The parts that cut into what a system produces are dirt and snow on the glass, plus an inverter with a shorter working life than the panels above it. Understanding those pressure points ahead of time means you can budget for the decades of bills instead of getting hit with them one at a time.

How dirt and snow cut into what panels produce

Dust and other debris settle on a panel’s surface and block or scatter some of the sunlight hitting the cells underneath, Sandia National Laboratories’ modeling guide explains. Some of that layer washes off on its own the next time it rains. The rest needs a manual wash, the same guide notes, since caked-on grime doesn’t rinse away the way a light dusting does.

Snow does more damage than dust in the places that get it. Sandia has compiled published estimates of snow-related energy losses ranging from 1% to 12% a year, with some individual months losing close to 100% of production while a buried panel sits idle. A single storm rarely matters much on its own. If your roof holds snow through a whole winter instead of shedding it, that lost production shows up directly on your electric bill until spring clears the panels.

EnergySage, a solar marketplace, suggests you clean panels a couple of times a year, and notes that enough rain often does the job for free. A sloped roof usually gets that rinse on its own between storms. A flat array in a dry climate is the one you’ll end up hosing down, or paying someone to, every season.

What a professional visit involves and costs

An annual check by a technician covers the wiring connections and mounting hardware, plus the inverter, the parts likely to develop a fault between cleanings and the ones you can’t safely inspect yourself from a ladder. Budget a few hundred dollars a year if you sign up for one of these bundled maintenance plans, EnergySage reports, since it folds the cleaning and the basic check into one visit.

Utility-scale operators budget on a different scale entirely. Lawrence Berkeley National Laboratory found that operators now plan for $5 to $8 per kilowatt of DC capacity a year in operations and maintenance spending, a figure for big solar farms, several orders of magnitude larger than the single rooftop array on your own roof.

The other bill that can catch you off guard is a roof replacement. Pulling panels off and reinstalling them once the job is done runs about $200 per panel, EnergySage says, on top of whatever the roofing job itself costs. Timing a re-roof to happen before your solar install, where possible, avoids that bill entirely.

How fast panels actually lose output

Panel output declines a little every year, and DOE says most systems sold over the last several decades lose about 1% of their power a year. That’s slow enough to miss from your own yard, since a system won’t visibly dim and the drop shows up when you compare a full year’s production against the year before. A broader literature review that pooled nearly 2,000 field-measured systems found a median decline of 0.5% a year, lower than DOE’s rounder estimate partly because it draws on newer, better-performing hardware alongside decades-old panels.

Either rate lands in roughly the same place by the 25-year mark DOE treats as a healthy system’s baseline, keeping at least 80% of its original output, the same number that shapes how panel warranties are written.

What the warranty behind a panel actually promises

Panel warranties run in two layers, an equipment warranty against physical defects and a performance warranty guaranteeing a minimum output. REC’s warranty fact sheet guarantees its Alpha series at 92% of rated power in year 25. Its other two lines, TwinPeak 5 and N-Peak 3, guarantee less. Those panels are rated to keep 86% of their output at the same 25-year mark.

Every REC panel carries a 25-year product warranty underneath those performance numbers. Other makers land in a similar range with different terms of their own. Qcells backs its residential panels with a product warranty of up to 25 years and a performance warranty that runs to 30. Silfab offers a 25-year product warranty alongside a 30-year linear power warranty.

A performance warranty pays out when a panel underperforms the curve its maker guaranteed in writing. It doesn’t cover storm damage or outright neglect. Keep your own notes on cleanings and inspections, in case you need to prove the panel was maintained when you file a claim.

Why the inverter usually needs attention before the panels do

The inverter is the one part of a home solar system built almost entirely from switching electronics, and its own maker warranties show the difference from a panel’s slow, steady decline. Enphase backs its IQ8 microinverters with a limited warranty running up to 25 years, and SolarEdge’s Home Wave inverter carries a 12 to 25 year warranty range, depending on whether you pay to extend past the standard term.

Even the shorter end of that range falls short of a panel’s 25-year product warranty. Skipping the extension is the practical reason an inverter swap, not a panel replacement, tends to be the one big maintenance bill a system generates in its lifetime. Set money aside for that swap somewhere around the middle of your system’s life. The inverter swap is the maintenance bill that shows up, even though the panels get all the attention.

Why year 15 to year 30 is still a guess for most US systems

The industry’s own planning assumptions keep moving because the evidence keeps arriving late. Lawrence Berkeley National Laboratory found that the average project-life assumption utility-scale developers use grew from about 21.5 years in 2007 to about 32.5 years in 2019. Most of the operators LBNL surveyed now plan on 30 years or more for a project’s working life.

DOE runs its own reliability and safety research for the same reason, to give owners and manufacturers a clearer picture of how a system degrades and what its upkeep costs over time as more of the fleet ages into its second and third decade. Roughly 70% of the panels in the US fleet went up since 2019. If your system is part of that group, most of what a panel does in the back half of its working life hasn’t happened yet, including on your own roof.

What the percentage on a panel means

The number on a panel’s spec sheet is its rated efficiency. The Department of Energy defines it as the share of sunlight a cell turns into usable electricity. Inside the cell, sunlight knocks electrons loose in a wafer of silicon, and the wafer’s own electric field pushes those electrons into a current your house can run appliances on.

Not all the sunlight that hits a panel gets that chance. Bare silicon reflects away more than 30% of the light that lands on it, which is why every commercial panel gets an anti-reflective coating before it ships. Of the light that does get through, some passes straight through the cell or turns into heat instead of current, which is a separate loss a coating can’t fix.

