The EV Cell Denominator Is Shrinking Without Capacity Disappearing

Three US battery facilities originally announced for EV cell production have disclosed or signaled application migration toward stationary storage, data-center supply, or blended automotive/ESS use since late 2025. Combined original nameplate: approximately 184 GWh. None of it has vanished from capacity trackers. It has changed what it will produce, and for whom.

Anyone carrying "announced US battery capacity" as a proxy for future domestic EV cell supply is working with a number that overstates the EV-available share by a quantity that is growing. Cancellation subtracts from the total. Conversion leaves the total nominally intact while redirecting output to a different buyer, but the EV denominator contracts either way. Only cancellation shows up cleanly in aggregate attrition data.

E2 reports 216 cancelled, closed, or downsized clean energy projects since January 2025, representing $68.2 billion in foregone investment and approximately 9 GW of battery storage capacity affected. That figure captures the projects that die outright. Projects that survive but change purpose fall outside its scope. E2 tracks battery storage cancellations in GW, not battery manufacturing losses in GWh, a unit mismatch that makes their data useful as background weather but not as a facility-level denominator tool. For additional context, Clean Energy Associates reported that more than 20 GWh of planned US ESS cell manufacturing had been cancelled by Q2 2025, including KORE Power in Arizona (9.6 GWh/year) and FREYR in Georgia (10.2 GWh/year). Those are outright cancellations. The conversion pattern requires different evidence entirely, and the three cases below sit at materially different evidence states.

BlueOval SK Has the Cleanest Evidence File

Ford's original BlueOval SK commitment covered 129 GWh of annual capacity across three plants: two in Glendale, Kentucky (up to 43 GWh each, 86 GWh combined) and one in Stanton, Tennessee (43 GWh). All three were announced for EV battery production. The DOE project page used a lower aggregate phrasing of "more than 120 GWh" supporting "Ford and Lincoln EVs."

Ford's May 21, 2026 Form 8-K restructured the entire commitment. Ford Energy Battery LLC acquired all BlueOval SK interests in the two Kentucky plants. Ford assumed a $3.81 billion DOE note tied to the single Kentucky facility for which advances had been made; total advances under the BlueOval SK DOE loan were $7.84 billion. The loan exhibit defines "Product" broadly enough to include battery and battery energy storage products, systems, and services for both stationary storage and advanced-technology vehicle use.

Ford has not left the Kentucky application to inference. Ford Energy's public materials describe the Glendale facility as a dedicated LFP prismatic battery energy storage systems production site with 20 GWh annual production capacity, serving utilities, data centers, and large industrial and commercial customers. First customer deliveries are planned for late 2027. The product is a 5.45 MWh DC block in a standardized 20-foot container.

An 8-K filing, a DOE loan restructuring, a named product with stated capacity. The stated BESS nameplate is 20 GWh. The original Kentucky EV-linked nameplate was up to 86 GWh. That gap is significant and unresolved in the public record. Kentucky Facility 1 had obtained a certificate of occupancy; Kentucky Facility 2's building had been constructed. Whether the second facility will eventually produce additional BESS capacity, produce EV cells, or sit underutilized is not disclosed. What is disclosed: 86 GWh of originally-announced EV capacity now sits under Ford Energy ownership with publicly stated BESS orientation, and at least 20 GWh of that is directly described as non-EV nameplate.

SK retained the Tennessee plant. As reported by the Tennessee Lookout, the Tennessee facility is targeting automotive production plus ESS from 2028, with SK describing a blended application orientation rather than the pure EV purpose originally announced. The original 43 GWh Tennessee nameplate is therefore signaled as blended. No application-specific GWh split has been disclosed. No timeline evidence beyond the 2028 target is public.

Samsung StarPlus Disclosed Conversion Without Quantifying It

Samsung SDI and Stellantis announced the StarPlus Energy JV in Kokomo, Indiana in two phases. The first plant, announced May 2022, carried 23 GWh initial capacity expandable to 33 GWh. The second plant, announced October 2023, added 34 GWh for a combined 67 GWh. Both were announced for EV cell and module production.

