Editorial Article
An EV Line Became a Storage Line in Five Months
The cheapest gigawatt-hour in this market is a conversion
In July 2026 Ultium Cells — the General Motors and LG Energy Solution joint venture — began producing LFP batteries for stationary storage at Spring Hill, Tennessee. It did not build a battery plant. It converted an EV battery line, in under five months, for USD 70 million, and it supplies the output through LG Energy Solution’s North American system-integration subsidiary Vertech, meeting the “U.S.-made” content requirements of the Inflation Reduction Act.
Set that beside what a new plant looks like in the same market. Zenergy signed a 50 GWh long-duration storage manufacturing plant in Suzhou in March 2026: CNY 5.2 billion, roughly USD 720 million, phase one production in Q3 2027 and phase two in 2028. Peak Energy is spending USD 71 million on a 4 GWh sodium-ion factory in Sacramento, with shipments from Q1 2027. iNVERGY opened a 3 GWh-a-year plant in Dasna, Uttar Pradesh, in June 2026 — 217,000 square feet, roughly USD 25 million, producing LFP packs, complete BESS units and solar inverters.
Which means this market’s supply response has a shape that does not appear anywhere in its forecast: the marginal gigawatt-hour is increasingly an existing gigawatt-hour that used to be destined for a vehicle.
Capex per annual gigawatt-hour, and what the spread actually measures
Divide announced capital by announced capacity and the table looks like a league table of efficiency:
| Project | Capacity | Announced investment | USD m per annual GWh | What is being counted |
|---|---|---|---|---|
| iNVERGY, Dasna, India | 3 GWh/yr | ~USD 25m | 8.33 | Packs, complete BESS units, inverters |
| Zenergy, Suzhou, China | 50 GWh | ~USD 720m | 14.40 | 588 Ah long-duration cells |
| Peak Energy, Sacramento, USA | 4 GWh | USD 71m | 17.75 | Sodium-ion cells |
| Ultium Cells, Spring Hill, USA | not disclosed | USD 70m | not computable | A converted EV line |
The range is 2.13 times — Peak is 2.13 times iNVERGY per annual gigawatt-hour, and 1.23 times Zenergy. It would be a mistake to read that as an efficiency ranking. iNVERGY’s USD 8.33 million buys pack assembly and system integration, not cell manufacturing; Zenergy’s USD 14.40 million buys cells. The spread is not telling you who is good at building factories. It is telling you that “announced capacity” and “announced investment” are counted against different scopes in almost every press release, and that a capex-per-GWh comparison without the scope is meaningless.
What the Ultium number does tell you, precisely because it is a conversion and not a build, is the price of not building: USD 70 million and five months, against a greenfield that was signed in March 2026 and reaches first output around eighteen months later. The cheap capacity is cheap because it buys no new building, no new cell design and no new permits. It buys a line that already exists.
Twelve thousand cycles is not a vehicle specification
Here is the part of the conversion that the capex line hides. The products this market now specifies are storage products: cells above 300 Ah, at least 12,000 cycles at 80 per cent discharge, durations of four hours and beyond; Zenergy’s headline part is a 588 Ah cell rated for four to twelve hours of storage duration. Those are duty cycles written in full cycles. A traction cell installed in a car is warranted in years and kilometres, and its chemistry, electrode design and formation protocol were optimised for energy per kilogram and cost per kilowatt-hour, not for a ten-thousand-cycle life at a fixed state of charge.
So a converted line delivers volume on a schedule and inherits a qualification problem on a different schedule. The capex clears in five months. The requalification — cycle-life evidence at the storage duty cycle, safety testing at the system level, and a utility or an EPC willing to underwrite a warranty on a cell that was designed for something else — does not. That is where a storage project’s real lead time sits, and it is invisible in every table of announced capacity.
Why the timing forces this route
The forecast is front-loaded: USD 64.54 billion in 2025 to USD 247.68 billion by 2033, an 18.31 per cent baseline CAGR, with the strongest annual gains between 2028 and 2030. Capacity that is meant to serve that peak has to be ordered before the peak, which is now. A greenfield signed in 2026 is a 2027 or 2028 asset. A conversion is a 2026 asset.
Which produces the conclusion I would defend hardest: this forecast’s supply side is a claim about how fast the car industry’s lines can be reassigned, and the forecast does not contain that variable. The 18.31 per cent is a demand curve with an assumed supply curve behind it. If every incremental gigawatt-hour has to come from a greenfield, the peak slips. If it can come from a converted vehicle line, it does not. The market’s growth rate is therefore hostage to a decision made inside a different industry, by firms whose primary business is losing money per vehicle.