A rooftop panel’s percentage has nothing to do with the numbers chased in a lab. Experimental cells built for satellites and other niche uses have reached nearly 50% efficiency, a figure no panel on your house gets anywhere close to.

Today’s panels run from about 21% to nearly 25%

Four current spec sheets show where that range sits. LONGi’s Hi-MO 9 line rates up to 24.80% efficient, and REC’s Alpha Pure-RX reaches 22.6% using a different cell design called heterojunction, or HJT. Maxeon’s Maxeon 6 line tops out at 22.8% in its March 2024 datasheet. Canadian Solar’s standard PERC panels run from 20.8% to 21.3%.

Source: Manufacturer datasheets and product pages (LONGi, Maxeon, REC Group, Canadian Solar), live as of September 2026.

These ratings also fade over time, and the makers put a number on that too. Maxeon warrants its Maxeon 6 line to keep at least 98% of rated output in year one and lose no more than 0.25% a year after that. REC warrants its Alpha Pure-RX to still deliver 92% of its rated power in year 25.

A few percentage points of rated efficiency mostly decide how much of your roof a given amount of power needs. They don’t decide how close any of these panels gets to its own number once it’s installed on your house. Heat, shade, dust and orientation do that.

Heat cuts more output than most people expect

Solar cells actually work better in the cold. The Department of Energy explains that higher cell temperatures cause a slight increase in current but a much larger decrease in voltage, and since power is current multiplied by voltage, the net effect is less electricity on a hot day, not more.

Every panel’s spec sheet lists a temperature coefficient, the percentage of power it loses for each degree Celsius above its 25°C test rating. LONGi’s Hi-MO 9 loses 0.260% of its power per degree. Maxeon’s Maxeon 6 loses 0.29% per degree. A rooftop runs far hotter than that lab bench on a sunny afternoon, so the gap between two panels’ coefficients matters more in a hot climate than a cool one.

If your own roof runs hot in summer, checking the temperature coefficient next to the efficiency rating before you buy shows how much of that heat will show up as lost output instead of staying on the spec sheet.

A shaded panel loses more than its own share of power

Manufacturers don’t treat shade as a minor problem. Maxeon’s Maxeon 6 datasheet says its wiring uses an “electrical architecture that mitigates the impact of shade and prevents hot-spot formation”, and lists three bypass diodes built into the panel’s junction box. LONGi built a separate “soft breakdown” design into its Hi-MO 9 line specifically to prevent localized overheating from shade.

A shaded cell can heat up in addition to losing output, and that’s the failure bypass diodes and shade-tolerant designs exist to stop. A tree branch that has grown in since installation or a new structure next door can turn a small patch of shade into a bigger problem than the shaded area alone suggests. If a tree in your own yard has filled in since your panels went up, that one shaded panel could be pulling down every panel it shares a string with.

One system-level fix is inverter choice. DOE notes that a microinverter allows independent operation of each panel, “which is useful if some modules might be shaded,” instead of one shaded panel pulling down every other panel wired to the same inverter.

Dust and dirt take a real but uneven bite

Soiling is the industry’s word for what builds up on a panel’s glass over time. Sandia National Laboratories’ PV Performance Modeling Collaborative describes it as dust and other particulates that accumulate on module surfaces, absorbing and scattering some of the sunlight before it reaches the cells.

There’s no single loss percentage that holds on every roof. Sandia’s own modeling guide calls local soiling assumptions a major source of uncertainty in performance models, because how much a panel loses depends on local dust and rain and on the panel’s own surface coating. How much collects on your panels depends on your climate and how often it rains, so a roof in a dry, dusty region needs more attention than one that regular weather rinses clean.

A 2026 study in IEEE Journal of Photovoltaics, hosted through DOE’s research library, tested two commercial panel models from the same manufacturer side by side and found that one lost roughly twice as much power to soiling as the other, a gap the researchers traced partly to differences in the panels’ surface coating. Sandia’s guide notes that some soiling washes off naturally in rain, while other types need manual washing to remove.

Orientation sets the ceiling before any of this happens

Everything above assumes a panel is already pointed the right way, and orientation is fixed at installation. DOE says that for the highest annual energy output, panels in the Northern Hemisphere should point due south and tilt to the local latitude. The EIA adds that panels produce the most power facing the sun directly, which is why most US systems face south at a tilt chosen to optimize output.

If your own roof faces east or west instead of south, none of the fixes below close that gap, because orientation sets the ceiling before heat, shade or dust get a chance to subtract from it.

None of that changes a panel’s rated efficiency. It changes how close a system gets to that rating over a full year, a question of a system’s capacity factor more than the number printed on the panel.

What you can still change once the panels are up

Orientation is fixed the day the installers leave, and so is your local climate. Shade and dust are not. If a tree or a new structure has grown into your panels’ sightline since installation, ask an installer whether microinverters would stop that one shaded panel from dragging down the rest of the array.

If your roof collects more dust than the local rain washes off, a manual wash clears soiling that otherwise sits there. Neither costs anything close to a new set of panels, and both push your real output back toward the number printed on the spec sheet.

The test protocol California adopted from Sandia National Laboratories produces a second, lower “weighted” number built from how an inverter performs across a full day. How hot the box runs and how the array is sized against it move the real number even further from that headline figure.

What the percentage on a spec sheet measures

Efficiency here means one specific ratio. Inverter efficiency is the ratio of AC output power to DC input power, the way Sandia National Laboratories defines it in the test method the industry uses. Whatever DC power doesn’t come out the other side as AC left the box as heat instead of current. That’s the gap between the sunlight hitting the panels on your roof and the power your meter counts.

