Every figure below comes from the same data that runs redo.energy's commercial calculators: our tariff database of demand charges and time-of-use (TOU) windows, our commercial solar calculator's battery model, and each utility's commercial rate from the U.S. Energy Information Administration (EIA).
Part 1 was about the one decision a restaurant can make alone, the water heater, and we took two restaurants through it. The first has a 120-seat dining room: 5,000 square feet, a gas line under the range, and a peak draw around 35 kilowatts (kW) on a demand-metered rate. The second is a counter-service shop: 2,500 square feet in a strip mall, on a small rate with no demand charge, with a landlord who owns the roof. On National Grid in Massachusetts, the shop came out with an eight-year payback on a heat pump water heater, under three years on Mass Save's small business offer, and a green light. The dining room came out with a plan for the day its gas heater dies.
The rest of the bill is harder, because it runs through the utility's tariff and the landlord's roof, and because of one fact about both restaurants. A house's peak is a cold morning. A restaurant's is dinner, after the sun has gone down.
Demand charges: your summer peak sets a winter floor
Commercial electric bills carry a line that home bills don't: a demand charge, billed on the highest kilowatts you draw in any short window during the month. Our calculator's default for a restaurant depends on where it is: about $4 per kilowatt in Georgia, $13.50 in Boston, $30 at Southern California Edison. At $18, the dining room's 35 kW peak costs roughly $630 a month before a single kilowatt-hour is billed.
Some utilities add a ratchet. Your highest peak of the year, usually a summer dinner service with the kitchen and the air conditioning both running, sets a floor under the demand you're billed for the next eleven months. In our database, Alabama Power's floor is 90 percent of the summer peak, Mississippi Power's 75, Green Mountain Power's 60, Idaho Power's and Central Maine Power's 50, and the Tennessee Valley Authority's local distributors' 30. The investor-owned utilities in California, New York, and eastern Massachusetts have none.
Under a ratchet, a heat pump for winter heat is close to free on the demand line. The dining room's 35 kW summer peak sets an Alabama floor of 31.5 kW. If winter service draws 20 kW and the heat pump adds 8, the meter reads 28, under the floor, and the demand charge doesn't move. What breaks this is resistance backup. Fifteen kilowatts of electric strip heat on a January morning, on top of the same 28, sets a new peak and a new floor. A heat pump sized to carry the building without strips is the whole discipline, and the ratchet rewards it more than any rebate does.
The shop's version of this problem is the opposite one. Small restaurants often sit on a small general service rate with no demand charge at all: Pacific Gas and Electric's B-1 under 75 kW, Xcel Colorado's C-TOU under 50 kW, Con Edison's small time-of-day rate up to 10 kW. Not every utility has one: Santee Cooper bills demand from the first kilowatt, even on its small general service rate. Add a heat pump, or a bank of electric fryers, and a shop near the line can cross into demand metering and start paying for its peak for the first time. Before electrifying anything in a small restaurant, find the threshold on your rate, and size so you stay under it. Our commercial heat pump calculator shows the demand charge and the ratchet for your utility on the results page.
Batteries are what's for dinner
Restaurants are the best daytime match for solar of any building type our planner models. Walk-ins, ice machines, hood make-up air, and the air conditioning all run while the sun is up, so the planner expects 70 to 85 percent of midday production to be used on site. That's the shop's story: lunch is its biggest meal and it happens under the panels. The dining room's biggest meal happens at 7 PM.
Dinner is also the expensive hour. Of the 108 business TOU rates in our database with published peak hours, 81 put the peak over dinner service all year and 24 more for part of it: 4 to 9 PM at PG&E, SDG&E, and Southern California Edison, where it's mandatory even for the smallest commercial customer; 5 to 9 PM on weekdays at Xcel Colorado, where it's the default for a business under 50 kW, and at Central Maine Power; 6 to 10 PM at Rocky Mountain Power in Utah; noon to 8 PM on Eversource's western Massachusetts rate. The gap isn't always wide. On PG&E's B-1, summer peak power costs about 47 cents and off-peak about 40. Unfortunately, restaurants can't change everyone's dinner schedule (and not everyone is a retiree who wants to sit down for supper at 4:30 PM).
