The largest battery fleet in the United States is being asked to do the one thing it was never bought for: hold the grid up when there is nothing to follow.
ERCOT counts 14.2 GW of operating storage, ahead of CAISO's 12.4 GW. Its April board update also recorded that no operating storage resource had advanced grid support, the capability the market calls grid-forming.
The gap now has a price. Storage with an interconnection agreement signed on or after 1 April 2026 must demonstrate advanced grid support, and ERCOT has proposed $1,500/MW for older projects that retrofit, under a $25 million cap.
Largest US battery fleet
ERCOT passed CAISO in the second quarter of 2025, according to S&P Global Commodity Insights. pv magazine USA, October 2026
Outside the new requirement
ERCOT's estimate of existing storage resources not covered by the advanced grid support rule. NPRR 1333, May 2026
Transmission headroom
Potential improvement in generic transmission constraints, including West Texas, McCamey and the Panhandle. ERCOT assessment, 2026
The spec is about voltage, not megawatt-hours
Most batteries on the grid are grid-following. They measure the voltage around them and push current in phase with it. That works while a synchronous machine or a strong import sets the reference.
Grid-forming inverters set their own reference. They hold a voltage waveform, absorb small disturbances and can run a section of the network without a generator attached. No two projects are alike; the tests are not.
ERCOT's rule leans on model quality and unit validation instead of a hardware list. A developer submits an electromagnetic model, passes quality tests and demonstrates the response. Section 2.14 of the operating guide and Section 6.2 of the planning guide carry the detail.
| Characteristic | Grid-following | Grid-forming | Synchronous condenser |
|---|---|---|---|
| Voltage reference | ✗ Tracks the grid | ✔ Sets its own | ✔ Sets its own |
| Fault response | ◐ Can trip on weak voltage | ✔ Rides through and limits current | ✔ Physical inertia |
| Black start | ✗ Needs an external source | ✔ With enough stored energy | ◐ Possible, slow to start |
| Energy shifting | ✔ Core function | ✔ Core function | ✗ No storage |
| Main cost driver | Battery and PCS | Controls, PCS sizing, validation | Civil works and maintenance |
Requirements and capabilities as described by ERCOT, AEMO and NERC documents, 2025–2026
The engineering difference shows up in the first few cycles of a fault. A grid-following plant can trip when the voltage it is tracking collapses. A grid-forming plant rides through, limits current and helps the voltage recover.
That behavior is why ERCOT calls it advanced grid support and why the test protocol runs at full charge power, with reactive steps and angle jumps. It is also why the first projects are being watched closely.
A deadline that has already passed
ERCOT filed NOGRR272 in October 2024. The request was tabled twice, sent to two working groups and finally approved by the Public Utility Commission of Texas on 6 November 2025, effective 1 December 2025.
Then came the cliff. Any storage resource with a standard generation interconnection agreement signed on or after 1 April 2026 must provide advanced grid support. In April, ERCOT reported zero operating storage resources with the capability. One was working through the process.
The queue explains the urgency. ERCOT had roughly 462 GW of generation and storage in its interconnection process at the end of March 2026, and about 410 GW of large-load requests. It logged 198 new large-load requests in the first quarter alone.
Not everyone wanted the requirement scoped to storage. Jupiter Power, Plus Power, Spearmint Energy, NextEra Energy Resources and HGP Storage filed comments through 2024 and 2025. Several argued that singling out storage tilted a competitive market. ERCOT kept the scope and added the retrofit incentive to reach the exempt fleet.
The Texas rule is not the only clock. NERC's PRC-029-1, approved by FERC in July 2025, adopts IEEE 2800 ride-through curves and becomes enforceable for bulk-system inverter-based resources on 1 October 2026.
The retrofit arithmetic
ERCOT puts roughly 50 GW of existing storage outside the new requirement. NPRR 1333, filed on 26 May 2026, offers those resources $1,500/MW multiplied by an availability factor. A plant that stays above 90% availability for its first 12 months keeps a factor of one.
The programme caps total awards at $25 million. At full availability that buys about 16.7 GW of nameplate, or a third of the exempt fleet.
Payment arrives 12 months after the capability is in service, and the award carries an 18-month implementation window. First come, first served.
For a 100 MW battery, the cheque is $150,000 before the availability adjustment. That is real money for a controls update and small against a full power-conversion swap. ERCOT's own documents stress that many existing inverters may need no substantial modification, which is the assumption the incentive is built on.
