
• Evaluate efficiency using unit economics. "Knowing your unit economics -- IT cost per business transaction -- is the key for optimization. It will give you a clear picture [of] on-prem vs. cloud efficiency," Gurskis said.

Data center optimization

• Improve cooling systems. "For on-prem environments, retrofitting cooling and power usually yields better ROI and enables [more] growth than new build-outs," Gurskis said.
• Optimize power usage effectiveness. PUE measures how efficiently a data center converts incoming electricity into useful compute. Organizations running a PUE above 1.5 have meaningful room to improve before investing in new capacity.

Cloud optimization

• Right-size workloads. Oversized cloud instances run at partial capacity, meaning organizations pay for power that delivers no output.
• Choose energy-efficient regions. Cloud providers vary significantly in the energy mix and efficiency of their regional infrastructure.
• Optimize AI training cycles. Training that runs during off-peak periods and avoids redundant retraining can reduce compute and energy overhead. "You must really look at the total compute cost of your full business use case, including all of its assets and models, in order to understand that use case's cost," Olsen said.

Sustainable IT architecture

• Invest in renewable-powered data centers. Sourcing power from renewables is a foundation of sustainable IT infrastructure and reduces both cost exposure and the organization's carbon footprint over the long term.
• Adopt energy-aware computing strategies. Smaller, task-specific AI models that run on standard employee hardware consume less energy than large, centralized deployments. "Technologies based on lower-power devices or distributed computing can be more energy-efficient," Jacobs said.
• Implement carbon-aware workload scheduling. Shifting non-urgent workloads to times and regions where clean energy is available reduces emissions without affecting output.

Energy consumption should be an explicit line item in enterprise IT budgets and a factor in total cost of ownership calculations when evaluating new tools.

"IT leaders must treat energy as a dimension of service delivery, not an afterthought," Sanker said.

Sean Michael Kerner is an IT consultant, technology enthusiast and tinkerer. He has pulled Token Ring, configured NetWare and been known to compile his own Linux kernel. He consults with industry and media organizations on technology issues.