KenGen Risks Losing Olkaria Wells Over 18-Year Licence Blunder Kenya Electricity Generating Company (KenGen) faces a serious legal and operational risk at Olkaria after an error in the coordinates of its geothermal licence left part of the company’s wells, steam infrastructure and other assets outside its legally recognised area. The licensing mistake, made nearly two decades ago, remains unresolved despite the Ministry of Energy and Petroleum acknowledging the problem in 2017. The dispute has raised questions about the security of KenGen’s investments, the government’s management of geothermal resources and the potential consequences for Kenya’s electricity supply. Olkaria is the country’s most important geothermal centre and accounts for the overwhelming majority of Kenya’s installed geothermal generation capacity. Licence Error Exposes Strategic Assets The problem dates back to September 19, 2008, when the Ministry of Energy and Petroleum issued KenGen a licence covering ...
Geothermal Data Centers for AI: How Hyperscale AI Data Centers Use 24/7 Carbon‑Free Baseload Geothermal Power for Cooling, Reliability and Sustainable Energy
Geothermal Data Centers: Why AI Is Driving the Next Wave of Geothermal Demand
Explore how hyperscale AI data centers are creating demand for 24/7 carbon-free electricity and why geothermal is becoming an attractive solution. Cover power requirements, cooling, reliability, and opportunities for geothermal developers.
Artificial intelligence is changing the physics of the internet, turning data centers into some of the most energy‑hungry buildings on the planet and they now need clean, round‑the‑clock power that solar and wind alone can’t easily provide. Geothermal is suddenly moving from niche to strategic, becoming a compelling way to power and cool hyperscale AI data centers with 24/7 carbon‑free baseload electricity.
The AI Data Center Energy Crunch
AI data centers are not like traditional server farms that mostly handle web traffic and cloud storage. They run intensive GPU clusters for training and inference, drawing massive and constant loads of power. In the U.S. alone, data centers are expected to rise from roughly 4.5% of total electricity consumption in 2024 to 7–12% by 2028, driven heavily by AI workloads.
Three forces are converging:
- Hyperscale energy demand
AI training clusters can require hundreds of megawatts on a single campus, and large regions are planning gigawatt‑scale expansions just to keep up with AI, cloud and semiconductor growth.
- 24/7 utilisation
Unlike typical enterprise IT, AI clusters often run close to full utilisation for long periods, which flattens the load profile into a near‑constant demand instead of peaks and lulls.
- Carbon‑free obligations
Tech giants have committed to 24/7 carbon‑free electricity, not just annual offsets, which means every hour of data center operation must be matched with zero‑carbon power.
This combination creates a problem: intermittent renewables and fossil-heavy grids were not designed for this kind of hyperscale, always‑on demand.
Why Geothermal Fits AI’s 24/7 Carbon‑Free Needs
Geothermal is uniquely positioned to solve this challenge because it behaves more like a conventional baseload power plant—only without the emissions. It taps heat from beneath the earth’s surface to produce continuous electricity and, in many configurations, useful thermal energy for cooling.
Key attributes that match AI data center needs:
- Baseload power and high capacity factor
Geothermal plants can run at 90%+ capacity factors, meaning they deliver nearly their full rated output most hours of the year, in stark contrast to solar and wind which vary by weather and time of day.
- 24/7 carbon‑free electricity
Because geothermal output is constant and renewable, it can directly support 24/7 decarbonization goals without relying as heavily on large-scale batteries or backup fossil plants.
- Grid stability and firmness
Firm geothermal power reduces grid stress in regions where AI data center growth is pushing peak demand and causing congestion, helping utilities manage both reliability and clean energy targets.
Analyses suggest next‑generation geothermal could meet up to 64% of expected data center demand growth by the early 2030s, and even 100% in scenarios where facilities are strategically sited in geothermal‑rich regions. That makes geothermal not just a green option, but a realistic backbone for AI infrastructure in certain geographies.
