Grids Without Inertia

Established The prize is a power system whose second-to-second stability survives after the last large spinning machine is switched off, so that decarbonisation does not have to stop at the point where synchronous inertia runs out. Today a bulk grid holds its frequency because gigawatt-scale turbines store rotational energy that resists change for the first few hundred milliseconds after a fault; a grid-forming (GFM) inverter reproduces that voltage-source behaviour from a battery. Hornsdale Power Reserve, a 150 MW battery in South Australia, was cleared in July 2022 as the first big battery in the world to sell grid-scale inertia, and it can present roughly 2,000 MW.s of synthetic inertia to the network — about 15 per cent of the state’s predicted inertia shortfall from one site.

Frontier The magnitude is the whole synchronous fleet: a fully inverter-based grid removes the requirement to keep coal, gas or nuclear turbines spinning purely for stability, which is the single largest reason a decarbonised system still burns fuel at 3 a.m. A grid can already run at more than 70 per cent instantaneous renewables in ERCOT and 90 per cent on Kauai; the last tranche — from roughly 90 per cent to 100 per cent, with no synchronous machine online for hours — is the part nobody has demonstrated at metropolitan scale, and it is the part where the entire case for retiring the last thermal plant lives.

Speculative If, and only if, a metropolitan grid can be black-started and run for hours on inverters alone — which no operator has shown, and the required-experiments material says exactly what would settle it — the gains accrue broadly: every jurisdiction with cheap solar and wind but no domestic gas keeps the lights up without importing stability. Until that demonstration exists the upside is bounded by a quantity nobody reports: the minimum share of grid-forming capacity a real network needs, and the fault current inverters can actually deliver. The bottleneck material names both, and the honest reading is that the technology is proven in megawatt-scale islands and unproven in cities.

Established Inertia is the kinetic energy stored in the synchronised rotating mass of generators and motors, and it is what buys an operator the fraction of a second between a fault and a control response. When a large generator trips, the remaining machines slow down; the rate at which frequency falls — the rate of change of frequency, or RoCoF, measured in hertz per second — is set by how much rotating inertia is left. More inertia means a gentler RoCoF and more time for reserves to arrive. Established A typical large interconnection holds several seconds of inertia constant, and grid codes have long set a maximum tolerable RoCoF — historically around 0.5 Hz/s, relaxed toward 1 Hz/s as fleets changed — because protection and generators trip above it. This is Newtonian mechanics applied to a shaft, and it has run every large grid for a century.

Established A solar panel, a wind turbine’s converter and a battery connect to the grid through a power-electronic inverter, which has no rotating mass and therefore no inherent inertia. An inverter is a switching device that synthesises an AC waveform; left to itself it contributes nothing to RoCoF. As inverter-based resources (IBRs) displace synchronous generators, system inertia falls, RoCoF for a given loss of generation rises, and the century-old assumption — that there is always spinning metal to lean on — stops holding. Established The decline is already operational, not hypothetical: system operators in Ireland, Britain and South Australia now schedule minimum-inertia and system-strength constraints that routinely bind, sometimes forcing renewable curtailment or the dispatch of thermal units purely to keep enough inertia online.

Established Inverters divide into two control families, and the distinction is the axis this entire brief turns on. A grid-following (GFL) inverter uses a phase-locked loop (PLL) to measure the grid’s voltage and inject current in step with it; it is a current source that needs an existing, stiff voltage to follow. A grid-forming (GFM) inverter instead holds an internal voltage phasor that stays constant or nearly constant in the sub-transient to transient time frame; it behaves as a voltage source behind an internal impedance — a Thevenin source — and can establish voltage and frequency itself. Established A grid-following inverter is fast and efficient but fundamentally a follower; a grid-forming inverter is the reference other devices can follow, which is why the two are not interchangeable at high penetration.

Established The consequence is categorical, not incremental: a grid cannot be operated at 100 per cent GFL penetration, because current sources need a voltage to follow and if every source is following there is nothing to follow. A grid can, by contrast, be operated at 100 per cent IBR provided some of those inverters are grid-forming. GFM controls can operate stably at very low grid strength — short-circuit ratio below one — where GFL controls lose synchronism. That single asymmetry is why “grids without inertia” is really shorthand for “grids whose voltage source is silicon, not a shaft.”

Frontier Synthetic inertia (also called virtual inertia) is the deliberate programming of an inverter to inject or absorb active power in proportion to RoCoF, mimicking the swing of a real machine. In a grid-forming controller the emulated inertia constant H and a damping term have distinct roles: inertia shapes the RoCoF and the depth of the frequency excursion, while damping governs how fast the response settles. This is genuinely inertia-like within the inverter’s current and energy limits — and those limits, not the control law, are where the argument gets hard.

