W.157Google will buy 50 MW of superhot geothermal from Quaise for more than $500M
Quaise Energy said on November 17 that Google will buy 50 MW from Project Obsidian, its first plant at Newberry Volcano in Oregon, beginning as early as 2030. The release gives no price or term; Canary Media reported that Quaise confirmed the agreement is worth more than $500M, and that Quaise said the plant will cost about $350M. That is at least $10M per megawatt of contract against $7M of build cost. Spread over 15 to 25 years at a 90% capacity factor it is at least $51 to $85 per MWh, below the roughly $106 that recent Nevada geothermal offers carried. Quaise has not yet drilled a commercial well with its millimeter-wave method.
The agreement
Google will buy 50 MW from Quaise’s first plant. Quaise’s release says Google will purchase 50 MW from Project Obsidian in Central Oregon, beginning as early as 2030, to support its data center operations in The Dalles. OPB places the project on the southern flank of Newberry Volcano. Quaise calls it a power purchase agreement and says it is signed. No utility is named as a party or as the delivery route; the release mentions Northern Wasco County PUD only in describing local rates.
The dollar figure is not in the release. The release states no price, value or term, and OPB reported that the companies did not disclose the sale price. Canary Media reported that the agreement is worth more than $500 million, per Quaise’s confirmation; the Wall Street Journal, which we could not read, also reported the figure, according to a summary. We take it as the total over the term. Whether it is nominal or includes escalation is not stated.
Canary Media reported in March that the first 50 MW was already contracted. It reported on March 3 that Quaise had signed a power purchase agreement for the first 50 MW with an undisclosed customer. No source says whether the November announcement names that customer or describes a new contract.
The plant costs about $350 million, Quaise says. On 50 MW that is $7 million a megawatt, the same per kilowatt as Fervo’s stated cost for the first phase of its Cape Station in Utah. The agreement value is at least 1.4 times the plant cost. Quaise had raised $280 million by August and has been selected for up to $25 million from the Department of Energy, subject to negotiation.
The agreement is worth more than $500M; the plant costs about $350M
Millions of dollars. Pink: the value of the agreement, more than $500 million, per Quaise, as reported by Canary Media; the bar is drawn at $500 million and the true figure is higher by an amount not stated. Black: the cost of the 50 MW plant, about $350 million, per Quaise. Strong grey: the funding Quaise had raised to date at the final close of its Series B on August 27. Grey: the Department of Energy support for which Quaise was selected, up to $25 million. The first is revenue over an undisclosed term and the others are capital, so the bars are for scale.
What price that implies
The term decides it, and the term is not disclosed. At a 90% capacity factor, 50 MW produces 394,200 MWh a year. $500 million spread evenly over 15 years is $85 per MWh; over 20 years, $63; over 25 years, $51. Each is a floor, because the value is more than $500 million. The capacity factor and the terms are our assumptions.
Recent geothermal prices are higher than those floors. CTVC reported Fervo’s Corsac Station agreement with NV Energy, the 115 MW project that supplies Google, at $107 per MWh for 15 years, and Ormat’s Heber 1 extension at $86 with a 1.5% annual escalator. Three geothermal offers in NV Energy’s 2026 resource plan averaged $105.56, Oilprice.com reported from the filing. At $105.56, 50 MW would earn about $41.6 million a year and pass $500 million in about 12 years.
So either the term is short or the price is below Nevada’s. A value near $500 million at Nevada prices implies a term of about 12 years. A term of 20 years or more at Nevada prices would imply a value well above $500 million, which “more than” allows. The disclosed facts do not separate the two.
Spread over 15 to 25 years, $500M is $51 to $85 per MWh, below recent geothermal prices
Dollars per megawatt-hour. Dashed pink: $500 million divided by an assumed term and by 394,200 MWh a year (50 MW at a 90% capacity factor); the term is not disclosed, the capacity factor is our assumption, and each bar is a floor because the value is more than $500 million. Strong grey: geothermal prices reported for other agreements and offers, by CTVC and by Oilprice.com from NV Energy’s filing; none is a company-published contract price that we read directly.