Samsung SDI's March 2026 disclosure states that StarPlus Energy has been:

since Q4 2025, with LFP battery mass production targeted for Q4 2026 alongside existing high-nickel NCA output. The same disclosure announced a KRW 1.6 trillion LFP cathode supply agreement with L&F.

"Part" is the only scope language. No line count, no GWh figure, no yield or customer-acceptance evidence accompanies the disclosure. The denominator against which "part" is measured depends on whether conversion extends across the full JV or is confined to the first plant's lines: 33 GWh on the narrow reading, up to 67 GWh on the broad one. The public record does not resolve this. The direction is clear: capacity originally committed to Stellantis EV supply is being redirected to ESS, and the redirection has been underway for approximately nine months without a quantified scope entering the public domain.

Samsung SDI is Korea's sole prismatic cell maker, a structural fact relevant to the BESS market it is entering. But no StarPlus output, utilization, or yield data is public. The conversion is disclosed. Its scale is not.

SK Commerce Is Stressed and Signaling, Not Converted

SK Battery America's Commerce, Georgia facility carries 22 GWh of announced capacity on $2.6 billion of capital investment. The Georgia governor's office described it as enough to power approximately 300,000 EVs annually.

The WARN Act notice filed for Commerce tells a different story: 958 affected employees out of 2,566 total, with first separations on May 6, 2026. Manufacturing Dive reported the reductions brought headcount to approximately 1,600 and quoted SK Battery America saying it remained committed to Georgia while pursuing future customers including BESS. SK's own US operations page still describes Commerce as currently operating two EV battery plants, a statement that supports physical operating status but discloses nothing about utilization, output, or whether any production has shifted to ESS.

Commerce does not belong in the same evidentiary category as Ford Kentucky or StarPlus. No line conversion has been disclosed. No BESS product has been announced. No utilization or output data is public. The observable evidence is a 37% workforce reduction at a facility that has lost its primary customer base, paired with company language explicitly naming BESS as a target market. The facility is better described as stressed and signaling than as converted. But the signal points in the same direction as the other two, and 22 GWh of originally-announced EV capacity is operating at an undisclosed fraction of nameplate with its owner publicly pursuing non-EV customers.

Panasonic's De Soto, Kansas facility, covered in the adjacent facility-level audit, represents an additional data point in this pattern: operational and ramping EV cell production, but with its owner publicly describing partial conversion toward data-center battery applications.

Bounding the Migration

The three named facilities plus the Tennessee blended-orientation plant represent an original EV-linked capacity pool of approximately 184 GWh using StarPlus first-plant scope (86 + 33 + 22 + 43), or 218 GWh using the full StarPlus Kokomo JV scope. The evidence distributes unevenly across that pool:

The floor is 20 GWh. That is the single directly stated BESS figure, traceable to Ford Energy's own materials. The ceiling is 184 GWh, the sum of Ford Kentucky, StarPlus first-plant, Commerce, and Tennessee at full nameplate. The ceiling assumes every GWh at every facility with any migration signal has fully left the EV denominator, which overstates the case for Tennessee's blended orientation and Commerce's signaling-without-conversion. A midpoint of this range carries no analytical meaning because the facilities sit at five distinct evidence states and the tiers are not interchangeable. Even the floor, 20 GWh, represents capacity that was in the EV denominator eighteen months ago and is no longer there, while the aggregate US battery capacity figure it belonged to has not been revised downward.

What Aggregate Trackers Miss

Capacity that migrates rather than cancels remains invisible to every aggregate tracker I am aware of. E2 captures the projects that die outright. The application migration visible at facility level, through 8-K filings, DOE loan restructurings, and company product announcements, requires individual tracking that no aggregate dataset currently performs. That gap in coverage is itself a material data problem.

The EV cell supply denominator is smaller than the battery capacity numerator suggests. By how much depends on how many of these signaled conversions reach production, at what utilization, on what timeline. The 20 GWh floor is conservative and traceable. The true figure is almost certainly higher. And the gap between "US battery capacity" and "US EV cell supply" will continue widening for as long as BESS and data-center demand offer these facilities a commercially viable alternative to waiting for EV volume that has not materialized on the originally announced schedule.