Geothermal Energy for AI Data Centers: Why It's Becoming the Unexpected Baseload Power

As an AI researcher specializing in renewable energy, I've been closely tracking the escalating energy demands of artificial intelligence infrastructure, and what I've discovered is a fascinating and often overlooked trend: AI's insatiable appetite for continuous, reliable power is quietly pushing geothermal energy, particularly Enhanced Geothermal Systems (EGS), into the spotlight as a critical baseload solution. This isn't just about 'more renewables'; it's about the right kind of renewables to keep the AI revolution running 24/7.

I've seen the numbers, and they are stark. Data centers, the physical homes of AI, are consuming electricity at an unprecedented rate. Their share of total U.S. electricity consumption more than doubled from 1.9% in 2018 to a staggering 4.4% in 2023, and projections indicate it could reach between 6.7% and 12% by 2028. Globally, the International Energy Agency projects data centers will consume between 650-1,050 TWh by 2026, with AI-focused facilities growing even faster. My research shows that companies like Microsoft are already reporting significant increases in emissions, with their energy use surging by 168% since their pre-ChatGPT baseline, largely driven by AI and cloud expansion. This isn't a future problem; it's a current crisis, and intermittent sources like solar and wind, despite their growth, simply cannot meet this relentless demand alone.

The AI Energy Conundrum: Beyond Intermittency

The core challenge I've identified is that AI operations require uninterrupted power. Unlike traditional computing, AI training and inference demand constant, high-density energy flows. Relying solely on solar or wind, which are inherently variable, introduces significant instability and necessitates expensive, large-scale energy storage or, more often, a fallback to fossil fuels. This is where the unexpected angle emerges: geothermal energy, with its unique ability to provide continuous, carbon-free power regardless of weather or time of day, is perfectly positioned to fill this baseload gap.

I've observed that the traditional grid infrastructure is already struggling. Grid connection bottlenecks and insufficient transmission lines are delaying new power generation projects for years, creating a critical hurdle for data center expansion. Data center operators are desperate for solutions that can come online faster and provide the unwavering power their systems demand. This urgency is what's driving the pivot to less conventional, but more stable, renewable sources like geothermal.

Geothermal's Underestimated Ascent

What truly excites me about geothermal is the rapid advancement of Enhanced Geothermal Systems (EGS). Historically, geothermal power was limited to regions with naturally occurring hot water or steam reservoirs near tectonic plate boundaries. However, EGS technology, which leverages advanced drilling techniques adapted from the oil and gas industry—including horizontal drilling and hydraulic fracturing—can create geothermal reservoirs in hot, dry rock formations where they don't naturally exist. This innovation significantly expands the geographic potential for geothermal development, making it a viable option for powering data centers in many more locations.

My data shows that the global geothermal energy market was valued at USD 74.4 billion in 2025 and is projected to grow to USD 79.06 billion in 2026, with forecasts indicating it will reach USD 118.81 billion by 2034, exhibiting a compound annual growth rate (CAGR) of 5.22%. Even more telling is the projected growth in installed capacity, expanding from 17.35 GW in 2025 to 17.97 GW in 2026, and an impressive 29.5 GW by 2031, at a CAGR of 10.42%. EGS, specifically, is showing the fastest growth, with an 18.80% CAGR through 2031. This isn't just incremental growth; it's a strategic ramp-up driven by the urgent need for reliable clean energy.

Big Tech's Quiet Bet on Deep Earth Energy

I've noticed that major technology companies, facing immense pressure to meet their ambitious net-zero commitments while simultaneously fueling their AI expansion, are making significant, targeted investments in geothermal. For instance, Google partnered with Fervo Energy to bring a 3.5-MW EGS geothermal plant online in Nevada, specifically to power its data centers. This initial project is part of a larger contract that aims to deliver up to 50 MW in 2025-2026, scaling to 100 MW between 2027 and 2036. Furthermore, Google has signed a long-term agreement with Ormat Technologies for up to 150 MW of geothermal electricity in Nevada, with projects slated to come online between 2028 and 2030, directly supporting its growing digital infrastructure.

This isn't an isolated case. My research indicates that tech giants collectively signed 14 geothermal Power Purchase Agreements (PPAs) totaling 635 MW in 2025 alone, representing a threefold increase from 2024. Data centers are now driving 60% of new geothermal capacity, with a long-term target of 120 GW by 2050. This suggests a strategic shift, where tech companies are not waiting for battery breakthroughs to provide 24/7 carbon-free baseload power, but instead are directly investing in proven, dispatchable clean energy sources like geothermal.

Beyond Power: Geothermal for Direct Cooling

Another fascinating, often overlooked aspect I've uncovered is geothermal's potential beyond just electricity generation. Data centers require massive cooling to prevent servers from overheating, with cooling systems accounting for as much as 40% of their total energy consumption. Geothermal energy offers a solution here too, through technologies like Cold Underground Thermal Energy Storage (Cold UTES). This involves injecting cold water into the subsurface, storing it, and then drawing it back to the surface to offset peak cooling demands, potentially reducing the need for new power plant capacity specifically for cooling. Microsoft, for example, has even filed patents for geothermal cooling systems that circulate fluid through deep underground wells to absorb waste heat from servers, effectively using the Earth as a massive radiator. This dual benefit—power and cooling—makes geothermal an even more compelling solution for AI infrastructure.

What to Watch

I believe the convergence of surging AI energy demand and technological advancements in EGS is creating a unique investment and development opportunity. Watch for continued partnerships between hyperscalers and geothermal developers, particularly in regions with high geothermal potential and supportive regulatory frameworks like Utah, which is already being eyed for enhanced geothermal-powered data center clusters. The U.S. Department of Energy aims to reduce the levelized cost of electricity (LCOE) for EGS by 90% by 2035, targeting around $45/MWh, which would make it highly competitive with other energy sources. The scalability and rapid deployment potential of modular EGS plants, with project completion possible in as little as 18 months, will be crucial in meeting the urgent timeline of AI expansion. This is a fundamental shift in how we power our digital future, moving towards a truly continuous, clean energy backbone.

Bottom line: The AI energy crisis is accelerating geothermal's rise from a niche renewable to a cornerstone of the future power grid, providing the 24/7 clean baseload that intermittent sources simply cannot offer alone. I see this as a critical, under-appreciated development that will reshape renewable energy investment for the next decade.