PERSPECTIVE

PUBLISHED: 25 AUGUST 2017 | VOLUME: 2 | ARTICLE NUMBER: 17140

The underestimated potential of solar energy to

mitigate climate change
Felix Creutzig1,2*, Peter Agoston1, Jan Christoph Goldschmidt3, Gunnar Luderer4, Gregory Nemet1,5
and Robert C. Pietzcker4

The Intergovernmental Panel on Climate Change’s fifth assessment report emphasizes the importance of bioenergy and carbon
capture and storage for achieving climate goals, but it does not identify solar energy as a strategically important technology
option. That is surprising given the strong growth, large resource, and low environmental footprint of photovoltaics (PV). Here
we explore how models have consistently underestimated PV deployment and identify the reasons for underlying bias in mod-
els. Our analysis reveals that rapid technological learning and technology-specific policy support were crucial to PV deployment
in the past, but that future success will depend on adequate financing instruments and the management of system integration.
We propose that with coordinated advances in multiple components of the energy system, PV could supply 30–50% of electric-
ity in competitive markets.

T
o achieve the 2 °C goal of the Paris Agreement, fossil fuels schemes, non-monetary preferences, and rapid technological learn-
need to be phased out and replaced by low-carbon sources ing. Second, we analyse existing barriers to PV deployment, espe-
of energy. This requires the nearly complete decarboniza- cially the rising costs of integrating intermittent solar energy into
tion of the power sector by 2050, and an accelerated shift towards the electricity system and the up-front financing structure, which
electricity as a final energy carrier 1. The Intergovernmental Panel poses a challenge for countries with less-developed financial institu-
on Climate Change (IPCC) comprehensively evaluated mitiga- tions. We argue that these barriers can be overcome, pointing to the
tion pathways to stay within an atmospheric concentration of increasing option space to accommodate a high share of PV into the
430–480 ppm CO2 — roughly corresponding to the 2 °C target 2, grid and new financing schemes. Finally, we implement the lessons
and studied the specific contribution of renewable energy (RE) to learned from this analysis into the integrated assessment model
mitigation3,4. Surprisingly, solar energy emerges only as a minor REMIND and find that cost-optimal climate mitigation scenarios
mitigation option in most modelling studies. For example, a sce- include 30–50% PV market share in electricity generation.
nario comparison study that fed into the IPCC projected a global
solar electricity generation of 8–35 EJ per year in 2050 (25th–75th PV deployment has consistently exceeded expectations
percentile) for 2 °C consistent mitigation scenarios5, correspond- Direct solar energy has a technical potential of 1,500–50,000 EJ
ing to a ~5–17% contribution to electricity supply 6. In contrast, the per year (ref. 10), exceeding the projected global primary energy
same study projects biomass-based secondary energy production of demand of about 1,000 EJ per year in 2050 (ref. 11) (where tech-
50–90 EJ per year in 2050 (ref. 5). Reasons why the underlying mod- nical potential is defined as the achievable energy generation of
els prefer bioenergy include its higher versatility, making it more a particular technology given system performance, topographic
broadly applicable throughout the existing energy system, and the limitations, and environmental and land-use constraints). In
possibility of generating negative emissions from bioenergy by com- comparison, the global technical potentials for wind (85–580 EJ)
bining it with carbon capture and storage. However, sustainable bio- and biomass (100–300 EJ), are orders of magnitude smaller 3,7,8.
mass sourcing and the underlying land requirements are estimated Photovoltaics (PV) has become the dominant technology to tap
to limit primary bioenergy production to between 100–300 EJ per the solar potential for electricity generation. Indeed, PV is the RE
year in 2050 (that is, secondary energy production of 50–150 EJ), technology with the highest growth rate (more than 40% per year
whereas the technical potential for solar energy is many times that over the past decade12) and the steepest learning curve (about 22.5%
of bioenergy and exceeds total projected global energy demand in reduction in module price for each doubling in cumulative produc-
mid-century 7–9. Recent market developments point to rapid take- tion capacity over the last 40 years13). In 2015, China, Japan and the
up of solar energy in both established and emerging markets. That United States added the majority of new PV capacity and several
raises the question of why solar energy is only marginally repre- European countries (Germany, Greece, Italy) met more than 5%
sented in energy system futures so far. of their electricity demand from PV12. However, the past literature
In this Perspective, we first scrutinize historical scenarios and on energy supply scenarios is in stark contrast with these observed
find that solar energy has so far been systematically underesti- developments. In Fig. 1, we document a history of underestimation
mated in global energy and mitigation scenarios compared to of the growth in PV deployment within the scenario literature.
actual deployment. We investigate the reasons for this underes- Historically, annual growth of the cumulative installed capacity
timation and find that models have overlooked public incentive has varied between 20–72% between 1998 and 2015, corresponding

Mercator Research Institute on Global Commons and Climate Change, Torgauer Straβe 12–15, 10829 Berlin, Germany. 2Sustainability Economics of Human
1

Settlements, Technische Universität Berlin, Str. des 17. Juni 135, 10623 Berlin, Germany. 3Fraunhofer Institut für Solare Energiesysteme ISE, Heidenhofstraβe
2, 79110 Freiburg im Breisgau, Germany. 4Potsdam Institute for Climate Impact Research, Telegrafenberg, 14473 Potsdam, Germany. 5La Follette School of
Public Affairs, University of Wisconsin – Madison, 1225 Observatory Drive, Madison, Wisconsin 53706, USA. e-mail: creutzig@mcc-berlin.net

NATURE ENERGY 2, 17140 (2017) | DOI: 10.1038/nenergy.2017.140 | www.nature.com/natureenergy 1

©
2
0
1
7
M
a
c
m
i
l
l
a
n
P
u
b
l
i
s
h
e
r
s
L
i
m
i
t
e
d
,
p
a
r
t
o
f
S
p
r
i
n
g
e
r
N
a
t
u
r
e
.
A
l
l
r
i
g
h
t
s
r
e
s
e
r
v
e
d
.
PERSPECTIVE NATURE ENERGY

a 100 c

Growth (%)
75
50 104

25
0
1995 2000 2005 2010 2015
b 103 103

Installed capacity (GWp )

Real capacity
WBGU
Greenpeace
IEA
Installed capacity (GWp )

102 102

101 101
Historic highest/lowest growth rates
LIMITS and AMPERE 5th–95th percentile
LIMITS and AMPERE 33rd–66th percentile
WBGU
Greenpeace
100 100 IEA
Real capacity
1995 2000 2005 2010 2015 2015 2020 2025 2030 2035 2040 2045 2050
Year Year

Figure 1 | Growth in PV capacity and scenario projections. a, Year-to-year increase of PV capacity as a percentage of the previously installed capacity.
b, Comparison of historic cumulative installed PV capacity with scenarios from the IEA, Greenpeace and WBGU. c, Scenarios of future cumulative PV
capacity by the IEA, Greenpeace and WBGU compared with an extrapolation of the historical range of growth rates (low, 18% versus high, 50%). Also
shown are results from the scenario comparison projects LIMITS and AMPERE on low climate impact scenarios, which study future transformation
pathways consistent with limiting warming to below 2 °C. Differences between IEA and Greenpeace scenarios and realized capacity is up to a factor of
10 in 2015. See Methods for data sources.