The arithmetic by term
| Assumed term | $M a year | $K per MW-year | $ per MWh, floor |
|---|---|---|---|
| 15 years | 33.3 | 667 | 84.6 |
| 20 years | 25.0 | 500 | 63.4 |
| 25 years | 20.0 | 400 | 50.7 |
The technology and the size
The first phase does not use the new drill. Quaise’s method uses millimeter waves from a gyrotron to melt and vaporize rock. The November release says phase 1 uses conventional drilling to reach superhot conditions and that phase 2 is expected to be the first to use millimeter-wave drilling. Canary Media reported that the method is still being tested in Texas. Quaise’s April release put its two well systems at average temperatures of about 315°C and 365°C, about 5 km down.
A confirmation well was being drilled in September. That is from Quaise’s Department of Energy announcement of September 22. We found no later statement of depth or temperature reached. Quaise must also build a 10-mile line to a substation in La Pine, OPB reported.
50 MW is small for Google. It is 0.14% of the nearly 35 GW of clean energy agreements the release says Google signed from 2010 to 2025, and an eighth of the 396 MW Google agreed to buy from Fervo’s Cape Station in September. It is 1.9% of the 2.7 GW of US geothermal capacity in the EIA’s February article.
Quaise plans more at the site. Its April release describes a second phase targeting 250 MW, and Canary Media reported it is seeking buyers for a further 200 MW. Its chief executive estimates the project could generate about 1 GW in the next decade, OPB reported.
At 50 MW, the Quaise agreement is the smallest of Google’s firm-power deals shown
Megawatts. Pink: Quaise’s Project Obsidian. Strong grey: five other agreements Google has announced for power that runs around the clock, from the reports listed under Sources, with the date announced. Duane Arnold is the size of the plant under a 25-year agreement; Kairos is a total by 2035; the others are the contracted amount. Technologies and start dates differ.
What this changes on inthesingularity.ai
- inthesingularity.ai has no tracker of firm-power agreements for data centers. Lead: a table by megawatts, start date, term and price where disclosed. This one is 50 MW, as early as 2030, more than $500 million, term undisclosed.
- W.154 compared data center leases by value per megawatt-year. A power contract can be put on the same footing once its term is known: $500K per megawatt-year at 20 years here, against $1.67 million for Giga’s lease, which is for data center capacity.
- W.104 covered Oklo’s fuel contracts, which disclosed neither quantity nor price. This agreement discloses a quantity and a floor on value, and not the price.
- A figure to watch for: the term, which would turn the $500 million into a price.
Sources and method
| Item | Figure | Source |
|---|---|---|
| The agreement | Google will purchase 50 MW from Project Obsidian in Central Oregon beginning as early as 2030, for its data center operations in The Dalles; no price, value or term stated; 200 to 400 construction jobs and more than 50 permanent jobs; Google signed over 240 agreements for nearly 35 GW of clean energy from 2010 to 2025 | Quaise Energy release, Nov 17 |
| The agreement | Worth more than $500 million, per Quaise’s confirmation; the plant will cost about $350 million to build; Quaise is seeking buyers for a further 200 MW at the site; millimeter-wave drilling is still being tested in Texas | Canary Media, Nov 17 |
| The agreement | Original report of the agreement and its value | The Wall Street Journal, Nov 17 (not read; as summarized by Nasha Niva) |
| The agreement | The companies did not disclose the sale price; Quaise must build a 10-mile line to a substation in La Pine; three Google data centers in The Dalles; Quaise’s chief executive estimates the project could generate about 1 GW in the next decade; the project is on the southern flank of Newberry Volcano | OPB, Nov 17 |
