Resources / References / 03
Fervo Energy: Big Tech’s Clean Energy Bet Is Unraveling
E² review · 2026-10-06
Review
Summary
Support Distribution
Claims Provenance
- Report inference or forecast 13
The report’s interpretation requires further measurement, a matched comparator or evidence of intent.
C04 · C05 · C12 · C14 · C19 · C23 · C24 · C25 · C26 · C29 · C37 · C43 · C45
- First-party source checked 9
Company, technical-paper or grid-planning disclosure supports an attribution. Publication is not independent performance validation.
- Conditional calculation 8
Published model inputs, disclosed averages or report screenshots. Correct arithmetic does not authenticate the underlying assumptions or contracts.
- Operator submissions reproduced 7
Nevada hosts the plant operator’s monthly entries. Calculations reproduce those submissions, not independent sensor measurements.
- Anonymous testimony 4
Morpheus reports interviews; identity, date, access and recordings were not authenticated here.
- Company catalogue reproduced 2
Frozen public company-origin event records; no independent operational attribution.
- Independent seismic record checked 2
USGS regional records support counts, event location and magnitude, with catalogue uncertainty; causation remains open.
- Supplier model 1
Inspectable multi-well mechanism, with a commercial relationship; field validation remains separate.
Replicability
- Arithmetic matched 12
- C06 Monthly injection increases 33.42%; daily and monthly quantities differ.
- C07 Matched permit balances reproduce 21.17% and 56.86%.
- C10 The six-calendar-month difference annualizes to 24°F/year; trend validity remains open.
- C11 24 / 1.5 = 16; the economic benchmark remains unvalidated.
- C13 2.1 / 5.1 = 41.18%; the ratio is not a capacity factor.
- C17 June’s injection–production balance is 34.68%; July falls to 22.68%.
- C18 Signed monthly balances total 192,872,191 gallons; source anomalies remain flagged.
- C28 7,000 to 3,000 requires 57.14% reduction.
- C31 115 MW × $700/MW-day × 7 days = $563,500, conditional on the screenshot.
- C32 The stated supply and nameplate imply 86.57%.
- C33 80% of supplied annual energy implies 69.26% of nominal annual nameplate energy.
- C38 58 yellow events; 36 in November–April, matching the company disclosure.
- Discrepancy identified 2
Fallacies
Findings
- Faulty comparison 1
- D01 · Net pilot output compared with gross capacity
Unresolved cases (4)
- Hasty generalization 1
- U03 · Pilot limitations transferred to commercial failure
- Faulty comparison 1
- U04 · Illustrative levelized cost treated as a project loss
Passages and reasoning
D01 · Net pilot output compared with gross capacity
Similarly, based on the Google/NV Energy PPA, Fervo will trigger a default event if it operates at less than ~70% of the contracted capacity. This might turn out to be problematic for Fervo, since its own pilot plant operated at less than 50% of its nameplate capacity, according to our analysis of Fervo public statements.[14]
The conclusion depends on a net/gross ratio and a transfer from a different pilot design to a commercial contract. Keeping power definitions consistent changes the pilot ratio to 70%, before calendar accounting and contractual exclusions. This identifies a defect in the comparison, not a prediction that the commercial contract will be met.
Article passage · Faulty comparison criterion
Reasoning and repair
- The report’s pilot numerator is 1.4 MW net.
- Its 3 MW reference is described as gross generation in the prospectus.
- The resulting less-than 50% ratio is compared with a commercial delivery threshold.
Conclusion: The pilot comparison indicates difficulty meeting the commercial delivery threshold.
Counterreading: The passage says only that default “might” become problematic and intends a warning about execution risk. That softer wording remains reasonable, but its numerical comparison still depends on incompatible power definitions.
Repair: Use comparable net nameplate and calendar energy, then apply the actual commercial contract conditions.
U01 · Different temperature presentations treated as concealment
Fervo has never discussed this test or its implications with investors, but we believe it explains both the substantial temperature fluctuations that Fervo seems to have hidden from investors, and reveals another operational challenge that Fervo withheld.
A discrepancy and temporal association do not settle deliberate alteration. Whether the inference improperly excludes innocent explanations depends on the unmatched measurement definitions and unavailable data history.
Article passage · False cause criterion
Reasoning and repair
- Two presentations show different temperature behavior.
- The report associates the difference with an attempt to boost production.
Conclusion: The historical series was altered to hide poor performance.
Counterreading: The report uses “we believe” and “may” and presents concealment as a hypothesis supported by interviews, not a deductive certainty. That is a plausible abductive reading.
Repair: Reconcile sensors, filters and timing, obtain original files and edit history, and state the alternative explanations still possible.
U02 · A decline comparator promoted to an economic limit
Through our public records request to NDOM, we discovered that thermal decline at Project Red has now accelerated to an even faster rate of 24°F per year. Further, in June, production temperatures inexplicably plummeted by 90°F in a single month.[4]
The numerical comparison is reproducible. The scope of the claimed economic limit is not. A hydrothermal heuristic may or may not apply to this EGS design and cash-flow structure.
Article passage · Accident criterion
Reasoning and repair
- A selected interval annualizes to 24°F/year.
- Experts cite roughly 1–1.5°F/year as acceptable decline.
Conclusion: The observed pilot decline is uneconomical and undermines commercial viability.
Counterreading: The report may be using expert experience as a warning threshold rather than asserting a universal rule. No universal numerical cutoff has been independently tested here.
Repair: Show a design-specific cash-flow model across thermal trajectories, with a justified comparator.
U03 · Pilot limitations transferred to commercial failure
Despite this, Fervo doubled down by claiming its approach is “proven at scale” and raising ~$2.2 billion through its May IPO to build a much larger commercial-scale EGS site in Utah, known as “Cape Station.” Investors believe the structural failures of the pilot plant will somehow disappear at scale.
The pilot identifies risks worth testing. The transfer to a different design needs evidence about the changes. The report examines some changes and interviews relevant experts, so lack of representative sampling alone does not establish a fallacy.
Article passage · Hasty generalization criterion
Reasoning and repair
- The pilot shows lower output and substantial short-term water imbalance.
- Cape Station uses a different resource and multi-well design.
Conclusion: Commercial scale will reproduce the pilot’s failures rather than resolve them.
Counterreading: A pilot in related geology can materially update beliefs about shared failure mechanisms. The report’s skepticism may be a conditional risk assessment rather than a claim that every design is identical.
Repair: Specify which mechanisms transfer, predict changed-design outcomes and test them against commercial records.
U04 · Illustrative levelized cost treated as a project loss
This appears to be less than Fervo’s overall cost to produce energy, even taking its claims at face value. Financial advisory firm Lazard estimates a cost of $107/MWh for “new geothermal,” and the footnotes of its report reveal inputs of $7,000/kWh in CapEx and $160/kW-per year in O&M to estimate that cost. These inputs match the figures disclosed by Fervo.[11]
Price and cost are comparable units, but they are not matched project cash flows. Tax treatment, financing, capacity factor and escalation remain untested. The report explicitly distinguishes LCOE from a GAAP metric in footnote 11; that qualification prevents a confident equivocation finding.
Article passage · Faulty comparison criterion
Reasoning and repair
- A PPA is reported at$99/MWh.
- Lazard’s illustrative new-geothermal point is$107/MWh with matching capex and O&M assumptions.
Conclusion: Fervo loses money on every MWh unless costs decline.
Counterreading: The author may mean lifetime economic cost exceeds price under a stated illustrative scenario, which is a valid conditional concern.
Repair: Reconcile executed price schedules and project cash flows, including tax credits and a transparent capital-recovery assumption.