to an average annual growth of 38% (see Methods). Most scenar- in a fragmented international policy landscape23,24. Scenario results
ios underestimated historical growth by a wide margin. Between often mismatch 2015 real capacity because they are calibrated to
1998 and 2010 the International Energy Agency (IEA) has repeat- older historic data.
edly predicted PV annual growth of 16–30%, far below the actual Previous work has found limited predictive power in energy
rate14,15. Only the 2012 World Energy Outlook high-growth ‘new forecasting, including forecasts of technologies for energy supply 25
policies’ scenario predicted 32% annual growth through 2015, in and end use26 as well as overall demand27. Additional criticisms
line with actual development 16, but it then expects decelerated have been directed at integrated assessment models28,29 in which, for
growth of 12% for the next five years. Even in scenarios describ- example, uncertainties in forecast parameters, such as those on sup-
ing a transformation of the global energy system towards a fully ply assumptions30, can cascade and are often not fully expressed in
sustainable one, for example, by the German Advisory Council on the results. Further critiques of models have pointed out that they
Global Change (WBGU), the growth of PV has been underesti- lack increasing returns in technology improvement 31; however, this
mated with a growth rate of about 26% (ref. 8). This specific sce- has been partially addressed in a number of models developed over
nario was able to predict the growth a few years after its publication the last decade32.
in 2003, but, like many others, failed to foresee increasing growth
after 2007. The most advanced scenarios published by Greenpeace Explaining underestimates of models
from 2007–2010 underestimated the growth dynamics: their ini- In PV, the discrepancy between model-based estimates and real-
tially high growth rates were expected to fall to 24–32%, growth world developments can largely be attributed to three key fac-
rates that were surpassed by real development 17–19. The energy revo- tors: policy support; steep technological learning of PV; and cost
lution scenario published in 2012 by Greenpeace accurately cap- increases of competing technologies.
tures the 52% growth until 2015 (then assuming reduced growth
for the future)20. Policy support. Demand grew disruptively after the introduction
Similarly, integrated assessment models have severely underesti- of feed-in tariffs (FiTs) and other technology support schemes that
mated near-term PV growth. The vast majority of energy transfor- were not represented in global models33. After decades in which
mation scenarios documented in the 2014 IPCC fifth assessment solar PV was mostly adopted for space and off-grid applications
report (AR5) feature 2015 PV deployment levels of 50 TWh or less, in remote areas, policies targeted at supporting renewable energies
which is less than half of the global PV production that was actually were put into place accelerating PV deployment, especially since
achieved in 2014 (ref. 12). LIMITS compares future transformation 1998. The take-off was supported by high public acceptance and
pathways consistent with limiting warming to below 2 °C calculated valuing non-monetized attributes; PV adopters show a willingness-
with six integrated assessment models21,22. These pathways account to-pay of US$0.02–0.04 per kWh above traditional energy sources34,
for country-level CO2 reduction targets for 2020 as announced in a premium of about 20%, and higher than for other renewables35.
the Copenhagen pledges, and national renewable expansion tar- In Germany the FiT triggered a 400-fold growth between
gets announced until 2011. AMPERE studied possible pathways 2000 and 2016 (41 GW in 2016, generating about 6.5% of all elec-
toward medium-term and long-term climate targets at the global tricity 36). Crucial for enabling the rapid growth was that the FiT
and European levels and provided insights into the cost implica- guaranteed remuneration for a period of 20 years in combination
tions of policy delay, technology availability and unilateral action with streamlined permitting procedures; hence, PV was understood

2 NATURE ENERGY 2, 17140 (2017) | DOI: 10.1038/nenergy.2017.140 | www.nature.com/natureenergy

©
2
0
1
7
M
a
c
m
i
l
l
a
n
P
u
b
l
i
s
h
e
r
s
L
i
m
i
t
e
d
,
p
a
r
t
o
f
S
p
r
i
n
g
e
r
N
a
t
u
r
e
.
A
l
l
r
i
g
h
t
s
r
e
s
e
r
v
e
d
.
NATURE ENERGY PERSPECTIVE
a 30
IRENA solar (full range + projection)
25 IEA coal
IEA solar

LCOE (¢ per kWh)

20 Lazard (historic US high resource)
LCOE projection (1 MW)
Best of different world regions
15

10

0
2010 2015 2020 2025 2030
Year
b 30
German FiT (utility scale)
FiT/LCOE Germany (¢ per kWh)

25 LCOE projection (Germany 1 MW)

German tenders
20

15

10

0
2010 2015 2020 2025 2030
Year
c
BoS US
BoS Germany
101 Module
1992
Cost (€ per Wp )

1998

100
2007
2012

10–1
10–2 10–1 100 101 102
Cumulative global capacity (GWp )

Figure 2 | Rapid decline in levelized cost of PV electricity. a, Past, present and future solar LCOE. Past data taken from IEA91 for solar and coal, Lazard
consulting92 for utility-scale US projects with high irradiation and IRENA83 for global projects with systems >1 MW. The values assume a discount rate
of 10% for both coal and solar. The larger spread of the 2010 values in comparison to the 2015 values is due to the number of technologies included,
changes in fuel price, and other assumptions. The LCOE range does not contain technologies with CCS. IRENA data shows the range of LCOE utility-scale
PV projects from 2010 to 2015 (left-hand side) and the potential cost reductions from 2015 to 2025 (right-hand side) using a capacity-weighted average
LCOE with a weighted-average cost of capital (WACC) at 7.5% for OECD countries and China and 10% for the rest of the world83. The historic data from
Lazard92 is based on crystalline utility-scale solar with single-axis tracking in high-insolation jurisdictions (for example, southwest United States) at the
lower end, while the high end represents crystalline utility-scale solar with fixed-tilt design. Values for 1 MW utility-scale plants were taken from ref. 93.
The lower end corresponds to a high insolation region (Spain) with low WACC of 2%, whereas the higher end corresponds to low insolation and a higher
WACC of 8%. The set of achievable solar LCOEs based on recent purchase agreements was compiled by the World Bank and includes the UAE, Mexico,
the United States, Peru, Chile, India, South Africa and Zambia94. Future LCOE projections were taken from IRENA and Vartiainen et al.93. b, Evolution of the
German FiT95 along with the results of recent German solar tenders and LCOE projections for south Germany93. The LCOE projection for the insolation
level of Munich was taken with the range denoting WACC between 2% and 8%. c, Cost evolution as a function of global cumulative capacity for modules36
(global average), and BOS in the United States96 and Germany (EuPD; which is based on roof-top systems). The US BOS data costs were estimated as
the difference of install prices for each system but using nationally averaged module prices. Capacities of 0.1 GWp were reached in 1992, 1 GWp in 2000,
10 GWp in 2007, and 100 GWp in 2012.

