SK On reported KRW 821.8 billion in Q2 operating profit against 36.4% H1 utilization. Those two numbers cannot coexist unless non-manufacturing income dominated the quarter. The H1 DART filing adds facility-level capacity data and a utilization methodology that blocks the GWh conversion most readers will attempt, but it does not resolve the profit-layer opacity flagged in our initial Q2 decomposition. The balance sheet is better after the BlueOval dissolution. Whether SK On made or lost money producing and selling cells remains undeterminable from public disclosure.
Decomposition table
Q1 2026 revenue was restated from KRW 1.791T to KRW 1.870T after SK changed how it presents US production credits. Credits that affect customer price or transaction terms now sit partly inside revenue rather than entirely in other operating income, so Q2 and restated Q1 revenue carry an AMPC component that the Q3 and Q4 2025 figures do not. Q3 and Q4 were not restated. That breaks direct revenue comparability across the trailing series: the apparent KRW 1.076T Q1-to-Q2 revenue increase is partly a presentation artifact sitting on top of real volume and price movement, and the two cannot be separated from what is disclosed.
Q2 2026 average exchange rate: approximately KRW 1,380/USD.
| Q3 2025 | Q4 2025 | Q1 2026 | Q2 2026 | |
|---|---|---|---|---|
| Revenue | KRW 1.808T | KRW 1.457T | KRW 1.870T¹ | KRW 2.946T (~$2.13B) |
| Operating profit/(loss) | (KRW 124.8B) | (KRW 441.4B) | (KRW 349.2B) | KRW 821.8B (~$595M) |
| AMPC (disclosed) | KRW 173.1B | KRW 101.3B | Not disclosed | Named as increasing; amount not disclosed |
| Customer compensation | — | — | — | Named; amount, counterparty, trigger undisclosed |
| BlueOval exit effect | — | — | — | Not separately identified in battery OP |
| Recurring mfg result | ~(KRW 298B)² | ~(KRW 543B)² | Not isolable | Not isolable |
| ASP direction | —³ | —³ | —³ | —³ |
| EV vs. ESS mix | —⁴ | —⁴ | —⁴ | —⁴ |
| Utilization | — | 48.7% (FY 2025) | — | 36.4% (H1 2026) |
¹Restated; originally KRW 1.791T. ²Approximate: operating loss minus disclosed AMPC. Assumes AMPC was the only non-manufacturing contributor, which SK's disclosure does not confirm. ³SK On does not disclose ASP per kWh; the revenue restatement further prevents inference from revenue over volume. ⁴SK On does not disclose an EV vs. ESS revenue or volume split in reviewed public materials.
Three of four profit layers — AMPC, customer compensation, and BlueOval exit effects — are named but not quantified in Q2. The recurring manufacturing result cannot be isolated from Q1 2026 onward.
45X isolation
SK disclosed quarterly AMPC through Q4 2025: KRW 173.1 billion in Q3, KRW 101.3 billion in Q4. From Q1 2026 the quarterly figure is absent from reviewed public materials. The disclosure stopped one quarter before the profit quarter, not alongside it, which rules out the simplest reading. I cannot determine whether the change reflects IR formatting or a deliberate reduction in granularity. The consequence is the same: the 45X contribution to Q2's KRW 821.8 billion cannot be isolated.
The last measurable AMPC movement was a Q3-to-Q4 decline of KRW 71.8 billion. Q2 materials describe AMPC as increasing, without a figure. At Q3 2025 levels — roughly KRW 170 billion — it would account for about 21% of Q2 operating profit. Under the restated presentation methodology it would run higher, and the share with it. Neither bound is confirmable.
SK's Q2 materials name three contributors to the KRW 1.171 trillion swing from Q1 loss to Q2 profit: increased AMPC, a one-time customer compensation payment, and higher Asian sales volume. None is quantified separately. A fourth candidate, effects from the BlueOval SK dissolution that closed May 20, does not appear in the battery segment result at all.
