Predictions vs Reality: RethinkX's energy forecasts
Sixteen dated claims from three RethinkX energy reports, set against independent data from 2025 to 2026. What is supported, what is not, and what is still open.
Part of the Future & Foresight collection.
Cheap clean energy is changing the system. How much of RethinkX's abundance forecast is proven? Independent evidence supports rapid deployment of solar and batteries, competitive renewable generation costs and growing use of surplus output. It does not establish the reports' full claims about universal competitiveness, the speed at which conventional assets collapse, or inevitable energy abundance.
This page checks sixteen dated claims from three RethinkX reports against independent evidence, as of 7 November 2026. Full-year 2026 outcomes do not exist yet, so we use 2025 annual data and clearly dated 2026 partial-year data. For how we read forecasts, see Tony Seba and RethinkX. For the energy argument itself, see The future of energy: from scarcity to abundance?.
What this page is, and is not
- It tests the original reports, not RethinkX's later claims about its own record. RethinkX's retrospective assessments are not treated as independent verification.
- It is not a rerun of the models. We did not rebuild RethinkX's models, value individual power plants, or verify every reference and input across the reports' 208 pages. The independent data comes from an editorial evidence assessment. We have not re-checked each figure against its original publication, so verify each before relying on it.
- "By 2030" is still a future test. Many deadlines have not arrived.
- A competing institutional forecast is a comparator, not a disproof. And a model result using historical technology costs is still a model result.
- US results stay in US dollars and US geography. Nothing here is converted to Australian dollars or applied to Australia unless stated.
The three reports
| Key | Report | Published |
|---|---|---|
| R20 | Dorr and Seba, Rethinking Energy 2020–2030 | October 2020 |
| R21 | Dorr and Seba, The Great Stranding | February 2021 |
| R25 | Dorr, Arbib and Seba, Understanding Stellar Energy | May 2025 |
Page numbers below are the reports' printed pages. For R20 the PDF page is the printed page plus 2, and for R21 plus 1. R25's printed and PDF pages coincide.
Status labels
- Open: the deadline has not arrived, or the evidence cannot yet settle it.
- Direction supported: independent evidence points the same way, but the exact numbers are not confirmed.
- Not validated: independent data does not match the claim as stated.
- Unresolved: it depends on definitions or evidence we do not have.
- Open forecast: a future claim about what will happen, not yet testable.
- Scenario assumptions: inputs to a model, not outcomes to be scored.
Claims about technology and cost
| # | Original claim (source) | What independent evidence shows | Status |
|---|---|---|---|
| 1 | By 2030, a 100% solar, wind and battery (SWB) electricity system is physically possible and economically affordable across the continental US and most populated regions (R20 pp. 7, 11–12). | R20 p. 11 itself calls the analysis an illustrative limit scenario, not a forecast. Current renewable shares cannot falsify feasibility, and national reliability is unverified by this model. IRENA's 2026 firm-renewables study models a related mechanism for favourable sites (S4). | Open (conditional feasibility claim) |
| 2 | From the report's baseline to 2030, solar, onshore wind and lithium-ion battery capacity costs fall a further 70%, 40% and 80% (R20 p. 9). | Direction is supported. IRENA reports solar at US$0.043/kWh and wind at US$0.034/kWh in 2024 (S1), and the IEA reports average battery prices down 8% in 2025 (S3b). But global generation costs and battery-pack prices are not the same series as US installed capacity costs, so the exact endpoints cannot be scored yet. | Direction supported, endpoints unverified |
| 3 | SWB system costs fall about 70% by 2030, and the visual summary shows electricity under US 3 cents/kWh (R20 p. 8). | The report's own Table 1 (p. 13) gives 3.1, 3.5 and 6.1 cents for the lowest-capital-cost system with none of the surplus used, and the text conditions "under 3 cents" on at least 50% of surplus being used. Networks are excluded (p. 29). IRENA's firm-renewables study (S4) gives US$54–82/MWh, but its reliability target, geography, price year and delivery profile differ from RethinkX's. It therefore neither directly validates nor disproves RethinkX's price. | Conditional, overgeneralised in the headline |
| 4 | Least-cost 2030 systems generally need 35–90 hours of average demand in battery energy capacity (R20 p. 12, Table 1 p. 13). | Model result: the case studies give 37, 49 and 89 hours. The IEA notes most batteries are around two hours' duration (S3), but that is rated power, not regional average demand. Neither number disproves the other. | Model result, not observed |
| 5 | A continental-US 100% SWB build in the 2020s costs under US$2 trillion, "roughly 1% of GDP" (R20 p. 12). | An unverified deployment scenario. It is not an observed total or a complete system budget, since networks are excluded (p. 29). A ten-year total is not the same thing as 1% of one year's GDP. | Unverified scenario |