But, a battery can move the electricity. Our commercial solar calculator prices a four-unit Powerwall 3 stack, 54 kilowatt-hours (kWh) with about 49 usable, at $37,700 installed, $26,400 after the 30 percent federal credit, and about $18,400 for a taxpaying owner after bonus depreciation. Charged from the roof or from off-peak power and discharged into dinner, it earns four ways.
- Shifting the hour. Moving about 46 kWh a night from off-peak into the peak is worth about $1,100 a year at Southern California Edison and $1,500 at SDG&E. At Xcel Colorado, where a small business pays two and a half times the off-peak price from 5 to 9 PM on weekdays, it's about $2,700. At Con Edison, on an optional TOU rate, it's about $800. At PG&E it's worth nothing: B-1's peak is only a few cents above off-peak, less than a kWh loses on its way through the battery. National Grid's and Eversource's eastern Massachusetts business rates barely price energy by the hour, so there's nothing to shift there. In the Southeast, where spreads are a few cents, our calculator doesn't count any.
- Storing the roof's surplus. With panels on the roof, the battery fills up on the midday power the kitchen doesn't use, instead of exporting it, and serves it at dinner. At PG&E, where shifting earns nothing, that's worth $4,100 to $5,300 a year with about 60 kW on the dining room's roof, depending on your hours. It's worth less where the utility already pays well for surplus: $800 to $1,000 in Massachusetts, and nothing beyond shifting at Xcel Colorado, where net metering credits it at retail.
- Shaving the peak. A battery discharges exactly when the peak hits, which solar can't promise. Over a three-hour window it can hold the dining room's demand down by about 15 kW, worth about $3,200 a year at $18 per kilowatt. The shop, on a rate with no demand charge, earns nothing here.
- Selling flexibility. Utility programs like Massachusetts' ConnectedSolutions pay for dispatch during their summer events. Our calculator keeps that money out of the headline payback, since it's paid only for events the battery actually joins, and shows it as a second payback instead. That figure discounts the demand saving for the summer months, because the same kilowatt can't shave your peak and serve the grid at once.
In California the answer depends on the utility. At Southern California Edison and SDG&E the battery is the solar case for a dining room. On its own it earns about $4,300 and $4,700 a year at $18 per kilowatt, on an $18,400 net cost, five years to pay back. Our calculator's own default demand charge there is $30, which lifts that to $6,400 to $6,900 a year and three years. Without a battery, a restaurant's midday surplus sells for four to six cents while dinner buys back at 31 to 47 cents, so on the dining room's roof the battery pays back in two to four years. PG&E is the exception. A 35 kW dining room fits under B-1's 75 kW line, so it has no demand charge to shave and a peak only a few cents above off-peak. There a battery alone earns nothing and doesn't pay back within 25 years. On the dining room's roof it stores the midday surplus instead, for $4,100 to $5,300 a year, and pays for itself in four or five years.
In Massachusetts the roof already gets a fair price for daytime surplus, so the battery is mostly a demand-charge play. That's about $3,200 a year at $18 per kilowatt and six years to pay back. Joining ConnectedSolutions adds roughly $1,700 a year after the summer demand savings it gives up, which brings payback to about four years. At Boston's $13.50, our calculator's default there, it's about $2,400 a year and nine years, or five with ConnectedSolutions. On the dining room's roof, storing surplus trims those paybacks to five and six years, or four with ConnectedSolutions. At Con Edison, shifting adds about $800, for about $4,000 a year and five years, or about $4,700 and still five at the $21.50 default there. New York's value stack pays a roof's surplus far less than dinner costs, so on the dining room's roof the battery earns about $6,300 to $7,400 a year and pays back in three. Con Edison's demand response pays about $1,400 a year, but joining gives up about as much in summer demand savings, so it doesn't change the payback.