The customer compensation is where the measured and the inferred need separating. SK identified a one-time payment and disclosed nothing about counterparty, trigger, or amount. Ford's Q2 10-Q separately records $0.5 billion for settlement of pre-existing claims related to the BlueOval dissolution, without naming the recipient. The two disclosures are temporally coincident with the May 20 closing. Neither filing confirms the connection, and I am not asserting one. At prevailing rates $0.5 billion is roughly KRW 690 billion, which would account for most of Q2 operating profit if SK On were the counterparty. That inference requires verification against a document neither company has published. What holds without inference: a payment large enough to move a KRW 1.171 trillion quarterly swing is a solvency event, and if it is dissolution-related, it also records the end of a customer relationship.
The dissolution itself does not surface as a separate line in battery segment operating profit. The H1 DART filing shows SK transferred KRW 5.939 trillion of BlueOval PP&E, already impaired by KRW 3.467 trillion, together with KRW 5.378 trillion of long-term borrowings, to Ford. SK's Q2 presentation puts forward structural savings at approximately KRW 300 billion per year in depreciation and KRW 200 billion per year in interest. Those are future cost reductions. They are not disclosed components of Q2 profit.
For Q3 and Q4 2025, subtracting disclosed AMPC from operating loss gives approximate recurring manufacturing losses of KRW 298 billion and KRW 543 billion. The trajectory was worsening through the last quarters where the arithmetic was possible. From Q1 2026 the calculation breaks, and nothing in Q2's reported profit constitutes evidence that the underlying manufacturing result reversed.
Utilization trend
H1 2026 utilization is 36.4%, down from 48.7% for full-year 2025. The rate of decline is not decelerating.
Both forces are operating at once. On the demand side, SK consolidated Georgia EV production from two plants into one in April 2026, which is not a move a company makes when orders fill both buildings. On the denominator side, nameplate capacity went from 71.5 GWh at year-end 2024 to 94.3 GWh at mid-2026, a 32% increase, driven by Hungary Plant 3 (ramping, at 67% of planned capacity) and Yancheng Plant 3 (at 8%). The disclosure does not allow disaggregating demand-driven decline from denominator-driven decline. The Georgia consolidation is direct evidence of insufficient EV orders; the ramping plants are adding capacity into a market that is not absorbing what already exists.
The utilization figure resists the conversion most readers will reach for. Multiplying 94.3 GWh of nameplate by 36.4% to estimate H1 production produces a methodologically wrong number. SK defines utilization as load hours divided by operating hours, and reports production in cells — 89.475 million in H1 — not in GWh. The filing supplies neither energy content per cell nor the hour-level data. A GWh output estimate is not derivable from public disclosure, and any figure circulating on that basis was constructed, not read.
The 94.3 GWh capacity stock remains useful for geographic and eligibility mapping:
| Region | GWh/yr | Notes |
|---|---|---|
| Korea (Seosan) | 4.7 | FEOC-compliant; not 45X-eligible |
| Hungary (Plants 1–3) | 37.8 | Plant 3 at 67% of planned capacity; FEOC-compliant; not 45X-eligible |
| China (Yancheng 1–3) | 29.75 | Plant 3 at 8% of planned; FEOC-ineligible (covered country) |
| US — Georgia (Plants 1–2) | 22.0 | FEOC-compliant and 45X-eligible |
| Total | 94.3 |
The table excludes SK On Tennessee, which SK retained at 100% ownership after the dissolution. Original project documents assigned 45 GWh to Tennessee. Its absence from the capacity stock means it is not producing. When it produces depends on customer commitments that have not been disclosed. At nameplate it would add 45 GWh of FEOC-compliant, 45X-eligible capacity, but that is a project figure carried over from a JV that no longer exists.