Inside the box, transistors switch the incoming DC current’s direction back and forth rapidly, and a filter smooths that switching into the clean 60-hertz wave a refrigerator or a laptop charger expects. There’s no fuel and nothing that spins. That efficiency number sits downstream of how well the panels themselves turn sunlight into DC power.

Peak efficiency and weighted efficiency are different numbers

Manufacturers advertise peak efficiency, one point on the curve. SolarEdge’s SE7600H-US home inverter lists a maximum efficiency of 99.2%, while its own CEC weighted efficiency, the figure built from performance across a full range of loads, comes in at 99%. On a microinverter the gap runs a little wider. Enphase’s IQ8 and IQ8+ models list a 97.7% peak efficiency against a 97% CEC weighted efficiency.

Both figures come from the same California Energy Commission protocol, which makes a CEC weighted rating from one brand directly comparable to a CEC weighted rating from another. If two quotes for your own roof list peak numbers a fraction of a point apart, ask which CEC weighted number backs each one. That’s the figure closer to what gets credited to your account over a year of changing sun.

What CEC weighted efficiency counts that a European rating doesn’t

CEC weighted efficiency and the older “European efficiency” number share a formula and differ in their assumptions about how much time an inverter spends at each power level. Both come from the protocol California adopted from Sandia National Laboratories in 2004, which sums an inverter’s measured efficiency at seven power levels, each multiplied by its own weighting factor.

The two weighting sets pull in different directions because they model different climates. The CEC weighting is built on irradiance and temperature data for the Southwest US, where clear, high-sun days push an inverter toward 75% of its rated output more often. The European weighting assumes cloudier, lower-sun conditions and leans instead on the 50% load point. If your town gets the kind of clear, high-sun afternoons the Southwest does, the CEC weighted number on a spec sheet is the closer match to what your own system will produce.

Source: California Energy Commission inverter test protocol, developed by Sandia National Laboratories, Endecon Engineering and BEW Engineering, 2004.

The same inverter tested under each method can end up with two different official efficiency numbers. Neither one is wrong. They’re measuring the same box under two different assumptions about the sky.

Why arrays get built bigger than the inverter

Installers routinely wire more DC panel capacity into a system than the inverter can output in AC, a practice measured as the DC-to-AC ratio. US solar plants typically report 10% to 30% more DC capacity than AC capacity, the US Energy Information Administration found. The capacity-weighted average for large US plants moved from about 1.17 in 2010 to about 1.26 in 2016, and climbed further to a 1.33 median by 2018, Lawrence Berkeley National Laboratory reported. In that 2018 data, fixed-tilt projects ran a 1.41 median ratio against 1.31 for projects on single-axis trackers.

The tradeoff is clipping, where the inverter caps its output once the array sends more DC power than it can convert. Modeling of a fixed-tilt system in Oak Ridge, Tennessee found a 1.25 ratio cost about 1% of potential annual generation to clipping, a small price on a roof built to hold more panels than the inverter alone would use. Pushing the ratio to 2.0 cost 16% of potential generation on a fixed-tilt array and 22% on a single-axis tracker. That gap between clipped output and full potential runs alongside a plant’s annual capacity factor, the separate measure of how much of its rated capacity a plant delivers over a year.

Regulators don’t let this run unchecked everywhere. Australia’s Clean Energy Council requires a grid-connected inverter’s AC rating to be at least 75% of the array’s DC rating, which caps the DC-to-AC ratio at about 1.33 for systems the guideline covers.

What heat does to that rating

None of the weighted numbers above get measured at temperature extremes. Efficiency drops once the electronics get hot enough to trigger derating, where an inverter steps down its own output to protect itself. SMA’s own technical documentation puts it plainly: “the power is reduced in steps,” and “in extreme cases, the inverter will shut down completely.”

There’s no single temperature that triggers this across every model or install. It depends on the DC voltage the array runs at and how much bigger the array is than the inverter. In one of SMA’s worked examples, an inverter fed at 800 volts DC with an array 40% larger than its AC rating can deliver full rated power at any ambient temperature up to 31°C before it has to start backing off. That box is often mounted on an outside wall of your house or garage, so the same summer heat radiating off the siding is what it has to shrug off to keep production up.

Operating range is a separate promise from full-power range. SolarEdge rates its Home Hub inverter to operate from -40°C to 60°C, though surviving a temperature and delivering full rated output at that temperature are different claims, and only the second one is what a hot rooftop tests.

String inverters and microinverters trade efficiency for shade tolerance

A string inverter wires many panels into one circuit feeding a single box, and it’s the cheaper, more common setup. Shading or a fault on one panel drags down the output of the whole string, the US Department of Energy notes, because the panels in a string share one electrical path.

A microinverter sits behind each panel and converts that panel’s power on its own. Shading or damage to one panel won’t touch what the others produce, though microinverters cost more per panel than a single string inverter serving the whole array. If a chimney or a tree shades one corner of your roof for part of the day, a string inverter lets that shaded panel drag down the whole array and a microinverter doesn’t. You pay for that protection in cost per panel.

Microinverters post a slightly lower CEC weighted efficiency in the numbers above, 97% for Enphase’s IQ8 line against 99% for a comparable SolarEdge string inverter. Both ratings come from the same CEC protocol, so that two-point gap reflects a real difference between the tested products.

