3 SEP 2026 — Google has contracted 396 megawatts from Fervo's Cape Station in Utah, with an option on a further 600 through June 2030, taking the potential total to almost a gigawatt. Cape Station comes online in 2028. The same industry installing gas turbines this year is signing geothermal contracts for the end of the decade, and both are responses to the same shortage.
The agreement
The power purchase agreement covers 396 megawatts from Cape Station, expected to begin generating in 2028. Google holds an option through June 2030 to buy an additional 600 megawatts, which would bring the total to 996.
Fervo is permitted to develop up to 2 gigawatts at the site, and its engineers believe the resource could support twice that. Sarah Jewett, the company's senior vice-president of strategy, discussed the site's heat capacity in May.
The two companies have worked together for more than five years. Google bought some of Fervo's first electricity and invested in its final private funding round before the company listed in May 2026. The purchase supports Google's stated net-zero target for 2030 while supplying power for AI data centres.
Firm power is why this is different from solar
Google has contracted a great deal of wind and solar over the past decade, and neither suits a data centre particularly well. Both are intermittent, which means matching them to a constant load requires storage, overbuilding, or grid power filling the gaps.
A geothermal plant produces at a steady rate around the clock. That profile matches a data centre's demand: a near-constant load every hour of the year, with no useful ability to shift work to sunny periods.
This makes 396 megawatts of geothermal worth considerably more to Google than the same capacity in solar, and it is why the sector has become interested in enhanced geothermal at exactly this moment. The technology has been possible for years; the buyer who needs firm carbon-free power at scale is new.
Enhanced geothermal is a drilling technique, not a heat discovery
Conventional geothermal requires a place where heat, water and permeable rock coincide naturally, which is why the industry has been confined to a small number of geologically fortunate regions.
Enhanced systems create the permeability instead, fracturing hot rock and circulating water through it, using techniques and equipment developed by the oil and gas industry. This technique changes geothermal from a resource discovery problem into an engineering one, which is why the number of possible sites is suddenly so large.
The very large figures quoted for national potential should be read as technical potential rather than as a pipeline. What is contracted here is 396 megawatts at one site, and the distance between a resource assessment and a delivered megawatt is measured in permits, drilling programmes and grid connections.
The option is worth more than the firm commitment
Google has committed to 396 megawatts and taken an option on 600 more expiring in June 2030. The option is the more revealing half of the agreement.
Holding it costs Google very little and gives it first call on most of a gigawatt if demand materialises. Granting it costs Fervo the ability to sell that capacity to anyone else while the option runs, which is a real constraint on a developer trying to finance further phases.
That asymmetry shows who holds the leverage. For all the talk of desperate demand for firm power, this looks like a buyer's market. A developer with several bidders does not hand out four-year options on two thirds of a site, which suggests that in 2026 the scarce thing is still a creditworthy offtaker rather than a megawatt.
The date is the whole problem
Cape Station generates from 2028. An operator with a capacity shortage in 2026 gets no help from this contract for two years, which is why the industry is installing gas turbines in the meantime.
We reported that operators are installing on-site gas turbines to bypass multi-year interconnection queues, and that a study of one Virginia site estimated 3.4 to 6.5 additional premature deaths a year from the emissions. Those turbines are the answer to the 2026 problem. Geothermal is the answer to the 2030 problem. Both are being pursued by the same companies at the same time, which is what a capacity shortage looks like from inside.
A power purchase agreement of this kind is also how a plant gets financed. A developer with a signed offtake from a creditworthy buyer can raise construction capital; without one it cannot. The contract is therefore not just a purchase but the instrument that lets the plant be financed and built, which is worth remembering before reading the 2028 date as a delay.
Why this is not the regional answer
Indonesia and the Philippines are among the largest geothermal producers in the world, sitting on conventional resources that most countries lack. That makes the enhanced-systems story less relevant here than it is in Utah, because the constraint in this region has never been the resource.
It has been transmission, land and the commercial structures that get a plant built. A gigawatt of geothermal potential in Sumatra does not help a data centre in Johor unless something carries the electricity there, and the same interconnection problem that pushes American operators toward on-site generation applies with more force across an archipelago.
What does transfer is the financing pattern. A hyperscaler signing a long-dated offtake is the mechanism that turned an undeveloped site into a funded project, and it is available to any regional developer who can find a buyer willing to commit to 2030.
View author profile → · Editorial policy
