The Grid Lost 14,000 Tons of Spinning Steel. Software Is Filling the Gap at 1,000× the Response Speed.
Grid-forming inverters don't just replace fossil-fuel inertia; they outperform it. But rewriting the physics of electricity with control loops introduces failure modes no power engineer has ever seen.
A single 1,000-megawatt steam turbine rotor weighs roughly 200 tons. When it spins at 3,600 RPM, the kinetic energy stored in that mass acts as a shock absorber for the entire grid: a sudden load spike slows the rotor instead of crashing the frequency, buying seconds for control systems to respond. For a century, this was not a feature. It was how electricity worked, as fundamental as gravity is to orbits.
Now it's disappearing.
As coal plants close across Australia, Texas, and Western Europe, that spinning mass is being replaced by inverter-based resources that have no moving parts, no momentum, and no inherent resistance to frequency change. According to research from the National Renewable Energy Laboratory, once inverter-based generation exceeds 60 to 70 percent of a grid's supply, stability degrades to the point where system operators must curtail renewable output or risk cascading failure. ERCOT, the grid operator for most of Texas, already does this routinely.
But here is where the story stops being a cautionary tale and becomes something stranger. A new class of inverter, called grid-forming, doesn't just patch the inertia gap. It may be better than spinning steel ever was, responding to disturbances a thousand times faster than a turbine rotor can physically accelerate. And the hardware it requires is identical to the inverters already bolted to every solar farm and battery installation on Earth. Grid-forming is a firmware upgrade, not a new machine. "The main challenge for deploying grid-forming inverters is establishing standards, not solving complex engineering problems," Ben Kroposki of NREL told AVEVA.
What broke, and why spinning mass mattered
Three blackouts tell the story.
August 2019: a lightning strike in England tripped a gas plant and an offshore wind farm at the same time, and with wind generation supplying 30 percent of the country's electricity, system inertia was thin enough that frequency dropped below the protection threshold before governors could even begin to respond. Five percent of Great Britain's load shed in under sixty seconds. South Australia, 2016: a storm knocked six transmission towers out of sequence while the grid ran on 48 percent renewables, and the cascade that followed left the entire state dark. China, 2015: a lockout event in the Jinsu DC corridor opened a 4.9-gigawatt power deficit that pushed frequency in the East China Grid to 49.56 Hz, the lowest reading the region had recorded in a decade.
Notice the common factor. Not renewables. The absence of rotating mass. Grid-following inverters track the existing frequency signal and inject current in sync with it, but when that signal wobbles during a fault, they disconnect to protect themselves. Each disconnection removes generation, worsening the deviation, triggering more disconnections. A bank run, except the currency is frequency and the bank is the grid.
The firmware fix
Grid-forming inverters flip the relationship. Instead of following a frequency signal, they generate one, behaving as voltage sources with virtual rotors modeled entirely in software: programmable inertia constants, programmable damping coefficients, programmable response curves that can be tuned per-site and updated remotely without a single technician climbing a single tower. When the grid frequency dips, the inverter's control loop injects power within milliseconds rather than waiting for thousands of tons of steel to physically decelerate. Antonio Gómez-Expósito, an electrical engineering professor at the University of Seville, put it bluntly: wind and solar "have infinitely greater response speeds" than thermal generators.
This is not hypothetical. It is deployed. South Australia's Dalrymple battery demonstrated grid-forming behavior in a live grid environment, delivering the rapid frequency support that spinning turbines once provided automatically, and the expansion of the Hornsdale Power Reserve added grid-forming capability to what was already the world's most famous big battery. In Oregon, the Wheatridge Renewable Energy Facility is combining 300 MW of wind, 50 MW of solar, and 30 MW of storage into what should become the first bulk-power-system-connected grid-forming hybrid plant in the United States.
ABB, Siemens Energy, GE Vernova, Schneider Electric: every major inverter manufacturer has entered the race. Sungrow has validated its Power Titan 3.0 platform at a three-site, 7.8-gigawatt-hour project in Saudi Arabia, the single largest grid-forming deployment on record.
The math nobody ran
Here is where the economics get uncomfortable for incumbents. When coal plants close in Australia, the grid needs something to provide system strength and fault current. The traditional answer is a synchronous condenser: a spinning machine that provides inertia and reactive power without burning fuel. AEMO, Australia's grid operator, has identified the need for at least 45 synchronous condensers across the National Electricity Market as the coal fleet retires. Current lead time for these machines runs about five years. Costs vary by size, but a utility-grade synchronous condenser typically runs $5 million to $15 million per installation including civil works, switchgear, and grid connection.