Power Requirements: From Megawatts to Gigawatts
AI data centers are hyperscale by design. A single large campus can draw hundreds of megawatts, and long‑term build‑outs are increasingly described in gigawatts. As power demand climbs, developers and utilities face a set of hard constraints.
How big is “hyperscale energy”?
- A single hyperscale AI data center can rival the load of a mid‑sized city, especially as GPU clusters scale and more workloads shift to AI inference.
- Projections show total data center electricity demand could double in some regions within the decade, driven by AI, cloud services and high‑performance computing.
Traditional grid‑connected strategies—simply buying more power from whatever mix the grid supplies—clash with 24/7 carbon‑free goals when the underlying grid still relies heavily on fossil fuels.
Why geothermal is appealing at this scale
Geothermal offers several structural advantages for hyperscale energy:
- Co‑located generation
Geothermal plants can be built near or directly tied to AI data centers, reducing transmission losses and easing grid congestion.
- Long asset life
Geothermal plants typically have lifetimes spanning decades, which aligns with the long‑term nature of data center campuses and gives investors predictable cost structures.
- Competitive economics
Levelized cost of electricity (LCOE) for geothermal is commonly reported in the range of roughly $61–102/MWh today, with enhanced geothermal systems (EGS) expected to fall closer to $50/MWh by the mid‑2030s as technology matures and incentives apply. For hyperscale buyers, that means stable pricing without the volatility of fuel markets.
One study estimated that a theoretical 1‑GW geothermal‑powered data center could save billions of dollars in operating costs over 30 years compared to conventional fossil‑heavy power, largely due to reduced backup infrastructure and fuel volatility.
Keeping AI chips cool is as much of a challenge as powering them. High‑density racks of GPUs produce enormous heat, and conventional data centers rely on power‑intensive chillers and evaporative cooling systems that consume both electricity and water.
Geothermal introduces a different approach: use the earth itself as a thermal battery.
Underground thermal energy storage
Aquifer thermal energy storage (ATES) and other underground geothermal cooling concepts use subsurface formations to absorb and release heat over time.
- In cooler months, data centers can “charge” the aquifer by depositing waste heat underground.
- In warmer months, the stored cool water or rock is tapped for heat exchange, offsetting mechanical cooling needs.
Researchers have found that such systems can significantly reduce peak cooling demand, cut electricity consumption, and dramatically reduce water use compared with traditional evaporative cooling. This is especially attractive in regions already facing water stress—an issue that is gaining attention as AI‑related data center build‑outs expand in semi‑arid areas.
Cooling cost reductions
Geothermal systems can leverage stable underground temperatures to reduce cooling costs by an estimated 30–40%, creating a dual win: lower operating expenses and improved environmental performance. For AI data centers, where cooling can represent a large fraction of total power use, those savings compound over time.
This is why you are starting to see designs for “geothermal data centers” that integrate power generation and thermal management into one combined infrastructure, rather than treating cooling as a separate, bolt‑on system.
AI workloads are unforgiving. Training runs can stretch over weeks, and a single unplanned outage can waste millions of compute hours and electricity. AI inference systems that power real‑time products—from search to autonomous tools—also demand extremely high uptime.
Geothermal reinforces reliability in several ways:
- Minimal weather dependence
Unlike solar and wind, geothermal output is largely independent of surface weather conditions, delivering predictable baseload even during long cloudy or calm periods. [9][10]
- Reduced need for backup
Because geothermal is firm, data center operators can rely less on diesel generators or gas peakers for resilience, further reducing emissions and operating costs.
- Long‑term performance
Modern geothermal systems are engineered for decades of operation with stable output, which aligns with data center uptime commitments often measured in “five nines” (99.999%) of availability.
This reliability is one reason large technology companies and chip makers are starting to actively explore geothermal as part of their energy mix, including partnerships that bring AI expertise to geothermal exploration and optimization itself.
Opportunities for Geothermal Developers
For geothermal developers, the rise of AI data centers represents one of the most significant new demand centers since the advent of electric vehicles and utility‑scale renewables.