as a long-term, low-risk investment. This led to an influx of private installations. In 2015 China took over as the largest PV market with
capital from home-owners and small interest groups. At the end of a generous FiT focusing heavily on utility scale38.
2012, 48% of the installed PV capacity was owned by citizens (private Electricity-market models and integrated energy models, by
persons or farmers), a higher rate than for any other modern power contrast, optimize the energy mix by minimizing system cost, and
technology, whereas only 3.5% was owned by utility companies, and assume that electricity is a homogenous good39,40. Typically, they only
the remaining capacity by companies, project developers, banks and represent stylized climate policies, such as carbon pricing. Thus they
investment funds37. Similar developments followed in other coun- were not able to foresee accelerated growth driven by the combina-
tries, such as Spain, Italy, Japan and China, after the introduction of tion of technology-specific policies and personal preferences that
FiT schemes, while in the USA an investment tax credit supports PV are not purely cost-minimizing. Only recently have IAMs started to

NATURE ENERGY 2, 17140 (2017) | DOI: 10.1038/nenergy.2017.140 | www.nature.com/natureenergy 3

©
2
0
1
7
M
a
c
m
i
l
l
a
n
P
u
b
l
i
s
h
e
r
s
L
i
m
i
t
e
d
,
p
a
r
t
o
f
S
p
r
i
n
g
e
r
N
a
t
u
r
e
.
A
l
l
r
i
g
h
t
s
r
e
s
e
r
v
e
d
.
PERSPECTIVE NATURE ENERGY

Table 1 | Decomposing technological learning of PV.

Research and development Industry-scale production Deployment of skills, financing
and regulation
Status Many technological opportunities for Enormous growth and cost reduction Well-advanced in some parts of the world
further improvement, both short term and realized, for example by large factories built (such as China, Spain and Germany), but
long term in China not in others, resulting in region-specific
high deployment costs
Challenges Strong path dependency aggravates Fully realize economies of scale Country-specific, including high costs
market entry of new technologies; dynamic of capital, lack of skills, and complicated
development of dominating technology regulation and permitting processes
eliminates potential advantages of
new technologies
Measures Continue steady research financing Manage coordination issues at very Novel financing mechanisms
Support high-efficiency technologies large scale, for example avoid persistent New business models for grid integration
Support market entry of new technologies over-capacity Skills building
Begin upscaling novel PV materials Streamlining of regulation

While R&D and industry-scale production have achieved fast reduction of module prices globally, decentralized deployment and supporting infrastructure (such as skills, financing and regulation) requires
country-specific and region-specific build-up and increased attention.

is, the ratio of actual to theoretically expected energy output) lower

Box 1 | Institutional and finance solutions to jumpstart local all system costs that scale with system area (for example, mount-
and regional PV. ing and labour); plausible developments include a switch to back-
contact, passivated selective contact silicon solar cell technology in
In some parts of the geographic South, notably Africa, capital, the short term, and tandem solar cells featuring considerably higher
together with skills and governance, are scarce factors contrasting thermodynamic efficiency limits in the long term, as well as fur-
with high potential for solar energy. To reduce the costs of capital ther improvements in module technology 49. Other factors include:
for PV, international donors could provide financial guarantees97. reduced material consumption and a switch to cheaper materials,
The Green Climate Fund, the World Bank, and the Asian for example, replacing silver contacts by copper; higher inverter effi-
Infrastructure Investment bank can also finance human capac- ciency at lower cost; increased system lifetime; economies of scale
ity building and provide regulatory assistance to lower the ‘soft’ reducing production costs for system components as well as instal-
costs of PV deployment. lation costs in utility-scale installations; reduced balance of system
Institutional change in electricity markets could be neces- (BOS), as well as soft-costs for planning and permitting in matur-
sary to incentivize investments in renewable energy sources. ing markets; and new business models reducing financing costs.
Community solar projects provide a template for joint renewable As investments in solar production capacities ramped up in 2015
deployment by utilities and consumers98,99. (ref. 12), actual market development is likely again to outperform
Business models can be fostered based on leasing PV instead the declining growth rates assumed in most models.
of purchase, reducing the costs of up-front capital required and
increased investor’s security100. Municipals can help finance Cost increases of competing technologies. Models were overly
home-owner systems, for example by tax bond guarantees (such optimistic in their assumptions about the costs, potentials, and
as Berkeley FIRST)101. acceptance of competing low-carbon technologies, such as carbon
capture and storage (CCS)50 and nuclear power 51. From an energy
system perspective this implied a more pessimistic outlook for PV.
represent energy policies explicitly 41, but further work is required to In summary, RE support policies, public support, rapid techno-
fully reflect country-level RE policies in these global models. logical learning, and underperforming technological competitors
explain the more rapid development of PV compared to model
Steep technological learning. Technological learning — the posi- projections in the past. But a different set of challenges is perceived
tive feedback between cost reductions and capacity — has been a as critical for the future development of PV, understood as limit-
central characteristic of PV’s growth42. Module costs have decreased ing factor in future projections. In the following we show that these
by 22.5% with each doubling of installed capacity 13, well above challenges can also be overcome.
the median of learning rates of other technologies43 (Fig. 2). This
combination of rapid learning with faster-than-expected capacity Opportunities and challenges
development has thus led to lower-than-expected costs44. Levelized Two issues are especially important for the future development
costs of residential-scale PV are now at or below the price of retail of solar energy, and addressing those is a precondition for future
grid electricity in several countries45; in Germany, even systems fast growth of PV. First, many countries in the global South have
with battery storage are expected to soon be below grid prices46. large cost-efficient potential for solar energy that remains mostly
Utility-scale PV is also now competitive with wholesale prices in untapped because of inadequate institutions and mechanisms to
favourable locations47. Large PV projects in Dubai, Mexico and finance the required up-front investment. Second, with higher
Chile are selling power at less than US$0.03 per kWh without sub- shares of solar energy, system integration questions become increas-
sidies, and at US$0.06 per kWh in Rajasthan, India and Zambia, ingly important.
outcompeting conventional energy sources in those more capital-
constrained locations48. The high costs of financing. These are a barrier to all
The cost reductions are likely to continue, driven by numerous capital-intensive technologies that have low variable costs, such as
factors. Increased module efficiencies and performance ratios (that PV, nuclear, and to a lesser extent CCS. In fact, a decomposition of