Frontier A vocabulary distinction the field itself only recently standardised is worth stating plainly: “grid-forming” is a control mode, not a device type. The same physical battery inverter can be shipped as grid-following and reconfigured as grid-forming in firmware; conversely a poorly tuned grid-forming controller can behave worse than a good grid-following one. The label on the datasheet is a claim about behaviour that only field measurement settles, which is one reason the standards material below spends so much effort defining what the words must mean.

Established On 28 April 2025 continental Spain and Portugal suffered a total blackout, and the ENTSO-E expert panel’s final report, published 20 March 2026, is the first continental-scale forensic dataset on the collapse of a low-inertia, inverter-heavy grid. A factual report had appeared on 3 October 2025; the final report added causation. Its central finding is that the blackout was a voltage collapse driven by fast overvoltage and cascading generation disconnections, not a shortfall of energy or a shortage of generation. This is the anchor for the whole brief, and it holds — with one important qualification the brief takes up below.

Established The panel attributes the event to a combination of interacting factors: oscillations, gaps in voltage and reactive-power control, differing voltage-regulation practices, rapid output reductions and generator disconnections in Spain, and uneven stabilisation capabilities across the system. Two oscillation episodes preceded the collapse: a converter-driven forced oscillation at 0.63 Hz between 12:03 and 12:08, and a classic inter-area oscillation at 0.2 Hz — the East-Centre-West mode — between 12:19 and 12:22. The first of these is a power-electronics phenomenon, not a rotating-machine one. Frontier That a converter-driven oscillation appeared at all is the single most technically significant detail for this brief: it is the failure mode that scales with inverter share and that a synchronous fleet would have damped.

Established The terminal cascade ran in roughly 78 seconds. At 12:32:57 a 400/220 kV transformer near Granada tripped on overvoltage protection, severing 355 MW; at 12:33:16 two substations near Badajoz lost 727 MW of photovoltaic and thermo-solar capacity; over 12:33:17–18 a further 928 MW of wind and solar disconnected; a further 500 MW of large renewables reduced output and 208 MW of distributed units ramped down or tripped. More than 2.5 GW of generation vanished. At 12:33:19 Spain and Portugal began losing synchronism with the rest of Europe; at 12:33:21 the AC interconnections to France tripped on out-of-step protection; at 12:33:24 the HVDC link to France was the last connection to sever.

Established Voltage, not frequency, was the driver: voltages exceeded 435 kV as generation disconnected and reactive-power balance was lost. The panel found that key voltage-control equipment was switched manually, slowing response; that operators lacked real-time visibility of reactive-power margins; and that conventional generators failed to meet their reactive-power set points, delivering less than 75 per cent of required output at critical moments. Many renewable plants operated under fixed power-factor schemes that limited their ability to respond to voltage swings, and a significant number disconnected automatically before their connection-point thresholds were reached, some with overvoltage protection set below regulatory limits. Established The investigation was hampered by incomplete data: distribution operators lacked production data from generators below 1 MW, and the panel could not determine the cause of some disconnections at all — a data-visibility failure that is itself one of the report’s findings.

Frontier The most important honesty check in this brief: the commission calls the Iberian event a “low-inertia collapse,” and that framing is only partly right. The panel and its expert commentators are explicit that this was a voltage and reactive-power failure that would have happened regardless of the generation mix. José Luis Domínguez, reviewing the report, stated plainly that a system with greater inertia would have faced the same problem, because the oscillations and overvoltage evolved too fast and in the wrong variable for inertia to help. ENTSO-E’s chair Damián Cortinas summarised it as “the problem is not renewable energy, but voltage control, regardless of the type of generation.” So the Iberian blackout is the best forensic record we have of how an inverter-rich grid fails — and its lesson is that the first thing to break was voltage support, not RoCoF. A brief about inertia that mis-sold this event as an inertia failure would be committing exactly the error the flag system exists to prevent.

Established The reliability record of grid-following inverters under fault, by contrast, is unambiguous and measured. On 4 June 2022 a phase-to-ground fault on the 345 kV system near Odessa, Texas — caused by a failed lightning arrestor at a synchronous plant, cleared normally in three cycles — triggered the loss of 2,555 MW: 844 MW of synchronous generation tripped and 1,711 MW of solar PV reduced output, about 15.8 per cent of ERCOT’s solar at the time. The solar loss decomposed into inverter AC overcurrent (459 MW), passive anti-islanding on phase jump (385 MW), AC overvoltage (295 MW), DC-bus voltage imbalance (211 MW), incorrect ride-through configuration (135 MW) and momentary cessation (130 MW). Most of the same facilities had tripped in the 2021 Odessa event, and only one had implemented mitigations in the interval. Frontier The recurrence is the damning part: this is not a freak event but a repeatable response of grid-following fleets to a normally cleared fault, and NERC has tracked a growing list of comparable disturbances.

