A claim that building 1 gigawatt-hour of utility-scale battery storage consumes roughly 450 gigawatt-hours of energy before it delivers a single kilowatt-hour back to the grid sounds extreme until it is checked against the literature. It isn’t fabricated.
A widely cited 2016 review of lithium-ion battery life-cycle studies found cumulative energy demand ranging from 316 to 2,318 megajoules per kilowatt-hour of capacity, with 960 megajoules per kilowatt-hour as the likeliest figure, a range that converts to roughly 88 to 644 kilowatt-hours of embodied energy for every kilowatt-hour of nameplate storage. A 450 to 1 ratio sits comfortably inside that band, toward its upper half rather than at its center. That single data point, drawn from an essay arguing that renewable energy systems carry a systematically undercounted upfront energy cost, holds up reasonably well.
The essay’s broader argument, that solar and battery systems return far less net energy over their lifetimes than conventional or nuclear generation once embodied energy is properly counted, rests on energy-return-on-investment figures that diverge sharply from the published mean. The calculation applied to Masdar’s Al Dhafra round-the-clock project in Abu Dhabi, which pairs 5.2 gigawatts of solar with 19 gigawatt-hours of battery storage to deliver roughly 1 gigawatt of continuous output, produces an eROI of 5 to 6, falling to an estimated 2 to 3 if the same system were built in Germany. Those numbers rely heavily on two inputs: an assumed embodied energy of 4 megawatt-hours per kilowatt of solar capacity, and the Prieto and Hall estimate of eROI near 2.45 for Spanish photovoltaics, both several times higher on the cost side than mainstream assessments.
The most comprehensive meta-analysis of solar eROI in the literature, a 2015 review of 232 papers published between 2000 and 2013, put the mean for polysilicon panels at 11.6, with estimates clustering between roughly 6 and 16 depending on vintage and methodology. A 2024 Nature Energy analysis using a different, useful-stage framework reported a median eROI of 11.4 for solar and 23.6 for wind. The IEA’s own embodied-energy assumption for solar, cited and explicitly disputed within the essay itself as likely understated by an order of magnitude, is closer to 1.2 megawatt-hours per kilowatt, roughly a third of the figure used in the Al Dhafra calculation.
The Prieto and Hall estimate that anchors the low end of this range is not obscure, but it is not uncontested either. A Department of Energy-published rebuttal describes its system boundary as inconsistent and its inclusion of a large set of non-energy inputs as arbitrary, a criticism echoed elsewhere in the eROI literature. A separate widely cited claim of eROI 0.8 for Swiss solar, from a 2016 paper by Ferroni and Hopkirk, prompted a formal peer-reviewed response identifying what its authors called critical methodological and calculation flaws in the original analysis. None of this means low-eROI estimates for solar are wrong by construction, since system boundaries genuinely differ across studies and the field lacks a unified calculation standard, a point the essay itself makes explicitly and correctly. It does mean that the specific figures used here sit at the disputed edge of a wide published range rather than representing settled consensus, and presenting them as a definitive verdict on solar’s net energy return understates how contested that particular corner of the literature is.
The project data underlying the calculation checks out independently of the eROI dispute. Al Dhafra’s RTC development is real, financed, and under construction: Masdar reached financial close on $5.1 billion in debt financing in mid-2026 against a total project cost of $6.1 billion, broke ground in October 2025, and expects commercial operation in 2027, with the plant expected to offset roughly 5.7 million tonnes of carbon dioxide annually once running. The 5.2 gigawatt solar array and 19 gigawatt-hour battery pairing, along with the roughly 1 gigawatt of round-the-clock output it is designed to deliver, match public reporting closely. Where the essay extends beyond verifiable data is in extrapolating a raw-material and coal-equivalence comparison, estimating 30 to 35 million tonnes of material extraction for a 20-year Al Dhafra-scale system against 40 to 45 million tonnes of coal for an equivalent coal plant, a comparison that depends on the same contested embodied-energy assumptions feeding the eROI figure and should be read with the same caveat.
There is a genuine, underappreciated point buried inside the more contestable numbers. Battery replacement cycles of roughly a decade against multi-decade turbine or reactor lifespans do compound embodied energy in ways operational-efficiency metrics alone don’t capture, and the essay’s insistence that decommissioning, refurbishment, and repeated replacement belong in any full accounting is consistent with how life-cycle assessment methodology has evolved over the past decade, including in the battery LCA literature cited above. The 2024 Nature Energy paper makes a complementary point from the opposite political direction, finding that fossil fuels’ own useful-stage eROI, once conversion losses are counted rather than stopping at the power-plant gate, falls to around 3.5 to 1, well below the final-stage figures usually quoted and below even some of the more conservative solar estimates in the literature. That finding complicates any simple ranking of technologies by net energy return regardless of which side of the debate is doing the ranking.
The essay’s author, an energy economist and longtime coal commodity trader who co-founded a German-listed coal trading firm and is affiliated with the CO2 Coalition, brings a particular vantage point to a field where system-boundary choices already produce order-of-magnitude swings in the headline number. That affiliation doesn’t invalidate the underlying question, which researchers across the ideological spectrum treat as legitimate and methodologically unresolved. It does mean the specific eROI figures presented as a calculation, 5 to 6 for a UAE desert solar-and-battery project and 2 to 3 for an equivalent German one, are one entry in a published range spanning roughly 2 to 16, not a corrective to a consensus that itself remains genuinely unsettled.