Guidance versus actuals
SK On's Q1 presentation published no numerical Q2 targets for revenue, operating profit, utilization, or ASP. Forward language ran to improving long-term profitability through European sales growth and North American ESS expansion. Two of the three contributors to the Q2 result — customer compensation and increased AMPC — were not part of that outlook in any quantified form. A guidance-versus-actuals comparison is therefore not possible. This is the second consecutive quarter without numeric forward guidance, which leaves this pass with one structural observation: SK On is not publishing the markers that would let an outside party track execution against plan.
ESS pivot — physical gate crossed, commercial gate not
SK On's ESS position rests on three disclosed elements: a 1 GWh firm supply commitment to Flatiron Energy beginning H2 2026, a right of first offer covering up to 6.2 GWh of additional Flatiron projects through 2030, and a 20 GWh global ESS order target presented at CLEANPOWER 2026. The ROFO is not contracted backlog. The 20 GWh target has no disclosed pipeline behind it.
On the manufacturing side, SK consolidated Georgia EV production into Plant 2 in April 2026 and is converting the freed capacity to LFP for ESS. Trade reporting describes construction of an LFP line targeting October 2026 mass production. Line count, converted GWh, and remaining EV assignments on Plant 2 are undisclosed.
An LFP line under construction is committed capital and committed engineering time, expensive to walk back. That physical gate has been crossed. The commercial side has not kept pace: customer acceptance of LFP cells from the Georgia line, sustained production data, and yield figures are all absent from public disclosure, and contracted volume stands at 1 GWh against a 20 GWh aspiration. A procurement team looking at SK On for ESS supply is qualifying against a line that has not yet produced at scale, from a supplier whose capacity is available partly because its EV order book contracted. For a buyer able to carry early-qualification risk, that is a genuine opening. It also means the availability and the credit risk have the same cause.
Sourcing decision frame
The dissolution improved SK On's balance sheet by moving more debt than net asset value off it, and the forward savings of roughly KRW 500 billion per year in depreciation and interest will lift the operating loss floor. Manufacturing capability is intact: 94.3 GWh across four countries, 22 GWh of it FEOC-compliant and 45X-eligible in Georgia, plus a 45 GWh Tennessee facility retained but idle. What contracted is customer attachment. The offering circular names Hyundai, Volkswagen, and Ford as Georgia customers; the dissolution removed Ford as a JV partner, the compensation payment may mark further volume loss, and the consolidation into one EV plant reflects an order book that did not fill two. Hyundai is the anchor and increasingly the only one visible. Two readings of this quarter are consistent with the disclosed data: solvency improvement plus freed capacity as a window to negotiate with a motivated supplier, or 36.4% utilization as the leading edge of a durability problem over a multi-year qualification horizon. The evidence that would separate them is the recurring manufacturing result, which SK On disclosed through Q4 2025 and no longer does.
This analysis relies on SK Innovation's English-language IR materials, the Korean-language H1 DART filing, Ford's SEC filings, and English-language trade press. No primary Korean-language earnings call transcript was reviewed.
- Tennessee facility customer commitments: SK On retained 100% ownership of the Tennessee plant after the BlueOval dissolution, but its 45 GWh project capacity is absent from the H1 capacity stock, meaning no customer or production timeline has been disclosed.
- Georgia LFP line acceptance: Trade reporting targets an October 2026 mass-production start for the converted LFP ESS line, which would be the first observable test of whether SK On's ESS pivot crosses from physical construction to customer-accepted output.
- Ford settlement recipient: Ford's Q2 10-Q records $0.5 billion for settlement of pre-existing claims tied to the BlueOval dissolution without naming the counterparty, and SK On's Q3 disclosure may clarify whether this corresponds to the unquantified customer compensation.
- SNE Research H1 share erosion: SNE estimated SK On's non-China xEV battery usage at 18.9 GWh in H1 2026, down 6.7% YoY, with share falling from 9.5% to 7.0% — a demand trajectory that the ESS pivot has not yet visibly offset.