Frontier Against that failure record sits a working existence proof: grid-forming batteries in commercial service. Hornsdale Power Reserve runs Tesla’s Virtual Machine Mode; ARENA’s testing found that with final settings the expanded site would provide 2,070 MW.s of inertia at an equivalent H of 11.02 MW.s/MVA, and that for a 0.5 Hz/s RoCoF event about 41.4 MW of inertial response is expected, with peak active-power output reached before the frequency nadir — the signature of a genuine inertial contribution rather than a delayed reserve. A 30 MW ABB grid-forming battery has run at Dalrymple in South Australia since 2018, including in islanded operation. As of the ESIG survey the operating Australian grid-forming fleet totalled 230 MW/277 MWh. Frontier ARENA has since committed to eight further projects of 200–300 MW each, so the operational evidence base is about to grow by an order of magnitude in a jurisdiction that already runs at record instantaneous renewable share.

Speculative What has not been shown is a metropolitan grid running for hours on inverters with zero synchronous generation, black-started from inverters, and instrumented for RoCoF and fault-current contribution. Simulation reaches that far: an NREL electromagnetic-transient model of Maui at 97 per cent IBR and an inertia constant of H = 0.48 s shows a single grid-forming device materially improving both frequency deviation and voltage stability over its grid-following counterpart, and an IEEE 39-bus study replaced all ten synchronous generators with grid-forming controls and stayed inside stability limits. But a validated simulation is not a demonstrated city, and the gap between the two is the brief’s central open question. Speculative The simulations also tend to assume ideal, coordinated controls and complete models; the real fleets that actually exist are heterogeneous, partly black-boxed and independently tuned, which is exactly the condition under which the Iberian oscillation appeared.

Frontier What minimum share of grid-forming capacity does a real network need, and does that number depend on where the GFM units sit? NERC states that a grid can run at 100 per cent IBR provided some inverters are grid-forming, but neither NERC nor the DOE specification names a threshold; NERC explicitly lists “levels of GFM IBR capacity needed to obtain the desired benefits” as an open research question. Simulations suggest a modest GFM fraction can stabilise a largely GFL system, but the figure is network-specific and unmeasured on any large grid. Speculative Estimates in the literature range from a few per cent to a third of capacity depending on the network and the disturbance, which is a wide enough band that no operator can yet write it into a connection standard with confidence.

Frontier How much fault current can an inverter-based grid actually deliver, and will century-old protection schemes still work on it? Protective relays were designed to detect faults by the several-per-unit surge of current a synchronous machine dumps into a short circuit. A grid-forming inverter is explicitly not expected to inject that much: the DOE specification states a GFM “is not expected to inject a high magnitude of fault current (such as 6 per unit fault current obtained from rotating machines).” If inverters cap at roughly 1.1–1.5 p.u., a fault may not look like a fault to the relay that is supposed to clear it. This is arguably the deepest unsolved problem in the field, and it is a protection problem, not a control problem. Frontier Synchronous condensers or a residual population of synchronous machines can supply the missing fault current, which is why the honest near-term answer is a hybrid rather than a pure inverter grid.

Frontier Can grid-forming inverters remain stable when many of them interact, at scale, with no synchronous machine to set the reference? The Iberian 0.63 Hz converter-driven oscillation is a warning: power-electronic controls can excite forced oscillations that spinning machines would damp. ENTSO-E’s interim requirements specify a minimum 5 per cent damping factor for electromechanical oscillations precisely because inverter fleets can under-damp. Whether a heterogeneous fleet of GFM units from different vendors, each with proprietary firmware, remains collectively stable is not established at grid scale.

Frontier Is synthetic inertia from a battery equivalent to real inertia, or only similar? Real inertia is instantaneous, involuntary and free; synthetic inertia is a control response bounded by the inverter’s energy reservoir and by measurement and actuation delay. ENTSO-E’s parameterisation is concrete — a 2 Hz/s df/dt with a ramp-up time of 25 leads to a 1 per unit active-power variation — but the response is programmable rather than physical, which means it can be tuned, mis-tuned, or exhausted. The equivalence is a frontier claim, not an established one. Frontier Hornsdale’s measured response — peak active power before the frequency nadir — is the strongest field evidence that the emulation can be genuinely inertial rather than merely a fast reserve, but one site is not a proof for a system.

