Moved my builds to LFP. Here's what the …

I spent three years building NMC packs before I switched everything to LiFePO4. I want to give an honest technical breakdown because I see a lot of hype in both directions — people treating LFP as uniformly inferior because of energy density, and people treating it as magic because it's "safer." Neither framing is accurate.

The chemistry in plain English
Lithium Iron Phosphate (LiFePO4 / LFP) and Nickel Manganese Cobalt (NMC) are both lithium-ion chemistries, but the cathode material is fundamentally different. NMC uses a layered oxide structure — high energy density, but less thermally stable. LFP uses an olivine structure — lower energy density, but extremely stable even at elevated temperatures.

Where NMC wins
Energy density: NMC 811 cells hit roughly 220–320 Wh/kg depending on manufacturer and cell design; most quality commercial cells land in the 250–300 Wh/kg range. DIY-available LFP cells — standard EVE and CATL prismatic — are typically 120–165 Wh/kg, with CATL's latest-generation production cells approaching 200 Wh/kg. One thing comparison tables often miss: check volumetric energy density (Wh/L) too, not just gravimetric. LFP fares even worse there (~350–450 Wh/L vs. ~700 Wh/L for NMC). For a conversion where the battery compartment is a fixed size, volumetric density is often the more operationally meaningful number. If you're building an e-bike battery and you need 500Wh in a water bottle, NMC is your answer.

Where LFP wins, and it's not close
Cycle life is where LFP absolutely dominates — though both chemistries have improved since the early comparisons you'll see cited online. Modern NMC 622 cells achieve around 1,500 cycles to 80% capacity; NMC 811 reaches approximately 1,200 cycles. The old "500–1,000 cycle" figure for NMC is outdated. That said, LFP isn't close: the EVE LF280K is rated at 6,000 cycles on its current spec sheet, and REPT's Wending cells are rated at 10,000 cycles in manufacturer testing. Real-world DIY conditions yield less than lab spec, but the gap is enormous regardless. For a home storage battery cycling daily, LFP offers a decade-plus life advantage even after discounting the spec sheet numbers.

One underrated LFP advantage: you can charge to 100% routinely without meaningful degradation. NMC users who follow the standard guidance of limiting daily charging to 80% to preserve cycle life are effectively operating with a smaller usable pack. Factor this into any kWh-per-dollar comparison — LFP's real-world usable energy is higher than raw capacity numbers imply.

Thermal stability is the other big one. NMC onset temperatures for thermal runaway are typically 150–210°C depending on the variant — NMC 811 sits at the reactive end. LFP onset is more commonly reported at 220–270°C in the peer-reviewed literature; the 270°C number is near the top of the published range rather than the center. The safety advantage is real and substantial: NMC fires propagate roughly 9× faster than LFP, NMC peak runaway temperatures (~800°C) are dramatically higher than LFP (~620°C), and NMC cathodes release oxygen during runaway which feeds the combustion. That said — LFP cells do catch fire under severe enough abuse conditions. "Far less likely to go into runaway, far less severe when it does" is accurate. "Safe" is not. Don't treat the relative advantage as a reason to skip proper fusing, venting, and installation practices.

The voltage curve problem with LFP — and why cheap BMS hardware makes it worse
One genuine weakness, and it's more significant than most write-ups let on. LFP has an extremely flat voltage discharge curve. Between 20% and 80% state of charge, the cell voltage barely moves (~3.2–3.3V). Voltage-based SoC estimation is essentially useless in this range — your BMS must rely on coulomb counting, which accumulates drift errors over time. Cheap BMS units compound this: some have current sensing thresholds set to 1–3A, meaning small loads simply aren't tracked at all, and low voltage resolution can leave fewer than 10 measurable steps across 60% of the SoC range. Industry data shows SoC estimation errors of ±15% are common with budget BMS hardware on LFP; errors exceeding ±25–30% happen regularly when configuration parameters aren't tuned for the actual cell capacity. Your BMS choice on an LFP build is not an afterthought — it's the most critical component in the pack.

Cold weather — the LFP limitation nobody puts in the headline
LFP loses 10–20% capacity at 0°C and approximately 40% at -20°C. NMC retains roughly 70–80% of capacity at -20°C. LFP also throttles DC fast-charging heavily until the pack warms up — a major practical issue for EV conversions in cold climates. If you're building storage for outdoor installation in a northern climate, or an EV conversion for winter driving in a cold state, this is a real consideration that favors NMC or requires planning for pack heating. An indoor or well-insulated installation substantially reduces the problem.

My current recommendation
For home storage, EV conversion projects, and anything stationary in moderate climates: LFP. For high-performance compact applications (racing, aviation, e-bikes where every gram matters) or cold-climate installations where NMC's superior low-temperature retention matters: NMC still makes sense. On price: Grade A LFP cells (EVE, CATL, REPT) are now available at roughly $65–85/kWh from reputable importers; China wholesale has dropped well below $60/kWh. If you're seeing prices in the $90–110/kWh range today, shop around — that was accurate 2–3 years ago but the market has moved.

What chemistry are you running, and what's driven your choice? Especially curious to hear from people who've pushed either chemistry hard in real-world conditions.

Running a 20kWh LFP pack (32 EVE 200Ah cells in 16S2P — 51.2V nominal, 400Ah) in my garage for solar storage. Hit 18 months and just did a capacity test — sitting at 99.1% of original capacity. I know that's not the thousands of cycles where LFP supposedly shines, but the early data is encouraging.

