Why Are AI Data Centers Turning to Geothermal? The Stable, Clean Power Solution Nobody Saw Coming

The artificial intelligence revolution is running headlong into an uncomfortable truth: its insatiable energy demands are pushing global power grids to their breaking point. As an AI researcher specialized in renewable energy, I’ve been tracking this looming crisis, and what I found is a surprising, yet incredibly potent, solution emerging from deep within the Earth: geothermal energy. While solar and wind often grab headlines, it's geothermal's unique ability to provide 24/7, baseload clean power that is making it the dark horse candidate for powering the future of AI.

I’ve seen projections that global data center electricity consumption is set to double by 2030, with AI-focused data centers tripling their power usage in the same timeframe. In the United States alone, data centers consumed 4.4% of total electricity in 2023, and this is projected to surge to between 6.7% and 12.0% by 2028. This isn't just about more power; it's about reliable power. AI workloads operate around the clock and cannot tolerate the intermittency inherent in traditional solar and wind farms without massive, and often insufficient, battery storage. This critical need for dispatchable, always-on clean energy is why tech giants are quietly making significant moves into geothermal.

The AI Energy Crunch and the Baseload Imperative

I believe the core issue lies in the nature of AI itself. Training large language models and running complex AI inference tasks requires continuous, high-density computational power. Unlike a manufacturing plant that might pause for maintenance, an AI data center is a relentless machine, demanding a steady flow of electricity. My research indicates that AI operations consume 3-5 times more energy than traditional computing, driving a projected 160% surge in data center power demand by 2030. This creates a massive challenge for grid operators trying to balance supply with demand, particularly as older, carbon-intensive baseload plants like coal are retired. The gap is real, and it’s accelerating the search for reliable, carbon-free alternatives.

Traditional renewables like solar and wind are vital for decarbonization, but their output fluctuates. When the sun isn't shining or the wind isn't blowing, data centers still need power. This intermittency necessitates either fossil fuel backup or gargantuan battery storage, both of which come with their own environmental and economic drawbacks. This is precisely where geothermal energy, with its ability to provide constant, weather-independent power, steps in as a game-changer. It’s a resource that truly operates 24/7, making it an ideal partner for AI's relentless energy appetite.

Big Tech's Bets on Geothermal Acceleration

What truly surprised me was the scale and speed of recent corporate investments in geothermal, particularly Enhanced Geothermal Systems (EGS). I've observed a definitive shift from small-scale technology validation to commercial deployment between 2025 and 2026.

Google has been a pioneer here, initially partnering with Fervo Energy for a 3.5 MW EGS pilot project in Nevada that became operational in late 2023. Building on this success, Google expanded its commitment significantly in 2025-2026. They signed deals for 150 MW of conventional geothermal capacity with Ormat Technologies and a substantial 115 MW EGS project with Fervo Energy, both in Nevada. These landmark agreements were facilitated by an innovative financing mechanism called the Clean Transition Tariff (CTT), developed with utility NV Energy. This tariff structure allows corporate buyers like Google to absorb the technology risk and provide the financial certainty needed for utilities to approve large, long-term clean energy assets, creating a scalable pathway for new geothermal projects. This is a crucial, unexpected angle: innovative financial models are proving as important as technological breakthroughs.

Fervo Energy itself has become a poster child for this burgeoning sector. The company's Cape Station project in Utah is expected to deliver its first 100 MW by early 2027 and aims to scale to 500 MW by 2028, with power purchase agreements already secured with major utilities and energy companies. In May 2026, Fervo made headlines with a blockbuster IPO, raising $1.89 billion and achieving a valuation exceeding $10 billion, signaling strong investor confidence in next-generation geothermal.

Not to be outdone, Meta also announced a deal with XGS Energy in July 2025 for 150 MW of advanced geothermal power to fuel its expanding data centers in New Mexico, with operations projected to begin by the end of the decade. These are not speculative ventures; these are concrete, multi-million and multi-billion dollar investments by the world's largest tech companies, explicitly targeting geothermal for their AI infrastructure needs.

Unexpected Synergies and Stalled Ambitions

One of the most fascinating, and perhaps unexpected, aspects of EGS development is its reliance on technologies and expertise from the oil and gas industry. Advances in drilling performance and reservoir engineering, originally developed for unconventional oil and gas production, are now being leveraged to create and enhance geothermal reservoirs where they don't naturally exist. This cross-industry transfer of knowledge is significantly accelerating EGS commercialization, bringing decades of drilling innovation to the renewable energy sector.

However, the path isn't without its hurdles. A vivid example is the recent stalling of a planned $1 billion Microsoft and G42 geothermal-powered AI data center in Kenya. Announced in May 2024, the project aimed for an initial 100 MW phase to be operational by this year, with a long-term goal of 1 gigawatt. Yet, by May 2026, the Kenyan government indicated that meeting the full 1 GW demand would require