Chapter 4　Comparing generation costs

Key points

• The Agency for Natural Resources and Energy's Working Group on Generation Cost Verification (February 2025) estimated the generation cost of perovskite solar newly installed in 2040 at 15.3 yen/kWh (16.4 yen including policy expenses). In the same review, utility-scale solar in 2040 is 7.9 yen (8.4 yen) and residential solar 9.9 yen (10.1 yen)A · EV-011, EV-048, EV-049
• The 2040 perovskite figure is higher than both utility-scale solar (10.0 yen) and residential solar (14.0 yen) as of 2023A · CL-024
• The assumptions are a construction cost of 204,000 yen/kW (143,000 for equipment plus 61,000 for installation), a 14% capacity factor, a 20-year lifetime and a 250 kW rooftop system. The 20-year lifetime is the median of six manufacturers' outlooks, and the review itself notes that those outlooks do not assume a 250 kW rooftop systemA · EV-047
• The first product on the market, launched in March 2026, has a stated durability of 10 years. Replacing 20 years with 10 raises the capital cost about 1.7-fold and the generation cost to around 24 yen (our own mechanical calculation)C · CL-025
• The strategy's self-sustaining level of 10–14 yen is above the same government estimate for utility-scale silicon in 2040 (7.9 yen). "Self-sustaining" does not mean reaching the cost of silicon; it means a price at which a certain number of customers buy without subsidyA · EV-002, EV-053

1. The government's own estimate

As reference material for energy policy towards 2040, the working group estimated, on a common set of assumptions, "the cost per kWh of a generating plant newly built and operated in 2040" for each power source. Perovskite solar was included for the first time, treated as a reference value because "the technology is still under development and the predictability of its technology and generation costs is not necessarily high"A · EV-047.

Yen per kWhA · EV-011, EV-048, EV-049. The ranges depend on whether equipment costs converge to world levels, on deployment scenarios, on learning rates and on whether panel degradation is included. No range is given for perovskite.

Perovskite at 15.3 yen in 2040 is about 1.9 times utility-scale solar and about 1.5 times residential solar in the same year. It is also higher than silicon as of 2023 (10.0 yen utility-scale, 14.0 yen residential)A · CL-024.

2. Reading the assumptions

The assumptions behind the perovskite estimate are as followsA · EV-047.

Three points stand out.

First, most of the numbers come from "the cost outlooks of six manufacturers on the council", the six companies that presented at the second meeting (Chapter 3). The review notes that the companies' outlooks "do not necessarily assume a 250 kW rooftop installation" and that their assumptions "vary". The median of manufacturers' forecasts of their own future product costs has become the government's reference value.

Second, the 14% capacity factor is a rooftop value. The review states that "installation on walls or windows receives less irradiation than on roofs, so the capacity factor falls". The strategy's demand estimate applies a factor of 0.1875 to the wall and window share of domestic demand (Chapter 1), but wall and window costs are outside the scope of the review. The demand-study report gives present costs of 28.0 yen on roofs, 42.9 yen on walls and 57.5 yen on windowsA · EV-007.

Third, integration costs (the cost of handling output variability) were reviewed for utility-scale solar and wind but not for perovskite, which "remains a reference value"A · EV-047.

3. Ten years or twenty

Operating life is one of the assumptions with the largest effect on generation cost. The review's 20 years is the median of six companies' outlooks; in the same review, utility-scale solar is given 25 years and 30 years in a reference case.

The demand-study report (January 2025) uses a 10-year lifetime and an initial cost of about 251,000 yen/kW to calculate present costsA · EV-007. The first product launched (Sekisui Chemical, March 2026) has a published durability of 10 yearsB · EV-025.

What happens when 20 years becomes 10 can be worked out mechanically from the review's breakdown. Of the 15.3 yen for 2040, capital cost (construction 11.2 yen, property tax 1.0 yen, decommissioning 0.2 yen) is 12.4 yen and O&M is 3.0 yenA · EV-047. Capital cost is recovered over the operating life, so at a 3% discount rate, shortening the life from 20 to 10 years raises the annual recovery about 1.74-fold. The 12.4 yen of capital cost becomes 21–22 yen and the total 24–25 yen (the decimals shift with how property tax and decommissioning are treated). This is our calculation using the review's breakdown, not a government estimateC · CL-025.

In other words, at the lifetime of the product now on the market, perovskite in 2040 comes to about 1.7 times residential silicon in 2023 (14.0 yen) and about 1.7 times the upper end of the self-sustaining level, 14 yen. For 15.3 yen to hold, the manufacturers' outlook of a doubling of lifetime by 2040 must come true.