The industry asked for an updated test in 2018 and is still waiting

The test protocol behind both weighted efficiency numbers dates to 2004, built by Sandia National Laboratories before power optimizers and battery inverters were common on US roofs. The Solar Energy Industries Association asked the California Energy Commission in November 2018 to update the protocol with test procedures for battery inverters and power optimizers.

No revised protocol has replaced it. Every spec sheet you compare today, whether it’s a SolarEdge or an Enphase, is still scored against the same seven load points Sandia chose when it wrote the protocol, and whatever inverter ends up on your roof keeps the rating it shipped with for as long as it’s there.

About 49% of Oahu’s single-family homes have solar panels on the roof, and Hawaiian Electric had connected 118,841 small-scale solar systems across its five islands by the end of September 2025. Renewables supplied 37% of the power Hawaiian Electric delivered in 2025. If you’re one of the roughly half of Oahu homeowners without panels yet, there’s a decent chance your neighbor already has some on the roof and a battery in the garage.

Hawaii got here by shutting down its early export incentive and rewriting the payback rules twice, and homeowners answered by keeping more of that power for themselves.

How much of Hawaii’s power comes from solar right now

Renewables made up 37% of the electricity Hawaiian Electric supplied in 2025, up one percentage point from 2024. Solar specifically is harder to pin to one number. SEIA puts solar’s share of Hawaii’s generation at 23.67%, with 2,313 megawatts DC installed in its latest count, spread across utility-scale plants and rooftop systems on homes and businesses.

Nearly half of Oahu’s single-family homes carry part of that capacity on their own roofs, among the 118,841 solar systems Hawaiian Electric had connected across the islands by the end of September 2025. The share runs lower elsewhere in the state, 47% of homes in Maui County and 30% on Hawaii Island, so which island you live on still shapes the odds that your own street already has panels. Hawaii’s utilities generated a combined 9,160,979 megawatt-hours of electricity in 2024, split across five separate island grids instead of one connected network.

Why a new Hawaii solar system almost always comes with a battery now

Rooftop adoption kept climbing after Hawaii cut its rate for exported solar power, and the industry adapted by pairing batteries with new systems instead of selling into the grid at a discount. In Honolulu County, 92.4% of new residential solar permits included a battery in 2025, up from 26.8% in 2017 and 62.7% the year after that. A home battery has become the default part of a new system.

Hawaiian Electric encouraged that shift with Battery Bonus, a program that paid solar customers $850 per kilowatt of battery capacity up front, plus $5 per kilowatt every month for 10 years, delivered as a credit on the power bill. The program closed to new applicants on July 1, 2024. Similar pay-for-flexibility programs have since appeared in states like California and Vermont, all trying to get batteries that are already on a homeowner’s wall to help the shared grid instead of only backing up one house.

For a homeowner today, a battery bank does the job net metering used to do. It stores the power your roof makes at midday so your house can run on it after sundown, instead of sending it to the grid for a lower rate.

What Hawaii’s highest power bills have to do with your roof

None of this is cheap. If you’re paying Hawaii’s average residential bill, you’re handing over $213 a month for power, the highest bill of any state. The national average that year was $144. The reason isn’t a lack of sunshine.

EIA points to petroleum. Most of Hawaii’s grid-delivered electricity still comes from oil-fired plants, even on homes where rooftop solar already cuts what a customer draws from the grid. Hawaii’s average retail electricity price was 38.00 cents per kilowatt-hour in 2024, the highest of any state, well before the grid started running out of room for more rooftop systems.

How Oahu’s circuits pushed Hawaii to end net metering

Picture Oahu’s grid on a clear afternoon. Rooftop panels across the island all push power onto the wires within the same couple of hours around midday, flooding lines that were built mainly to carry electricity one way, from the power plant to the house. By 2015, solar penetration on some circuits was nearing 20%, a level Hawaiian Electric and its regulators worried could threaten the grid’s reliability. The Hawaii Public Utilities Commission responded by ending net metering for new solar customers in October 2015.

Many of Hawaii’s more than 400 circuits on Oahu can still only take on new rooftop solar during peak daytime hours, so where your house sits on the grid can still decide how fast a new system gets approved. Under the old program, homeowners had been credited at the full retail rate, about $0.27 per kilowatt-hour, for solar power they sent back to the grid. Its replacement, Customer Grid Supply, paid new solar customers up to $0.15 per kilowatt-hour for exported power.

The tariffs that replaced net metering, and why batteries won

Two years later, Hawaiian Electric rolled out a second round of tariffs. Smart Export, built for systems paired with a battery, pays zero from 9 a.m. to 4 p.m. for power sent to the grid. Outside that window, it pays about $0.15 per kilowatt-hour, capped at 25 megawatts statewide. Customer Grid Supply Plus, for systems without storage, pays roughly $0.101 per kilowatt-hour on Oahu, capped at 35 megawatts.

If you’re comparing the two, the math explains why nearly every new system on Oahu now ships with a battery. Exporting at midday for about ten cents a kilowatt-hour is a worse trade than storing that same power and using it yourself after dark, when Hawaii’s average retail rate ran 38.00 cents per kilowatt-hour in 2024. Readers curious how net metering works in states that still offer it can see how net metering works.

Retiring coal and reaching for 100% renewable by 2045

Hawaii shut down its last coal plant, on Oahu, in 2022. The AES-owned plant had supplied up to 20% of Oahu’s electricity before it closed. Hawaii is now working toward a legally required target of 100% renewable electricity by 2045, a law signed in June 2015.