The flat voltage curve thing is absolutely real and bit me hard at first. My first BMS (a cheap Daly) was showing 50% SoC when I was actually at 22% and triggered a low-voltage cutoff while I was using the pack. Swapped to an Overkill Solar BMS and calibrated it properly — no issues since.

The flat voltage curve issue compounds with age. As cells degrade, the plateau at 50% SoC becomes even flatter — what little voltage signal existed for SoC estimation disappears entirely. A Daly that was off by ±15% new will be off by ±25% at 500 cycles. Proper coulomb counting with a calibrated BMS is non-optional on LFP, not a nice-to-have.

Great writeup. One thing I'd add on the thermal stability front: it's not just about the catastrophic runaway event. In day-to-day use, NMC cells generate significantly more heat during high-rate charging and discharging than LFP. For a large stationary pack, that heat has to go somewhere — either you design a serious thermal management system or you accept that you're stressing the chemistry.

For a 20–30kWh home storage system, the thermal management requirements for NMC add real cost and complexity. LFP sidesteps most of this. It's one of those "hidden" advantages that doesn't show up in spec sheets.

The heat generation differential is real and the number most builders miss is the peak cell temperature, not the pack average. In a well-balanced 20kWh NMC pack at 2C discharge, peak cell temp runs 8–12°C above average. That delta is what your cooling system must handle — not the average readout on your BMS dashboard.

Converted a Beetle using salvaged Chevy Volt modules (NMC) and an older Honda Insight using new LFP cells. The Volt pack is compact and powerful — fits nicely in the rear floor well. The LFP pack in the Insight is bigger but I have more confidence parking it in the garage. Both have been running well, but I sleep easier with the Insight.

Question for the battery nerds: at what point does the energy density advantage of NMC actually matter for an EV conversion? My Beetle weighs maybe 180 lbs more than it would with an equivalent LFP pack. Is that meaningful for a ~60 mph neighborhood EV?

Running BMW i3 33kWh cells on my 1979 Beetle build — the volumetric density point is what sold me on NMC for a vehicle conversion. My rear floor pan has about 15L of usable battery space. LFP would need a larger enclosure I simply don't have room for. For a stationary garage pack I'd go LFP every time, but in the car the density wins.

The Beetle weight calculation deserves more precision, because the physics regime matters a lot for how large the penalty actually is.

At 60 mph on flat terrain, the dominant losses are aerodynamic drag and rolling resistance. Rolling resistance scales linearly with mass — so yes, 9% more mass gives you 9% more rolling resistance loss. But aerodynamic drag is mass-independent, and a Beetle's aero profile (Cd roughly 0.48 for the original) means drag is a disproportionately large fraction of total losses compared to a modern vehicle. When you blend the drag and rolling resistance contributions at 60 mph in a Beetle, the net range penalty from 180 lbs of extra mass is closer to 4–6%, not 8–12%, because the aero component dilutes the mass-related losses.

Where the mass penalty actually lands harder is in city duty — frequent acceleration from rest. F=ma is unambiguous there, and the motor has to work meaningfully harder on every launch. If the Beetle is doing pure neighborhood EV use with lots of stops, the LFP weight penalty is more like 7–10% in energy per mile than the highway estimate suggests. For a mixed-use profile the truth is somewhere between those bounds.

The practical conclusion doesn't change — "noticeable but not a dealbreaker" is right. And I'd still choose LFP for this application on garage safety alone, independent of the weight math. But the number is worth being precise about, because conversion builders regularly use range estimates from highway cycle data to make decisions about cars that mostly do city driving.

@garageconv The answer depends more on the vehicle weight than most people realize. In a large vehicle (3,000+ lbs), 180 lbs extra is 5–6% of total mass — genuinely not meaningful. But a Beetle conversion is a different situation: your finished weight is probably 2,200–2,600 lbs, which means 180 lbs is 7–9% of total vehicle mass. At those fractions, physics starts to show up more noticeably — 0–30 mph probably 0.35–0.45 seconds slower, and the range penalty closer to 8–12% rather than 3–4%, because rolling resistance (the dominant loss at 60 mph) scales directly with mass. For a neighborhood EV where you're not chasing lap times, this is still perfectly livable. But I'd frame it as "noticeable but not a dealbreaker" rather than "not meaningful."

The charge-to-100% point was the thing that finally moved me. I was running NMC at 80% limit religiously and always felt like I owned a 25kWh pack instead of a 32kWh pack. Three LFP builds in now — the flat voltage curve is an annoyance for BMS calibration but it's a manageable one, and getting to actually use the capacity I paid for makes up for it easily.

Coming at this from a solar install background: the industry has shifted decisively toward LFP for residential storage over the past 3–4 years — it's been a gradual transition, not a sudden one. Powerwall 3 is LFP, Enphase IQ Battery is LFP, and most Chinese manufacturers have moved there. LFP now accounts for the large majority of new residential installations. NMC products like the LG ESS Home 8 are still actively sold by some installers, so "impossible to buy" overstates it — but when I price a new residential job today, LFP is the default and NMC is the exception. The cycle life and thermal safety argument won, and the price gap closing sealed it.

Forty years in electrical engineering. I watched the same shift happen with lead acid to VRLA in the 90s — the industry consensus moved when the cycle data made it unavoidable, not when the marketing landed. Your point about Powerwall 3 adopting LFP is the real tell. Tesla uses whatever chemistry its own data supports. That's a louder signal than any spec sheet.

Moved my builds to LFP. Here's what the cycle data actually showed.