Grid-Forming Inverters Become the Default Requirement

For two decades, most inverters were grid-following: they measured grid voltage and phase angle and injected current in phase. That architecture fails when inverter-based resources dominate a network and there is no synchronous machine left to define voltage and frequency. Grid-forming (GFM) inverters instead synthesise an internal voltage reference using droop control or a virtual synchronous machine model, providing inertia and short-circuit strength. According to NREL, GFM control is now considered essential for high-renewable grids, and IEEE 2800-2022 defines the interconnection performance requirements that procurement teams increasingly write into tenders.
Regulatory pressure accelerated through 2024 and 2025. Australia's AEMO has introduced grid-forming requirements and system-strength obligations for new large-scale storage in weak parts of the network. Britain's National Energy System Operator has run grid-forming procurement and stability studies as synchronous generation retires. In the United States, ERCOT's ride-through and voltage-support expectations, plus FERC-driven interconnection reforms, push storage assets toward GFM capability even where it is not yet mandatory. The practical result: GFM is shifting from a premium option to a baseline specification for utility-scale projects above roughly 20 MW.
Technically, GFM demands more from the power stage than nameplate megawatts suggest. Inverters must deliver 1.5 to 2 times rated current for short periods, limit fault current through virtual impedance, ride through very weak grid conditions (short-circuit ratios below 2), and support black start. That requires oversized silicon, fast current control loops running at tens of kilohertz, and careful co-design with the battery management system. Buyers should ask for documented weak-grid test results and third-party certification rather than datasheet claims.

Silicon Carbide and the Wide-Bandgap Power Stage

The most visible hardware change in 2025 is the migration from silicon IGBTs to silicon carbide (SiC) MOSFETs in the DC-AC stage. SiC devices switch roughly two to three times faster with markedly lower switching and conduction losses, which lifts inverter peak efficiency into the 98.5–99.0 percent range and reduces the size of magnetics, heatsinks and cabinets. Major device suppliers including Infineon, Wolfspeed, onsemi and STMicroelectronics have scaled 1200 V and 1700 V SiC portfolios, and the cost premium over silicon has narrowed enough that payback now comes from efficiency and enclosure density, not just prestige.
The system-level consequences matter more than the semiconductor itself. Higher switching frequencies shrink inductors and filters, so a 2025-era 200 kW string power conversion system (PCS) can occupy a smaller footprint than a 2021 unit of half the rating. Lower losses mean less heat to remove, which enables sealed, liquid-cooled enclosures rated for desert and coastal environments. That improves round-trip efficiency for the whole storage system: a DC-coupled LFP asset using a SiC PCS typically achieves 88–92 percent AC-to-AC round-trip efficiency, with the inverter accounting for a meaningful share of the residual loss. Our own [inverter engineering overview](/tech/inverters) covers how these conversion stages are matched to battery voltage windows.
Gallium nitride (GaN) is following a similar trajectory in smaller products. Residential hybrid inverters in the 5–15 kW class increasingly use GaN or hybrid SiC/GaN designs to reach 98 percent-plus European weighted efficiency in a wall-mounted chassis. IEEE and IEC working groups are still harmonising test methods for wide-bandgap topologies, so specifiers should insist on CEC, European or Chinese weighted efficiency figures rather than a single peak number measured at one operating point.

1500 V DC, 2000 V Pilots and Modular String PCS

On the DC side, 1500 V has completed its transition from novelty to default for utility-scale storage. Higher bus voltage reduces current for a given power, cutting cable cross-section, DC losses and balance-of-system cost. Modules and racks are certified accordingly, with 1500 V DC insulation coordination, creepage and clearance requirements governed by IEC 62477 and UL 1741 SB. A parallel architecture shift is underway: modular string PCS units of 100–250 kW are displacing 2–4 MW central inverters in many new builds, because string-level maximum power point tracking and fault isolation improve availability and simplify augmentation.
Battery cell evolution is pulling inverter design along with it. The 280 Ah cell gave way to 314 Ah, and 2025 shipments increasingly use 587 Ah and 628 Ah cells in 20-foot containers delivering 5 MWh to 6.25 MWh. Larger cells and higher energy density mean the PCS must handle wider DC voltage windows, higher continuous currents per rack, and more aggressive thermal management. Containerised systems from CATL, Sungrow, Fluence and Tesla now integrate PCS, fire suppression, HVAC and controls, so the inverter is a subsystem rather than a standalone box.
A 2000 V DC class is emerging as the next step. NREL and DOE-funded programmes have studied 2000 V architectures as a route to further BOS savings, and early utility demonstrations are running. The barriers are practical: fewer certified components, new arc-flash and clearance rules, and limited field experience. For 2025 procurements, 1500 V remains the bankable choice, with 2000 V worth monitoring for 2027-plus pipelines. Our [battery storage technology pages](/tech/battery-storage) track these architecture transitions in more detail.

