US Battery Capacity Is Migrating From EV to ESS. The Aggregate Numbers Hide It.

LGES posted a KRW 207.8 billion operating loss in Q1 2026 and simultaneously called its North American ESS production network "now in place." Samsung SDI has been converting StarPlus Energy lines in Kokomo from EV to ESS since Q4 2025. Neither company has closed a facility. Both have changed what those facilities produce, for whom, and in what chemistry. The GWh attached to these plants still appears in aggregate US capacity figures, even as the market those GWh serve has fundamentally changed.

Anyone pulling a US battery capacity number for a sourcing recommendation or market-entry assessment needs to understand that the figure they're citing now contains two distinct failure modes. The first is familiar: announced capacity that was never built. KORE Power's Buckeye plant, FREYR's Georgia greenfield. Projects that existed in press releases and conditional DOE commitments and nowhere else. Trackers handle these with a lag, but they handle them.

The second failure mode is harder to catch. Built capacity, with equipment installed and workers on payroll, has been silently migrating from automotive to stationary storage. The lines are operational. The GWh count hasn't changed. But the cell format, the chemistry, the customer qualification pathway, and the credit eligibility profile may all be different. A tracker that counts a converted line as "US battery capacity" is technically accurate. A procurement lead at an automaker who treats that line as available EV supply is working with a number that describes something that no longer exists. Announced ≠ built, and built capacity can silently change its market identity. Both gaps matter.

LGES's Five-Facility Network Claim and What Sits Underneath It

LGES's January 2026 annual results laid out the conversion logic explicitly: reallocate capacity between EV and ESS to reduce new investment and improve idle-line utilization. The company reported a 140 GWh ESS order backlog at year-end 2025, set a target of more than 90 GWh in new ESS orders for 2026, and said it would build more than 60 GWh of global ESS capacity with more than 80% in North America.

By the Q1 earnings release in April, the language had shifted from plan to assertion. LGES named five facilities as its North American ESS production network: Holland, Michigan; Lansing, Michigan; Windsor, Ontario; Ultium Cells Tennessee; and L-H Battery Company Ohio. The near-term target was revised to more than 50 GWh of North American ESS production capacity by year-end 2026. ESS represented a mid-20% share of Q1 revenue.

What sits underneath that network claim is thinner than the headline suggests.

Holland is LGES's longest-standing US operation and has been producing LFP ESS cells. The company has not disclosed Holland-specific production volumes or utilization rates. It is the one facility in the network where ESS production predates the conversion strategy, making it the baseline rather than evidence of conversion.

Lansing is the facility designated for Tesla Megapack 3 supply. The strongest public timeline anchor appeared in June when a DOI release described a $4.3 billion LFP prismatic cell supply agreement with Tesla, with production launching in 2027. That is a supply-agreement announcement and a timeline target. It is not evidence of commissioning, equipment qualification, or yield as of late June 2026.

Windsor has completed the most legible identity migration of any facility in the network. NextStar Energy, the former Stellantis JV, announced June 25 that it had started production on a new battery pack line to support ESS demand, adding pack manufacturing to cell and module operations that began commercial production in November 2025. NextStar described itself in that release as a wholly owned LGES subsidiary. The Stellantis JV language from January is gone. This is a facility that has moved from OEM joint venture producing EV cells to wholly owned subsidiary producing ESS packs, reaching one million cells by February 2026. The GWh capacity at Windsor has not changed. Everything else about it has.

Windsor is in Canada and is not eligible for Section 45X manufacturing credits. LGES's own network framing blurs this distinction: the Vertech DTE contract explicitly says cells will come from "Michigan and other U.S. and Canadian facilities," treating the cross-border network as a single supply base. For anyone mapping LGES's capacity against US production-credit economics, the aggregate number includes GWh that cannot generate US credits. That disappears in most capacity counts.

The two JV facilities present the murkiest picture. LGES's January release described "temporary ESS use" of Stellantis JV and Honda JV production lines. The Q1 release listed Ultium Tennessee and L-H Ohio as part of the ESS network without the qualifier "temporary." Whether this reflects a change in plan or a change in disclosure language is not determinable from public sources. LGES has not published line-level production data for either facility, and neither Ultium nor L-H has independently disclosed ESS output volumes. The word "temporary" quietly disappeared between two quarterly filings, and the underlying operational reality may not have changed at all.

The customer-side evidence is more concrete than the production-side evidence. LGES Vertech signed a 6 GWh agreement with DTE Energy in May 2026 for eight Michigan storage projects, with delivery over two years. LGES has separately disclosed additional named ESS contracts beyond DTE, though line-level production sourcing for those contracts is not publicly mapped. The demand pipeline is real and named. What remains unverifiable from public sources is how much of the production capacity serving that pipeline was originally designated, equipped, and announced as EV supply.

StarPlus Energy's Line Conversion and the L&F Cathode Signal

Samsung SDI's conversion at StarPlus Energy in Kokomo follows a similar trajectory with one notable addition. Samsung disclosed in March 2026 that StarPlus had been gradually converting EV lines to ESS since Q4 2025, with LFP mass production alongside existing high-nickel NCA expected by Q4 2026. StarPlus is a Stellantis JV, making it part of the same pattern of OEM-affiliated joint ventures pivoting away from their original automotive mandate.

In the same March release, Samsung disclosed two large unnamed US ESS supply contracts: approximately KRW 2 trillion ($1.30 billion) signed in late 2025 and approximately KRW 1.5 trillion ($0.98 billion) announced March 16, 2026. Neither counterparty was named. Both figures come from Samsung's own disclosure in a single release; independent confirmation of the late-2025 contract through a separate filing has not surfaced.

