The inverter decade, in the figures that survive a source check.
Between 2015 and 2026 the solar inverter changed jobs: from a passive box that converted DC to AC and reported a kilowatt-hour count, into the device every other decision in a distributed energy system now passes through. That narrative is sound. Many of the specific figures circulating with it are not: they trace to vendor blogs, a component distributor, and one predatory journal rather than to primary data. This is the decade rebuilt on primary sources only: what is documented, what is measured, and what we had to leave on the floor.
Two firms now ship more than half the world's inverters, and that part is fully documented.
In 2015 the global inverter market was a genuine contest. Huawei led the shipment ranking with a 13.3% share (IHS Technology, reported via PV Tech, 2016), though on a revenue basis SMA still led that year, a distinction worth keeping rather than flattening into "Huawei was number one." SMA, ABB, TMEIC and Sungrow filled out a genuinely multipolar top tier.
By 2024 that field had collapsed into a duopoly. Global PV inverter shipments reached 589 GWac, up 10% year on year; Huawei shipped 176 GWac and Sungrow 148 GWac, a combined 55% of the entire market (Wood Mackenzie, July 2025). Wood Mackenzie records this as the first time in its ten years of inverter rankings that no vendor beyond the top two exceeded a 5% share, and the tenth consecutive year Huawei and Sungrow held the top two positions. Nine of the top ten vendors by shipment volume were headquartered in China; Ginlong Solis and Growatt held third and fourth.
What the hardware didThe shift from silicon IGBT toward wide-bandgap devices is real and measurable: the market-share numbers attached to it mostly aren't.
The efficiency story is genuine and primary-sourced. Peer-reviewed testing of wide-bandgap devices, silicon carbide (SiC) and gallium nitride (GaN), in a residential PV inverter found SiC cut overall energy losses by roughly 40% versus silicon at the highest current condition (MDPI Electronics 14(6):1061, 2025). In practice that shows up as a ceiling shift: modern silicon-IGBT string inverters peak around 98%, while SiC designs reach 99% and above; a SiC reference design has been documented at up to 99.5% peak (Wolfspeed). GaN, long confined to low-power electronics, reached solar in a specific 2025 event: Infineon's CoolGaN switches powering Enphase's IQ9 microinverter, the first GaN-based microinverter, announced November 2025.
What we could not source to any primary: the widely-repeated claims that silicon carbide reached 35% of new utility-scale string-inverter shipments in 2025 (up from 18% in 2023), that silicon IGBT holds exactly 65% of utility-scale string and 95% of central-inverter shipments, or that silicon carbide hit system-level cost parity precisely at 250 kW. No Yole, TrendForce or Wood Mackenzie report we could find carries those splits, and the "18%" figure appears to originate as an electric-vehicle silicon-carbide statistic, not a solar one. What the primary analysts do say is consistent and less precise: silicon IGBT remains the cost-driven default for central and most utility-scale inverters; silicon carbide has taken the premium residential and commercial tiers; and gallium nitride is a nascent slice concentrated almost entirely in microinverters (Wood Mackenzie; Yole Power GaN 2025).
The parallel pathModule-level electronics won on shade resilience, not peak efficiency: the honest yield figure is smaller than the one usually quoted.
Microinverters and power optimisers solved a problem string inverters structurally can't: one shaded or mismatched module dragging a whole string down with it. The figure usually attached to that advantage, a 5% to 25% yield gain, overstates the top end. NREL's own shading testbed, running identical 8 kW arrays, measured the microinverter system producing 3.7% more energy under light shade, 7.8% under moderate shade and 12.3% under heavy shade (NREL/TP-5200-62471). The "25%" appears in vendor material, not in NREL's data. On pure conversion efficiency the two architectures are now within about a percentage point: Enphase's IQ8M is rated 97.8% peak and 97.5% CEC-weighted (Enphase datasheet), against roughly 98–98.5% CEC for premium string inverters. The microinverter's edge was never peak conversion; it was resilience to real-world, uneven irradiance.
The battery momentWhen storage arrived, the winning design removed a conversion stage: the mechanism is certain even where the exact percentages aren't.
