Commercial Battery Storage in California: Demand Charges, Costs, and the 2026 Incentive Picture

SGIP is closed. Plan around that, not around a reopening.

The Self-Generation Incentive Program's Large-Scale Storage category — the non-residential bucket that funded most commercial battery projects in California for a decade — is closed to new general-market applicants. SelfGenCA's own program metrics page shows it sitting at Step 5, the final incentive tier, with status "Closed" as of the September 2026 data pull. The CPUC's program page independently confirms the same thing, listing Large-Scale Storage funding as "available through 2025."

There is money still sitting in the program's books by territory — $1.79 million reserved for CSE (San Diego), $2.30 million for SCE, $1.96 million for SoCalGas, $10.74 million for PG&E — but that is committed, reservation-stage funding for projects already in the pipeline, not open capacity for a new application. If your project isn't already in that queue, it doesn't reach that money. The only SGIP budget still accepting new applications is the income-qualified Residential Solar and Storage Equity program under AB 209 — residential-only, irrelevant to a commercial project.

As of this writing, no CPUC decision or legislative action has authorized a fresh round of general-market commercial storage funding. That could change, but nothing in the record supports promising a reopening date to a client, and neither should a vendor's proposal. Before the closure, the step-5 rate was $0.25/Wh standard, reduced to $0.18/Wh for projects also taking the federal investment tax credit — cite that as history, not as money you can still get.

What actually pays for the battery: demand charges, not backup

Most commercial electric bills have two components: a volumetric energy charge (cents per kWh consumed) and a demand charge (dollars per kW based on your single highest sustained draw in the billing period, usually measured over a 15-minute window). Demand charges are disconnected from total energy use — a facility that runs lean most of the month but spikes hard for twenty minutes during a shift change or equipment startup can carry a demand charge as large as its entire energy bill.

Demand charges commonly account for 30% to 70% of a California commercial or industrial customer's total electric bill — a figure corroborated across NREL-sourced analysis and repeated consistently in industry rate literature. A battery that discharges during that peak window and recharges when the facility's draw is low doesn't need to store much energy relative to the facility's total consumption; it needs to be sized to the gap between peak and average demand, which is why demand-charge management is a fundamentally different sizing exercise than backup power or solar self-consumption.

This is also why backup power is the wrong frame for evaluating one of these systems. A battery sized to ride out a demand spike for 15-60 minutes a few times a month is not sized to keep a building running through a multi-hour outage, and a battery sized for meaningful backup duration is usually oversized and uneconomical for demand-charge work alone. Decide which problem you're solving before a vendor sizes the system for you.

The peak-shaving math, worked through

There's no way to give a single California-wide demand-charge number that means anything, because PG&E, SCE, and SDG&E each publish their own commercial tariffs with different rate structures, and the current tariff sheets for PG&E's B-19/B-20 and SDG&E's AL-TOU commercial schedules were not obtainable in readable form for this page — that number needs to come from your actual bill or a fresh tariff pull, not a blog average. One hard data point exists in the record: SCE's TOU-GS-3, Option R facilities-related demand charge was $18.71/kW per month as of a mid-2022 tariff filing. SCE restructured its commercial time-of-use windows in a 2024-2025 rate case, so treat that figure as historical context, not a number to plug into a 2026 proposal without verifying against a current tariff sheet.

The calculation method matters more than any single rate. Take a generic illustrative example (not a California-specific rate): a facility peaking at 1,200 kW that shaves 30% off its peak, down to 840 kW, at an example demand rate of $15/kW, cuts its monthly demand charge from $18,000 to $12,600 — a savings of $5,400 a month, or $64,800 a year. Run that same formula with your actual utility's demand rate and your actual peak-reduction target, and you have a real number instead of a marketing one.

A real, named deployment illustrates the pattern at scale: Stem Inc. built a 350 kW / 800 kWh system (roughly 2.3 hours of duration) for Penske Truck Leasing using its Athena software platform, and reported a 40% reduction in peak demand. Stem's broader "Electrodes" portfolio — 345 MWh across 25 large commercial and municipal customers in the Los Angeles area, including office buildings, water treatment facilities, and medical campuses — reported over 30% in additional monthly savings after switching to Athena-managed dispatch versus a prior software provider, though the dollar figures behind that percentage weren't independently verified for this page.

What a system actually costs

Commercial battery storage in California runs $400-700/kWh installed, turnkey, behind-the-meter — the more commonly repeated band across multiple industry sources for projects sized to demand-charge management. The wider range cited across the same sources spans $250-700/kWh, with larger systems (above roughly 1 MWh) trending toward the low end on an economy-of-scale basis. A 100 kWh system, at that range, runs roughly $40,000 to $65,000 before any tax credit.

These figures come from industry blogs and vendor-adjacent sources rather than a manufacturer spec sheet or a utility-scale procurement filing, so treat them as directional rather than a quote. No current Tesla Megapack, Fluence, or Generac list pricing was independently verified for this page. If the number needs to hold up in a board presentation or a loan application, get an actual bid — the range above is a starting point for budgeting, not a substitute for one.