Several opportunity areas stand out:
1. Long‑term power purchase agreements (PPAs)
AI data center operators need certainty on both price and carbon content. That opens the door for:
- Multi‑decade PPAs tied to geothermal plants, providing stable revenue streams for developers and price certainty for data center operators.
- Contracts structured around 24/7 carbon‑free delivery rather than annual averages, which fit geothermal’s baseload profile perfectly.
Enhanced geothermal systems, which can be developed in more locations than conventional hydrothermal resources, further expand the addressable market for such deals.
2. Co‑located and modular geothermal data centers
Developers and infrastructure funds are increasingly exploring co‑location—placing data centers directly next to geothermal plants.
- Modular geothermal data centers use standardized building blocks and pre‑engineered energy modules to speed deployment, making it easier to align new AI capacity with new geothermal capacity.
- Co‑location reduces transmission build‑out needs and can simplify permitting by tying data center expansion to a clearly green energy source.
In some cases, the geothermal plant itself becomes a central design feature of an “AI campus,” integrating high‑performance computing with renewables‑first branding that appeals to investors and customers.
Geothermal developers with access to suitable aquifers or subsurface formations can offer specialized services focused on cooling:
- Designing and operating aquifer thermal energy storage systems for data centers.
- Providing performance guarantees around reduced peak cooling demand, lower water use, and improved power usage effectiveness (PUE).
This effectively turns geothermal developers into both power providers and thermal infrastructure partners, deepening their role in the data center ecosystem.
4. AI‑enabled geothermal exploration and operations
There is also a technological feedback loop: AI itself is being used to unlock more geothermal potential.
- AI models help identify promising geothermal sites by analysing seismic, geological and operational data.
- Machine‑learning tools optimize drilling strategies and reservoir management, lowering costs and improving output stability.
Partnerships between AI leaders and geothermal firms show growing confidence that geothermal can be scaled more rapidly than in the past, driven by better subsurface intelligence and targeted investment.
Challenges and What Needs to Change
Despite its promise, geothermal is not a plug‑and‑play solution everywhere. There are clear challenges:
- Resource location and exploration risk
Conventional geothermal requires favorable geology, and exploration carries technical and financial risk, especially in new regions.
- Upfront capital intensity
Drilling and plant construction demand high upfront investment, even if operating costs are low over the asset’s lifetime.
- Regulatory and permitting timelines
Geothermal projects can face lengthy permitting processes, which may not align with the rapid build‑out timelines AI companies are accustomed to.
To fully capture the AI data center opportunity, several shifts are needed:
- Policy support and incentives for enhanced geothermal systems, including tax credits and streamlined permitting, to broaden the geographic footprint.
- Standardized contract structures between data center operators and geothermal developers, focusing on 24/7 carbon‑free commitments rather than generic renewable procurement.
- Integrated campus planning where energy and cooling infrastructure are designed together with AI workloads from the start, rather than added as separate components later.
These changes would lower development risk, accelerate timelines, and make geothermal a default consideration in hyperscale planning rather than an exotic add‑on.
The Road Ahead: Geothermal as Strategic AI Infrastructure
As AI reshapes industries—finance, health, manufacturing, science—the hidden infrastructure behind it will increasingly determine who can scale and who stalls. Energy and cooling are no longer just operating costs; they are strategic constraints.
Geothermal stands out because it offers:
- Firm baseload power aligned with 24/7 carbon‑free goals.
- Integrated solutions for both electricity and cooling, cutting power and water use.
- Long‑term cost stability and resilience amid rising demand and grid stress.
For AI data centers, choosing geothermal is not just an environmental statement; it is a competitive advantage. The operators that secure reliable, clean baseload energy—and cool their machines more efficiently—will be the ones able to train bigger models, serve more users, and do so at lower risk and cost.
For geothermal developers, the message is equally clear: AI is driving the next wave of geothermal demand. Those who move now—building projects, partnerships and expertise aimed at AI data centers—can help define the energy backbone of the AI era.
Comments
Post a Comment