Physical evidence of the conversion at StarPlus remains limited to Samsung's own characterization. Indiana's WARN notice table, reviewed through June 18, 2026, shows no StarPlus Energy or Samsung SDI entries. The absence of mass-layoff filings suggests the conversion has not involved large-scale workforce reductions, but it does not confirm production readiness, equipment installation status, or customer qualification milestones.

The L&F cathode deal is the detail that has drawn the most external attention. Samsung signed a mid- to long-term supply agreement with L&F, the Korean cathode producer, worth approximately KRW 1.6 trillion (~$1.04 billion) over three years starting in 2027, with an option to extend for another three. L&F is building 60,000 tons per year of LFP cathode capacity. Samsung explicitly tied the deal to reduced China dependence and US supply-chain pressure related to prohibited foreign entities.

The deal tells you Samsung SDI is spending real money to build an alternative cathode pathway, and that the company views PFE exposure as a commercial risk worth approximately $1 billion over three years to mitigate. But the distance between that signal and actual credit eligibility needs to be stated precisely.

The 45X credit requires that eligible components not include material assistance from a prohibited foreign entity, and Notice 2026-15 requires taxpayers to determine direct material costs and PFE direct material costs for constituent materials. The MACR safe harbor tables that would simplify compliance calculations remain unpublished, with a deadline of late 2026.

The gap between "secured non-Chinese cathode" and "eligible for production credits across all constituent materials" is real. No public evidence closes it.

Where Capacity Actually Vanished

Against the conversion pattern, KORE Power and FREYR represent capacity that genuinely vanished from the pipeline.

KORE Power received a conditional commitment for up to $850 million from DOE's Loan Programs Office in June 2023 for the KOREPlex in Buckeye, Arizona: 1.33 million square feet, NMC and LFP cells, 6 GWh initial capacity expandable to 12 GWh. Canary Media reported in February 2025 that KORE confirmed cancellation, that the DOE loan never closed before the administration change, and that the company planned to sell the Buckeye property and search for an existing factory to retrofit. As of late June 2026, no public DOE source confirms formal withdrawal, reassignment, or conversion of the conditional commitment. No public source confirms KORE has identified or acquired a retrofit site. The $850 million sits in bureaucratic suspension, which is itself a data point about how DOE tracks its own pipeline.

FREYR's Georgia plant followed a cleaner exit. AP reported in February 2025 that FREYR told Newnan officials it would not build the planned $2.6 billion facility. Construction never began on the 368-acre site. No equipment was ordered and no production workforce was hired. Nothing physical existed to convert. Georgia had provided a $7 million grant for site acquisition, and the state and company were working on repayment. This is the analytically simple case: a project that never progressed past land acquisition leaves nothing behind to reclassify.

What the Conversion Pattern Means for Sourcing

The Korean conversions are analytically interesting precisely because they resist the binary framework that most capacity tracking imposes. The lines stayed open, kept their workers, and shifted to a different market entirely.

Three things follow.

Aggregate US battery capacity figures now require an end-market qualifier to be useful. A GWh number without an EV/ESS split, and ideally a chemistry breakdown, tells you less than it did eighteen months ago. LGES alone is targeting 50+ GWh of North American ESS capacity by year-end 2026, drawn substantially from lines that were announced for EV production. Samsung SDI has disclosed approximately $2.3 billion in ESS supply contracts with unnamed US counterparties, plus a separate ~$1.0 billion cathode supply agreement with L&F, and is converting StarPlus lines to serve the ESS pipeline. On the LGES side, a rough bracket is possible: the company's 50+ GWh North American ESS target, with 80%+ of 60+ GWh global ESS capacity in North America, implies 48–50+ GWh in the region, though how much of that represents converted EV capacity versus purpose-built ESS (Holland) is not disclosed at the line level. Samsung SDI has not published a GWh conversion target for StarPlus, making the total across both companies impossible to bracket precisely. What is clear is that the aggregate capacity number can look roughly stable while the composition shifts dramatically underneath.

The conversion reflects EV offtake weakness as much as ESS demand strength. LGES posted a Q1 operating loss and attributed it partly to ramp-up costs at ESS sites and weak North American pouch-type EV battery sales. All three Korean cell makers posted Q1 2026 operating losses. None disclosed Q1 GWh shipments. Utilization is not directly calculable from public disclosures, but triangulating from reported revenue, disclosed pricing trends, and nameplate capacity across the three producers suggests a range of approximately 47–50%, a figure consistent with the scale of conversion activity observed. EV offtake has not materialized at the volumes their JV partners originally contracted, and idle lines are more expensive than repurposed ones. The ESS order books are real, but they exist in the context of automotive programs that underperformed their contracted volumes.

The credit eligibility landscape for converted capacity is almost entirely opaque. LGES discloses aggregate North American production incentives in earnings (an estimated KRW 189.8 billion in Q1 2026) but does not break this out by facility or by EV versus ESS application. Samsung SDI has disclosed nothing about StarPlus's 45X posture beyond the L&F cathode deal. For converted facilities producing ESS cells, the relevant credit is 45X (manufacturing) rather than 30D (clean vehicle), and PFE restrictions under Notice 2026-15 apply. The MACR methodology requires tracking PFE-sourced constituent material costs against total direct material costs, with thresholds that tighten over time. Whether the Korean producers have navigated that transition at the component level across anode, electrolyte, separator, and current collectors is not publicly knowable. No facility-level 45X claims are disclosed in any public registry, because no such registry exists.

The capacity remains on the books. Its market, chemistry, customer, and credit profile have all changed. If you're still counting it the same way you counted it when these lines were announced for GM and Stellantis EV programs, you're measuring a market that no longer exists in the form your number describes.