Large computing facilities seek reliable electricity. Where hydroelectric or geothermal generation is abundant, operators may see an opportunity for lower-carbon power, competitive long-term contracts and a naturally attractive place to build. Yet a country's renewable electricity mix tells us very little about a specific site's water consumption, local hiring, ratepayer protections or long-term financial contribution. These deserve separate evidence.
Iceland: geothermal and hydropower with a cooling advantage
Iceland has an electricity system supplied largely by hydropower and geothermal generation, and a cool climate that can support efficient cooling. Operators including atNorth, Verne and Borealis have established computing facilities there. At some sites, heat-reuse projects may offer another local benefit. But it is still necessary to ask what electricity was contractually allocated, how much was used in operation, what cooling water was consumed and whether excess heat was delivered to actual users. An operator's power capacity is not its annual energy use.
El Salvador: geothermal potential is not proof of a site's power source
El Salvador's geothermal sector offers an interesting national energy example. The country also hosts commercial data-centre infrastructure. However, those two facts do not establish that any named data centre runs directly on geothermal electricity. To make that claim, we would need an operator disclosure, a utility contract or an independently documented supply arrangement. Public records should also tell us the size and operation of the facility, its cooling requirements and its lasting local contribution.
Paraguay: hydroelectric abundance meets competing computing uses
Paraguay's large hydroelectric resources have attracted energy-intensive computing, including cryptocurrency operations. Some operators are exploring or announcing AI and high-performance-computing conversions. The distinction matters: a mining campus with an announced AI strategy is not yet an operating AI data centre. The public accounting should compare contracted electricity, actual workload, prices and permanent employment before and after conversion.
The pattern worth testing
The common proposition is that readily available renewable electricity can attract computing investment. That may be an advantage when capacity is genuinely available, infrastructure costs are allocated transparently and local benefits materialize. It may be less advantageous if competing demand raises electricity costs, a grid connection crowds out other projects, or public incentives exceed delivered economic value. Neither outcome should be presumed.
What we want every operator and regulator to disclose: metered electricity and its generation source; hourly or annual energy mix as available; contracted power prices and protection of other customers; measured water withdrawal and consumption; permanent local jobs; public incentives; infrastructure costs; and evidence of any heat-reuse or community programme.Why this matters for our Observatory
If only a small fraction of facility records contains documented resource information, that is first a measure of our research coverage. It is not a verified finding that companies withheld the rest. We need to distinguish information we have not researched, information never published, commercially confidential material and claims for which the operator provides only estimates. Once the coverage improves, we can test whether geothermal, hydroelectric and conventional grid-powered facilities differ in measured community outcomes.
Our method is to publish the source, date and uncertainty alongside each claim, correct errors when evidence emerges, and compare projects on the same questions. This allows readers to evaluate the trade-offs themselves.
Further reading: IEA, Key Questions on Energy and AI (2026); see the Iceland, El Salvador and Paraguay entries in the Observatory for facility sources and open questions.