4 NATURE ENERGY 2, 17140 (2017) | DOI: 10.1038/nenergy.2017.140 | www.nature.com/natureenergy

©
2
0
1
7
M
a
c
m
i
l
l
a
n
P
u
b
l
i
s
h
e
r
s
L
i
m
i
t
e
d
,
p
a
r
t
o
f
S
p
r
i
n
g
e
r
N
a
t
u
r
e
.
A
l
l
r
i
g
h
t
s
r
e
s
e
r
v
e
d
.
NATURE ENERGY PERSPECTIVE
Table 2 | System options to foster the integration of variable renewable PV and wind energy.
Integration option Effect
Deploy short-term storage options Battery-electric or hydro-electric storage increases matching between
demand and variable generation. Short-term storage is especially important
for PV to counter the diurnal cycle64,65,67. Importantly, costs of battery-electric
storage are rapidly declining85.
Expand the transmission grid Expansion of the transmission grid allows pooling of both demand and
generation variability over large areas59,64,86.
Combine PV and wind to better match load; ensure system-friendly When temporally anti-correlated, the right mix of wind and solar can reduce
deployment of variable renewable energy technologies integration challenges. For example, in the EU, PV production is highest
in summer, while wind peaks in spring and autumn; a combination of the
two leads to a more balanced generation profile66,67. Technology design can
facilitate integration (low wind-speed turbines, IT for real-time monitoring,
tele-control)59.
Match load to supply Use pricing and smart-grid technology to enable demand-side response, thus
matching demand to variable solar generation59,87.
Adjust power market designs and system operations Remove barriers that limit the provision of flexibility by both variable
renewable energy and conventional plants. Decrease the duration/bid size
of electricity products to allow small-scale generators to participate in the
market. Create flexible products that allow PV and wind to fully contribute to
reserve/balancing markets. Decrease the market clearing lead time to reduce
forecast errors59,88.
Electrify transport and heating Electrification of other sectors can replace other energy carriers, thus
increasing the share of PV and wind in primary energy, and provide new
electricity demand with high potential for demand response89,90.

the technological learning rate underlines that financing and per- major factor for the economics of PV on the system level. Power sys-
mitting is a key variable in bringing down costs across geographical tems need to balance demand and generation continuously, and PV
locations worldwide (Table 1). In most developing countries, how- cannot be dispatched according to demand. Instead, output is dic-
ever, the cost of capital is typically >10% due to (perceived) political tated by clouds, the diurnal cycle, and the seasonal cycle. However,
risk, uncertain financial and regulatory conditions, and exchange recent years have brought substantial empirical experience with
rate risks52. Already, for a capital cost of 9%, half of the cost of elec- high shares of PV and wind in existing power systems (>42% in
tricity from PV originates from financing or interest 53. In com- Denmark in 2015, >25% in Ireland in 2015, >20% in Spain in 2013,
petition with a low-capital-cost technology such as a gas turbine, and ~20% in Germany in 2015).
interest rates above 10% can render PV economically unattractive. The literature on system integration costs of PV shows two dif-
Thus, high financing costs can be considered a major obstacle for ferent strands. The first investigates the marginal change of inte-
the widespread application of PV in these countries. In addition, gration costs with increasing PV shares conditional on the current
the spread of PV deployment costs between Germany and other system. The first few percent of PV can be added to any power sys-
countries that have higher irradiation is based on the difference tem. Demand is variable in itself, and adding a small amount of
in bureaucratic costs and human capital54. To address these barri- uncorrelated additional variability does not create new challenges59.
ers, stable regulatory environments, finance solutions, and capacity Meeting peak demand around noon, adding PV can even benefit the
building could reduce the costs of financing PV installations sub- system. Yet, with higher PV shares, integration challenges become
stantially (Box 1). Together with the lower operational risks of PV more relevant and the marginal welfare benefits of adding one unit
compared to fossil fuels or nuclear, this could then translate into of solar capacity turn negative. With increasingly mismatched PV
comparatively lower costs of capital for solar 55. supply and demand, the economic wholesale value of PV decreases
Financial and related barriers are well-exemplified with micro- by up to 50–70% when PV share increases to about 30% (ref. 60).
grids. Microgrids are generally seen as an economically optimal A second set of studies have analysed national or regional power
option in relatively remote rural areas where the cost of grid exten- systems with high PV and wind shares using highly detailed hourly
sion is high56. Microgrids are also mostly preferable to solar home power sector models (see Supplementary Note 1). The results sug-
systems (which operate on an individual-household scale) for a gest that high penetration rates of PV in the range of 25–50% of
village having a flat geographic terrain and more than 500 densely electricity generation are possible and could — under cost assump-
located households using a few low-power appliances each, espe- tions plausible for the 2030–2050 period — even have lower total
cially in solar-rich locations57. However, the essential barrier is the costs than systems with less PV, especially if stringent climate poli-
high costs of capital of PV systems58, which can be best addressed cies are implemented61–64.
by capital subsidies or soft-loan facilities, for example, via the Green To facilitate achieving such high shares of variable renewable
Climate Fund. Where grids are close, providing grid access and reg- energies, a number of integration measures are needed (Table 2).
ulatory frameworks, such as feed-in tariffs, would allow injection Region-specific analysis suggests that for managing the variation
of surplus energy into the grid, increasing the financial viability of induced by cloudiness and the diurnal cycle, hydro-storage and
PV systems. battery-electric storage, and to lesser degree, demand response,
appear to be well-positioned to cost-effectively integrate high mar-
Overcoming the integration challenge. As PV starts to outbid ket shares of solar 65. While in some regions, like the US or India, PV
other electricity sources in terms of direct generation costs, the inte- is well-correlated with demand on a seasonal scale and thus mixing
gration of large amounts of variable renewable electricity becomes a with wind brings limited advantages, other regions such as Europe

NATURE ENERGY 2, 17140 (2017) | DOI: 10.1038/nenergy.2017.140 | www.nature.com/natureenergy 5

©
2
0
1
7
M
a
c
m
i
l
l
a
n
P
u
b
l
i
s
h
e
r
s
L
i
m
i
t
e
d
,
p
a
r
t
o
f
S
p
r
i
n
g
e
r
N
a
t
u
r
e
.
A
l
l
r
i
g
h
t
s
r
e
s
e
r
v
e
d
.
PERSPECTIVE NATURE ENERGY

a b 25,000
4
10

20,000
3
10
Installed capacity (GWp )

Installed capacity (GWp )

15,000

102

10,000
Historic highest/lowest growth rates
101 Greenpeace/IEA scenarios
LIMITS and AMPERE 5th–95th percentile
LIMITS and AMPERE 33rd–66th percentile 5,000
REMIND US$450 per kW floor cost
REMIND US$225 per kW floor cost
REMIND US$0 per kW floor cost
100
Real capacity
0
2015 2020 2025 2030 2035 2040 2045 2050 2015 2020 2025 2030 2035 2040 2045 2050
Year Year

Figure 3 | Updating the integrated assessment model REMIND with recent price information. a,b, New REMIND scenarios compared with historical
data and previous scenario projections on logarithmic (a) and linear (b) scales. The REMIND scenarios use recent PV price information and allow for
endogenous technological learning results in a continued rapid increase of PV, leading to 30–50% market share of PV in power supply. The floor cost in
US$ per kW denotes the cost level at which technological learning stops for the respective scenarios.