The law set an early checkpoint of 30% renewable by 2020, already passed, and the next one due is 40% by 2030. A later checkpoint calls for 70% by 2040, five years ahead of the full 100% deadline. Every checkpoint after this one still runs through Hawaiian Electric’s fuel mix, and eventually through your bill, since the push toward that next target leans on the same rooftop solar and batteries already covering close to half of Oahu’s homes.

Hawaii’s rooftop boom now has to outrun record demand growth

Hawaiian Electric’s 37% renewable milestone in 2025 came with a warning attached. The utility’s own announcement paired that gain with the state’s biggest jump in electricity demand since 2004, a 2.5% rise in electricity use in a single year. Rooftop solar and batteries have grown fast enough before to reach about half of Oahu’s homes. Whether they can outrun a demand jump this size, and keep your bill from climbing past the $213 a month Hawaii already pays, is the honest question these numbers leave open.

Net metering credits solar customers for the extra power their panels send back to the grid, and the math changed for good in California on April 15, 2023, when the state swapped its version for a new formula regulators call the Net Billing Tariff and the industry calls NEM 3.0. Most states still run some kind of net metering or net billing, though the rules differ by state and utility.

Feed-in tariffs and solar leases work on different mechanics than net metering, and mixing them up is an easy way to misjudge what your rooftop, or someone else’s, will earn. Community solar offers a third path, letting people without a usable roof buy into a shared array instead.

Net metering rules change by state, and some let you buy into a shared array

Net metering isn’t one national rule. Each state, and often each utility inside it, sets its own terms for how much credit you get and how long it lasts.

Some states also run a version called community solar or virtual net metering. It lets you subscribe to a share of a solar array you don’t host on your own roof and get credited on your bill for your portion of what it produces.

Say your own roof faces north or sits under a stand of trees. A community solar subscription can still get you a share of a sunnier array a few miles away, credited on your regular electric bill instead of a separate payment. That’s the option for renters too, and for anyone whose lease or homeowners association rules out panels on the roof.

Rooftop solar has grown fast across the US. By September 2026, SEIA put total US solar capacity at enough to power more than 50 million American homes. That figure counts capacity, the amount of power the fleet could produce, and it doesn’t tally how many homes carry panels on their own roof. Energy Collective’s look at solar’s growth picks up where the homes count leaves off.

California swapped net metering for the Net Billing Tariff

California’s shift happened through CPUC Decision D.22-12-056, adopted in 2022. It took effect on April 15, 2023.

Under the previous rules, NEM 1.0 and 2.0 credited exports near the full retail rate. NEM 3.0 pays exports at a grid avoided cost value instead, usually lower than the retail rate, though it can climb above retail during peak demand hours. That gap is the core of the fight over what a utility owes a rooftop solar owner for power sent back to the grid. If you’re comparing what a new California system would earn against an older one grandfathered into the previous rules, that gap between the avoided cost value and the retail rate is where the difference on your bill comes from.

The new tariff requires an electrification time-of-use rate plan and bills monthly instead of annually. Anyone who enrolled when it started also keeps nine years of bill protection.

If you want to see how pairing solar with storage works instead of exporting at the avoided cost rate, the page on home battery backup with solar covers the mechanics.

Feed-in tariffs use a second meter and pay a fixed rate for everything

A feed-in tariff works differently from net metering. Picture two meters bolted side by side on the wall. One tracks the power your house pulls from the grid, the same as always. The other counts what your panels push out and nothing else, whether or not your household used any of that power itself.

That second meter is what a feed-in tariff pays for. It pays a fixed rate for every kilowatt-hour a system generates, while net metering only credits power a system exports to the grid. If your own utility ran on this model, a sunny month would raise your payment even in a month your house used plenty of that power itself, since the two meters keep separate tallies.

Germany’s Renewable Energy Sources Act built on a feed-in tariff system the country introduced in 1990, guaranteeing a fixed tariff for 20 years plus priority grid access. Since the 2014 EEG reform, installations larger than 100 kilowatts have had to sell their power on the exchange instead, collecting only the market premium, the gap between the tariff and the wholesale price. The surcharge that had funded the program, the EEG-Umlage, came off consumer power bills on July 1, 2022, with funding shifted to the country’s energy and climate fund, paid for by emissions-trading revenue.

A lease or a loan decides who owns the panels

Buying your solar system outright, whether with cash or a solar loan, means you own the equipment and you’re the one who can claim any tax credit for it. A lease or a power purchase agreement works differently. A third party owns the panels, and you pay a fixed monthly lease payment or a rate for the power the system produces instead of paying for the hardware.

A lease payment is often lower than a loan payment, since you’re paying for power instead of equipment. A loan payment builds toward owning the system outright once it’s paid off, and a lease never does. Which one fits depends on how long you plan to stay in the house and how the monthly numbers compare for your own roof and rate plan.

Ownership decides which tax credit applies. The homeowner credit belongs to whoever owns the panels, so a lease or power purchase agreement was never eligible for it.

The tax credit for buying your own system already ended

If you bought your system instead of leasing it, the federal Residential Clean Energy Credit covered 30% of the qualified cost, including battery storage installed alongside the panels. That credit applied to systems installed anytime from 2022 through December 31, 2025.

There’s no version of the purchase credit left for a system installed after that date. That deadline follows the equipment’s owner. A cash purchase and a solar loan both count as owning the system, so both lose the 30% credit for anything installed from 2026 on, regardless of how the purchase was financed.

Leases and power purchase agreements sit on the other side of that line. EnergySage says a lease or PPA still qualifies for a tax credit, since the panels belong to the leasing company and the credit goes to that company instead of the homeowner.