For contrast: EnergySage's California residential storage data shows an average pre-incentive cost of roughly $1,074/kWh, with a typical 13 kWh residential system running about $13,958. Commercial storage is markedly cheaper per kilowatt-hour than residential, which is expected given equipment scale and the absence of a residential installer's per-job overhead spread across a small system — but that residential figure is not a substitute for a commercial quote and shouldn't be cited as one.

The federal 48E credit is the incentive that matters now

With SGIP closed to new commercial applicants, the federal Section 48E investment tax credit is the primary incentive lever left standing, and it's arguably a stronger one than SGIP ever was for storage specifically. Standalone commercial battery storage qualifies for the 48E credit at 30% through tax year 2033, stepping down to 22.5% in 2034 and 15% in 2035, contingent on national grid emissions staying under statutory targets. Unlike the residential 25D credit, which expired for good on December 31, 2025, 48E does not require the storage to be paired with solar — a storage-only commercial project qualifies on its own.

There's a compliance wrinkle worth knowing before you sign a contract: Foreign Entity of Concern ("FEOC") material-assistance rules require a minimum non-restricted-source cost share for the credit to apply — at least 55% for projects beginning construction in 2026, rising to 60% in 2027 and eventually 75% for projects starting in 2030 or later. Most of the commercial battery market still uses Chinese-sourced cells and modules, so a project that can't document its sourcing cost ratio risks losing the credit entirely. Ask your integrator directly how they document FEOC compliance before you sign anything.

One more flag: sources disagree on the exact safe-harbor cutoff date for pre-FEOC treatment — some cite July 4, 2026, others July 6, 2026, tied to the One Big Beautiful Bill Act's July 2025 enactment. Treat the cutoff as "mid-2026" rather than a specific date until you've verified it against IRS guidance or the statute directly, and don't let anyone quote you an exact date without a citation. Separately, whether standard MACRS depreciation still stacks on top of the 48E credit for storage assets was not confirmed in the research behind this page — verify that with a tax professional before it goes into a return model, rather than assuming it works the way solar MACRS historically has.

Sizing: follow the load, not the solar array

There is no verified industry-standard ratio for sizing battery capacity as a percentage of solar array size in California commercial projects, and any page that gives you one is making it up. The defensible engineering principle, repeated across multiple sources, is that storage should be sized to the gap between a facility's peak demand and its average or baseline demand — not to the size of a co-located solar system. A facility with a flat, steady load profile and no sharp demand spikes gets weak peak-shaving economics from a battery no matter how large the adjacent solar array is, because there's no meaningful peak to shave.

The one concrete sizing data point available — Stem's 350 kW / 800 kWh Penske deployment, roughly a 2.3-hour discharge duration — is consistent with what's commonly understood as typical for pure demand-charge management: short-to-medium duration systems, generally in the 2-4 hour range, sized to cover the peak window rather than to run a facility for an extended outage. A widely cited industry convention holds to a minimum roughly 2-hour discharge duration for projects seeking utility incentive programs, though that specific figure reflects general industry practice rather than something independently confirmed in current program rules for this page, and should be treated as background rather than a hard requirement.

Before sizing anything, pull twelve months of interval (15-minute) demand data from your utility account. If your monthly demand profile is flat, a battery is very likely the wrong tool for your building regardless of what a proposal says.

Vendors: what they actually do, and what they don't

Tesla's Megapack and Powerpack lines are proven at scale, but one industry-trend source characterizes them as optimized primarily for utility-scale deployment rather than commercial-and-industrial-specific projects — worth knowing so you don't assume Tesla is the default option for a mid-size commercial demand-charge project the way it might be for a utility-scale asset. (Arevon Energy's Nighthawk project — 329 Megapack units, 300 MW/1,200 MWh, commercial operation beginning August 2026 — is real, but it's a utility-scale asset, not a commercial-building case study, and shouldn't be cited as one.)

Fluence's Mosaic platform and Tesla's Autobidder are asset-management software layers for dispatch optimization and market bidding — relevant to understanding how a battery decides when to discharge, but they are software, not hardware vendors in the behind-the-meter commercial sense.

Stem Inc. is the most directly relevant name for demand-charge management specifically, with its Athena AI dispatch software paired to battery hardware and a real California track record (the LA-area Electrodes portfolio, the Penske deployment above). One flag worth passing to a client: industry trade coverage has reported Stem publicly pivoting from owning battery hardware toward a software-and-services model, which may matter for how a specific deal is structured — ask directly whether you'd be buying hardware from Stem or licensing dispatch software on top of someone else's battery.

Generac's Concerto and PWRcell lines serve smaller commercial accounts, and BYD and CATL are commonly referenced as cell suppliers integrated into other vendors' systems. No verifiable California-specific deployment data or current pricing was found for either in the research behind this page — mention them as market participants, not as a ranked recommendation. No vendor comparison table exists in the record; a defensible one would require direct RFP responses or spec-sheet pulls this page doesn't have.