On October 28, 2026, the California Independent System Operator’s Board of Governors approved the large-load interconnection framework it will file with the Federal Energy Regulatory Commission by November 16. It is the first of the six grid operators facing FERC’s June 18 show-cause orders to carry a proposal through board approval, and the clearest statement yet of the bargain the regions are offering the country’s largest new customers: speed in exchange for less firm service. A load of 50 megawatts or more at a single site — a data center, a factory, any large end user — will be able to connect to the transmission system before the network upgrades it needs are finished, on the condition that its service can be cut back until those upgrades are done.
That trade is the centerpiece of CAISO’s answer to Docket EL26-71, the proceeding FERC opened on June 18 when it preliminarily found that CAISO’s tariff appeared unjust and unreasonable for lacking adequate provisions for large loads. The filing CAISO will make on November 16 is a tariff revision under section 205 of the Federal Power Act, shaped by a stakeholder process that began on February 27. The board vote is not the end of that process — FERC must still decide whether the package satisfies the show-cause order. But the shape of California’s answer is now fixed, and the other five operators file on the same deadline.
What the board approved
The package has five main elements. First, a definition: a large load is an end user at a single site with a peak load of 50 megawatts or greater. Second, study procedures: the participating transmission owners must have load study processes that finish within 60 to 90 days, a timeframe the proposal attributes to FERC, plus a new monthly review window for transmission additions or upgrades tied to large loads that fall outside the regular planning cycle. Third, cost protection: transmission owners must set tariff terms preventing the costs of serving a large load from shifting to existing customers. Fourth, rules for co-located generation, which exempt on-site generation serving only the facility from an ISO study while requiring any generation that injects power into the grid to complete one. Fifth, the flexible interim service itself.
The draft final proposal behind the vote was published on September 24, with draft tariff language the same day; a revised draft tariff followed on October 26. Two days after the vote, on October 30, CAISO posted a second revised draft, adding ride-through provisions written by the initiative’s large load technical working group — the most recent tariff text ahead of the filing.
The 50-megawatt line
Fifty megawatts is not an arbitrary number, but it is a chosen one, and the choice was contested. FERC’s show-cause order suggested a definition — a new commercial or industrial customer at a single site with a peak load of 50 megawatts or more, interconnected above 69 kilovolts and not part of a co-location arrangement — that CAISO has effectively adopted in substance while keeping a single threshold. During the stakeholder process, Pacific Gas and Electric suggested a lower threshold of 20 megawatts. CAISO kept 50.
The comparison that matters is MISO, which proposed two classes: a 50-megawatt generic threshold and a 20-megawatt threshold for computational loads. CAISO’s single line is simpler, and simplicity has a cost: a 49-megawatt data center is nearly indistinguishable electrically from a 51-megawatt one, yet only the second falls under the framework. California’s data-center load alone is forecast by the California Energy Commission to grow by 1.8 gigawatts by 2030 and 4.7 gigawatts by 2035 — at 50 megawatts a site, 1.8 gigawatts is roughly 36 large loads.
FILI: the interim service, and what it demands
The main new feature is the Flexible Interim Load Interconnection service, or FILI: a curtailable interim connection. A large load can take power before the network upgrades it needs are complete, but its service can be cut back until those upgrades are done. The service level tracks what the system can safely deliver at that moment, and the load moves to firm service once the upgrades finish. Each arrangement is bespoke to the offtaker — flexible enough to fit real operating constraints, but harder to compare, audit, and judge as just and reasonable.
FILI is not free speed. The draft final proposal calls for participating customers to provide hourly load forecasts and telemetry data, and to install equipment that lets CAISO remotely disconnect the load when necessary to maintain reliability. That last provision is the one that converts an interim service into a reliability tool rather than a queue-jumping device. A load that can be disconnected remotely is a load the operator can shed in an emergency without the slow, blunt instrument of rotating outages. It is also a load whose operator must build its business plan around the possibility of interruption.
For illustration, consider what FILI changes for a data center developer. Under the old sequence, the developer applied, waited for studies, waited for upgrades to be built — a process measured in years — and only then energized. Under FILI, the developer can energize, at least in part, once the interim arrangement is in place, accepting curtailment risk in exchange for months or years of earlier operation. The upgrades still take years to build. What changes is the waiting.
The missing piece: no permanent flexible service
Here is the framework’s most consequential omission. The approved package contains only one flexible service — the interim one. An August straw proposal had floated a second option, a permanent flexible service known as FLIP, and the draft final proposal dropped it. A customer that wants firm service still has to wait for the upgrades to finish. There is no permanent product for a load willing to be flexible forever in exchange for a smaller upgrade bill.
That gap drew the sharpest stakeholder objection before the vote. Constellation Energy, in a public comment letter dated October 28, supported approval while urging the board to direct CAISO to include a permanent flexible transmission service option — FLIP or an equivalent — in the FERC filing. The letter points to the show-cause order itself: FERC specifically expressed concern that CAISO’s tariff may not provide transmission services for flexible large loads on both an interim and permanent basis.