Speculative Does declining inertia matter on its own, or only through the phenomena it enables? The Iberian event suggests inertia decline is a symptom-multiplier rather than a first cause: it shortens the time available to correct a voltage or oscillation problem without itself being the problem. If that generalises, the right policy target is not a minimum inertia floor but fast voltage support and well-behaved ride-through — a different and cheaper intervention than keeping turbines spinning.

Established Fault current is the binding physical bottleneck. An inverter’s current is limited by the thermal rating of its semiconductor switches, which have almost no overload capacity; a synchronous machine can deliver five to seven times rated current into a fault for a few cycles, an inverter perhaps ten to fifty per cent above rating. Existing overcurrent and distance protection assumes the former. This is not a control choice that better firmware fixes; it is a property of silicon, and it forces a rethink of protection wherever synchronous machines retire. Frontier Grid-forming controls can be tuned to inject somewhat more current during a sag to aid recovery, but “somewhat more” is a fraction of a per unit, not the multiple a distance relay expects, so the relay logic itself has to change.

Frontier Protection and EMT modelling are the engineering bottleneck that follows from the physics. Because inverter fault behaviour is set by firmware rather than by a physical machine curve, a system operator cannot study stability without accurate electromagnetic-transient (EMT) models of every inverter as configured. NERC’s Odessa recommendations centre on establishing EMT-model requirements for all IBRs. The problem is that these models are vendor intellectual property, often encrypted or black-boxed, so the operator studying its own grid may not be able to see how the fleet will behave — a governance bottleneck wearing an engineering costume.

Frontier Grid-forming controls are still early in maturity, and are most available in batteries. NERC describes GFM as “still in the early stages of technology maturity and adoption” and most commonly available in battery energy storage. Solar and wind inverters can in principle be grid-forming, but a GFM source must be able to inject or absorb power on demand, which a maximum-power-point solar array at full output cannot do without curtailment or storage. So the cheapest place to put grid-forming capability — a battery — is not where most inverter capacity actually is.

Frontier Energy headroom, not power rating, limits synthetic inertia. A battery presenting synthetic inertia must have charge to give and room to take; a unit sitting at full charge cannot absorb, and one near empty cannot inject. Hornsdale’s 2,070 MW.s figure is a capability at a state of charge, not a guarantee at every hour. The bottleneck is that the same battery is usually also being asked to arbitrage energy and provide reserves, and those uses compete for the reservoir that inertia provision requires. Frontier A grid-forming solar or wind plant faces a sharper version of the same limit: to inject on demand it must run below its available power, so grid-forming capability from renewables costs curtailed energy unless it is paired with storage.

Speculative Vendor interoperability at fleet scale is an unquantified risk. Australia’s grid-forming build-out mixes Tesla, ABB and others; a continental grid would mix dozens of firmware stacks. Whether controls tuned independently remain collectively stable under a large disturbance has not been demonstrated beyond simulation, and the Iberian converter-driven oscillation shows the failure mode is real. There is no equivalent of a synchronous machine’s universal, physics-guaranteed behaviour to fall back on.

Established Voltage and reactive-power support is a bottleneck the Iberian event made concrete. Fixed power-factor operation, manual voltage equipment and missing real-time reactive-power visibility were all named causes. None of these is exotic to fix — active voltage control and better telemetry are mature — but they were not deployed, which makes this a bottleneck of requirements and incentives rather than of technology.

Established This topic depends on cheap, cyclable storage, which is largely a solved and improving input. Grid-forming stability services ride on batteries, so the cost and cycle life of storage set the cost of stability. That dependency is tracked in Energy Storage Revolutions and Advanced Battery Technologies; the point here is that stability provision competes with energy arbitrage for the same cells.

Frontier It depends on EMT-grade, non-black-box models of every inverter on the network. Without them an operator cannot certify that a proposed dispatch is stable. This is a research dependency only in the weak sense that the models exist but are withheld; the missing input is disclosure, not discovery, which is why the brief treats it as institutional as much as technical. Established NERC has already made EMT-model requirements for all IBRs a formal recommendation, so the direction of travel is set even where enforcement is not.

Frontier It depends on a protection paradigm that does not assume high fault current. Travelling-wave protection, differential schemes and communication-assisted relaying can detect faults without a large current surge, but retrofitting them across a transmission and distribution estate is a decade-scale programme, not a firmware update. The dependency is real hardware in thousands of substations.