What Australia already did
Australia is the closest read on how fast a market can move once the spec is mandatory. AEMO made grid-forming batteries a priority action for the National Electricity Market and the South West Interconnected System in 2026.
About 74% of the battery storage pipeline in the NEM is confirmed to carry grid-forming inverters, according to AEMO. Neoen's 270 MW/540 MWh Western Downs stage one runs grid-forming. Its 238.5 MW/477 MWh Blyth battery was commissioned in April 2025.
Transgrid is planning 5 GW of grid-forming storage in New South Wales. It expects the fleet to cover nearly half of the state's system-strength need, equivalent to 17 synchronous condensers.
The hardware claims are getting concrete. Sungrow's grid-forming power conversion system restored voltage in 19 seconds during black-start testing at a 30 MW field site. Fluence counts more than 2,000 MVA of contracted grid-forming projects.
The model, not the datasheet, decides
ERCOT does not certify a product. It certifies a plant. The advanced grid support requirement runs through an electromagnetic model of the whole facility, then a set of quality tests, then unit validation before commercial operation.
The scope is narrower than the label suggests. ERCOT's proposal does not require extra short-circuit current, does not require the plant to hold energy in reserve and does not require hardware beyond what the resource already has. It requires the response to be demonstrable while the resource has charge and is inside its design envelope.
That is a testable standard, and a demanding one. The protocol exercises the plant at full charge power, applies reactive steps and angle jumps, and checks whether the controls recover without tripping. A model tuned for a study and never benchmarked against field data will not survive the comparison.
The order of work matters. The interconnection agreement comes first, then the model quality submission, then unit validation, then commissioning. Each step can send a project back to the vendor, and each step sits on the critical path to revenue.
The learning curve is visible in the numbers. In April, one operating storage resource was working through the process and none had completed it. By the time the first wave of post-cutoff projects reaches commercial operation, the protocol will have been run on a handful of plants at most.
Where the premium hides
Grid-forming is mostly a controls and validation problem until it is not. Plant-level controllers coordinate dozens of inverters, and the model that goes into the interconnection study has to match what the plant does in the field.
That is where schedules slip. Model quality tests, hardware-in-the-loop runs and site commissioning add work before commercial operation. The performance obligation then follows the asset for years, through warranty negotiations and availability penalties.
ERCOT's estimate of the payoff is 5–10% relief on generic transmission constraints, including West Texas, McCamey and the Panhandle. Constraints decide whether a battery earns in one location and idles in another, so relief is worth more than the energy arbitrage it enables.
Scale is not waiting for the retrofit debate. ERCOT counted 29.1 GW of storage in 2024 and projects 59.3 GW by 2029. Every project that signs after the cutoff enters the fleet with the capability already specified.
Is the grid-forming claim made at the plant controller, or only per inverter unit?
Who owns the electromagnetic model, and who pays when the model fails a quality test?
What does the plant deliver at the top and bottom of its state of charge?
How is availability measured for the incentive, and who carries the penalty for a missed month?
Those questions matter more than the brochure specification. A grid-following plant with a strong grid connection can look identical on a datasheet and behave differently on the worst day of the year.
Procurement is splitting along the same line. Battery suppliers sell energy capacity, power-conversion vendors sell the grid interface, and the plant controller ties them together. When the model fails, the argument over which layer is at fault is a commercial problem before it is an engineering one.
That is why developers are writing performance guarantees at the point of interconnection instead of at the inverter terminals. The grid sees the aggregate response, and the contract has to match what the grid sees.
The demand-side answer to the same constraint is already being funded. As we wrote in October, Voltus' $225M raise landed one day before FERC closed PJM's data-center fast lane: paying load to move instead of paying generation to appear.
Grid-forming storage is the supply-side version. The asset has to hold the voltage, ride through the fault and then go back to shifting energy for the rest of the day. That is a different product from a four-hour battery, and it is now a condition of interconnection in the fastest-growing storage market in the country.
The $25 million retrofit pool covers about a third of the 50 GW that sits outside the rule.
Australia's 74% grid-forming pipeline is the best guide to how quickly the spec becomes the default.
The next test is commercial. A 100 MW battery that can black-start a substation will be procured by someone who needs that feature at 2 a.m. The open question is whether the contract pays for the capability or only for the megawatt-hours.