Will Data Center Electricity Use Triple by 2028?

The numbers are staggering and only going to get bigger. Data centers already consume roughly 1–2% of global electricity — a figure that sounds modest until you realize it exceeds the annual power consumption of many mid-sized nations. Now, credible forecasts suggest that data center electricity usage could triple by 2028, driven by AI workloads, cloud expansion, and an insatiable demand for compute. For anyone operating in energy infrastructure, land development, or clean power — this isn't background noise. It's the defining demand signal of the decade.

Understanding Current Data Center Energy Use

A modern hyperscale data center — the kind operated by AWS, Microsoft Azure, or Google Cloud — can draw anywhere from 100 to 500+ megawatts of power continuously. That's equivalent to powering a small city around the clock, every day of the year. Unlike industrial facilities that cycle up and down with production schedules, data centers run at sustained load. They don't sleep.

The primary energy culprits are well understood: servers, cooling infrastructure, and power distribution losses. Cooling alone typically accounts for 30–40% of a facility's total energy draw, which is why Power Usage Effectiveness (PUE) has become the industry's signature efficiency metric. A PUE of 1.0 is theoretical perfection — every watt goes to computing. Most enterprise facilities hover between 1.4 and 1.7. Hyperscalers have pushed closer to 1.1–1.2 through aggressive engineering, but they're also operating at a scale that makes even small inefficiencies enormous in absolute terms.

What's changed in the last three years isn't the existence of data centers — it's what they're being asked to do. Training large language models requires orders of magnitude more compute than traditional workloads. A single large-scale AI training run can consume as much electricity as hundreds of average American homes use in a year. Multiply that across thousands of training jobs, inference clusters, and the proliferating AI features baked into consumer apps, and you start to understand why utility planners are scrambling.

Projected Trends: What to Expect by 2028

The tripling projection isn't an outlier view — it's becoming consensus. The International Energy Agency, Goldman Sachs research, and multiple grid operators have all issued reports pointing to a dramatic acceleration in data center energy consumption through the late 2020s.

Several forces converge to make this trajectory credible:

AI infrastructure buildout is accelerating, not plateauing. Every major cloud provider has committed to multi-year, multi-billion-dollar capital expenditure programs specifically targeting AI compute infrastructure. Microsoft announced over $80 billion in data center investment for 2025 alone. Meta, Google, and Amazon are not far behind. These facilities are coming online regardless of short-term economic conditions — the competitive pressure is simply too intense for any major player to pause.

Edge computing is a secondary but meaningful driver. As latency-sensitive applications proliferate — autonomous systems, real-time analytics, industrial IoT — smaller distributed facilities are being built closer to end users. Individually modest, collectively they add up.

The technology angle cuts both ways. Next-generation chips from NVIDIA, AMD, and custom silicon designers are dramatically more capable per watt than their predecessors. But capability improvements have historically led to *more* total compute demand, not less — a dynamic economists call Jevons' Paradox. More efficient hardware tends to unlock new applications and lower the cost per computation, which drives adoption and total energy consumption upward even as per-unit efficiency improves.

Impact on Energy Markets and Pricing

Here's where the implications get genuinely consequential for infrastructure investors and energy developers.

Data center operators are not passive utility customers. They consume at a scale that directly influences grid planning, transmission investment, and wholesale power pricing. When a 500 MW campus comes online in a market like Northern Virginia — already the densest data center concentration on earth — it doesn't just consume available power. It reshapes the regional generation mix, accelerates interconnection queues, and can stress transmission infrastructure that was never designed for that load profile.

The markets that win the next wave of data center development will be the ones that can credibly deliver large blocks of power, fast. That's a remarkably short list. Interconnection queue backlogs in PJM and other RTOs stretch five to seven years. Sites with existing transmission capacity, access to water for cooling, and proximity to fiber backbone infrastructure are commanding significant premiums — and that dynamic is only intensifying.

For energy pricing specifically, the demand surge creates pressure in two directions. On one hand, large data center tenants are sophisticated power purchasers who increasingly procure through long-term Power Purchase Agreements (PPAs) with renewable generators. This has been an enormous tailwind for utility-scale solar and wind development. On the other hand, the raw volume of demand — particularly in constrained markets — is pushing peak wholesale prices higher and creating reliability challenges for grid operators.

An underappreciated angle: data centers are increasingly being viewed as potential grid assets, not just grid loads. With large battery storage systems and controllable UPS infrastructure, they can participate in demand response programs, providing flexibility that helps balance variable renewable generation. This is early-stage but directionally important.

Sustainability and Efficiency Measures

The clean energy math is both encouraging and sobering. Major hyperscalers have made ambitious commitments — Google targeting 24/7 carbon-free energy matching, Microsoft pledging carbon negativity by 2030, Amazon aiming for 100% renewable energy across operations. These aren't just PR. They've translated into some of the largest corporate renewable energy procurement deals in history, funding gigawatts of new solar and wind capacity.

But commitments and reality diverge when demand grows this fast. Matching every megawatt-hour of consumption with clean generation, hour by hour, is a fundamentally harder problem than annual carbon accounting — and the buildout required is enormous.

On the technology side, several approaches are gaining serious traction:

• Liquid cooling and immersion cooling replace air-based thermal management with direct fluid contact, dramatically improving cooling efficiency. This is becoming standard in high-density AI compute environments where air simply can't remove heat fast enough.
• Waste heat recovery — routing data center thermal output to district heating systems or industrial processes — remains underutilized in North America but is advancing in Northern Europe where district heating infrastructure exists.
• Stranded renewable integration connects data centers directly to generation assets in locations where curtailment would otherwise waste clean energy, effectively monetizing both the compute facility and the renewable project simultaneously.
• Nuclear is back on the table. Microsoft's deal with Constellation to restart Three Mile Island Unit 1 is a bellwether moment. Small Modular Reactors (SMRs) are being actively evaluated by multiple hyperscalers as a long-term baseload solution that doesn't depend on transmission access or weather.

Tax policy adds another layer. Several states have moved to exempt data center equipment purchases — including computers and supporting hardware — from sales tax, using these incentives to attract facility investment. These exemptions matter at scale: a billion-dollar facility procurement carries significant tax exposure, and states without exemptions are simply losing deals to those that have them.

Navigating the Energy Future

What the tripling forecast really signals isn't just a supply challenge — it's a structural reorganization of how electricity is generated, transmitted, priced, and consumed.

For infrastructure developers and investors, the opportunity surface is real and specific: transmission-ready land in viable power markets, utility-scale solar and storage projects with offtake potential, water access for cooling, and the specialized expertise to navigate interconnection processes that have never been more competitive or congested.

For grid operators and regulators, the challenge is equally clear: existing planning models, built around relatively predictable load growth, are being stress-tested by demand curves that AI infrastructure is bending in ways nobody fully anticipated even five years ago.

The developers, investors, and energy producers who move decisively now — securing land, capacity, and PPAs before the next wave of data center announcements hits the interconnection queue — will be positioned for the next decade. Those who wait for certainty will find the best sites spoken for and the most attractive offtake agreements already signed.

The tripling isn't a prediction to debate. At this point, it's closer to a planning assumption. The question worth asking isn't whether it happens — it's whether your portfolio is positioned for when it does.

Ready to navigate the future of energy and data centers? Explore opportunities in the InfraSale Marketplace today! [Visit InfraSale Marketplace](https://infrasale.com/marketplace)