Battery Cost per kWh: What Drives the Number

Quoting a single battery cost per kWh is close to meaningless without four qualifiers: which segment, which chemistry, which region, and whether the number is a cell, a pack or an installed system. The spread across those four is larger than the year-on-year change that usually makes the headlines.

The current numbers

BloombergNEF's annual price survey, published in December 2025, put the volume-weighted global average pack price at 108 US dollars per kWh for 2025, an 8 per cent fall in real terms on 2024 and a record low.

Three of those rows deserve attention.

Battery electric vehicle packs ticked up two dollars in nominal terms, from 97 in 2024 to 99 in 2025, which BloombergNEF attributes to higher metal prices reaching the automotive segment first. In real terms the figure was still slightly down, but it is the first time in years that the headline EV number has not fallen outright.

Stationary storage became the cheapest segment for the first time in 2025, having fallen 45 per cent in a single year. That reordering matters commercially. Grid storage historically paid a premium relative to automotive because volumes were smaller and requirements were bespoke. It now benefits from exactly the opposite: LFP cells produced at enormous scale, standardised large-format products, and a buyer base that does not pay for energy density.

The regional spread is wider than the annual global movement. A pack costing 84 dollars per kWh in China costs meaningfully more in Europe or North America, which reflects local production costs, dependence on imports and the tariff environment. Anyone benchmarking a European project against a global average number is benchmarking against the wrong number.

What the price is actually made of

Cost modelling for cells is a well-developed field, and Duffner and colleagues surveyed the main approaches. The components fall into a few blocks.

Materials are the largest single block for a mature, well-utilised plant, and their share moves with metal prices and with chemistry. Cathode active material dominates within that block, commonly a third or more of total cell cost for nickel-based chemistries, followed by anode material, electrolyte, separator, current collector foils and housing.

Depreciation on equipment and buildings, spread across whatever volume the plant actually produces. This is why utilisation matters so much: the same capital cost spread across half the volume doubles per-unit depreciation.

Yield loss. A cell scrapped after formation carries the full cost of everything already done to it, including its materials. During a plant ramp this is a large number, and the gap between a plant at mature yield and one still climbing exceeds any plausible difference in procurement terms.

Energy, dominated by drying and by the dry room, and labour, which varies by region and by degree of automation.

Overheads, warranty provision and margin, which in a market with severe overcapacity have been compressed to the point where several manufacturers have been selling at thin or negative margins.

Then pack integration adds module hardware, the battery management system, wiring, contactors and fusing, thermal management, the enclosure and assembly labour. Cell-to-pack designs exist largely to shrink that increment.

Why 2026 broke the usual pattern

For most of the last decade the story was straightforward: metal prices fell or stayed moderate, scale increased, and pack prices fell. Late 2025 into 2026 separated those two threads.

Battery-grade lithium carbonate bottomed at around 8,000 US dollars per tonne in mid-2025 and then rebounded sharply. Chinese spot prices traded between roughly 150,000 and 181,500 yuan per tonne (about 21,000 to 25,500 dollars) through the first half of 2026, according to Shanghai Metals Market, with the peak in January. The International Energy Agency's Global EV Outlook 2026 noted that lithium prices at the start of 2026 stood at more than double their level a year earlier, while remaining roughly 70 per cent below the 2022 peak. Cobalt was affected separately by export quotas introduced by the Democratic Republic of the Congo, and price reporting agencies recorded broader increases across several battery inputs.

Pack prices did not move up in step. The industry absorbed the increase through long-term supply contracts, hedging, continued migration to LFP, and margin compression under overcapacity. China alone held an estimated 557 GWh of annual cell capacity aimed at stationary storage, roughly double global demand in that segment. BloombergNEF's near-term outlook at the time of the survey still expected a further decline in 2026, to just under 105 dollars per kWh, about 3 per cent.

Whether that holds is the open question of the year. Long-term contracts reprice. Hedges expire. Overcapacity eventually consolidates. A forecast built on the assumption that input costs never pass through is a forecast with a specific and identifiable failure mode, and anyone building a business case in 2026 should test it against a scenario where they do.

The point people miss about system cost

At 70 dollars per kWh for a stationary storage pack, the battery has stopped being the dominant line item in an installed grid storage project.

An installed system adds the power conversion system, transformers and switchgear, the enclosure and its thermal and fire systems, site engineering, procurement and construction, grid connection, land, permitting and financing costs. As the cell and pack line falls, everything else becomes proportionally more important, and the levers that improve project economics move away from cell procurement towards interconnection queues, construction efficiency, cost of capital and revenue stacking.

The same shift is visible in vehicles, where the pack is no longer the single overwhelming cost driver it was in 2015, and where powertrain integration, software and manufacturing scale increasingly determine the delivered price.

This is a general pattern worth internalising: when one component of a system falls in cost by an order of magnitude, the interesting engineering and commercial problems move somewhere else.

Using cost numbers without embarrassing yourself

State the boundary. Cell, pack or installed system, and for installed systems say whether grid connection and land are inside or outside.

State the segment and chemistry. An LFP stationary pack and an NMC automotive pack differ by more than 50 dollars per kWh, which is larger than several years of average decline.

State the region and the year, and say whether the figure is nominal or real.

Distinguish price from cost. Survey figures are transaction prices, and in a period of overcapacity and thin margins they can sit below the production cost of some manufacturers. Prices that reflect a price war are not evidence of what a plant can sustainably produce for, and treating them as a cost floor for a business case is a mistake that has caught several new entrants.

Check the date. Prices in this sector move enough that a figure eighteen months old belongs in a history chart rather than in a model.

Informational and educational content only. Not professional, financial, legal, or engineering advice.