| 6 | The build assumes 50% annual solar growth, 12% wind growth and 85% battery-storage growth during the 2020s (R20 p. 59, note k). | These are assumptions that drive when spending occurs. Rapid global battery growth (S3) does not validate this particular regional trajectory, and growth in annual additions is not growth in cumulative capacity. | Scenario assumptions |
Claims about conventional energy
| # | Original claim (source) | What independent evidence shows | Status |
|---|---|---|---|
| 7 | Coal, gas and nuclear assets become stranded during the 2020s, and no new investment is rational from 2020 (R20 p. 7). | Asset risk is credible, but "universally irrational" would need project-level revenue, financing and policy analysis. Global coal generation fell about 0.5% in 2025 while gas rose about 0.5% and nuclear reached a record (S7). AEMO's plan still includes new flexible gas (S5). | Universal wording unsupported |
| 8 | US coal utilisation falls from 40% in 2020 to 10% by 2035, so a new 2030 coal plant has a modelled cost of 65.3 US cents/kWh (R21 p. 14). | US EIA reports a 2025 coal capacity factor of 48.8% against 40.5% in 2020, although the coal fleet shrank from about 220,600 MW to 169,900 MW (S8). Survivors can run harder as others retire, so fleet averages are not a full test. The 65.3-cent figure is a scenario output. | Assumed path not confirmed, 2035 open |
| 9 | US gas utilisation reaches 60% in 2022, then falls to 10% by 2035. A new plant's modelled cost reaches 18.4 US cents/kWh in 2030 (R21 p. 15). | For combined-cycle gas, EIA reports 56.6% in 2022, 60.5% in 2024 and 58.7% in 2025 (S8). A strict "2022 maximum" does not fit. The long-run collapse is unresolved. Combined-cycle data should not be compared with all gas technologies pooled. | Not validated, 2035 open |
| 10 | US nuclear utilisation declines from 93% in 2021 to 10% by 2035. New-build cost becomes 105 US cents/kWh in 2030 (R21 pp. 16–17, Figure 10). | EIA reports 91.1% in 2025 (S9), so no collapse is observed. A later nonlinear fall is not ruled out. The US$1.05/kWh figure is a model output, not an observed tariff. | Collapse not observed, 2035 open |
| 11 | US hydro utilisation declines from 43% in 2021 to 10% by 2035. New-build cost becomes 49.4 US cents/kWh in 2030 (R21 p. 18). | EIA reports 36.0% in 2021 and 34.4% in 2025 (S9). Lower hydro output cannot be assumed to come from solar competition, since hydrology and operating roles matter. | Causal claim unresolved |
| 12 | The conventional-energy valuation bubble could exceed US$1 trillion by 2030 (R21 p. 20). | "Could" is conditional. We found no comparable realised-loss inventory. Investment spending, lost market value, write-downs and stranded assets are different measures. It would be wrong to say US$1 trillion has already been stranded. | Open risk claim |
The 2021 report's point that fixed costs spread over fewer units of electricity raise unit costs is sound. But its "corrected" historical costs combine observed data with projected future declines in use (R21 pp. 14–18). They are not measured past errors and are not verified financial losses.
Claims about abundance
| # | Original claim (source) | What independent evidence shows | Status |
|---|---|---|---|
| 13 | SWB has an overwhelming economic advantage in virtually every inhabited region by 2030 (R25 p. 32, Figures 6–8). | The broad direction is supported by IRENA's data that 91% of new renewable capacity in 2024 was cheaper than the cheapest new fossil option (S1). Regional financing, resource quality and integration prevent global averages from proving every-region competitiveness. | Direction supported, universality unverified |
| 14 | Clean-energy superabundance by 2035 in leading regions and before 2045 for most of the world (R25 p. 49). | A future conditional claim. "Leading regions", "most" and "superabundance" need operational definitions before it can be scored. | Open |
| 15 | By the 2030s, total decentralised SWB cost falls below the cost of transmission alone across much of the world (R25 p. 54). | Needs location-specific comparisons with equivalent reliability, maintenance and replacement, and a clear definition of avoidable network costs. Cheap rooftop modules or low wholesale prices do not establish it. | Unresolved |
| 16a | By the 2040s, humanity uses at least as much energy per person as today's wealthiest societies (R25 pp. 9, 64–66). | A future claim, not observed in 2026. It would need defined measures of energy services per person, by region, before it could be scored. Access, affordability and distribution need their own measures. | Open forecast |
| 16b | The scenarios fix per-person consumption levels (about 6,150 kWh a year for residential, commercial and industrial electricity, and 4,200 kWh for road transport, plus climate-dependent heating and cooling electricity) and use UN 2040 population projections (R25 pp. 64–66). | These are chosen inputs, not an econometric forecast that prosperity follows from equipment. They define the scenario and are not outcomes to be scored. | Scenario assumptions |
Where the model assumptions matter most
2020: a constrained experiment. Two weather years for three US regions cannot show resilience to every rare multi-day event or future climate. Removing useful flexibility (imports, rooftop solar, vehicles, demand response) can make the model conservative, while leaving out networks can make a whole-system price optimistic. These effects do not cancel by assertion (R20 pp. 24–29, 59).
