Many Minded · The cost curves · ENERGY

How much does a solar panel cost per watt?

Solar module price per watt · $/W · measured 1975–2025 · fitted live from 51 observations by the Many Minded cost-curve engine.

The last measured value is $0.12/W (2025, OWID (IRENA / Nemet / Farmer & Lafond)). Over the fitted window the series falls 16.3% a year (80% interval −24.1% to −7.7%) — a halving every 3.9 years (80%: 2.5–8.6). Carried forward on that fit, 2030: $0.05/W (80%: $0.02/W–$0.12/W).

$0.12/Wlatest measured (2025)−16.3%fitted, per year3.9yper halving51observations, 1975–2025measuredbasis · 1 live

the curve

solar module price — measured history and fitted projectionsolar module price: measured 1975–2025 in $/W, last value $0.12/W, fitted at −16.3% a year, projected to 2033 inside an 80% interval. Logarithmic vertical axis.$/W$0.0100$0.100$1.00$10$100197519821989199620032010201720242031$0.12/W (2025)
Vertical axis: $/W, log scale. Solid: measured observations. Dashed: the fitted central projection. Shaded: the 80% interval, which widens with horizon because shocks accumulate.

what this series measures

global average solar PV MODULE price per watt — not the installed-system cost.

what the fit says

quantityvaluehow it is computed
fitted rate−16.3%/yr (80%: −24.1% to −7.7%)Farmer–Lafond drift over the trailing window · n=11 points, 10 years
halving time3.9 years (80%: 2.5–8.6)implied by the fitted drift
Wright's law25% cheaper per doubling of cumulative deploymentfirst-difference regression on 2,600 GW installed deployed · OWID/IRENA measured through 2024; 2026 estimated
deployment tempodoubles every 4.2y (was 2.9y) → implies −6.6%/yrlearning holds while annual declines slow — each doubling of an ever-bigger base takes longer
Way scenario (scenario, not fit)2032: $0.08/W (80%: $0.04/W–$0.17/W)400 Monte-Carlo paths, deployment-conditional — growth 18%/yr decaying to 2%/yr (τ≈8y)
scale to the next halving2.4 doublings of deployment ≈ 10 years at today's pacejudged by scale, not calendar — scale survives regime noise better
curve checkaccelerating (−12% → −15%/yr)first half of the record versus the second, judged against the direction that helps this metric (falling)
regime break~2012 (−8% → −12%/yr, F=26)Chow-style best single breakpoint, kept only above a conservative sup-F threshold

the projection, with its interval

Log cost as a random walk with drift: the forecast variance grows with horizon (τ + τ²/m), which is why these bands widen instead of staying parallel. The middle column is the least useful number on this page; the interval is the claim.

yearcentral fit80% interval
2026$0.10/W$0.07/W to $0.14/Wnear horizon
2027$0.08/W$0.05/W to $0.14/Wnear horizon
2028$0.07/W$0.04/W to $0.13/Wnear horizon
2029$0.06/W$0.03/W to $0.12/Wthe band is already wide here
2030$0.05/W$0.02/W to $0.12/Wthe band is already wide here
2031$0.04/W$0.02/W to $0.11/Wthe band is already wide here
2032$0.03/W$0.01/W to $0.10/Wfar horizon — read the interval, not the middle
2033$0.03/W$0.01/W to $0.09/Wfar horizon — read the interval, not the middle

Projected to 2033, the horizon the whole engine uses.

questions this page answers

How much does a solar panel cost per watt?

$0.12/W as of 2025, the latest measured value in the series (OWID (IRENA / Nemet / Farmer & Lafond)). The fitted trend has it falling 16.3% a year, with an 80% interval of −24.1% to −7.7%.

How fast is the price of a solar panel falling?

−16.3% a year over the fitted window, an 80% interval of −24.1% to −7.7% — a halving every 3.9 years (80%: 2.5 to 8.6 years). Fitted from 11 observations spanning 10 years.

What will the price of a solar panel be in 2030?