Established It depends on standards that define what “grid-forming” means precisely enough to procure. Great Britain’s GC0137, the DOE/UNIFI specification, IEEE 2800 and ENTSO-E’s interim requirements are the substrate; without an agreed specification an operator cannot write a contract or a connection requirement. That dependency is being met, unevenly, right now.

Speculative It depends, ultimately, on the retirement schedule of synchronous plant being allowed to proceed. If regulators require thermal or nuclear units to stay online for stability — or pay them to — the whole question of an inverter-only grid becomes moot by fiat. The dependency is therefore partly political: the technology is only tested if the spinning fleet is actually allowed to leave.

Speculative The decisive test is a black start and multi-hour islanded run of a metropolitan grid carrying zero synchronous generation, instrumented for RoCoF and for fault-current contribution at every inverter terminal. This is the single result that would most change the brief’s assessment, because it converts the central claim — that silicon can be the voltage source for a city — from simulation into measurement. Nobody has run it at metropolitan scale, and no operator has scheduled or funded one; the components exist but the commitment does not.

Frontier The building blocks of that test are already demonstrated separately. Grid-forming batteries have performed frequency control during islanding, inertial response comparable to synchronous machines, grid energisation and load pick-up during restoration, and voltage control during disturbances — but on megawatt-scale islands and substations, not across a city with its full diversity of load and protection.

Frontier A black start from inverters is the sharpest sub-experiment. Black start has traditionally relied on a diesel or hydro unit energising a dead network; a grid-forming inverter can do it, but only if designed for it, and the DOE specification treats black start as an optional capability rather than a universal one. Demonstrating a purely inverter-based black start of a large network, with cold-load pick-up and transformer inrush handled, would retire a major objection.

Frontier A protection experiment must run in parallel: does a real relay fleet clear real faults when the only sources are current-limited inverters? A grid-forming inverter injecting perhaps 1.1–1.5 p.u. into a short circuit may not trip protection tuned for several per unit. This can be tested on an instrumented islanded network by staging faults and recording whether they clear correctly — a measurement, not a model.

Frontier A fleet-interaction experiment would stage a large disturbance on a network carrying grid-forming units from several vendors. The question is whether independently tuned controls remain collectively stable and adequately damped, given the Iberian converter-driven oscillation at 0.63 Hz. ENTSO-E’s 5 per cent minimum damping requirement is the pass/fail criterion this experiment would test in the field rather than on paper.

Established One natural experiment is already running: Australia’s grid-forming build-out. ARENA committed 176 million dollars to eight grid-scale battery projects, each 200–300 MW, totalling about 2.0 GW/4.2 GWh and commissioning between 2024 and 2026, on a grid that already runs at very high instantaneous renewable share. This is the closest thing to the decisive experiment being approached incrementally, and its results are being measured as they come.

Established The Iberian blackout itself was an involuntary experiment whose data now exists. The ENTSO-E final report is the first continental-scale forensic record of an inverter-heavy collapse, and its 22 recommendations are effectively a specification for what the next controlled experiment must get right: active voltage control, real-time reactive-power visibility, and overvoltage ride-through that does not disconnect prematurely.

Speculative What would falsify the optimistic case is equally clear. If an instrumented metropolitan island repeatedly failed to clear faults, or if vendor fleets oscillated under disturbance in ways no tuning fixed, the conclusion would be that some minimum synchronous or rotating presence — synchronous condensers, if not generators — is required, and the “zero-synchronous” goal would give way to a “minimum-synchronous” one.

Established The grid-forming battery is the workhorse, and its engineering is proven at the hundred-megawatt scale. Hornsdale’s 150 MW (100 MW/129 MWh plus a 50 MW/64.5 MWh expansion) running Virtual Machine Mode is the reference design; the engineering task is not inventing the device but qualifying it, since below-one-per-cent enablement of an inverter fleet still reads as grid-following at the site level, so the whole fleet must be converted for the site to behave as grid-forming. Frontier That threshold effect — below roughly full conversion the site still reads as a follower — is an under-appreciated procurement fact: half-measures buy almost nothing, which changes how a stability programme must be phased.

Frontier Synchronous condensers are the pragmatic bridge, and they are engineering, not research. A synchronous condenser is a spinning machine with no prime mover: it provides real inertia and several-per-unit fault current without burning fuel. Deploying condensers alongside grid-forming inverters — inertia and fault current from steel, voltage-source behaviour and fast response from silicon — is the hybrid most real low-inertia grids are actually building, and it complicates any pure “zero rotating mass” claim.