In the Southeast a battery alone earns only from the demand charge. At $18 per kilowatt that's a six-year payback. Our calculator's defaults there are lower, and at them the payback runs from seven years in Tennessee and South Carolina to 14 at Alabama Power and past 25 in Georgia. Surplus there sells at avoided cost or less, so on the dining room's roof the battery also stores it. That brings payback to four to six years at $18, and at the defaults to five in Tennessee, six or seven in South Carolina and at Alabama Power, and 11 or more in Georgia.
The physical limit holds everywhere. The dining room uses about 130 kWh between 5 and 9 PM and the battery holds 49, so one stack carries about a third of dinner. Three stacks would carry all of it, at three times the cost for the same spread, which is why the economics favor a battery sized to the demand peak rather than to the whole evening.
Two dates matter. The federal investment credit for solar ends for anything not in service by December 31, 2027, unless construction began by July 4, 2026, per IRS Notice 2025-42. Storage is not included in that termination, so a battery keeps its 30 percent credit even after a late solar array would lose its own. And 100 percent bonus depreciation is back permanently for property placed in service after January 19, 2025, which is what turns $26,400 into $18,400 for an owner who pays tax.
One more line for Massachusetts: a private system over 25 kW gets market net metering credits at about 60 percent of retail, under 25 kW gets full retail. A dining room's roof can hold enough panels to cross that line; a shop's can't, and in a strip mall the roof isn't the shop's to use.
The roof and the rooftop units belong to the landlord
The water heater, the kitchen equipment, the walk-in, and the lighting are usually the tenant's, which is why the shop could act on its own in part 1. The roof, the rooftop heating and cooling units, and the service entrance are the landlord's, and everything above runs through them. Massachusetts' small business offer pays up to 70 percent of a heat pump or weatherization project, and up to 100 percent for a renter using under 1.5 million kWh a year who occupies more than half the building, with the landlord signing one qualification form. That form is the conversation to have. The dining room is likelier to own its building or hold a long lease, which is where the solar math and the space heat pump belong. For the shop in a strip mall, solar is the landlord's decision, and the tenant program is the shop's lever.
Rebates and grants worth chasing
- Municipal light plants. Concord Municipal Light Plant pays businesses $2,500 a ton for air-source heat pumps and $4,500 a ton for ground-source, up to $50,000 per customer every three years, five times its residential cap. Gas-heated buildings go to Mass Save instead. Ask your light plant for its commercial schedule; several are far above the investor-owned utilities' offers.
- Mass Save. Custom review for heat pump projects over 35 tons, and the 70 and 100 percent small business offers above.
- Rural restaurants. The Rural Energy for America Program (REAP) pays grants of up to half of a solar or efficiency project's cost, up to $1 million, for small businesses outside towns of 50,000. That half came from the 2022 climate law's funding; the statutory baseline without it is 25 percent. The grant window is shut at the moment. USDA paused new grant applications in June 2025 to work through a backlog, and says it will rewrite the program rules before reopening them, though guaranteed loans are still being made. Have the paperwork ready for the day it reopens.
Where you are changes the answer
After dark, three things set the answer: whether your rate has a peak window and a ratchet, what the utility pays for your midday surplus, and what your sun is worth.
Massachusetts and New York. No ratchets on the investor-owned utilities, and full retail credit for a roof under 25 kW in Massachusetts, so the roof there already gets a fair price for daytime surplus. Solar goes on an owned roof before the 2027 clock. In Massachusetts the battery is mostly a demand-charge play: about $3,200 a year for the dining room, six years to pay back or four with ConnectedSolutions, and five with panels on its roof. At Con Edison it earns about $4,000 a year and pays back in five years alone, or three on the dining room's roof, because the value stack pays far less for surplus than dinner costs. Joining its demand response adds almost nothing. The shop has no demand charge to shave and no roof of its own, so its battery case is thin.