Your utility’s tariff decides what your export credit is worth

Program rules are the part that moves. SEIA frames it as most states offering some form of net metering or net billing, while the details vary by state and utility. A program summary, including this one, is a starting point and never a guarantee. DSIRE tracks current net metering and billing rules by state as programs change, including any state that redoes its export rate the way California did in 2023.

Start with your own usage. The average U.S. household used 10,791 kWh of electricity in 2022, according to the EIA. Pull that figure up against your own bill, then set it next to what your utility’s own tariff pays for a kilowatt-hour sent back to the grid. That comparison shows what your own roof would earn.

Solar power costs $38 to $78 per megawatt-hour today, unsubsidized. A newly built US nuclear reactor costs $141 to $220 for the same unit of power, according to Lazard’s 2025 cost analysis. If your utility picks nuclear over solar for its next plant, that gap is most of what ends up in your rate case. Solar panels only run 20% to 30% of the time in a year, while nuclear reactors run closer to 90% of the time, so these two price tags aren’t measuring the same thing on their own.

What a megawatt-hour costs today

Lazard, the financial advisory firm that publishes the industry’s most closely watched cost benchmark, put out its 2025 unsubsidized cost report in June. The ranges below strip out every state and federal subsidy, so they show what it costs to build and run each kind of plant:

Source: Lazard LCOE+, June 2025.

Solar and wind land in roughly the same range, both well under half of nuclear’s cost. Pairing solar with batteries costs more than solar alone, but even its top end, $131 per megawatt-hour, undercuts new nuclear’s low end by $10 per megawatt-hour. That’s the choice a utility faces on every new project, solar’s price or nuclear’s, for each megawatt-hour it puts on your line.

What EIA’s numbers show, and why they don’t match Lazard’s

The US Energy Information Administration publishes its own annual outlook, and its 2025 projections for plants entering service in 2030 show solar at $29.58 per megawatt-hour and a solar-plus-battery hybrid at $53.44. Advanced nuclear comes in at $81.45 in the same projections. Those numbers sit much closer together than Lazard’s, and that’s because they answer a different question.

EIA’s figures include federal tax credits from the Inflation Reduction Act, while Lazard’s strip every subsidy out. The EIA table shows what a developer might pay after credits. Lazard’s shows what a plant costs to build and run before any credits apply. Whether your utility bill ends up reflecting the subsidized number or the unsubsidized one depends on which of these two tables it used to plan the project.

What the last two US reactors cost

Lazard’s nuclear range is built on the only new US reactors completed in the past three decades, Vogtle Units 3 and 4 in Georgia. The project’s budget was $14 billion, targeting service around 2017. Instead, the first reactor started producing power in mid-2023, and the second followed in April 2024. Both arrived six to seven years behind the original target.

The project’s total cost passed $31 billion. Add Westinghouse’s $3.7 billion settlement for walking away from the build, and the running total climbs to about $35 billion. Lazard’s nuclear range assumes a follow-on reactor gets a 30% cost discount from lessons learned on Vogtle, spread over a 70-year operating life. Skip that assumption, and a new reactor built today would likely cost even more than Lazard’s range already shows.

Why cost per megawatt-hour isn’t the whole story

A megawatt-hour of solar and a megawatt-hour of nuclear aren’t delivered the same way. Lazard assumes utility-scale solar runs at a 20% to 30% capacity factor. New nuclear runs at 89% to 92%.

If you flip a light switch after dark, something other than a solar panel is keeping the current flowing, because panels stop producing the moment the sun goes down. That gap is what the industry calls capacity factor, and it’s the reason a cost-per-megawatt-hour number can’t stand alone.

Pairing panels with battery storage is how a solar plant keeps delivering power after dark, and that added equipment is a large part of why the storage line in Lazard’s table, $50 to $131 per megawatt-hour, costs more than solar by itself. Storage still costs less than new nuclear at either end of the range. None of these levelized-cost figures capture the extra grid costs, like backup capacity and transmission upgrades, that show up once solar makes up a large share of what one grid delivers.

The 2010 Duke study that started this argument

This question isn’t new. In July 2010, Energy Collective writer Osha Davidson covered a Duke University economist’s finding that solar had crossed nuclear on price in North Carolina.

Duke economist John Blackburn’s study, commissioned by the advocacy group NC WARN, found commercial-scale solar developers in the state were already offering power at 14 cents per kilowatt-hour. Two utilities were pushing ahead with nuclear plants projected at 14 to 18 cents per kilowatt-hour for the same unit of power. Phys.org’s coverage of the same study put solar at 16 cents per kilowatt-hour when it crossed nuclear’s line, close to Blackburn’s number.

If you lived in North Carolina in 2010, this was the study that started the argument this page settles nationally with 2025 numbers. The 14-cent solar figure only held with subsidies included, and Blackburn predicted unsubsidized solar would compete within about a decade. It took closer to fifteen years, and the eventual gap turned out bigger and more national than his state-specific comparison described.

Was Gemasolar’s storage worth the price?

A related question came from a different direction in a companion piece by writer Nathan Wilson, about whether pairing solar with storage could compete on cost the way plain solar panels do. Gemasolar, a concentrated solar power tower in Spain operated by Torresol Energy, stores heat in molten salt so it can keep generating after dark. Physicist Tom Murphy’s independent analysis at UC San Diego put its capacity factor at 63.1%, well above Lazard’s range for ordinary solar panels.

That capacity factor came at a cost. The same UC San Diego analysis put Gemasolar’s capital cost at roughly $33,000 per kilowatt of average output, a different way of pricing a plant than the per-megawatt-hour figures above, but one the analysis itself called expensive at the time. Ordinary solar paired with batteries is priced today at $50 to $131 per megawatt-hour unsubsidized, well below what Gemasolar cost to build.