2021: a useful warning built around a disputed path. It substitutes fleet-average utilisation for how a new plant would actually run. A low-use "peaker" plant can still be valuable, and high use alone does not guarantee a good investment. Project-level cash-flow tests are needed, not a static cost figure from either side (R21 pp. 14–20).
2025: better weather data, with large exclusions. Ten years of NASA reanalysis weather data is a better sample, but demand is synthesised, costs are US 2024 benchmarks used worldwide, and land and network costs are left out. No Australian case study is included (R25 pp. 63–68).
What we say, and what we do not
We can say with reasonable confidence: solar and batteries are growing rapidly; new renewable generation is competitive; the generation-versus-storage trade-off is real; utilisation risk matters to conventional generators; and surplus output offers a real opportunity.
We qualify prominently: the 2030, 2035 and 2045 dates; the under-3-cent cost; the 35–90 hour storage figure; worldwide cost extrapolations; conventional-asset stranding; and predictions of universal prosperity.
We do not say: "Seba's energy predictions have all come true"; "100% renewable Australia is proven at 3 cents/kWh"; "US$1 trillion has already been stranded"; "the IEA or IRENA has verified RethinkX"; or "clean energy needs no ongoing materials". None follows from this evidence.
How we will keep score
This page is a baseline, not a final verdict. To score the forecasts fairly we will keep the original wording and update these tests:
- Compare 2030 installed solar, wind and battery costs with the original US baseline, matching currency year, system boundary and battery duration.
- Look for reproducible hourly model runs with wider weather sampling, outages, transmission limits, degradation and explicit reliability standards. An energy-balance study is not a full grid-security study.
- Track US capacity-factor paths each year, separating new plants, survivors and retired assets.
- Define "stranded" and build an auditable asset-loss inventory before scoring the trillion-dollar claim.
- Define abundance milestones for named regions: energy services per person, affordability, reliability, access and environmental performance.
- Test Australian applications with Australian demand, network, financing and land assumptions rather than transferring a US cost directly to the NEM.
Independent sources used
All are as compiled in the editorial evidence assessment of 7 November 2026. We have not independently re-checked each figure against the original publication. Years refer to the data period.
| Ref | Source | Data period | What it was used for |
|---|---|---|---|
| S1 | IRENA, Renewable Power Generation Costs in 2024 | 2024 | Solar US$0.043/kWh, wind US$0.034/kWh, 91% cheaper than new fossil |
| S3 | IEA, Global Energy Review 2026, battery storage | 2025 | 108 GW additions, up 40%; typical duration about two hours |
| S3b | IEA, Global EV Outlook 2026, Electric vehicle batteries | 2025 | Average battery prices down 8% |
| S4 | IRENA, 24/7 Renewables | May 2026 model | Firm renewable cost US$54–82/MWh at favourable sites |
| S5 | AEMO, 2026 Integrated System Plan | Published 25 June 2026 | Least-cost NEM pathway includes storage and gas |
| S7 | IEA, Global Energy Review 2026, electricity supply | 2025 estimate | Global generation mix and changes |
| S8 | US EIA, Electric Power Annual, Table 4.08.A | 2020–2025 | Coal and combined-cycle gas capacity factors |
| S9 | US EIA, Electric Power Annual, Table 4.08.B | 2020–2025 | Nuclear and hydro capacity factors |
We have used only the sources we cite above. The assessment also draws on IRENA 2025 costs, CSIRO GenCost 2025–26, IEA Electricity 2026, AEMO's Q3 2026 market report and IEA PVPS, which appear in the energy feature.
About this page
This is an editorial assessment, not a replication. We read the three RethinkX reports directly and confirmed the quoted figures against the PDFs. We have not independently re-checked the independent evidence against its originals, did not rerun any model, and did not verify every number in the reports. The assessment did not inspect the IRENA 2025-cost PDF in full. This is an early version, published so that it can be corrected and improved.