How fast do EV batteries degrade?

Two-component fade per chemistry, both expressed as fractions of original capacity:

SoH = 1 − calendar(t, climate) − cycle(years, chemistry, driving + V2G)

Calendar fade uses a power-law in time: a · years0.7 with a = 0.020 (NMC) and a = 0.012 (LFP), ×1.4 for hot climates. Square-root-of-time would also fit the early data; t^0.7 matches the modest mid-life acceleration that fleet data shows.

Cycle fade = per-cycle loss × years × EFCdrive/yr · dcMult(dcShare) + EFCV2G/yr. Per-cycle loss is 0.008% (NMC) and 0.005% (LFP) at temperate / mostly-AC charging. dcMult is (1 + 0.5 · dcShare) for NMC and (1 + 0.2 · dcShare) for LFP — LFP is much less DC-sensitive because its operating-voltage window doesn’t include the high-voltage stress region. V2G/V2H throughput uses the same per-cycle loss with no DC-fast multiplier (modelled as slow AC cycling).

EFCV2G/yr = (v2gKwhPerDay × 365 × 2) / packSizeKWh — net discharge to grid/home per day, doubled for the matching recharge, divided by pack size.

One equivalent full cycle (EFC) is one full pack discharge. At your settings: —.

The constants are calibrated so the curves at default settings (15k km/yr, 20% DC, temperate) land on the published fleet averages. They are not derived from a battery physics simulation. If you have better data, the constants live at the top of battery.js.

• Geotab 2024 EV battery degradation report. ~5,000 connected EVs across 21 makes/models. Average ~1.8% capacity loss per year for cars from 2017–2019, falling to ~1.0–1.5%/yr for newer packs as thermal management has improved. By year 10 the average vehicle has ~12–18% loss (i.e. 82–88% SoH).
• Recurrent 2024 (n ≥ 20,000). Owner-network telemetry. Battery replacement rate <2.5% across all model years, dominated by recall-driven replacements (Bolt, Hyundai/Kona NMC). Voluntary replacement for capacity loss alone is rare.
• Tesla 2023 Impact Report. Model S/X fleet at 200,000 mi (322,000 km) retains ~88% capacity on average. Linear extrapolation suggests >15-year lifespans for typical use.
• LFP cycle life (BYD, CATL): manufacturer-rated 3,000–6,000 full cycles to 80% SoH at 25°C. Independent testing (e.g. Wang et al., J. Electrochem. Soc.) confirms 4,000+ cycles is realistic for well-managed packs. NMC sits at 1,500–2,000 cycles for the same threshold.

Vehicle-to-grid (V2G) and vehicle-to-home (V2H) discharge add equivalent full cycles on top of driving: same fade mechanisms as the model above (calendar time still ticks; every kWh in and out counts toward cycling). Use the V2G / V2H net discharge (kWh/day) control in the pack panel to fold a rough average into the curves; it is a throughput knob, not a grid-market simulator.

• NREL / TP-5400-69017 — Critical Elements of Vehicle-to-Grid Economics (2017). Operating V2G can increase wear because of added charge/discharge cycles. Across the studies it cites, the main levers are cycle count, depth of discharge per cycle, and temperature; limiting DoD (e.g. capping usable window to ~80% of nameplate capacity) is the most common mitigation called out to keep added fade acceptable.
• Peterson, Apt & Whitacre, Journal of Power Sources (2010). Lab cells under combined driving + V2G-style duty cycles still showed >95% of initial capacity after multi-thousand-day simulations. In their analysis, slow galvanostatic V2G-type cycling produced less capacity loss per kWh moved than aggressive vehicle acceleration/deceleration profiles, i.e. not all cycling stresses the pack equally.
• Later modelling work (reviewed e.g. in grid-service papers, 2020s): reported incremental capacity fade from chronic V2G ranges from minor (shallow daily arbitrage within a mid-SoC window) to double-digit percent extra loss over a decade for aggressive strategies (high power, deep daily DoD, hot cells). Treat any single percentage as scenario-specific, not a universal “V2G tax.”

V2H whole-home backup behaves like occasional deep discharge events: a short outage adds modest EFCs; repeated multi-day discharge stacks up faster. Manufacturer cycle-life claims for bidirectional hardware still matter more than rules-of-thumb from the grid literature.

Most manufacturer warranties guarantee ≥70% capacity at 8 years / 160,000 km. The 80% figure comes from second-life / grid-storage repurposing thresholds, not drivability. A 60 kWh pack at 80% SoH is a 48 kWh pack — equivalent to buying a base-model EV today. Range drops, but not catastrophically.

At year 10, buyers usually care whether remaining range still covers daily use more than whether the pack touched 80% SoH on a schedule. For many owners (commute under 100 km/day), NMC and LFP both clear that bar.

• Per-model variation. Model 3 vs Atto 3 within the same chemistry differs by ~±2 percentage points at year 10. Chemistry, climate, and DC share dominate.
• Catastrophic failures (recall-driven Bolt/Kona). These are recall risks, not fade curves.
• Calendar-only sitting in storage (e.g. cars parked at 100% SoC in 40°C). Avoid this and the curves above hold.
• BMS-driven step changes (some cars re-balance and report a sudden drop after a long-distance trip; this is measurement, not real loss).
• V2G/V2H beyond a single average kWh/day (DoD window, export power, tariff-shaped profiles, cell-to-cell imbalance).