Established Standards now define the interface. IEEE 2800-2022 sets performance for IBRs rated above 20 MVA at 69 kV and above: Category III units must ride through zero-per-unit voltage for 0.15 s and frequency dips to 57.5 Hz for at least 0.3 s, operate to a power factor of plus or minus 0.95, and provide fast frequency response within about 0.5–1 s. It permits but does not mandate grid-forming; the IEEE P2800a amendment, whose project authorisation was approved on 10 December 2025, exists specifically to reduce barriers for grid-forming equipment.

Established Great Britain went further and defined grid-forming in the grid code itself. GC0137, implemented on 14 February 2022, made Britain the first jurisdiction to specify GB Grid Forming (formerly Virtual Synchronous Machine) capability, requiring phase-jump power, inertia power and damping power, with inertia as a programmable value. Its RoCoF withstand differentiates by technology — 1 Hz/s for synchronous plant, 2 Hz/s for power-park modules, 2.5 Hz/s for HVDC — and it feeds the Stability Pathfinder procurement through which the operator buys inertia and short-circuit level as services.

Frontier ENTSO-E’s interim requirements turn grid-forming into numbers a manufacturer can build to. A GFM unit is a voltage source behind an effective impedance with defaults of 0.25 p.u. at low voltage, 0.33 p.u. at medium and 0.48 p.u. at high voltage; it must reach 10–90 per cent of a step in about 10 ms and settle within roughly 60 ms, with synthetic-inertia ramp-up defaulting to 10 s. The implementation guidance was expected to follow the updated regulation in 2025. Established The convergence of GC0137, IEEE 2800 with its P2800a amendment, the DOE specification and ENTSO-E’s requirements means the world now broadly agrees on what a grid-forming inverter must do, even where it does not yet agree on making it compulsory.

Frontier This engineering connects directly to long-distance transmission and to demand-side infrastructure. An inverter-based grid still has to move power, and the converter stability problem recurs on the DC links analysed in Energy Corridors; conversely, the demand for firm, stable power from data centres analysed in The Physical Stack of AI is part of what makes zero-synchronous operation worth demonstrating rather than deferring.

Established Battery storage is the adjacent technology this topic leans on hardest. Every grid-forming stability service today is delivered by a battery, so advances in energy density and cycle life — the province of Advanced Battery Technologies — directly lower the cost of stability, while the systems view in Energy Storage Revolutions covers how storage is dispatched among competing uses.

Frontier HVDC and power-electronic transmission share the converter-stability problem. The same control-interaction and weak-grid questions that dominate grid-forming inverters recur on voltage-source-converter HVDC, which is why Energy Corridors is the nearest sibling: a corridor delivers power an inverter-based grid must then stabilise, and both fail through converter dynamics rather than through wires.

Frontier Advanced nuclear and firm generation are the competing answer to the same question. If firm, dispatchable low-carbon generation from Advanced Fission stays cheap, the pressure to run a zero-synchronous grid falls, because a nuclear unit is itself a synchronous machine. The two topics are substitutes at the system level: more firm rotating generation means less need to prove silicon can stand alone.

Speculative Space-based and planetary-scale generation would inherit the problem whole. Any concept in Space-Based Solar Power or Planetary Scale Energy Systems delivers power through inverters into a terrestrial grid, so the stability of an inverter-dominated network is a precondition for those concepts mattering at all, not a detail downstream of them.

Frontier Digital control and cyber-exposure are the uncomfortable adjacency. A grid whose stability lives in firmware is a grid whose stability is a software supply chain; the dependency mapping in Critical Dependency Atlas and the resilience framing in Infrastructure Resilience both bear on a system where the voltage source can be patched.

Established The central institutional requirement is a grid code that mandates grid-forming capability rather than merely permitting it. IEEE 2800 permits GFM; GC0137 specifies it but is non-mandatory; the DOE and ENTSO-E documents are specifications, not mandates. As long as grid-forming is optional, it is a cost a developer avoids, and the fleet fills with grid-following inverters that cannot, collectively, hold a grid up.

Frontier Operators need a market that pays for inertia and system strength as services. Britain’s Stability Pathfinder and Australia’s emerging arrangements are the templates: stability is unbundled from energy and procured explicitly, which is what let Hornsdale sell 2,000 MW.s of inertia at all. Without such a market the cheapest stability — a grid-forming battery already on the system — earns nothing for providing it, and the service goes unbought. Established Hornsdale earning revenue for inertia is the proof of concept that such a market can clear; the open question is whether every high-renewable jurisdiction builds one before its synchronous fleet retires.