4. What is the 10–14 yen self-sustaining level compared with?

The strategy sets the generation cost at which perovskite solar "can become self-sustaining" at 10–14 yen/kWh and explains the 20 GW as demand at that levelA · EV-002.

That level is 1.3–1.8 times the same government estimate for utility-scale silicon in 2040 (7.9 yen), and equal to or slightly above residential silicon (9.9 yen). "Self-sustaining", then, does not mean reaching the cost of silicon. It refers to the price band at which a certain number of respondents to the demand survey said they would install. Those respondents presumably answered with sites in mind where silicon cannot be installed (low load-bearing roofs, walls, windows), but the survey questions have not been published.

The secretariat paper at the tenth meeting (May 2026) states, in its roadmap illustration, that perovskite solar "is expected to remain higher-priced than existing silicon solar cells" beyond 2040A · EV-053. The government is building the policy on the premise that perovskite will still cost more than silicon in 2040.

The strategy's demand estimate carries another note: "In an estimate under the assumption that deployment proceeds only where the economics of perovskite exceed those of silicon, demand of a certain scale emerges after generation cost falls to about 16 yen/kWh"A · EV-044. Where silicon can be installed, perovskite will not be chosen unless it is cheaper. The 16 yen figure reads as the level at which demand emerges at sites where silicon cannot be installed.

5. Where things stand, and the interim targets

The strategy and the secretariat papers set out the following cost pathA · EV-053, EV-026.

"Technology established" does not mean a product sold at that price in that year; it means that the technology enabling that price exists. The first manufacturer to ship describes its 2030 generation-cost outlook as "about 20 yen"A · EV-026. The secretariat's "technology for 14 yen established by 2030" and the manufacturer's "about 20 yen in 2030" are different numbers for the same year.

6. A third-party view

In a November 2025 commentary, the US-based Institute for Energy Economics and Financial Analysis (IEEFA) set the government reference value of 15.3 yen beside 7.6–10.4 yen for residential and 6.6–8.4 yen for utility-scale solar, cited manufacturers' published durability of 5–12 years, and recommended that perovskite be confined to applications where silicon cannot be installed and to tandem cells, with priority given to deploying siliconA · EV-027. IEEFA's recommendation and the situation of overseas manufacturers are covered in Chapter 7.

7. What we know so far, and what we do not

What we know

• The government's 2040 reference value of 15.3–16.4 yen is about 1.5–1.9 times silicon in 2040 and higher than silicon in 2023
• Most of the assumptions behind the reference value are medians of six manufacturers' outlooks, including the 20-year life. Today's product life is 10 years
• Recalculated at a 10-year life, the figure is around 24 yen (our calculation)
• The 10–14 yen self-sustaining level does not mean parity with silicon; the secretariat paper assumes perovskite stays more expensive than silicon beyond 2040
• Wall and window costs and integration costs are outside the government estimate

What we do not know (being checked)

• The individual cost outlooks of the six companies and their assumptions (life, efficiency, scale)
• Whether the secretariat has recalculated at a 10-year life
• The basis for the 10–14 yen self-sustaining level (open questions Q02 and Q03)

Sources

Unless stated otherwise, costs are in yen per kWh. URLs and retrieval dates for every document are in the evidence register and primary-source list (Japanese).

Pages in this series

• Series home
• Chapter 1 Where did 20 GW come from?
• Chapter 2 What the council discussed (Japanese)
• Chapter 3 Who decided, who received (Japanese)
• Chapter 4 Comparing generation costs
• Chapter 5 Demand created by subsidies (Japanese)
• Chapter 6 How the press reported it (Japanese)
• Chapter 7 What is happening abroad
• Chapter 8 What the same money could have bought
• Chapter 9 Open questions to the government (Japanese)
• Policy timeline (Japanese)
• Public support register (Japanese)
• Evidence register and primary sources (Japanese; CSV)

Related pages

• Rethinking perovskite solar from zero (Japanese) — Checking the technical, economic and institutional premises one by one.
• Reviewing MRI's perovskite industrial-policy column (Japanese) — The 'Team Japan' argument against primary sources.
• JC-STAR and grid cyber-resilience: an effectiveness check — An earlier I-S3 investigation using the same evidence-register method.

Citation and reuse

The CSV and Markdown source files (index) are released under CC BY 4.0. You may copy, adapt and use them commercially if you credit I-S3 Co., Ltd. and link to this page. Quotations from the text with attribution are welcome.

Anyone named in this series who wishes to rebut or correct a statement may write to our contact form or s_masuda@i-s3.com; we publish rebuttals at equal length.

Suggested citation

Masuda, S. (2026). "Chapter 4　Comparing generation costs". I-S3 Co., Ltd., published 2026-10-29, https://www.i-s3.com/perovskite-policy-investigation-ch4-cost-comparison-en.html