The central fit says $0.05/W, inside an 80% interval of $0.02/W to $0.12/W. The interval is the forecast; the middle number is only its midpoint. Bands widen with horizon because shocks accumulate — a constant-width band would be overconfident.

Is the decline in the price of a solar panel accelerating or slowing?

Splitting the record in half, the fitted rate went from −12% to −15% a year — accelerating. A Chow-style test finds a regime break around 2012 (−8% → −12% a year, F=26). Regime breaks, not window choice, are what dominate this method's errors — which is why every projection here carries an interval.

What is the learning rate for the price of a solar panel?

25% cheaper per doubling of cumulative deployment, fitted as a first-difference regression against 2,600 GW installed (OWID/IRENA measured through 2024; 2026 estimated). Deployment currently doubles every 4.2 years, against 2.9 years in the prior era, which implies −6.6% a year. Wright's law prices per doubling, not per year — as the installed base grows, the same learning rate yields smaller annual declines.

Where does this data come from?

OWID (IRENA / Nemet / Farmer & Lafond). 51 observations spanning 1975–2025, 1 of them from the live feed. Curated benchmark history, extended by a weekly authoritative fetch and by news figures fact-checked against their source before they may touch a fit. Both the observation ledger and the fitting code are public, and the engine publishes its own calibration score and its misses.

the raw numbers

Every observation behind the fit, unrounded by us and unsmoothed. This table is here on purpose: graphs make people underestimate exponential change, and the raw series beside the curve is the one correction shown to work.

year$/Wchangegrade
1975$128.3/W—curated
1976$96.5/W−24.8%curated
1977$70.4/W−27.0%curated
1978$49.8/W−29.3%curated
1979$41.9/W−16.0%curated
1980$35.5/W−15.3%curated
1981$28.4/W−20.0%curated
1982$25.5/W−10.2%curated
1983$20.6/W−19.3%curated
1984$19.1/W−7.1%curated
1985$16.7/W−12.7%curated
1986$13.8/W−17.3%curated
1987$11.7/W−15.0%curated
1988$11.0/W−6.6%curated
1989$11.3/W+3.4%curated
1990$11.7/W+3.4%curated
1991$10.8/W−7.4%curated
1992$10.1/W−6.8%curated
1993$9.5/W−6.5%curated
1994$8.9/W−5.4%curated
1995$8.3/W−7.5%curated
1996$7.7/W−6.5%curated
1997$7.7/W−0.3%curated
1998$6.9/W−9.9%curated
1999$6.4/W−7.6%curated
2000$6.3/W−1.9%curated
2001$6.1/W−3.3%curated
2002$5.6/W−8.5%curated
2003$5.3/W−4.9%curated
2004$4.6/W−14.0%curated
2005$4.6/W+1.1%curated
2006$5.0/W+9.0%curated
2007$5.1/W+0.7%curated
2008$4.6/W−8.9%curated
2009$3.1/W−33.2%curated
2010$2.44/W−20.7%curated
2011$2.00/W−17.9%curated
2012$1.08/W−46.1%curated
2013$0.83/W−22.8%curated
2014$0.77/W−7.7%curated
2015$0.72/W−6.9%curated
2016$0.66/W−7.5%curated
2017$0.56/W−16.0%curated
2018$0.50/W−10.9%curated
2019$0.45/W−8.3%curated
2020$0.36/W−20.8%curated
2021$0.32/W−10.2%curated
2022$0.36/W+10.2%curated
2023$0.31/W−12.0%curated
2024$0.26/W−17.6%curated
2025$0.12/W−53.4%feed

where this comes from

Other ENERGY curves:

Recent briefs from ENERGY, the daily record of what moved:

every ENERGY brief →

Every tracked curve: disk storage · human genome sequencing · transistors per chip · fastest supercomputer · DRAM memory · solid-state storage · space launch, to LEO · wheat yield · AI task horizon (METR) · frontier AI training compute · modular housing, turnkey · frontier-grade intelligence · AI video, finished minute · humanoid robot, capable unit · cheapest humanoid, list price · people still offline · people outside a mobile-broadband signal · farm labor share · healthcare spend per person

← the whole engine: build clock, cascade, calibration · the archive · how we know