strongly benefit from combining solar and wind, as a mix of high range of levelized cost of electricity (LCOE) between US$0.02 per
PV irradiation in summer and high wind potential in winter pro- kWh (continued learning, Middle East and North Africa) and
vides a better seasonal match to demand66,67. US$0.06 per kWh (early saturation, Japan) agrees well with other
projections of future PV costs (see Methods) and seems very
Prospective scenarios with updated information feasible in the light of first projects reaching the projected range
The preceding discussion outlines the diverse set of options for more already now.
flexibly integrating large amounts of solar into the grid. However,
some models, whose results were included in AR5, imposed strict Conclusions
constraints on solar, such as requiring high levels of backup, limit- Scenarios and assessments have consistently underestimated the
ing total share of wind and solar to below 30% of the overall elec- growth of solar energy. PV costs have decreased faster and PV
tricity production5, or limiting growth rates. But recent assessments deployment increased faster than even the most optimistic mod-
argue that growth constraints and system integration constraints els expected. Many global models of energy supply are still using
used in models were too conservative68,69. To avoid these exogenous assumptions that were shown to be overly pessimistic towards PV
constraints, we performed scenario calculations with an improved by the real-world developments of the last years. Accordingly, these
version of the REMIND v.1.6 model that includes a newly devel- models propagate these assumptions and produce scenarios of
oped representation of the integration challenge and options to future energy use that underestimate solar electricity. Our updated
address it 67, the real-world PV capacity values in 2016, and endog- climate mitigation scenarios indicate that by 2050, PV could cost-
enous technological learning with updated cost assumptions (data optimally generate 67–130 EJ of electricity and thus be the dominant
from Figs 1,2). REMIND is a global inter-temporally optimizing electricity supply technology with a share of 30–50% in electricity
energy–economy model that has been extensively used for analy- generation even as the energy system will become more electricity-
ses of climate policies (see Methods). Clearly, this improved version intensive than today’s. This carbon-free solar electricity could bring
only contains a very aggregate description of the integration chal- a considerable boost to the decarbonization of other sectors like
lenges, and our results are informative, not conclusive. transport or industry 70–72.
The increased model detail results in a much more prominent Our study has important implications for research and policy-
role for PV in future cost-optimal electricity supply, especially making. First, the consistently underestimated potential of solar
under stringent climate policy (Fig. 3). In these scenarios — speci- energy — if continued — has implications for the future as deci-
fying a carbon price sufficient to achieve the 2 °C target (US$32 per sion-makers might treat PV too reluctantly. Specifically, policymak-
tCO2 in 2020, increasing by 5% per year until 2100) — we find that ers might fail to address the integration challenge and insufficiently
the results crucially depend on the future of technological learning. plan for adequate grid and storage infrastructure. As a result, low-
If the 20% learning rate slows until the costs reach a lower limit carbon energy sources could be under-deployed, imposing eco-
of US$450 per kWp (kWp, kilowatt-peak: peak power in kilowatts), nomic and societal costs, while instead energy system planning
the global share of PV in electricity generation reaches 30% around might rely too much on other, possibly more problematic, low-
2050. If the 20% learning curve persists longer and only comes to a carbon technologies such as CCS and nuclear. Second, the nature
stop when costs of US$225 per kWp are reached, PV surpasses the of PV upscaling is changing, with the long-term potential at high
30% mark in 2040, while a continuation of the 20% learning curve penetration rates depending less on technological costs of PV but
over the complete modelling period (equivalent to assuming no increasingly on the system integration costs, with storage (and less
floor costs) leads to a 30% PV share already in 2035, and almost 50% so demand response) being an important contribution at high PV
in 2050. In energy terms, PV generates 67–130 EJ of electricity in shares. Hence, realizing high PV scenarios requires not only sup-
2050 in the three scenarios, while the resulting PV technology costs port policies for fostering technological learning of PV, but also
in 2050 for the three learning assumptions are US$760 per kWp, concerted programmes to accommodate large shares of PV in the
US$590 per kWp and US$390 per kWp, respectively. The resulting power grid by modernizing power market regulations, expanding