Affordable storage forced the question the rest of the decade answered: where should DC-to-AC conversion happen once a battery sits in the system? AC-coupled designs keep a solar inverter and add a separate battery inverter, so stored energy is converted DC→AC→DC to charge and DC→AC again to discharge: two extra conversion stages. DC-coupled hybrid inverters combine PV and battery conversion in one unit, moving energy DC-to-DC between panels and battery with a single DC-to-AC stage for the grid. The efficiency consequence follows from stage count, not from any one component: DC-coupled round-trip typically lands a few points ahead of AC-coupled (EnergySage; SolarEdge). The exact bands often quoted for this (commonly 95–97% DC-coupled versus 90–94% AC-coupled) vary by product and are best read as illustrative rather than as a measured benchmark. The same SiC power density that shrank the power stage is what made a single combined PV-plus-battery enclosure practical without excessive cooling.
The device becomes infrastructureBy 2024 the inverter was a grid resource in policy, not just in principle.
Two shifts turned the inverter from a passive grid-follower into grid infrastructure. Technically, grid-forming inverters, which establish a voltage and frequency reference themselves rather than synchronising to an existing one, moved from research toward deployment; this is the capability that lets grids with few spinning generators hold stability (NREL research on grid-forming inverter controls). Commercially and legally, 2024 was the year the virtual power plant became policy rather than pilot: 38 states and the District of Columbia took 105 distinct policy actions on VPPs and distributed-energy-resource aggregation that year (SEPA / NC Clean Energy Technology Center, "50 States of Virtual Power Plants," 2024), and the US Department of Energy's Pathways to Commercial Liftoff frames aggregated behind-the-meter fleets as dispatchable capacity on the order of 80–160 GW. None of that aggregation works unless each inverter carries the metering, communications and control-response payload to be individually addressable inside a fleet of thousands, a payload no 2015 string inverter carried.
What we left outThree figures we checked, couldn't source, and didn't use.
Evidence-first only means something when the discipline is visible where it costs something. Three widely-circulated figures would have made this piece cleaner. None survived a primary-source check, so none appear above.
"Silicon carbide reached 35% of utility-scale string-inverter shipments in 2025." No primary semiconductor or PV market analyst we could find publishes a silicon-versus-silicon-carbide split segmented to string inverters, and the paired "18% in 2023" appears to be an electric-vehicle statistic. Cut.
"Sungrow held 27% and Huawei 22% of the utility-scale segment in 2025," attributed to S&P Global. S&P Global's 2025 inverter output was a Tier-1 bankability list, not a market-share split. The 27/22 pairing traces only to secondary aggregators. Cut.
"AI-orchestrated virtual power plants cut renewable curtailment by 65–75%." This traces to a single low-credibility journal. Credible reviews report gains that are real but more modest and differently defined: higher renewable utilisation, lower operating cost, with no 65–75% curtailment figure behind them. Cut.
The through-line survives all three removals: the inverter did accumulate responsibility across the decade (converter, battery manager, grid-service provider, VPP node) and the market did concentrate as the engineering bar rose. Those claims stand because the primary data stands behind them.
Where it pointsFrom converter to control layer.
Read end to end, the documented threads describe one accumulation: a device that in 2015 converted power and stepped out of the way now sits where tariff signals, battery state, weather forecasts and grid-dispatch instructions all meet. Grid codes in a growing number of jurisdictions are moving toward mandating grid-forming capability, which keeps pushing firmware and control sophistication, not raw conversion efficiency, as the differentiator between products. That is the decade's real signal, and it holds without the numbers we cut.
Where a figure could not be traced to a primary source, it was omitted and noted above.
Sources13 references
- Wood Mackenzie: Global PV inverter shipments grew 10% in 2024 to 589 GWac (10 July 2025)
- PV Tech: Huawei retains top spot in global inverter rankings, latest IHS data (2016)
- MDPI Electronics 14(6):1061: Effects of wide-bandgap devices on inverter performance and efficiency for residential PV applications (2025)
- Wolfspeed: CRD-60DD12N SiC reference-design documentation
- Infineon: CoolGaN powering Enphase IQ9 microinverters (November 2025)
- Yole Group: Power GaN 2025
- NREL/TP-5200-62471: Photovoltaic shading testbed for module-level power electronics (2016)
- Enphase: IQ8M/IQ8A microinverter datasheet
- EnergySage: Comparing efficiencies: AC vs DC-coupled batteries
- SolarEdge: DC vs AC-coupled batteries
- NREL research on grid-forming inverter controls
- SEPA / NC Clean Energy Technology Center: 50 States of Virtual Power Plants (2024)
- US DOE: Pathways to Commercial Liftoff: Virtual Power Plants
Read our ownership, funding and editorial disclosures.
Stay across the benchmark.
New scorecards, score changes and market signals, published as they happen.