Take an illustrative 60-unit apartment building in the East Bay with a 1,600-amp, 208-volt service and a plan to put a Level 2 charger at every stall. Nameplate math says that is a 400 kW addition. The building has 176 kW of headroom, and the shortfall points to a PG&E transformer upgrade and a 12-to-24-month wait. The usual answer is a six-figure service project. On a building like this one, ours is more often a 100 kW cap and a controller. What separates those answers is a load forecast.
Multi-family is the hardest EV load to predict because the owner controls neither who buys a car nor when it plugs in. Here is how we build the forecast, and where the common shortcuts go wrong.
Nameplate is the wrong number
A Level 2 charger on a 40-amp breaker is limited to 32 amps continuous. On a 240-volt single-phase supply that is 7.7 kW. But most apartment buildings run 120/208-volt three-phase, where the same 32 amps delivers 6.7 kW. Sixty chargers is 400 kW. That is the right ceiling for the code and the wrong forecast for the building, because it assumes every car is plugged in, empty, and charging at the same moment.
An EV driven 30 miles a day uses about 10 kWh, which is roughly 90 minutes at 6.7 kW. Sixty vehicles need 600 kWh a night — 60 kW of average power across a ten-hour overnight window, or 15% of nameplate. The real load sits between 15% and 100%, and the forecast is deciding where.
What the code lets you assume
NEC 220.57 calculates each charger at the larger of 7,200 watts or its nameplate rating, with no diversity credit. A feeder sized by that rule carries the full 400 kW. The way out is load management: under NEC 625.42, a listed energy management system that caps total charging load can set the feeder size, provided the limit is documented and enforced by the system. That is what turns 400 kW on paper into 100 kW in the panel schedule.
Your inspector decides how far to trust it. Some jurisdictions accept a software limit; others want it locked in a listed control system. Ask before design, not at plan check.
When cars actually plug in
Left unmanaged, residents plug in between 5 and 8 PM. That is the building’s own evening peak and the 4–9 PM on-peak window under PG&E’s current rates, which we covered in PG&E’s 2027 Demand Charge Restructure. Coincidence also depends on fleet size: small groups overlap heavily, large ones average out. Our planning figures for unmanaged charging are 60–70% simultaneous draw below 15 vehicles, falling to 35–45% above 50.
Twelve EVs today (20% adoption) at 65% coincidence is about 52 kW. Sixty vehicles at full build-out is 140–180 kW at the 6:30 PM peak. At roughly $25 per kW-month, that is $3,500–4,500 a month in demand charges if it lands on the building’s peak, and it will, because nobody charges at 3 AM unless something tells them to.
Cap it, then do the math on the cap
Dynamic load management caps total EV draw and divides it among whichever cars are connected. Cap the 60 stalls at 100 kW and each car averages at least 1.7 kW even if every stall is occupied. Over ten overnight hours that is 16.7 kWh per car against a 10 kWh daily need — a 67% margin. A resident with a long commute still wakes up charged. Nobody gets 6.7 kW at 7 PM.
The cost of the cap
A 100 kW ceiling bills at about $2,500 a month in demand charges. An unmanaged 160 kW peak, the midpoint of the range above, bills at $4,000. That is $18,000 a year, and a managed load may keep the transformer upgrade off the critical path.
The tradeoffs are policy, not physics. Decide who gets priority when the cap binds, whether faster charging is a paid tier, and what the chargers do when the network drops. Specify units that fall back to a fixed low current on lost connectivity. A charger that defaults to full power when its controller goes offline defeats the whole design.
Check the service before you order chargers
NEC 220.87 lets you size new load from 12 months of utility interval data: the maximum measured demand times 125%, in place of a calculated load. In our example the building peaked at 320 kW last July, so the existing load counts as 400 kW. A 1,600-amp, 208-volt service is 576 kW, which leaves 176 kW. An unmanaged 400 kW does not fit. A 100 kW cap fits with margin, and so does a modest phase-in beyond it.
The panel is rarely the binding constraint. The utility transformer is, and its lead times are the subject of The True Cost of Delaying EV Charging at Commercial Facilities. Request the interval data first. If the managed load still stacks on the evening peak, a battery shaves it; the economics are in The Hidden ROI of Commercial Batteries.
Three numbers to ask for
Before approving any multi-family charging proposal, ask for the building’s measured peak over the last 12 months, the proposed EV cap in kW and the mechanism that enforces it, and the transformer capacity with the utility’s position on it. A proposal that multiplies chargers by nameplate and stops has not forecast anything. It has priced the worst case.
If you are planning charging for a multi-family property, talk to our team. We will run your interval data before you commit to a service upgrade.
A charger count is not a load forecast. Interval data, a documented cap, and a fallback setting are — and together they usually cost less than the upgrade they replace.