Why a nuclear plant can outlast a solar farm after dark

On unsubsidized cost, solar wins by a wide margin today. A reactor still has one advantage these price tags don’t capture. It can run without storage or extra pairing, day and night, for months at a stretch.

That case for nuclear gets stronger under different assumptions. The International Energy Agency and the OECD’s Nuclear Energy Agency, using a lower discount rate and a longer assumed plant life, rated nuclear the cheapest dispatchable low-carbon option in their 2020 cost report, a more favorable read than Lazard’s 2025 US figures give it. Nuclear’s price is far more sensitive to the discount rate and plant lifespan you assume than solar’s is.

No US utility has finished a large reactor since Vogtle, so Lazard’s 30% cost-learning discount for the next one is still an assumption, not a measured result. If you’re deciding which number to trust for the next plant on your grid, solar’s $38 to $78 per megawatt-hour has decades of installed projects behind it. New nuclear’s price range rests on two finished reactors, the pair at Vogtle.

The world added 452 gigawatts of new solar in 2024, and the International Energy Agency has undercounted solar growth at that pace for more than a decade. If you’re trying to guess when solar and batteries get cheap enough to matter for your own roof or your town’s grid, the agency’s own numbers have run behind reality the whole time. Its 2025 outlook lowered its five-year forecast by 5% instead of raising it, the first downward move in years, and it points to policy shifts in the US and China as the reason.

How much solar the world built in 2023 and 2024

By the numbers that matter, solar’s growth kept accelerating while the forecasts were catching up. Global solar capacity additions reached 345.5 gigawatts in 2023, based on IRENA data reported by pv magazine, then climbed to roughly 452 gigawatts in 2024, according to IRENA’s own tally. Total renewable capacity added in 2024, across solar and every other renewable technology combined, came to 585 gigawatts. Cumulative installed solar capacity crossed 1 terawatt in 2022 and then doubled to 2 terawatts two years later, according to Ember’s Global Electricity Review.

Generation is a different measure than capacity, since how much of it converts to electricity depends on average capacity factor. Output grew 474 terawatt-hours in 2024, up 29% year over year, which Ember called the largest source of new electricity generation worldwide for the third year running. If your utility added new generating capacity that year, solar was probably most of it. Growth kept climbing in 2025, when solar output rose a record 636 terawatt-hours, up 30%, enough on its own to meet 75% of the world’s increase in electricity demand that year.

What the IEA expects for 2025 through 2030

The IEA’s Renewables 2025 outlook projects 4,600 gigawatts of new renewable capacity worldwide between 2025 and 2030, with solar PV accounting for about 80% of that, or roughly 3,680 gigawatts. In the US, the lower number reflects an earlier phase-out of federal tax credits and new restrictions on solar and wind permitting on federal land. If you live in the US, those policy changes are likely to slow how fast new solar reaches your state’s grid over the next five years.

Al Jazeera covered the same revision under a sharper headline, calling it a sharp fall in the outlook for renewable power.

How far short the early forecasts fell

Adam Whitmore, an energy policy analyst who has tracked IEA forecasts since 2013, laid out the scale of the early miss in a 2019 post. The 2013 outlook expected around 28 gigawatts of solar installed per year by 2018, while actual installations that year were already running at 100 to 110 gigawatts annually. A 2025 academic review of IEA World Energy Outlooks published between 1993 and 2022 found that renewable energy growth was underestimated across nearly all of them, in an analysis by LUT University researchers reported by pv magazine.

Source: Adam Whitmore (2019). Carbon Brief. IRENA. IEA, Renewables 2025.

Why the models kept missing solar’s growth

Part of the miss came down to price. In his 2013 analysis, Adam Whitmore argued that the IEA had been slow to recognize solar’s falling costs, with cost assumptions in its own outlooks running well above the prices solar was selling at. The agency’s models also leaned on linear extrapolation of national plans at a time when solar costs and installations were both moving on an exponential curve.

That gap between what’s affordable and what the models expect still shows up in long-range scenarios. The IEA’s own 2022 net-zero pathway projects 15.5 terawatts of solar by 2050, while a LUT University analysis reported by pv magazine put what’s technically and economically feasible at 63.4 terawatts, about four times higher. If you’re weighing a solar or battery quote against a year-old projection, the agency’s own long-range scenarios still run understated by a similar margin.

Has the agency fixed the problem, or changed direction

The IEA did correct course for a while. Between 2020 and 2022, it raised its own five-year renewable capacity forecast in three consecutive editions, from 1,092 gigawatts to 1,496 gigawatts to 1,920 gigawatts. That’s a 76% increase in two years, according to Carbon Brief’s analysis of the agency’s own reports. Solar installers kept beating even those newly raised numbers in the years that followed, adding 345 gigawatts in 2023 and 452 gigawatts in 2024.

The 2025 outlook breaks that pattern. The IEA’s five-year forecast moved down for once, and it points to policy retrenchment in the US and China as the cause. That’s a different kind of miss than the one Whitmore and the LUT researchers described.

What the forecast gap has meant for policy and investment

Whitmore’s argument went beyond accuracy for its own sake. His case was that cost assumptions running well above real market prices produced projections that made solar growth look more dependent on subsidy than it was. Actual installations kept accelerating through the years that followed regardless, as the capacity and generation numbers above show.

