String inverters, microinverters and power optimizers

The efficiency comparison that does not work

We read California's certified equipment list directly, because it is the closest thing to a common yardstick in the US market. What it shows is real, and what proposals do with it is not.

Across the listings, microinverters from the largest manufacturer cluster at 97% weighted efficiency, with some models at 97.5%. String inverters from two major manufacturers cluster between 96.5% and 97.5% for residential units. And optimizer-system inverters from one manufacturer list as high as 99%.

Read that quickly and one architecture looks two points better. It is not a valid reading.

In an optimizer system, the DC power from each panel passes through that panel's optimizer first. The optimizer performs a conversion, and every conversion has a loss. Only then does the power reach the inverter whose 99% is being quoted. That 99% describes what the inverter does to power that has already been processed.

A microinverter's 97% describes a single, complete conversion from the panel's DC output to grid AC. There is no earlier stage whose loss has been excluded from the number.

So the two figures are measured at different boundaries, and the difference between them is not a system-level efficiency gap. It is partly a difference in what was measured.

We cannot close the gap for you, and we will not pretend otherwise. The equipment list does not publish optimizer conversion efficiency, and the optimizer data sheets were not reachable from our environment. So we can tell you the comparison is invalid as usually made. We cannot tell you which architecture is more efficient at the system level, and neither can anyone quoting you those two numbers.

What is actually on your roof, and what that costs later

The architectures differ physically in a way that decides what a repair looks like in year twelve, and it is worth being concrete about it because no efficiency figure captures it.

With a string inverter, almost nothing electronic is on the roof. The panels are wired together and DC runs down to a box on a wall. When the inverter fails, an electrician replaces a wall-mounted unit at ground level. That is the cheapest possible service call in this category, and it is the architecture's genuine advantage.

With microinverters, there is an inverter under every panel. High-voltage DC never leaves the roof, because conversion happens up there and what runs down is AC. That is a real safety and design benefit. The cost is that when one fails, someone has to get on the roof, lift a module and work under it. One failure in a twenty-panel array is a roof visit.

With optimizers, you get both. A device under every panel and an inverter on the wall. The optimizers are warranted longest and the inverter shortest, so the component most likely to need replacing first is the accessible one, which is the favourable arrangement of the two.

None of this is an argument for one architecture. It is an argument for asking a question that proposals do not answer: when the first component fails, what does that specific repair involve, and who is paying for the access? A twenty-five-year system will have at least one of these events, and the difference between a wall box and a roof lift is most of its cost.

The comparison nobody makes, and it is the one that matters

Set efficiency aside. Over twenty-five years the question that decides your experience is which components fail, when, and who pays to replace them. The warranty documents answer that, and they answer it in a way neither architecture's marketing leads with.

In a microinverter system, the microinverters themselves carry 25 years from activation. But the system needs a gateway to communicate, and that carries 10 years.

In an optimizer system, the optimizers carry 25 years. But the string inverter they feed carries 12 years.

The shape is identical. Both architectures place a shorter-lived electronic component inside a system sold on a twenty-five-year horizon. One puts it on the wall and calls it an inverter; the other puts it indoors and calls it a gateway. Either way, something in the system is warranted for less than half the life of the panels.

And in both cases the labour to replace it is excluded. Every manufacturer document we have read across this work, now five of them, excludes the cost of removing and reinstalling. So the honest question to ask about either architecture is not which is more efficient. It is: what is the shortest warranty in this system, what does that component cost to replace, and who is paying the labour in year eleven?

We could not obtain warranty documents for two major string inverter manufacturers, so we state no term for them. If you are considering a plain string inverter, that is a document to ask for rather than a figure to take from this page.

Module-level electronics are largely a code requirement

Both microinverters and optimizers are frequently sold as a performance upgrade over a plain string inverter, on the grounds that each panel is managed independently so shading on one does not drag down the string. That mechanism is real.

What is usually left out is that module-level electronics are also how a system on a building satisfies a safety requirement in the electrical code. Rapid shutdown exists so that firefighters can de-energise conductors on a burning building, and it is a safety obligation rather than a feature anyone chose to buy.

Manufacturers certify to it explicitly. One microinverter data sheet states the product is "UL Listed as PV rapid shutdown equipment and conforms with NEC 2014, NEC 2017, NEC 2020, and NEC 2023 section 690.12… for AC and DC conductors, when installed according to the manufacturer's instructions." Optimizer products carry equivalent listings.

So the honest framing is that both architectures satisfy a requirement, and both also happen to deliver per-panel management. A proposal that presents module-level electronics purely as a performance choice, without mentioning that the code effectively drives module-level shutdown on buildings, is telling you half of why the component is there.

One deliberate omission. We are not quoting the code provision's text, or the voltage thresholds usually attached to it, anywhere on this page. The code is copyrighted and paywalled, the publisher's free portal returns no text to automated retrieval, and the only version we located was a machine-generated summary on a republisher's site that the site itself disclaims. We can verify that manufacturers certify against the provision across four editions. We have not read it, so we do not quote it.