EV Efficiency Potential and California Fleet Impact

Three pressures are converging on a question the field has been able to ignore. Global EV electricity consumption reached roughly 180 TWh in 2024 and is projected to hit 780 TWh by 2030; in California alone, statewide EV load could approach 99 TWh by 2050 under a high-adoption pathway, which is a distribution planning obligation rather than an abstraction. Battery pack sizes have grown alongside vehicle size, so efficiency now decides critical mineral demand as well as electricity demand. And the market is not self-correcting: the spread in energy consumption between comparable models is widening, not closing.

Policy has begun to respond unevenly. China adopted the first legally enforceable minimum efficiency standard for battery-electric vehicles, effective 2026. The European Commission has proposed sales multipliers for smaller, more efficient models. In the United States, the Petroleum Equivalency Factor embedded in CAFE compliance rates a 230 Wh/mi vehicle at 357 mpg-equivalent — against 147 on a tank-to-wheel basis — which effectively decouples compliance credit from on-road efficiency altogether. So the question is not only how much efficiency is available, but how much of the electricity system's exposure it would actually remove.

The EPRI/NRDC study defines four technology packages, from a frozen 2025 baseline to an advanced case with lightweighting, and reports vehicle efficiency for each. Re-running those same parameter sets through NREL's FASTSim — a more detailed accounting of component losses and limits than the one-dimensional model behind the original — reproduces them closely, which is the check that matters: the ceiling is real.

The more useful result is what happens when the parameters are allowed to vary independently rather than moving in unison. Across Monte Carlo draws over the same ranges, energy consumption forms a right-skewed distribution centred near 140 Wh/km. The Reference scenario sits almost exactly at the mean; the Advanced case sits 2.8 standard deviations below it, and Advanced-plus-lightweighting 3.4 below. Those are attainable positions, but they are the tail — reaching them requires many parameters to succeed at once, which is a policy question rather than an engineering one. Tire rolling resistance turns out to be the single largest determinant of dynamic consumption, and it is the one manufacturers control least: they do not choose the replacement tires an owner fits, or whether those tires stay inflated.

The UC Davis Transportation Transition Model is a cohort stock-turnover framework on an ASIF backbone — activity times stock times intensity times fuel carbon. It starts from today's fleet, adds annual sales, retires vehicles by age against survival curves, computes age-specific travel, and converts miles into energy and emissions. It has been used across California and US pathway work for years, and it could not answer any of the questions this study needed to ask.

It could not answer them because it did not carry body class, it did not carry batteries as objects with a size and a chemistry, and it had no geography below the state line. Three modules were added for this project, and the model was rebuilt from a spreadsheet into Python so that forty-eight scenario combinations could be run and disaggregated rather than maintained by hand.

Five dimensions, varied independently. The full grid is forty-eight combinations; the report reads three benchmarks across it — a high-load case, a midpoint, and a low-load case. Adoption and body mix are exogenous policy and market inputs rather than model outputs, which is deliberate: the model projects what follows from a sales path, it does not predict the sales path.

Source: Hwang, Rabinowitz, Jamhar & Tal (2026), Table 3-1.

Both ACC II variants put the same 27 million electric light-duty vehicles on California roads by 2050 — roughly 86% of the fleet, against 8.2 million and 26% under the baseline. They satisfy the same mandate with the same vehicle count. What separates them is what those vehicles are.

Under High Shift, small and large SUVs and pickups make up about 90% of the 2050 electric fleet, with small SUVs alone accounting for more than half of every EV on the road. Under Maintain 2025, light trucks reach about 19 million — some five million fewer — and cars expand 28% to fill the gap, mostly as large cars. That single compositional difference is 15.3 TWh a year by 2050, which is more than the entire ambitious-efficiency package delivers.

1. 01

   Efficiency and vehicle size are the same size lever

   Moving to the ambitious efficiency path removes 13.8 TWh from 2050 demand. Freezing the electric fleet's body-class mix at its 2025 composition removes 15.3 TWh — slightly more. An efficiency standard written in kWh per mile captures the first and is silent on the second, which means the best available regulatory instrument reaches roughly half of the load reduction on the table.

2. 02

   Light trucks are not part of the problem, they are the problem

   In the high-load case, SUVs and pickups draw 92.0 of the 98.8 TWh — 93% — while every car on the road draws 6.8. Even in the best case, after both levers are pulled, light trucks are still about 78% of what remains. A policy that reaches only passenger cars is working on the smaller share of a problem that is already concentrated.

3. 03

   Efficiency shrinks the load without moving it

   Southern California Edison and PG&E together carry about 80% of statewide EV load in every scenario tested; the three investor-owned utilities carry just over 90%, and adding LADWP reaches roughly 94%. Los Angeles County alone falls from 27.2 to 20.2 TWh and remains nearly twice the next-largest county. Efficiency changes the size of the upgrades, not where they have to be built.

4. 04

   Where the efficiency gets spent decides the mineral bill

   The same efficiency gains can buy longer rated range at an unchanged pack, or the same range in a smaller one. Only the second shows up as avoided material: cumulative pack production falls from 3.17 to 2.07 TWh, and 2050 annual demand in the nickel-rich portfolio falls from 86.1 to 44.9 kt of nickel, 13.7 to 7.2 kt of lithium, 13.4 to 7.1 kt of cobalt and 124.0 to 64.6 kt of graphite. Holding capacity flat cuts projected new deposit openings from about 90 to roughly 62.

5. 05

   Fleet efficiency lags new-vehicle efficiency by years

   Cohort roll-forward weights newer, more efficient vintages against the older vehicles still surviving, and older vehicles are driven less. The consequence is that even sustained multi-percent annual gains in new-vehicle MPGe take many years to arrive as reduced statewide load — so the timing of an efficiency standard matters nearly as much as its stringency.

The policy conclusion is narrower than the headline number and more useful. If the objective is to reduce the electricity system's exposure to light-duty electrification, an efficiency standard indexed to kWh per mile is necessary and roughly half sufficient. The other half is a segment question, and the instruments that touch segment — footprint-based standards, ZEV credit structures, purchase incentives — currently either ignore vehicle size or actively reward it.

Two things follow for planning. Utilities can size their expectations to the efficiency case they think is realistic, but they cannot relocate the problem: the LA Basin, the Bay Area and San Diego remain the load centres under every scenario, so the sequencing of distribution upgrades is unchanged by any efficiency outcome. And the range-versus-pack-size decision, which no on-road efficiency metric can see, is where roughly a third of the cumulative mineral requirement is decided.

This is phase one. It establishes the technical ceiling and the fleet consequences; it does not estimate what the efficiency improvements cost, or how much of a point-of-sale gain survives real-world driving, tire replacement and HVAC use. Both are proposed for the next phase, and both would move the answer.