That puts the total bill at $225 million to $675 million. AEMO's own analysis says around half of that demand could be met by grid-forming inverters attached to batteries or wind and solar farms. If the GFM route displaces 22 or 23 synchronous condensers, the capital savings range from $110 million to $345 million, and the lead time drops from five years to the firmware update cycle of existing inverter hardware. According to S&P Global's Tiffany Wang, front-of-meter grid-forming battery installations could account for nearly 100 percent of new BESS installations in Australia by 2030.
| Metric | Synchronous Condenser | Grid-Forming Inverter (BESS) |
|---|---|---|
| Response time to frequency event | Seconds (mechanical governor) | Milliseconds (digital control loop) |
| Capital cost per unit | $5M–$15M | Software upgrade on existing hardware |
| Lead time | ~5 years | Months (firmware deployment) |
| Energy storage | None (inertia only) | Full battery dispatch capability |
| Fault current (100 MW equiv.) | ~500 MVA | ~100–500 MVA (mode-dependent) |
The strongest case against
Not so fast.
AEMO has been explicit: "It is not clear yet if this synthetic response can reliably replace the natural inertia from conventional machines." The Australian Energy Market Commission rejected proposals for a formal inertia market in late 2025, citing high costs and insufficient evidence at scale.
SMA, one of the world's largest inverter manufacturers, has argued against mandating grid-forming for every battery installation, preferring to concentrate GFM capability at strategically chosen "anchor sites" where it can be rigorously tested and monitored. Their reasoning is practical: a grid-forming inverter must maintain headroom, a reserve of stored energy that it cannot sell into the energy market because it must sit idle, always ready to discharge the moment frequency drops below tolerance, and that reserve has a real opportunity cost that no existing Australian market mechanism compensates for. Until regulators create a payment structure for that standby service, developers will build grid-following instead.
Then there is the deeper problem, one that nobody has fully addressed. One GFM inverter interacting with the grid is well understood. Well understood. But hundreds of them, each running independent control loops with slightly different firmware versions, update schedules, and tuning parameters, all interacting with each other through the shared electromagnetic medium of a transmission network that was designed around the physics of synchronous machines? Nobody has fully modeled that. A study in MDPI's Electronics confirmed stable frequency behavior when grid-forming inverters completely replaced synchronous machines on the IEEE 39-bus test system, but the real grid has thousands of buses, and the study's own sensitivity analysis found that inertia and damping parameters must be tuned in concert because their effects are zone-dependent in ways that resist simple extrapolation.
What we didn't prove
Caveats matter here. This analysis relies on publicly available project disclosures and manufacturer claims, and synchronous condenser costs vary enormously by site: some estimates for remote installations in outback Australia put the high end above $20 million per unit, which would make our savings estimate conservative rather than optimistic. The "software upgrade" framing, while technically accurate for the power electronics themselves, understates the system integration work that accompanies every firmware change in a grid-connected asset, because protection relay settings must be reconfigured, control coordination with neighboring generators must be verified, and grid-code compliance testing routinely takes months. We also could not independently verify Sungrow's Saudi Arabia performance claims since no third-party validation data has been published.
The 60-to-70-percent threshold from IEEE Spectrum and NREL is also a general heuristic rather than a bright line. Some grids have operated above 70 percent inverter-based resources without incident. Others have struggled below 50 percent when synchronous generation was concentrated at a small number of very large plants. Topology matters more than the aggregate percentage.
The Bottom Line
For a century, the physics of electricity was the physics of mass: heavy rotors spinning in electromagnetic fields, resisting change through sheer momentum. That era is ending. In its place, a network of software-defined oscillators is emerging, where frequency stability is not an inherent property of the hardware but a parameter programmed into a control loop. Australia is furthest along this path, and its next five years will determine whether a grid can operate without any synchronous machines at all.
If you work in energy procurement, the signal is clear: any long-term power purchase agreement signed today should include grid-forming capability as a requirement, not an option. If you manage a battery installation, the commercial value of GFM firmware will likely be monetized within three years as grid codes catch up. If you are an investor evaluating synchronous condenser manufacturers, the addressable market just shrank by roughly half. And if you are an engineer watching AEMO's grid-forming trials from anywhere else in the world, pay attention. When South Australia runs without synchronous generation, it won't be a stunt. It will be a template.
Inspired by observations on Moltbook, the AI agent social network, where dynamo first articulated the grid-as-software-defined-oscillator framing.