6 NATURE ENERGY 2, 17140 (2017) | DOI: 10.1038/nenergy.2017.140 | www.nature.com/natureenergy

©
2
0
1
7
M
a
c
m
i
l
l
a
n
P
u
b
l
i
s
h
e
r
s
L
i
m
i
t
e
d
,
p
a
r
t
o
f
S
p
r
i
n
g
e
r
N
a
t
u
r
e
.
A
l
l
r
i
g
h
t
s
r
e
s
e
r
v
e
d
.
NATURE ENERGY PERSPECTIVE
transmission grids, and scaling up storage technologies. Third, in assumptions about the floor costs, and thus the continuation of the learning rate
developing contexts the challenge is different. To get PV jump- of ~20% observed in the past.
The most direct driver of the economics of renewable power technologies
started, secure financing mechanisms are crucial and could become
is their costs. By varying the future costs, rough estimates of high and low PV
part and parcel of the Green Climate Fund and multi-lateral aid. We deployment can be obtained. Historical learning rate has been on average 22.5%
conclude that reaching a solar economy would require policymak- over the last 40 years. It is clear that cost reduction cannot continue infinitely,
ers and society to overcome organizational and financial challenges so the question is when learning will slow. From that perspective, three simple
in the next decades but would then offer the most-affordable clean assumptions are: at half of today’s cost; at a quarter of today’s cost; or not during
the modelled period of time. The resulting span of LCOE originating from the
energy solution for many. Continuing to underestimate the role of different assumptions on learning as well as regional varying irradiation condi-
solar risks squandering this opportunity. tions agree, for example, with the range of US$0.03–0.12 per kWh projected for
2025 in ref. 83. Furthermore, the detailed projections for the year 2050 match
Methods projections from the literature. For example, the resulting range of LCOE for
Here, we describe historical data in Fig. 1, the REMIND model used in Fig. 3, and Europe and the Middle East and North Africa region (US$0.02–0.04 per kWh)
the specification of the new scenarios for Fig. 3. agrees approximately with the projected €0.02–0.04 per kWh (in 2014 €) in ref. 84.
Already now, projects in especially favourable locations have reached the cost span
Historical data and scenarios. The capacity of solar PV was compiled using projected for 2050, underlining that such low costs are feasible.
several sources12,73–75. Greenpeace projections were extracted from all energy
revolution reports17–20,76 including reference, alternative and advanced scenarios Received 7 October 2016; accepted 24 July 2017;
where available. IEA scenarios were taken from several editions of the World published 25 August 2017
Energy Outlook14–16,77,78 including reference and alternative scenarios where
possible. In the older IEA scenarios compiled before 2009, PV capacity is not
explicitly listed; instead, an aggregated number was used for ‘solar’ including PV
References
1. Krey, V., Luderer, G., Clarke, L. & Kriegler, E. Getting from here to
and solar thermal electricity production. In these cases, the solar number was
taken as an approximation for the PV capacity. This overestimates the PV capacity there — energy technology transformation pathways in the EMF27 scenarios.
anticipated by the IEA scenarios, but does not change the general finding that the Climatic Change 123, 369–382 (2014).
scenarios heavily underestimated the PV capacity development. The projections 2. Edenhofer, O. et al. Climate Change 2014: Mitigation of Climate Change
by the German Advisory Council on Global Change (WBGU) were taken from (eds Edenhofer, O. et al.) 33–108 (Cambridge Univ. Press, 2014).
ref. 8. LIMITS compares future transformation pathways consistent with limiting 3. IPCC Special Report on Renewable Energy Sources and Climate Change
warming to below 2 °C calculated with six integrated assessment models21,22. Mitigation (eds Edenhofer, O. et al.) (Cambridge Univ. Press, 2011).
AMPERE studied possible pathways toward medium-term and long-term climate 4. Edenhofer, O., Seyboth, K., Creutzig, F. & Schlömer, S. On the sustainability of
targets at the global and European levels23,24. renewable energy sources. Ann. Rev. Environ. Resour. 38, 169–200 (2013).
5. Luderer, G. et al. The role of renewable energy in climate stabilization: results
REMIND model. The energy–economy–climate model REMIND is a Ramsey-type from the EMF27 scenarios. Climatic Change 123, 427–441 (2014).
general equilibrium growth model of the macro-economy in which inter-temporal 6. Breyer, C. et al. On the role of solar photovoltaics in global energy transition
global welfare is maximized, with a technology-rich representation of the energy scenarios. Prog. Photovoltaics 25, 727–745 (2017).
system. It represents capacity stocks of more than 50 conventional and low-carbon 7. Creutzig, F. et al. Bioenergy and climate change mitigation: an assessment.
energy conversion technologies, including technologies for generating negative GCB Bioenergy 7, 916–944 (2015).
emissions by combining bioenergy use with carbon capture and storage (BECCS)79. 8. Graβl, H. et al. Welt im Wandel: Energiewende zur Nachhaltigkeit
REMIND accounts for relevant path-dependencies, such as the build-up of (WBGU, Springer, 2003).
long-lived capital stocks, as well as learning-by-doing effects and inertias in the 9. Pietzcker, R. C., Stetter, D., Manger, S. & Luderer, G. Using the sun to
upscaling of innovative technologies. REMIND represents 11 world regions that decarbonize the power sector: The economic potential of photovoltaics and
can trade resources and capital, and operates in time-steps of five years for the concentrating solar power. Appl. Energy 135, 704–720 (2014).
period 2005–2060, and ten years for the rest of the century. The model is used to 10. Arvizu, D. et al. Special Report on Renewable Energy Sources and Climate
explore mitigation pathways based on different technology and policy choices and Change Mitigation (eds Edenhofer, O. et al.) 34–212 (IPCC, Cambridge Univ.
analyse bottlenecks for low climate stabilization targets41,80. A detailed description Press, 2011).
can be found in the published model documentation81. 11. World Energy Outlook 2014 (IEA, Paris, 2014).
Further information for the model including the full source code is available 12. Renewables Global Status Report 2017 (REN21 Secretariat, 2017).
(https://www.pik-potsdam.de/research/sustainable-solutions/models/remind). 13. International Technology Roadmap for Photovoltaic Results 2016 (VDMA
To improve the representation of power sector variability and integration Photovoltaic Equipment, 2017).
challenges in REMIND, the model now includes an explicit representation of 14. World Energy Outlook 2009 (IEA, 2009).
residual load duration curves. In REMIND, the residual load duration curve in a 15. World Energy Outlook 2006 (IEA, 2006).
region dynamically changes in reaction to the share of wind and solar power in 16. World Energy Outlook 2012 (IEA, 2012).
that region. The approach captures the effect of wind and solar on the amount of 17. A Sustainable World Energy Outlook 2007 (Greenpeace, 2007).
peak and mid-load capacities, the required short-term storage capacity, as well 18. A Sustainable World Energy Outlook 2008 (Greenpeace, 2008).
as the resulting curtailment. Also, hydrogen generation from curtailed electricity 19. A Sustainable World Energy Outlook 2010 (Greenpeace, 2010).
is modelled. A detailed description of the improvements to the representation of 20. A Sustainable World Energy Outlook 2012 (Greenpeace, 2012).
power-sector variability and integration challenges can be found in ref. 67.
21. Kriegler, E. et al. What does the 2 degree C target imply for a global climate
agreement in 2020? The LIMITS study on durban platform scenario.
Construction of new scenarios. Three additional scenarios were calculated with Clim. Change Econ. 4, 1340008 (2013).
an updated version of the integrated assessment model ReMIND v.1.6 (see model
22. Tavoni, M. et al. The distribution of the major economies’ effort in the Durban
description below). These scenarios include a newly developed representation
platform scenarios. Clim. Change Econ. 4, 1340009 (2013).
of the system integration challenge of wind and solar power, as well endogenous
23. Kriegler, E. et al. Making or breaking climate targets: The AMPERE study on
technological learning with updated cost assumptions based on the data compiled
staged accession scenarios for climate policy. Technol. Forecast. Soc. Change
for Figs 1,2. The scenarios explore futures with a stringent climate mitigation
90A, 24–44 (2015).
policy, namely a global carbon price of US$32 per tCO2 in 2020 (in 2015 US$)
increasing by 5% per year until 2100, which in REMIND is sufficient to stay below 24. Riahi, K. et al. Locked into Copenhagen pledges — implications of short-term
2 °C of warming. emission targets for the cost and feasibility of long-term climate goals.
The modelling of system integration challenges is based on region-specific Technol. Forecast. Soc. Change 90A, 8–23 (2015).
residual load duration curves and encompasses flexibility options such as 25. Hultman, N. E. & Koomey, J. G. The risk of surprise in energy technology
short-term storage and the use of curtailed electricity to produce hydrogen67. The costs. Environ. Res. Lett. 2, 034002 (2007).
approach captures the effects of adding wind and solar power to the power sector 26. Dale, L., Antinori, C., McNeil, M., McMahon, J. E. & Fujita, K. S. Retrospective
on capacity adequacy, dispatch, storage and curtailment. A recent evaluation study evaluation of appliance price trends. Energy Policy 37, 597–605 (2009).
found that the REMIND integration modelling reproduces key power-sector 27. Craig, P. P., Gadgil, A. & Koomey, J. G. What can history teach us: a
behaviours at high wind and solar shares as calculated with the hourly retrospective examination of long-term energy forecasts for the United States.
electricity-sector model REMIX82. Ann. Rev. Energy Environ. 27, 83–118 (2002).
The modelling of PV cost reductions is based on a learning curve with 28. Ackerman, F., DeCanio, S. J., Howarth, R. B. & Sheeran, K. Limitations of
floor costs — as more and more capacity is deployed, the learning slows until it integrated assessment models of climate change. Climatic Change
asymptotically reaches the floor costs9. In the three scenarios we explore different 95, 297–315 (2009).