Multi-Port Hybrid Inverters: One Chassis, Several Jobs

Hybrid and multi-port inverters are the fastest-moving segment in commercial and residential storage. Rather than separate PV inverter, battery inverter and transfer switch, a single chassis provides multiple MPPT inputs, a high-voltage battery port, a generator or grid input, and increasingly an AC-coupled EV charger port. Residential units now span 5–15 kW with two to four MPPTs, 97.5–98.5 percent peak efficiency, 200 percent DC oversizing ratios, and backup transfer in under 20 ms. Commercial three-phase models reach 50–125 kW with 1500 V battery inputs.
The DC-coupling versus AC-coupling debate has effectively resolved into a portfolio question. DC-coupled hybrid inverters win on round-trip efficiency and cost when PV and storage share a site and the battery is charged from solar. AC-coupled PCS units win on retrofit projects, independent sizing and simpler expansion. Multi-port designs blur the line by offering both paths in one device, at the cost of more complex firmware and a single point of failure. Battery compatibility is now the binding constraint: hybrid inverters must support the specific high-voltage stack in use, which is why pairing decisions should be made at design stage, not procurement stage.
For integrators, this convergence changes the bill of materials. Fewer enclosures, fewer transfer switches and one monitoring platform reduce installed cost and commissioning time, but they also concentrate risk. Buyers should verify independent certification of each port, confirm the battery list approved by the manufacturer, and check whether the unit supports the [lithium battery chemistries](/products/lithium-battery) — LFP, sodium-ion or high-nickel — expected across the asset's life.

Software-Defined Inverters, AI and Virtual Power Plants

Modern storage inverters are firmware platforms. Over-the-air updates now deliver new grid-support functions, ride-through profiles and market algorithms without truck rolls. Communication has standardised around SunSpec Modbus, IEEE 2030.5 and the Common Smart Inverter Profile, which lets aggregators dispatch thousands of units as a virtual power plant. FERC Order 2222 in the United States opened wholesale markets to distributed storage aggregations, and European and Australian market reforms are following similar logic. For asset owners, the practical question is no longer whether an inverter can be networked, but how securely and how granularly.
Artificial intelligence has moved from marketing to operations. Machine-learning models trained on DC current, temperature and voltage signatures now detect cell-level anomalies and string faults before they cause outages, and forecast-based dispatch optimises charge and discharge against day-ahead and balancing market prices. Vendors report double-digit reductions in truck rolls and unplanned downtime, though independent verification remains thin. Cybersecurity is the counterweight: inverters are now critical infrastructure, and IEC 62443 alignment, signed firmware, secure boot and role-based access control should be explicit tender requirements. Lessons from real deployments are collected in our [project portfolio](/projects).
Grid-forming and software-defined operation also change testing. A unit that passes factory acceptance may behave differently after three firmware revisions. Owners should require regression test evidence, keep a firmware register, and validate grid-support behaviour at commissioning — not merely at delivery.

Total Cost of Ownership: Where the Numbers Land

Market context matters for specification. BloombergNEF's energy storage outlook has tracked rapid growth, with annual global installations rising from roughly 50 GWh in 2023 toward the 100 GWh-plus range in 2025, and BNEF's battery price survey showed pack prices falling below USD 120/kWh in 2024 before supply-chain volatility re-emerged. Turnkey four-hour utility BESS pricing has compressed accordingly, with PCS typically representing 10–20 percent of the installed cost. IEA analysis in its renewables and battery reports consistently identifies storage as the fastest-growing flexibility resource, particularly in China, the United States and Europe.
Because the PCS often carries a 10–15 year design life against a 20-year asset life, replacement strategy belongs in the financial model. Efficiency matters more than it appears: a one-percentage-point gain in round-trip efficiency on a 100 MWh, 200-cycle-per-year asset approaches 2 GWh of avoided losses annually. Auxiliary loads — HVAC, pumps, controls — can consume 2–5 percent of throughput in hot climates, so liquid-cooled PCS with high ingress protection and low parasitic draw deserves weighting.
Finally, availability and serviceability separate good hardware from good assets. Modular string PCS allows single-unit replacement without a full-site outage; central inverters typically require more extensive isolation. Warranty terms, spare-part logistics, remote diagnostics and firmware support commitments should be scored alongside price. A storage inverter is now a 20-year software and service relationship wrapped around a power electronics chassis.

Specifying for 2025 and Beyond

A practical 2025 specification for utility-scale storage should require grid-forming capability certified to IEEE 2800-2022 or the applicable national grid code, SiC-based power stages with published weighted efficiency, 1500 V DC inputs with demonstrated 2000 V roadmap, modular string PCS architecture, IEC 62443-aligned cybersecurity, and an open Modbus or IEEE 2030.5 interface. For commercial and residential projects, prioritise multi-port hybrid designs with broad battery compatibility, sub-20 ms backup transfer, and manufacturer-published efficiency curves rather than peak figures.
The direction of travel is clear: inverters are becoming the intelligence layer of the power system, not just its conversion layer. Procurement teams that evaluate firmware governance, grid-support testing and lifecycle support as seriously as kilowatts and price will capture most of the value. Reviewing our [product range](/products) and technical documentation is a useful starting point for building a 2025-compliant specification.