California. Mandatory 4 to 9 PM peaks for every commercial customer and exports worth four to six cents. At Southern California Edison and SDG&E the battery is the solar case: about $4,300 to $4,700 a year for the dining room at $18 per kilowatt, five years to pay back alone and two to four with panels on its roof. At PG&E both restaurants fit on B-1, with no demand charge and a nearly flat peak, so a battery alone earns nothing and doesn't pay back. On the dining room's roof it stores the midday surplus and pays back in four or five years. The shop has no roof of its own to pair it with. The one thing to size for there is the 75 kW line that moves a restaurant onto B-10 and demand metering.
The Southeast and Tennessee. Ratchets of 30 to 90 percent, small TOU spreads, and exports at avoided cost. A space heat pump sized without strips rides under the summer floor for free, and that is the decision that matters. Solar is a self-consumption play. A battery alone earns only from the demand charge, so at our calculator's defaults it pays back in seven years in Tennessee, 14 at Alabama Power, and not within 25 in Georgia. On a dining room's roof it also stores the surplus, which brings those to five, six or seven, and 11 or more.
The Midwest. Cheap gas keeps space heat on gas in most dining rooms. Ask for a commercial interruptible or dual-fuel rate, which changes the space heat answer the way it does for homes. In Minnesota, commercial solar above 40 kW is credited at avoided cost rather than retail, so size the roof to the daytime base and no larger.
Northern New England and Idaho. Ratchets of 50 to 60 percent with better power prices than the Southeast. The ratchet rule applies. Efficiency Vermont and Efficiency Maine run the programs.
The Mountain West. Strong sun, and time-of-use peaks that land on dinner: 5 to 9 PM on weekdays at Xcel Colorado, where it's the default for a business under 50 kW, and 6 to 10 PM on Rocky Mountain Power's optional rate in Utah. A dining room under 50 kW at Xcel pays no demand charge, but its peak costs two and a half times off-peak, so the battery earns about $2,700 a year from shifting alone, more than at SDG&E, and pays back in about eight years. Panels on the roof don't add to that, because Xcel's net metering already credits surplus at retail. In Utah it's a demand-charge play again, about six years at our calculator's default, with a roof or without. Watch the windows against your dinner. In Xcel's Colorado territory, a business on a demand rate can also get Solar*Rewards, which pays 4 cents for every kilowatt-hour an 8 to 250 kW system produces, for 20 years, in exchange for its renewable energy credits.
Rural, anywhere. REAP pays up to half of a solar project, whenever the grant window reopens. Our calculators don't count the grant, so take it off the quoted cost yourself.
Some methodology honesty before you call anyone
Kitchen exhaust is the heating load a house doesn't have. Every cubic foot the hood pulls out has to come back in, usually heated, usually by gas, and the design guides say the 80 percent rule of thumb for make-up air oversizes it. Our calculators don't model that load. Demand-controlled ventilation and transfer air from the dining room cut it before any heat pump question arises, and they come first. Refrigeration heat recovery is real and not modeled either. And the battery figures assume the stack empties into the peak every day, which the dining room's roughly 130 kWh between 5 and 9 PM easily supports; a lunch place shifts less and needs a smaller battery.
The takeaway
Across both posts the two restaurants never got the same answer, and the reason was never the equipment. The shop's water heater is likely electric and its own, so it swaps it now, with Mass Save paying 70 to 100 percent. Its rate has no demand charge, so a battery earns little and a heat pump has to stay under the kilowatt threshold that would move the shop onto a demand-metered rate. Its roof is the landlord's, so solar is a conversation. The dining room's water heater is likely gas, so it waits for the end of that heater's life and plans the new unit's placement. Its rate has a demand charge and maybe a ratchet, so the space heat pump is sized without strips and the battery is sized to the peak. Its roof is its own, so solar goes up before the 2027 clock. The rate schedule, the lease, and the fuel under the range decide more than any brand does. Find those three things out first. Our calculators do the rest.
Our commercial planner takes a restaurant as a building type, with the time-of-use window and the 2026 tax rules for your state already loaded. It doesn't model demand charges, so price the battery on our commercial solar calculator, which does. Run your address before you sign a quote.