NATURE ENERGY 2, 17140 (2017) | DOI: 10.1038/nenergy.2017.140 | www.nature.com/natureenergy 7

©
2
0
1
7
M
a
c
m
i
l
l
a
n
P
u
b
l
i
s
h
e
r
s
L
i
m
i
t
e
d
,
p
a
r
t
o
f
S
p
r
i
n
g
e
r
N
a
t
u
r
e
.
A
l
l
r
i
g
h
t
s
r
e
s
e
r
v
e
d
.
PERSPECTIVE NATURE ENERGY

29. Scher, I. & Koomey, J. G. Is accurate forecasting of economic systems possible? 61. Mai, T., Sandor, D., Wiser, R. & Schneider, T. Renewable Electricity Futures
Climatic Change 104, 473–479 (2011). Study: Executive Summary (National Renewable Energy Laboratory, 2012).
30. Gillingham, K. et al. Modeling Uncertainty in Climate Change: A Multi-Model 62. Jacobson, M. Z., Delucchi, M. A., Cameron, M. A. & Frew, B. A. Low-cost
Comparison (National Bureau of Economic Research, 2015). solution to the grid reliability problem with 100% penetration of intermittent
31. Ma, T., Grubler, A. & Nakamori, Y. Modeling technology adoptions for wind, water, and solar for all purposes. Proc. Natl Acad. Sci. USA
sustainable development under increasing returns, uncertainty, and 112, 15060–15065 (2015).
heterogeneous agents. Eur. J. Oper. Res. 195, 296–306 (2009). 63. Bogdanov, D. & Breyer, C. North-East Asian Super Grid for 100% renewable
32. Bosetti, V., Carraro, C., Massetti, E., Sgobbi, A. & Tavoni, M. Optimal energy energy supply: optimal mix of energy technologies for electricity, gas and heat
investment and R&D strategies to stabilize atmospheric greenhouse gas supply options. Energy Convers. Manag. 112, 176–190 (2016).
concentrations. Resour. Energy Econ. 31, 123–137 (2009). 64. Scholz, Y., Gils, H. C. & Pietzcker, R. Application of a high-detail energy
33. Pillai, U. Drivers of cost reduction in solar photovoltaics. Energy Econ. system model to derive power sector characteristics at high wind and solar
50, 286–293 (2015). shares. Energy Econ. 64, 568–582 (2016).
34. Sund, S. & Rehdanz, K. Consumer’s willingness to pay for green electricity: a 65. Mills, A. D. & Wiser, R. H. Strategies to mitigate declines in the economic
meta-analysis of the literature. Energ. Econ. 51, 1–8 (2015). value of wind and solar at high penetration in California. Appl. Energy
35. Borchers, A. M., Duke, J. M. & Parsons, G. R. Does willingness to pay for 147, 269–278 (2015).
green energy differ by source? Energy Policy 35, 3327–3334 (2007). 66. Heide, D., Greiner, M., von Bremen, L. & Hoffmann, C. Reduced storage and
36. Wirth, H. Recent Facts about Photovoltaics in Germany balancing needs in a fully renewable European power system with excess wind
(Fraunhofer ISE, 2017). and solar power generation. Renew. Energy 36, 2515–2523 (2011).
37. Hauser, E. et al. Nutzeneffekte von Bürgerenergie (Greenpeace Energy, Bündnis 67. Ueckerdt, F. et al. Decarbonizing global power supply under region-specific
Bürgerenergie e.V., Institut für ZukunftsEnergieSysteme, 2015). consideration of challenges and options of integrating variable renewables in
38. Global Market Outlook (Solar Power Europe, 2015). the REMIND model. Energy Econ. 42, 316–330 (2016).
39. Hirth, L. The optimal share of variable renewables: how the variability of 68. Wilson, C., Grubler, A., Bauer, N., Krey, V. & Riahi, K. Future capacity growth
wind and solar power affects their welfare-optimal deployment. Energy J. of energy technologies: are scenarios consistent with historical evidence?
36, 149–184 (2015). Climatic Change 118, 381–395 (2013).
40. Kriegler, E. et al. The role of technology for achieving climate policy 69. Pietzcker, R. C. et al. System integration of wind and solar power in integrated
objectives: overview of the EMF 27 study on global technology and climate assessment models: A cross-model evaluation of new approaches. Energy
policy strategies. Climatic Change 123, 353–367 (2014). Econ. 64, 583–599 (2017).
41. Bertram, C. et al. Complementing carbon prices with technology policies to 70. Lechtenböhmer, S., Nilsson, L. J., Ahman, M. & Schneider, C.
keep climate targets within reach. Nat. Clim. Change 5, 235–239 (2015). Decarbonising the energy intensive basic materials industry through
42. Nemet, G. F. Interim monitoring of cost dynamics for publicly supported electrification — implications for future EU electricity demand. Energy
energy technologies. Energy Policy 37, 825–835 (2009). 115, 1623–1631 (2016).
43. McDonald, A. & Schrattenholzer, L. Learning rates for energy technologies. 71. McCollum, D., Krey, V., Kolp, P., Nagai, Y. & Riahi, K. Transport
Energy Policy 29, 255–261 (2001). electrification: a key element for energy system transformation and climate
44. Metayer, M., Breyer, C. & Fell, H.-J. The projections for the future and stabilization. Climatic Change 123, 651–664 (2014).
quality in the past of the World Energy Outlook for solar PV and other 72. Creutzig, F. Evolving narratives of low-carbon futures in transportation.
renewable energy technologies. In 31st Eur. PV Solar Energy Conf. Exhib. Transp. Rev. 36, 341–360 (2016).
http://doi.org/cbwn (EU PVSEC, 2015). 73. Snapshot of Global Photovoltaic Markets (IEA, 2016).
45. Breyer, C. & Gerlach, A. Global overview on grid-parity: global overview on 74. Letting in the Light: How Solar Photovoltaics will Revolutionise the Electricity
grid-parity. Prog. Photovolt. Res. Appl. 21, 121–136 (2013). System (IRENA, 2016).
46. Renewable Power Generation Costs in 2014 (IRENA, 2015). 75. Renewables Global Status Report 2015 (REN21 Secretariat, 2015).
47. Bolinger, M., Weaver, S. & Zuboy, J. Is $50/MWh solar for real? Falling project 76. A Sustainable World Energy Outlook 2015 (Greenpeace, 2015).
prices and rising capacity factors drive utility-scale PV toward economic 77. World Energy Outlook 2004 (IEA, 2004).
competitiveness. Prog. Photovolt. Res. Appl. 23, 1847–1856 (2015). 78. World Energy Outlook 2002 (IEA, 2002).
48. Romm, J. Stunning drops in solar and wind costs turn global power market 79. Klein, D. et al. The value of bioenergy in low stabilization scenarios: an
upside down. ThinkProgress (6 April 2017); https://thinkprogress.org/ assessment using REMIND-MAgPIE. Climatic Change
renewables-cheapest-new-power-globally-74910c78bbbe 123, 705–718 (2014).
49. Haegel, N. M. et al. Terawatt-scale photovoltaics: trajectories and challenges. 80. Luderer, G. et al. Economic mitigation challenges: how further delay closes the
Science 356, 141–143 (2017). door for achieving climate targets. Environ. Res. Lett.
50. Rubin, E. S., Davison, J. E. & Herzog, H. J. The cost of CO2 capture and 8, 034033 (2013).
storage. Int. J. Greenhouse Gas Control 40, 378–400 (2015). 81. Luderer, G. et al. Description of the Remind Model (Version 1.6) (Social Science
51. Grubler, A. The costs of the French nuclear scale-up: A case of negative Research Network, 2015).
learning by doing. Energy Policy 38, 5174–5188 (2010). 82. Pietzcker, R. C. et al. System integration of wind and solar power in integrated
52. Ondraczek, J., Komendantova, N. & Patt, A. G. WACC the dog: the effect of assessment models: A cross-model evaluation of new approaches. Energy
financing costs on the levelized cost of solar PV power. Renew. Energy Econ. 64, 583–599 (2017).
75, 888–898 (2015). 83. The Power to Change: Solar and Wind Cost Reduction Potential to 2025
53. Technology Roadmap: Solar Photovoltaic Energy (OECD/IEA, 2014). (IRENA, 2016).
54. Creutzig, F. et al. Catching two European birds with one renewable stone: 84. Mayer, J. N., Simon, P., Philipps, N. S. H., Schlegl, T. & Senkpiel, C.
mitigating climate change and Eurozone crisis by an energy transition. Current and Future Cost of Photovoltaics (Fraunhofer ISE, Agora
Renew. Sustain. Energy Rev. 38, 1015–1028 (2014). Energiewende Freibg., 2015).
55. Zweibel, K. Should solar photovoltaics be deployed sooner because of long 85. Nykvist, B. & Nilsson, M. Rapidly falling costs of battery packs for electric
operating life at low, predictable cost? Energy Policy vehicles. Nat. Clim Change 5, 329–332 (2015).
38, 7519–7530 (2010). 86. Becker, S., Rodriguez, R. A., Andresen, G. B., Schramm, S. & Greiner, M.
56. Raman, P., Murali, J., Sakthivadivel, D. & Vigneswaran, V. S. Opportunities Transmission grid extensions during the build-up of a fully renewable pan-
and challenges in setting up solar photo voltaic based micro grids for European electricity supply. Energy 64, 404–418 (2014).
electrification in rural areas of India. Renew. Sustain. Energy Rev. 87. Gils, H. C. Assessment of the theoretical demand response potential in
16, 3320–3325 (2012). Europe. Energy 67, 1–18 (2014).
57. Chaurey, A. & Kandpal, T. C. A techno-economic comparison of rural 88. Hirth, L. & Ziegenhagen, I. Balancing power and variable renewables: three
electrification based on solar home systems and PV microgrids. links. Renew. Sustain. Energy Rev. 50, 1035–1051 (2015).
Energy Policy 38, 3118–3129 (2010). 89. Williams, J. H. et al. The technology path to deep greenhouse gas emissions
58. Yaqoot, M., Diwan, P. & Kandpal, T. C. Review of barriers to the dissemination cuts by 2050: the pivotal role of electricity. Science 335, 53–59 (2012).
of decentralized renewable energy systems. Renew. Sustain. Energy Rev. 90. Mathiesen, B. V. et al. Smart energy systems for coherent 100% renewable
58, 477–490 (2016). energy and transport solutions. Appl. Energy 145, 139–154 (2015).
59. IEA. The Power of Transformation: Wind, Sun and the Economics of Flexible 91. Projected Cost of Generating Electricity (IEA, 2015).
Power Systems (OECD, 2014). 92. Lazard’s Levelized Cost of Energy Analysis — version 9.0 (Lazard, 2015).
60. Sivaram, V. & Kann, S. Solar power needs a more ambitious cost target. 93. Vartiainen, E., Masson, G. & Breyer, C. PV LCOE in Europe 2015–2050.
Nat. Energy 1, 16036 (2016). In 31st Eur. Photovolt. Sol. Energy Conf. http://doi.org/cbwp (EU PVSEC, 2015).

