Forecast retrospective · 2020 → 2025
"Solar's Future is Insanely Cheap," five years on
In May 2020, Ramez Naam projected solar electricity costs to 2050 using a 30% learning rate and 16% annual industry growth, split across four location tiers. Half a decade of actual data is now in. The verdict: the big call — solar keeps getting cheap enough to undercut existing fossil plants — held. But the smooth curves didn't: hardware over-delivered, financing under-delivered, and the two roughly cancelled.
The scorecard, tier by tier
Dashed gray lines are Naam's projected curves (digitized from the chart above; extended to 2026). Colored series are actual prices — auction awards and tariffs for the cheap tiers, benchmark LCOEs and auction averages for the rest. All in US cents per kWh, nominal. Hover any point for details; the full data table is below.
| Tier (Naam's examples) | Naam's implied 2025 | Actual, 2024–25 | Verdict |
|---|---|---|---|
| Ultra-low cost — best Gulf / Chile auctions | ≈1.9¢ | 1.3–1.6¢ (awards); record 1.04¢ in 2021 | Ahead on level — but flat since 2021, not still falling |
| Low cost — India, China | ≈2.5¢ | India tariffs 2.6–3.6¢; IRENA LCOE: China 3.3¢, India 3.8¢ | Behind ~0.5–1¢ — India's ₹2.00/kWh record (2020) still stands |
| Medium cost — California-type / global average | ≈3.6¢ | Global avg 4.3¢ (IRENA), BNEF benchmark 3.9¢; US unsubsidized ~5.8¢ (Lazard mid) | Behind — global close in slope, US far off track |
| High cost — Northern Europe | ≈5.8¢ | Germany auctions 5.3–5.5¢; UK CfD ~8.3¢; Japan ~5.4–6.3¢ | Mixed — Germany on/ahead, UK and Japan behind; all flat-to-up since 2020 |
What actually happened
The trough came early, then the curve broke. The record lows Naam's ultra-low tier anticipated arrived almost immediately — Qatar at 1.57¢ and Abu Dhabi's Al Dhafra at 1.35¢ in 2020, then Saudi Arabia's 1.04¢/kWh in April 2021, a price his curve didn't reach until the mid-2030s. Then the streak stopped. Polysilicon rose roughly sixfold into 2022, freight quintupled, and interest rates repriced every power contract on Earth. Auction prices rose 40–60% in Saudi rounds, German tenders peaked at 7.0 €c/kWh in April 2023, Lazard's US range blew out, and Chile's auctions spiked on curtailment risk.
Hardware recovered; financing didn't. China's manufacturing buildout crushed module prices ~50% during 2023 to below $0.10/W — modules now sell at or below production cost — and IRENA's global installed cost fell to $691/kW in 2024, down 22% from 2020. Global average LCOE resumed falling: 4.4¢ in 2023, 4.3¢ in 2024. But BNEF's benchmark ticked up 6% in 2025 (to $39/MWh) on cost of capital, curtailment and price cannibalization, and no market has revisited its 2020–21 record lows.
The model's inputs erred in opposite directions. The measured learning rate through 2024–25 is ~26% (ITRPV) — below the 30% Naam called conservative. But deployment grew ~25–31%/year, roughly double his 16% input, delivering nearly two doublings by 2025 instead of one. Faster scaling largely offset the slower learning rate, which is why the global-average trajectory landed close to his curve even though both inputs were wrong.
The blind spot: solar's cost is no longer mostly hardware. Naam's model priced a technology; by 2025 the binding constraints are money and grids. Identical panels produce 1.3¢ power under Gulf sovereign financing and ~7¢ power in the US (IRENA puts 2024 US LCOE at 7.0¢ vs China's 3.3¢). Cost of capital, land, interconnection, tariffs and curtailment now set the price — which is why the cheap tiers sit near his curve while high-cost markets drift above it.
The headline claim aged well anyway. IRENA finds utility solar averaged 41% cheaper than the cheapest fossil alternative in 2024; BNEF benchmarks solar at $39/MWh against $102 for new combined-cycle gas. In sunny markets, new solar at 1.3–1.6¢/kWh is comfortably below the operating cost of existing coal and gas plants — the crossover his chart's shaded band promised. And with 2025's ~664 GW build (a third of it added in a single year), the volume story beat even his optimistic case.
Data table — every point plotted above
| Panel | Series | Year | US ¢/kWh | Note |
|---|
Sources
- Ramez Naam, Solar's Future is Insanely Cheap (May 2020) — prediction curves digitized from the "Future Solar Costs by Year" figure.
- IRENA, Renewable Power Generation Costs editions 2020–2024 (global weighted-average utility PV LCOE: 5.7 → 4.3¢/kWh; installed cost $883 → $691/kW; country LCOEs 2024: China 3.3¢, India 3.8¢, US 7.0¢).
- BloombergNEF LCOE updates 1H2021–2026 (fixed-axis benchmark $48 → $45 → $41 → $36 → $39/MWh; 2035 forecast $25/MWh), via BNEF releases, PV Tech and pv magazine.
- Lazard, LCOE v14 (2020) through LCOE+ June 2025 (US unsubsidized utility solar $38–78/MWh, mid $58 in 2025; $20–45 with PTC).
- LBNL, Utility-Scale Solar 2025 edition (US levelized PPA prices ~$20/MWh 2020–21 → $29/MWh 2024, ITC-inclusive).
- Gulf auctions: pv magazine / PV Tech / MEED coverage of Al Kharsaah, Al Dhafra, Saudi NREP rounds 2–5 (Al Shuaibah 1.04¢ 2021; Al Sadawi 1.29¢ 2024), DEWA Phase 6, EWEC Al Ajban & Khazna.
- India: Mercom / SECI (₹2.00/kWh record, Nov 2020; ₹2.48–3.04/kWh 2024–25 tariffs).
- Germany: Bundesnetzagentur ground-mount tenders 2020–2025 (5.0–5.3 €c 2020 → 7.03 €c Apr 2023 → 4.7–4.8 €c 2025). UK: CfD AR4–AR7 strike prices (2024-money basis). Japan: METI FIP auctions.
- Deployment & hardware: IEA-PVPS Snapshots 2021–2025, SolarPower Europe GMO 2025 (~664 GW added 2025; >3 TW cumulative), InfoLink/Bernreuter module & polysilicon prices, ITRPV 2025/2026 (learning rate 25.6–26%).
Compiled July 2026. All prices nominal US cents per kWh; auction/PPA awards are as-bid levelized tariffs (Gulf IPPs benefit from sovereign-backed financing and free land; India figures converted at award-date exchange rates; UK CfD converted from 2024-money strike prices). Naam's curves digitized from the 2020 chart, so predicted values are approximate to ±0.1¢.