Frontier Regulators need EMT-model disclosure powers. The Iberian and Odessa investigations both ran into the same wall: the behaviour that mattered lived in vendor firmware the operator could not fully see, and data gaps prevented the Iberian panel from determining the cause of some disconnections. A requirement that non-black-box EMT models be lodged with the operator as a condition of connection is an institutional, not a technical, fix.

Established Standards harmonisation across borders is a live requirement, not a background one. The Iberian panel recommended harmonising operating voltage ranges across Europe toward 380–420 kV precisely because differing national practices contributed to the collapse. An interconnected continent cannot run inverter-heavy on incompatible voltage-control rules.

Speculative Someone must own the decision to retire the last synchronous machine. No operator will switch off the final turbine on faith; the institutional gap is that no jurisdiction has a defined process for certifying that a region is safe to run zero-synchronous, and until one does, the last thermal plant stays online as insurance regardless of its economics.

Established The most direct societal stake is that a blackout is a public-safety event, and the Iberian collapse showed how fast one propagates: more than 2.5 GW gone and two countries islanded in about 78 seconds. The ethical weight of getting low-inertia stability right is measured in the hospitals, water pumps and traffic systems that a cascade takes down, which is why honesty about what is and is not demonstrated matters more here than in most engineering fields.

Frontier There is a real risk of premature blame being assigned to renewables. The Iberian experts were explicit that the fault was voltage control “regardless of the type of generation,” and that the report should not be turned into single-cause evidence against renewable energy. A society that mis-reads a voltage-control failure as proof that solar and wind are inherently unsafe may keep fossil plant online for the wrong reason. Frontier The symmetric error is also possible: dismissing the genuine technical difficulty because the political attack on renewables is unfair. Both errors are live, and honest flagging is the only defence against either.

Frontier Concentration of a public-safety function in a few vendors is an equity and security concern. When the voltage source of a national grid is proprietary firmware from a handful of manufacturers — at Odessa, two inverter makers accounted for most of the affected capacity — a design flaw or a withheld model becomes a systemic risk, and the public bears a hazard it cannot inspect.

Speculative The distributional question is who pays for stability once it is a bought service rather than a free by-product. Under the old regime inertia came free with electricity; unbundled, it is a cost that lands somewhere — on developers, on consumers, or on the batteries that could arbitrage instead. Whether that cost is levied fairly is a policy choice that the market designs above will make, largely invisibly.

Frontier If zero-synchronous operation is demonstrated, the civilizational consequence is that deep decarbonisation stops having a stability ceiling. Every credible net-zero pathway assumes very high inverter penetration; if the last 10 per cent turns out to require permanent thermal or nuclear backing purely for stability, the achievable share of clean electricity is capped below 100 per cent for reasons of physics, not cost. The demonstration decides which world we live in.

Speculative A grid whose stability is software is a grid that can be upgraded — and attacked — at the speed of software. Rotating inertia was slow, dumb and reliable; programmable inertia is fast, flexible and dependent on a control and communication layer. That is a change in the character of critical infrastructure as much as a change in its fuel, and it couples the power system to the software supply chain in a way the twentieth-century grid never was.

Frontier The geography of power changes: any region with sun, wind and batteries could in principle be electrically self-sufficient without importing rotating stability. This matters most for jurisdictions with no domestic gas or hydro, for which the current stability requirement is an implicit fossil-fuel dependency. Removing it would decouple electrical reliability from fuel geopolitics in a way with obvious strategic weight.

Speculative The failure mode also changes shape. A synchronous grid degrades gracefully as machines slow; an inverter grid can trip cleanly and suddenly when protection thresholds are crossed, as Odessa and Iberia both showed. A civilization that runs on inverters may trade frequent small events for rare, fast, wide ones, and must build restoration — black start from inverters — to match. Speculative Whether that trade is favourable depends on numbers nobody has yet measured at scale: the frequency of the small events against the probability and cost of the rare wide one.

These horizons track demonstration and mandate, not device invention, which is the term least likely to be limiting.

• 10 yr: Frontier Grid-forming batteries at gigawatt scale become routine on high-renewable grids, led by Australia’s 2.0 GW build-out and Britain’s Stability Pathfinder; grid codes increasingly mandate rather than permit grid-forming capability. Speculative A first metropolitan-scale, multi-hour zero-synchronous islanded run is achievable in this window and is the single highest-value demonstration anyone could stage; it needs commitment, not new science, and connects to the corridor-stability question in Energy Corridors.
• 25 yr: Speculative Protection paradigms that do not assume high fault current are deployed widely enough that current-limited inverter grids clear faults reliably; synchronous condensers persist as a fault-current bridge rather than being eliminated.
• 50 yr: Speculative Large interconnected regions run for extended periods with no synchronous generators online, stability provided by a mix of grid-forming inverters and condensers, if the demonstrations of the 2030s succeed and the protection problem is genuinely solved.
• 100 / 250+ yr: Handwave The category “synchronous generation” becomes a historical curiosity, and the concept of grid inertia is taught as an artefact of an era when electricity came from spinning metal — contingent entirely on the earlier demonstrations holding.

