Analysis
Ormat Technologies, a geothermal power company with roughly six decades of operating history, is repositioning its business around a new primary customer: AI data centers, Fortune reported this week.
Why geothermal specifically is having a moment
The AI infrastructure buildout's most persistent bottleneck this year hasn't been chips -- it's been power. Data centers need enormous, reliable electricity supply, and the AI industry has scrambled toward every available source: solar and wind where land and interconnection allow, gas peaker plants for flexible generation, and a wave of nuclear startups like Apollo Atomics promising compact reactors years away from actual deployment. Geothermal offers something distinct from most of those alternatives: baseload, always-on power generation that doesn't depend on weather conditions the way solar and wind do, without the yearslong regulatory approval timeline nuclear projects face before a single reactor comes online.
“## Why geothermal specifically is having a moment The AI infrastructure buildout's most persistent bottleneck this year hasn't been chips -- it's been power.”
Ormat has operated geothermal plants for roughly sixty years, giving it operating infrastructure, permitted sites and technical expertise that a new entrant chasing the same AI-power opportunity would need years to build from scratch. That existing base is exactly what makes Ormat's pivot credible in a way a brand-new geothermal startup's promises wouldn't be -- the company isn't proposing to build new capacity from zero, it's repositioning proven, already-operating generation toward a new class of customer.
The real constraint: geography
Geothermal's core limitation is one gas and nuclear don't share to the same degree: viable geothermal resources exist only where geology cooperates, typically along specific tectonic and volcanic regions, rather than being sitable anywhere with land and grid access. That means Ormat's pivot benefits data centers that can be built near its existing geothermal plants far more than it offers a broadly deployable solution to the AI industry's power problem nationwide. A hyperscaler building a data center far from a viable geothermal resource gets no benefit from Ormat's repositioning regardless of how successful the company's AI-focused pivot proves to be near its own facilities.
Part of a broader pattern: legacy power meets AI demand
Ormat's move fits a pattern extending well beyond geothermal specifically -- established power generators of every type are repositioning existing assets and expertise toward AI data-center demand, which has emerged as one of the most reliable, long-duration power-purchase customers utilities and independent power producers have encountered in years. That's a meaningfully different dynamic than the AI-native energy startups getting more headline attention: Ormat isn't promising a future technology, it's offering to sell power it already knows how to generate reliably to a new class of buyer willing to pay a premium for guaranteed, dedicated supply.
The counterweight
A sixty-year operating history is a real asset, but it doesn't automatically translate into AI-data-center-scale power delivery -- Ormat's existing plant capacity was built and sited for its historical customer base, not for hyperscale AI demand, and expanding output at existing sites or developing new geothermal capacity both take years and significant capital regardless of how favorable current demand conditions look. Geothermal also remains a small share of total US power generation relative to gas, nuclear and renewables, meaning even a fully successful AI-focused pivot for Ormat specifically addresses a narrow slice of the AI industry's much larger power needs.