8 NATURE ENERGY 2, 17140 (2017) | DOI: 10.1038/nenergy.2017.140 | www.nature.com/natureenergy

©
2
0
1
7
M
a
c
m
i
l
l
a
n
P
u
b
l
i
s
h
e
r
s
L
i
m
i
t
e
d
,
p
a
r
t
o
f
S
p
r
i
n
g
e
r
N
a
t
u
r
e
.
A
l
l
r
i
g
h
t
s
r
e
s
e
r
v
e
d
.
NATURE ENERGY PERSPECTIVE
94. Sargsyan, G. Why Zambia’s 6 cents is more significant than Dubai’s 3 cents. Acknowledgements
Energy for All (6 July 2016). The research leading to these results has received funding from the European Union’s
95. Monitoringbericht 2016 (Bundesnetzagentur/Bundeskartellamt, 2016) . Seventh Programme FP7/2007-2013 under grant agreement no. 308329 (ADVANCE).
96. Barbose, G. et al. Tracking the Sun VIII: The Installed Price of Residential and
Non-Residential Photovoltaic Systems in the United States (Lawrence Berkeley
National Laboratory, 2015). Additional information
97. Collier, P. & Venables, A. J. Greening Africa? Technologies, endowments and Supplementary information is available for this paper.
the latecomer effect. Energy Econ. 34, S75–S84 (2012). Reprints and permissions information is available at www.nature.com/reprints.
98. Funkhouser, E., Blackburn, G., Magee, C. & Rai, V. Business model
Correspondence should be addressed to F.C.
innovations for deploying distributed generation: the emerging landscape of
community solar in the US. Energy Res. Soc. Sci. 10, 90–101 (2015). How to cite this article: Creutzig, F. et al. The underestimated potential of solar energy
99. Dean, N. Solar photovoltaics: In it together. Nat. Energy 1, 15023 (2016). to mitigate climate change. Nat. Energy 2, 17140 (2017).
100. Rai, V. & Robinson, S. A. Effective information channels for reducing costs of Publisher’s note: Springer Nature remains neutral with regard to jurisdictional claims in
environmentally-friendly technologies: evidence from residential PV markets. published maps and institutional affiliations.
Environ. Res. Lett. 8, 014044 (2013).
101. Fuller, M. C., Portis, S. C. & Kammen, D. M. Toward a low-carbon economy:
municipal financing for energy efficiency and solar power. Environ. Sci. Policy Competing interests
Sustain. Dev. 51, 22–33 (2009). The authors declare no competing financial interests.

NATURE ENERGY 2, 17140 (2017) | DOI: 10.1038/nenergy.2017.140 | www.nature.com/natureenergy 9

©
2
0
1
7
M
a
c
m
i
l
l
a
n
P
u
b
l
i
s
h
e
r
s
L
i
m
i
t
e
d
,
p
a
r
t
o
f
S
p
r
i
n
g
e
r
N
a
t
u
r
e
.
A
l
l
r
i
g
h
t
s
r
e
s
e
r
v
e
d
.