• Depends on This brief depends on Energy Storage Revolutions and Advanced Battery Technologies, because every grid-forming stability service on any grid today is delivered from a battery, and the reservoir that provides synthetic inertia is the same reservoir being asked to arbitrage energy and hold reserves. Nothing else on the map blocks the physics; the constraints below do.
• Requires (not on this map) Five constraints, in order. First, manufacturing of grid-forming batteries and inverters at the scale of a national fleet rather than a demonstration, since a site behaves as grid-following until essentially all of its inverters are converted. Second, a grid code that mandates grid-forming capability rather than permitting it, because while it is optional it is a cost developers avoid and the fleet fills with grid-following units that cannot collectively hold a grid up. Third, vendor-released, non-black-box electromagnetic-transient models for every inverter firmware revision, without which an operator cannot certify a dispatch is stable and the next forensic investigation hits the same data wall the Iberian panel did. Fourth, a market that pays for inertia and system strength as explicit services, of the kind that let Hornsdale sell its 2,000 MW.s at all. Fifth, the scientific result nobody is yet producing: a metropolitan-scale zero-synchronous black start actually demonstrated and instrumented, which is the decisive experiment this brief turns on.
• Enables Delivers to Planetary Scale Energy Systems and Space-Based Solar Power, both of which push power through inverters into a terrestrial grid and therefore inherit the inverter-stability problem whole; a zero-synchronous grid is a precondition for either concept mattering, not a downstream detail.
• Adjacent Adjacent to Energy Corridors, the nearest sibling, which shares the voltage-source-converter stability problem on its DC links; to Advanced Fission, a substitute that supplies synchronous rotating generation and so reduces the pressure to prove silicon can stand alone; to Infrastructure Resilience and Critical Dependency Atlas, for a grid whose voltage source is patchable firmware; and to Smart Cities, whose loads are what a zero-synchronous metropolitan grid would have to keep up.

Established “The 2025 Iberian blackout proves renewables cannot keep a grid stable.” This is the claim in widest circulation and it is wrong on the report’s own evidence. The ENTSO-E panel found a voltage and reactive-power failure, with conventional generators failing to meet reactive set points and delivering under 75 per cent of required output alongside renewables under fixed power factor; its chair stated the problem was voltage control “regardless of the type of generation.” The failure was in how sources were controlled, not in what fuel they used.

Frontier “More inertia would have prevented the Iberian blackout.” Reviewing experts were explicit that a system with greater inertia would have faced the same problem, because the overvoltage and oscillations evolved too fast and in the wrong variable for inertia to arrest. This is the misconception this brief most wants to correct, because it points policy at the expensive intervention (keep turbines spinning) instead of the effective one (fast voltage support and better ride-through).

Established “Grid-forming inverters give you inertia, so fault current is solved too.” False, and the conflation is dangerous. Synthetic inertia and fault current are different quantities: a grid-forming inverter can emulate inertia while still being unable to deliver the several-per-unit fault current a synchronous machine provides — the DOE specification says a GFM is not expected to inject 6 p.u. Protection, not inertia, is the harder unsolved problem.

Frontier “A grid can just run on grid-following solar and wind if there is enough of it.” Categorically not: a grid cannot operate at 100 per cent grid-following penetration, because current sources need a voltage to follow. Some grid-forming capability is not optional at very high IBR share; it is the thing that makes the rest possible.

Speculative “Zero-synchronous grids already exist, so the problem is solved.” Small island systems run at 100 per cent inverter share, and simulations reach 97 per cent on Maui and 100 per cent on an IEEE test system, but a multi-hour, black-started, zero-synchronous run of a metropolitan grid instrumented for RoCoF and fault current has not been done. Conflating a demonstrated island with a demonstrated city is the error the required-experiments material exists to name.

Speculative “Synthetic inertia is just a marketing name for a battery responding quickly.” The distinction is real and measurable: ARENA found Hornsdale’s grid-forming inverters injected active power proportional to RoCoF with peak output before the frequency nadir, which is the signature of an inertial response rather than a delayed reserve. It is genuinely inertia-like — bounded by the battery’s energy and by control delay, but not merely fast reserve.