CAISO’s answer, for now, is deferral. A permanent flexible interconnection option, along with arrangements for electrically proximate generation, is to be developed sometime in 2027. FERC’s test will be whether the interim stage is bounded tightly enough that it does not become the permanent rule by default. If FILI stretches indefinitely because upgrades lag, customers will have bought interim service and received, in practice, a permanent non-firm product without a permanent product’s tariff terms.
The study clock, and the skeptics
The framework’s second bet is on speed in the study phase. Participating transmission owners must have load study procedures finishing within 60 to 90 days, and CAISO adds a monthly review window for large-load transmission additions and upgrades outside the standard planning cycle. In a system where interconnection studies have historically taken years, a 60-to-90-day study is a genuine change in tempo — and the element stakeholders most doubt.
Microsoft and other stakeholders questioned whether transmission owners can finish the studies in that window, according to the trade press covering the October 30 board materials. Study teams sized for a trickle of large-load applications now face a pipeline of gigawatt-scale requests. The honest summary is the one this desk has offered before: the filings compress the study phase, not the construction phase. A faster study gives a developer an answer sooner. It does not pour concrete faster.
Cost shifting: the protection that has to work
The element that matters most to existing customers is the least visible: the cost-protection framework. The approved package requires transmission owners to set tariff terms preventing cost shifting — the danger that upgrades built for a large load end up in the rate base, paid for by everyone else, if the load shrinks, delays, or disappears after the wires are built.
The mechanism FERC keeps returning to across all six proceedings is the cost recovery agreement — a standard, pro forma agreement among the operator, the transmission owner, and the transmission customer taking service on behalf of the large load, backed by financial security, obliging the load to cover the costs incurred to serve it, including network upgrades, even if the project is canceled or operates below its anticipated demand. One limit should be stated plainly: even the best agreement is only as strong as its enforcement.
Co-location: metered separately, studied when it injects
The framework’s co-location rules split the difference between the two positions that have divided the proceedings. On-site generation serving only the facility would be exempt from an ISO study. Any generation that injects power into the grid must complete one. And co-located loads must be metered independently of offsetting generation, so they pay ancillary service charges on their gross demand rather than netting their own generation against it.
The netting question — whether a co-located load can subtract on-site generation from its demand — is one of the sharpest disputes in the six dockets, answered differently by region. CAISO’s answer is that it cannot, at least for ancillary service charges: the load is metered on its own, and the generation is treated as generation. For a hyperscaler weighing on-site generation as a way to shrink its grid-facing footprint, the metering rule changes the economics: the backup generator still has value as backup, but it does not shrink the bill for grid services.
The ride-through thread
The October 30 revision deserves its own attention because it ties this proceeding to the reliability story running in parallel. NERC’s computational-load standards package, approved on final ballot on October 15, would require large data centers and crypto facilities to model their disturbance behavior and answer operator instructions — but, as this desk reported, it sets no industry-wide minimum voltage and frequency ride-through requirement. Ride-through is the technical heart of the July 22 event in Northern Virginia, when a normally cleared fault triggered the loss of roughly 3,800 megawatts of data center load. PJM proposed its own ride-through requirements on September 10, and ERCOT implemented its own on August 1.
What it means for developers, and for everyone else
For developers, the framework changes the question a California site must answer. It is no longer only how long the queue is, but what service the site can buy while it waits, how often it can be interrupted, and what it must bring to convert interim service into firm service. A site that can tolerate curtailment — a training cluster that can pause, a facility with on-site generation or storage to bridge interruptions — gains real value from FILI. A site that cannot tolerate interruption gains little; it still waits for the upgrades.
For existing customers, the framework will be judged on the cost-protection terms the transmission owners write and on whether the monthly review window produces upgrades that are actually built rather than merely reviewed. The scale at stake is large: data centers already connected to CAISO’s transmission network have a combined capacity of approximately 8 gigawatts, according to the trade press, and the Energy Commission’s forecast adds 1.8 gigawatts by 2030. Each gigawatt that connects under FILI and is curtailed during stress is a gigawatt that does not become a reliability event.
What to watch
Three things will determine whether the October 28 vote becomes a working tariff. The first is the filing itself on November 16: whether the tariff language matches the board-approved proposal, and whether CAISO includes a permanent flexible option or files the interim-only package and defends the 2027 deferral. The comparison that will matter most is with ISO New England, which proposes a permanent non-firm product alongside its interim bridge, and SPP, whose conditional service runs up to seven years.
The second is the answer period, 30 days after the filing, when the weakest joints — the missing permanent service, the 60-to-90-day study clock, the transmission-owner tariff terms — will be tested by the same stakeholders who raised them before the vote. The third is FERC’s disposition: it can accept the filing, accept it subject to condition, set it for hearing, or prescribe a remedy itself.
Beyond the docket, the watch item is operational: the first FILI customer. An interim service exists on paper until a load actually connects under it, is curtailed under it, and converts to firm service when its upgrades finish. That first conversion — or the first time a FILI customer sits on interim service longer than anyone expected — will tell the industry whether California’s bargain works.