The sodium exception, which is the one that is not renting
Peak Energy is the only project in this set choosing a different cathode chemistry, and it is also the most expensive per annual gigawatt-hour at USD 17.75 million — 1.23 times Zenergy, 2.13 times iNVERGY. It is also the only one holding contracted demand in advance: more than 6 GWh of customer commitments through 2030 against 4 GWh of factory capacity, coverage of 1.5 times, including 4.75 GWh from Jupiter Power — 79.17 per cent of the announced book — and 1.5 GWh from Energy Vault.
Its capital is 9.86 per cent of Zenergy’s for 8 per cent of the capacity. That is not an accident of geography; it is the premium for not inheriting anyone’s cell design. The sodium route buys a chemistry chosen for stationary duty cycles — no cobalt, no nickel, no lithium price exposure — and it pays for that in years and in dollars per gigawatt-hour. Whether that premium is worth 1.23 times the LFP alternative in a market where the LFP cell is getting cheaper is the single most interesting open question on this market’s supply side, and nothing in the announced data resolves it.
The end of the chain is scaling from almost nothing
One more datapoint, because it calibrates how early this still is. Livium completed commissioning of a battery recycling plant at its Envirostream facility in Campbellfield, Victoria, in July 2026, funded by Sell & Parker under an April 2026 variation to an existing agreement. In July 2026 Envirostream expected to receive 55 tonnes of material from Sell & Parker, against a historical monthly average of about 7 tonnes between May 2025 and May 2026 — a step-up of 7.86 times in a single month. The contract carries anticipated value above USD 5 million.
That is not the same kind of number as the market’s growth, and the mismatch is the finding. The market takes eight years to expand 3.84 times in value. A single supply contract moved one recycler’s intake nearly eight times in one month — because it started from seven tonnes. The recycling layer of this industry is not growing; it is being switched on. Recycled content obligations arrive from 2031, five years after the first converted line started shipping cells that will one day need somewhere to go.
Three falsifiers
One. An ESS-only greenfield in North America or Europe announcing and reaching first output in under twelve months. That would break the claim that conversion is the only fast capacity, and with it the idea that the forecast’s supply curve is a reassignment rather than a construction programme.
Two. A publicly disclosed cycle-life failure at the storage duty cycle on a converted vehicle line. The cheap capacity would become expensive at exactly the moment the warranty is tested, and the 12,000-cycle specification would turn out to be the barrier rather than the qualification.
Three. Peak Energy shipping 4 GWh on schedule in Q1 2027 from a greenfield at 17.75 million dollars per annual gigawatt-hour. If sodium-borne capacity can arrive that early and at that capital intensity, the rental route has a competitor, and the premium was mispriced.
What this does not claim
No gigawatt-hour figure is attributed to Spring Hill, because none was disclosed — which is why the conversion has no capex per gigawatt-hour and why it is excluded from the comparison rather than estimated. This article does not re-argue the levelised cost reversal or the power-semiconductor bottleneck; that is the companion article’s ground. It takes no position on which chemistry wins. And it makes no estimate of what share of the USD 247.68 billion in 2033 is served by reassigned rather than new capacity — that number does not exist in any disclosed source, and inventing it would be the easiest thing in this market to do and the least defensible.
Frequently Asked Questions
Is a converted EV battery line really cheaper than building a new storage factory?
No. Ultium Cells converted an existing EV battery line at Spring Hill, Tennessee in under five months for USD 70 million and began producing LFP cells for stationary storage in July 2026. Zenergy's 50 GWh greenfield in Suzhou was signed in March 2026, cost roughly USD 720 million, and reaches phase one production in Q3 2027. A conversion buys no new building and no new cell design, which is why it can arrive in one year rather than two.
Does a lower capex per gigawatt-hour mean a more efficient plant?
No, and the range is a warning rather than a ranking. iNVERGY's USD 8.33 million per annual GWh buys packs, complete BESS units and inverters; Zenergy's USD 14.40 million buys 588 Ah cells; Peak Energy's USD 17.75 million buys sodium-ion cells. A 2.13 times spread across those scopes is measuring what each announcement counted, not who builds efficiently.
Why does 12,000 cycles matter if the line is already running?
Because a line is qualified for a duty cycle, and the duty cycles differ. The storage market specifies cells above 300 Ah with at least 12,000 cycles at 80 per cent discharge and durations of four hours and beyond. A traction cell is warranted in years and kilometres. The capex clears in five months; cycle-life qualification at the storage duty cycle, and a utility willing to underwrite a warranty on the back of it, does not.
Why does the conversion route matter to an 18.31 per cent market forecast?
Because the forecast is front-loaded, with the strongest annual gains between 2028 and 2030, so capacity intended to serve that peak has to be ordered now. A greenfield signed in 2026 is a 2027 or 2028 asset; a conversion is a 2026 asset. The supply curve behind the 18.31 per cent is therefore partly a claim about how quickly vehicle lines can be reassigned, which is a variable the forecast does not model.