Grid operators now face a question the original pieces didn’t have to answer in the same way. They have to manage power generated when demand is low and the sun is still out, largely through battery storage, and that balancing act is now as much a part of the growth story as the installation totals themselves. For you, it shows up as new batteries going up next to substations and solar farms, built to hold your afternoon’s cheap solar power until demand picks back up after dark.

Why 2025 already looks like every year before it

The first year of the IEA’s newly lowered five-year window is already in the books. Solar generation grew a record 636 terawatt-hours in 2025, 33% above the previous record year. China alone built 336 terawatt-hours of that growth, more than half the global total. That’s the same shape as 2013, 2020 and every outlook since, with real installations outrunning the agency’s own numbers before the forecast period is even a fifth over.

Solar was the biggest source of new power on the US grid for the fifth year running in 2025, with 43 gigawatts of new capacity installed, even though that total was 14% below 2024. Installs picked back up in 2026, with 11.4 GW in the second quarter alone, up 45% from a year earlier.

Worldwide, solar grew faster still. Global installations hit a record 647 GW in 2025, and solar generation jumped 30%, the fastest growth rate in eight years. A new US law ended the tax credit for home solar at the end of 2025. Wood Mackenzie has cut its US forecast for 2025 to 2030 by 4%, and its downside case sits 18% below that.

How much solar the US installed in 2025 and 2026

Solar made up 54% of all new generating capacity added to the US grid in 2025. Add battery storage and the two together made up 70% of everything the US built in the first half of 2026, with solar alone at 45%. In the first quarter of 2026, solar and storage were 91% of all new US power capacity.

The US installed 7.8 GW of solar in the first quarter of 2026, then 11.4 GW in the second quarter, a jump of 43% from the first quarter and 45% from the same quarter a year earlier.

Trackers count the 2025 total differently. SEIA’s 43.2 GW covers every segment, from rooftops to power plants, and is measured in direct current. EIA counts utility-scale plants only and found 27.2 GW of utility-scale solar came online in 2025, down from 30.8 GW in 2024. Ember’s provisional data put utility-scale additions 6% below 2024.

Further out, the growth curve levels off. SEIA and Wood Mackenzie expect the US to add 217 GWdc of solar between 2026 and 2031, about 44 GWdc a year, a pace the report calls flat even as the country’s solar fleet nearly doubles.

How much of America’s electricity comes from solar now

Installed capacity and the electricity a panel produces are two different numbers, and solar capacity factor is what connects them. Utility-scale solar supplied about 9.6% of all US electricity generation in 2025. Generation from utility-scale solar rose 34% from 2024, to 296,000 GWh.

Counting rooftop solar and wind, wind and solar together made up 19% of US electricity generation in 2025.

EIA counts 151 GW of utility-scale solar in 2025, up from 123 GW in 2024, and projects 181 GW in 2026. Rooftop and other small systems add another 59.5 GW in 2025, which puts the US total at about 210 GW. EIA expects solar generation to grow 21% in 2026 and 18% in 2027.

How the 2025 tax law is reshaping the outlook

The One Big Beautiful Bill Act, signed in July 2025, ended the 30% federal tax credit for home solar on December 31, 2025, and tightened the rules for the commercial and utility credits that remain.

Wood Mackenzie has put a number on what that means. In its base case, the firm expects 246 GWdc of US solar between 2025 and 2030, 4% (11 GWdc) less than it projected before the law passed. Its downside case sits 18% (44 GWdc) below its own base case, with 2026 and 2027 alone about 30% below base.

Residential solar takes the direct hit. Wood Mackenzie forecasts US residential installations falling 21% in 2026. The International Energy Agency publishes its own yearly solar forecasts.

How fast the rest of the world is installing solar

The world added a record 647 GW of solar capacity in 2025, up 11% from 582 GW in 2024. Solar generation grew faster, reaching 2,778 TWh in 2025, up 30% from 2,143 TWh in 2024. That same year, solar overtook wind generation worldwide for the first time.

Solar alone met 75% of the increase in global electricity demand in 2025. It supplied 8.8% of global electricity in the first half of 2025, up from 6.9% a year earlier and 3.8% in the first half of 2021.

A solar module cost $2.51 per watt in 2010 and $0.265 in 2024, both in 2025 dollars, a drop of about 89%.

Which country has the most solar power

China installed 67% of the world’s new solar in the first half of 2025, more than twice what every other country added combined. Ember counted 380 GW of new solar worldwide in those six months, with the rest of the world installing an estimated 124 GW.

By the end of 2025, China’s cumulative solar capacity reached 1.2 terawatts, up 35% in a year, China’s National Energy Administration reported.

Source: China from China’s National Energy Administration. United States from EIA’s Short-Term Energy Outlook, September 2026, and Table 7e. World total from IRENA, Renewables in 2024.

Comparing these numbers exactly is harder than it looks. The world row is a year older than the China and US rows, dating to before the record global additions of 2025. China’s energy agency also counts installations in AC terms, while global trackers like Ember use DC ratings for the same projects.

What would change this outlook

Developers plan to add 43.4 GW of utility-scale solar in 2026, 60% more than in 2025 if it all gets built. That would be about half of the 86 GW of new capacity planned across all sources in 2026.

The tax law’s deadlines come next. Commercial and utility projects keep their federal credits only if they began construction by July 4, 2026 or are in service by the end of 2027, so the number of projects that finish by then is the thing to watch.

Wood Mackenzie’s base case has the US adding 246 GWdc from 2025 to 2030, and its downside case lands 18% lower. SEIA and Wood Mackenzie’s joint outlook of 217 GWdc from 2026 to 2031 already describes annual installs as flat.