Google will pay extra for heat that never stops, and Nevada wrote it into a tariff

The premium is the part worth reading, not the megawatts. The Google arrangement runs through the Nevada utility under a tariff that lets the buyer pay extra for emerging technology ^[1], which means the cost of firmness is attached to the party that wants it rather than blended into everyone's rate. The source does not give a dollar figure per megawatt-hour, and the regulatory approval expected in the second half of this year ^[4] is where that number either becomes public or does not. Either way, the structure exists, and it is the first place a compute budget can look for what continuous supply actually costs against the cheap intermittent alternative ^[5].

Now the arithmetic that makes the timeline honest. Ormat's first data centre deal, with Switch in January, was about 13MW over 20 years ^[3]; the Google contract is roughly 11.5 times that volume ^[1]. Set it against the company's own build history of about 100MW a year ^[7] and one agreement equals eighteen months of everything Ormat has ever managed to add annually ^[3]. The published target compounds it: going from about 1.85GW today to 2.8GW by 2028 ^[6] requires 950MW, which at the historical pace is nine and a half years of drilling ^[2]. That is why the projects behind the contract are dated 2028 to 2030 ^[4]. The scarce input is not demand and not willingness to pay. It is holes in hot rock.

The equity market has already sorted the two halves of this business. Ormat's first-half 2026 revenue was $662.7mn, up 43%, with $71.2mn of net profit ^[9], earned from conventional plants and storage ^[6]. Fervo Energy, the enhanced geothermal specialist, listed in May at $10bn and is now worth about half that ^[11], roughly $5bn of valuation gone ^[4]. Ormat's own upside case rests on the same unproven technique, through pilots with Sage Geosystems and SLB due online in late 2027 and individual projects that could reach 500MW ^[8]. Chief executive Doron Blachar says "all the stars are aligned exactly on time" ^[10], which is a claim about schedule, and schedule is precisely what the 100MW-a-year record argues about.

Europe supplies the control case. Denmark has paused grid connections because data centres are overwhelming its grid, and the EU has asked households to move electricity use away from peak hours ^[12]. The resource is not the obstacle there: Iceland runs on geothermal, Italy has generated from it since 1911, and the deep rock under Germany and the Netherlands is the subject of national programmes ^[13]. What is missing is the instrument. Nevada produced a tariff and a counterparty; Europe is still deciding how geothermal replaces gas while 15-year hyperscaler agreements get signed somewhere else ^[14]. For anyone costing capacity through 2030, that is two different prices for the same attribute: a disclosed premium in one jurisdiction, and a queue in the other.