| Quaise | A power purchase agreement for the initial 50 MW was already signed with an undisclosed customer | Canary Media, Mar 3 |
| Quaise | Phase 1 of at least 50 MW as early as 2030; a second phase targeting 250 MW; two well systems averaging about 315°C and 365°C at about 5 km | Quaise release, Apr 22 |
| Quaise | Series B final close of $180 million, including $35 million from Nabors Industries; $280 million raised to date | ThinkGeoEnergy, Aug 27 |
| Quaise | Selected for up to $25 million of Department of Energy support, subject to negotiation; a confirmation well was being drilled | Quaise release, via Business Wire, Sep 22 |
| Quaise | The first wells are drilled conventionally; millimeter-wave drilling at the site as early as 2027 | Canary Media, Jul 7 |
| Benchmarks | Fervo’s Corsac Station agreement with NV Energy: $107 per MWh for 15 years. Ormat Heber 1: 52 MW, 25 years, $86 per MWh escalating 1.5% a year | CTVC, Sep 2, 2025 |
| Benchmarks | Nevada regulators approved the 115 MW Fervo agreement that supplies Google under the Clean Transition Tariff | ThinkGeoEnergy, May 2025 |
| Benchmarks | NV Energy’s 2026 resource plan: three geothermal offers at $105 to $107 per MWh, weighted average $105.56 across 180 MW | Oilprice.com, Oct 20, citing the filing |
| Benchmarks | Fervo Cape Station: $7,000 per kW for its first phase, per the company | Canary Media, Sep 28 |
| Google’s agreements | Fervo Cape Station: 396 MW, online in 2028 | Power Engineering, September 2026 |
| Google’s agreements | Kairos Power: 500 MW by 2035; first unit of 50 MW in Tennessee by 2030 | ESG Dive, August 2025 |
| Google’s agreements | Commonwealth Fusion Systems: 200 MW from its fusion project in Virginia | S&P Global, June 2025 |
| Google’s agreements | NextEra’s Duane Arnold: a 25-year agreement on a 610 MW plant restarting by early 2029 | World Nuclear News, October 2025 |
| Scale | US geothermal summer capacity of 2.7 GW; 2,695.5 MW in the Electric Power Monthly table the article links | EIA, Feb 19 |
What we read. Quaise’s release of November 17; Canary Media’s and OPB’s reports; Quaise’s earlier releases and reports of its funding; and reports of the comparison agreements. We could not read the Wall Street Journal article and found no statement published by Google itself or any utility or regulatory filing. Arithmetic. 500 ÷ 50 = 10.0; 350 ÷ 50 = 7.0; 500 ÷ 350 = 1.43. Output: 50 × 8,760 × 0.9 = 394,200 MWh a year. Price floors: 500,000,000 ÷ 15 ÷ 394,200 = 84.6; 500,000,000 ÷ 20 ÷ 394,200 = 63.4; 500,000,000 ÷ 25 ÷ 394,200 = 50.7. Per megawatt-year, $ thousand: 500 ÷ 15 ÷ 50 = 667; 500 ÷ 20 ÷ 50 = 500; 500 ÷ 25 ÷ 50 = 400. At Nevada’s average: 394,200 × 105.56 = $41.61 million a year; 500 ÷ 41.61 = 12.0 years. Shares: 50 ÷ 2,695.5 = 1.85%; 50 ÷ 35,000 = 0.14%; 50 ÷ 396 = 0.126. Conflicts between sources. The value is more than $500 million in Canary Media and in summaries of the Journal, a flat $500 million in one trade post, and undisclosed in OPB’s report and the release. Duane Arnold is given as 610 MW and as 615 MW. Quaise’s phase 1 is “at least 50 MW” in April and “more than 50 MW by 2030” in September. Limits. The term, price, escalator and conditions are not disclosed, so every price here is an assumption-driven floor. A 90% capacity factor is typical of geothermal plants but is not a figure from the parties. The benchmark prices come from a newsletter and from a report of a utility filing. The megawatts in the last chart are not on one basis. Whether power is delivered physically to The Dalles or matched on the grid is not stated.
Wire reports are written within hours of the news, on a six-hour schedule. Each lists its sources and shows the arithmetic behind its figures. Not investment advice.