PG&E’s Flexible Interconnection Program Has Seven Sites Operating After Two Years

Seven projects are operating under PG&E’s Flex Connect program two years after launch, and two of them have since converted to firm interconnections because the distribution upgrades they were waiting on caught up, Latitude Media reported. The first flexible data center site, 5 MW in the Bay Area, is scheduled for this fall.

Flex Connect trades a guaranteed capacity right for an earlier energization date. A customer accepts a limit on how much it can draw during constrained hours, and in exchange the utility connects it now rather than after a substation rebuild.

The rollout pace. Deployment has run behind PG&E’s original forecast. The utility has not published a revised schedule, and seven operating sites in two years is the only completed-project figure on the record.

The operating limit. The mechanism is a day-ahead cap: the customer receives an operating limit in advance and is responsible for keeping grid draw below it. How the site complies is the customer’s choice. Chargers can be throttled, individual units can be switched off while others run at full power, or an on-site battery can cover the difference between the cap and actual demand.

The enforcement question. That distinction is where two years of operating data land. A flexible connection is a capacity contract whose collateral is equipment, and PG&E identifies its hardest engineering problem as demonstrating to distribution planners that a customer given a day-ahead operating limit will in fact stay below it.

The alternative has a measured track record. IESO found that hourly demand response resources in Ontario delivered about 64 percent of their capacity obligations when activated during summer 2025, and that participants rarely revised their bids to match what they could actually provide, RTO Insider reported. IESO is rewriting the 2027 capacity auction rules in response.

A planner sizing a distribution feeder cannot underwrite a 64 percent delivery rate. A planner can underwrite a device whose behaviour has been demonstrated.

Germany’s connection queue. Bundesnetzagentur monitoring data counts 18,158 battery grid connection requests in 2025, a 194 percent increase over 2024, totalling 573.5 GW and 1,583 GWh. Network operators issued 1,937 commitments covering 54.2 GW and 142.4 GWh, up 119 percent and 207 percent respectively. On the distribution grid, 1,429 of 17,775 applications received a commitment, roughly 8 percent.

The agency cautions that developers file the same project at several candidate sites, so the request count overstates real project volume. Even discounted heavily, an 8 percent commitment rate at the distribution level describes a queue that cannot be cleared with steel alone. That arithmetic is what makes flexible connection schemes durable rather than experimental.

The Australian comparison. Energy-Storage.News reported that DC-coupled solar-plus-storage configurations in Australia reach grid connection approval as much as six months faster than AC-coupled equivalents, because the battery sits behind the same inverter and does not trigger a separate assessment of a new grid-facing device. Separately, Solmech completed a DC-coupled retrofit at an operating South Australian solar farm without modifying the existing plant.

The mechanism is the one operating in California, pointed at generation instead of load. Value accrues to the configuration that avoids opening a new regulatory file.

Integration cost. PG&E bootstrapped its early sites with a vendor doing heavy custom work, and the stated goal now is to make the capability repeatable and affordable enough to become standard practice. The tariff is not the constraint that goal describes; the cost of building the control interface is.

Bespoke controls integration is a fixed cost, and fixed costs sort customers by size. A 5 MW data center can absorb it. A small commercial building cannot, which means the customers with the least ability to pay for a substation upgrade are also the ones least able to pay for the engineering that would let them skip it. Until the interface is productised, flexible interconnection is likely to remain a large-customer instrument regardless of how the tariff is written.

That last point is interpretation rather than reported fact, and it comes with a caveat worth stating plainly. Flex Connect is a load-side product governing new load additions, and PG&E has not published a segment breakdown of its pipeline. A battery installed to manage demand charges is not automatically a qualifying enforcement device. Nothing in two years of PG&E data establishes that building-scale storage can play that role, only that the underlying grid problem is a control problem.

The two conversions. The conversions are the most instructive number in the set, and they cut both ways. They confirm the design intent, which is a bridge rather than a permanent arrangement, and they put a clock on the value.

Storage financed partly on the strength of an earlier energization date carries an asset whose interconnection benefit expires the moment the upgrade behind it is energized. Demand-charge savings survive that transition. The speed premium does not. Two of seven sites reaching that point inside two years is a short duration for an argument that has to be underwritten across a fifteen-year or twenty-year equipment life.

What the figure does not show. The value of avoided delay appears nowhere on a project pro forma, because there is no line item for a connection that would not otherwise exist. PG&E has not published the interval between flexible energization and firm conversion at the two converted sites, which is the number that would price the bridge. Absent it, the program’s core economic claim, that control equipment substitutes for waiting, remains asserted by the utility rather than quantified in public.

Sources