Phase Space AI

Financial Model Notes

T1 Energy [TE]

T1 Energy [TE] — Financial Model Notes

Model version 1.0. First build. Every figure re-derived from SEC EDGAR XBRL and the filed statements. Units $m unless stated.

Two framing decisions govern this model. First, T1 Energy is energy manufacturing, not semiconductors — SEC SIC 3674 is a misclassification and the terminal margin is the parameter it corrupts. Second, the Section 45X credit sits inside cost of sales by stated policy and the company does not disclose the amount, so the product margin can only be bounded, never computed. Every product-margin figure below is a bound with its direction stated.

---

1. Revenue build

Quarterly RevenueFromContractWithCustomerExcludingAssessedTax, keyed on the fact's own period-end.

Q4-2025 is derived as FY2025 ($755.295m) less the three filed quarters ($396.741m). No quarterly duration fact is tagged for it.

TTM to 2026-03-31 = 132.767 + 210.522 + 358.554 + 177.647 = 879.490, matching the screen exactly. The screen's TTM arithmetic was correct.

FY2024 total net sales were $2.942m (essentially pre-revenue), of which $2.942m related party. FY2025 revenue grew $752.4m off that base — which is why the screen's op_margin_delta_pp of 2,654.4 is an artefact rather than a number.

---

2. Gross margin by quarter — and the Q4-2025 liquidation

Quarterly gross profit sums to the filed annual figure: 17.781 + 32.761 + 21.148 + (16.109) = 55.581. ✓ The Q4-2025 gross loss falls out of an identity, not an estimate.

FY2025 MD&A explains it: "In order to help preserve 45X on inventory prior to OBBBA effective date of January 1, 2026, we sold all inventory that would not be in compliance after December 31, 2025, into a weaker than expected market at the end of 2025. Realizations on these sales were lower than expected given the market was absorbing industry sales of modules with non-compliant PFE cells."

The largest revenue quarter in company history was a deadline-driven liquidation at a gross loss — and that −4.5% is already net of the 45X credit inside cost of sales.

---

3. Income statement, as filed

Three things a reader must not miss.

(a) Q1-2026 "net income from continuing operations" of +$3.902m is entirely non-cash mark-to-market. Warrant and derivative fair-value gains contributed +$30.368m against an operating loss of $(22.505)m. In FY2025 the same two lines ran the other way, −$39.579m. A screen or a reader taking the positive continuing-operations figure at face value has the sign of the business wrong.

(b) The loss from discontinued operations exceeds the loss from continuing operations in Q1-2026: $(24.321)m against $(22.505)m of operating loss. The legacy FREYR European business is still unsold.

(c) SG&A of $235.316m in FY2025 is 31.2% of revenue, up $156.1m (+197%). The MD&A attributes it to "commissions, royalty fees, personnel costs and legal and professional fees following the Trina Business Combination," and quantifies "Commissions, royalty fees, and other selling costs of $74.5 million" — 9.9% of revenue paid to the group that is also the customer.

---

4. Section 45X — the bound, and why it is only a bound

4.1 Stated policy

FY2025 10-K, Note 1: "Government Grants. The Company recognizes grants over the periods in which we recognize the related costs for which the grants are intended to compensate. For income-based grants, we recognize a receivable and a reduction to the related cost of activities that generated the benefit. For grants related to the purchase or construction of property, we reduce the carrying amount of the property and equipment recorded…"

So §45X reduces cost of sales. Same structure as First Solar and Enphase. Unlike both, T1 discloses no amount in the 10-K or the 10-Q.

4.2 The receivable, which is the only trace

Cash-flow-statement movement (a working-capital outflow — recognised but not collected):

The 2026-07-28 preliminary release confirms no 2025-vintage credit was monetised before Q2-2026: "During Q2 2026, T1 monetized the balance of the Company's remaining 2025 Section 45X tax credits for $39.1 million, at a gross price of $0.93 on the dollar, which was higher than previously announced 2025 sales." $39.1m ÷ 0.93 = ~$42.0m of face value sold in that quarter alone.

4.3 The bound, and its direction

Recognised 45X = movement in the receivable + amounts collected in cash. Since collections in the periods below were nil or unknown-but-non-negative, the receivable movement is a floor:

Direction of the bound: wherever any credit was collected in cash, the recognised amount is HIGHER and the product margin is WORSE. The bound is therefore safe in the direction that matters, and no estimate is substituted for the missing disclosure.

Q1-2026 product gross margin is negative while the company reported +16.4%. The subsidy is doing at least 23.3 percentage points of work at the operating line.

The 2026 vintage is unsold, sitting in the $77.801m receivable, with "early-stage negotiations" under way at a ~93-cent realisation.

4.4 The 45X-relevant coefficient

Q2-2026 preliminary: ~835 MW of module volume on $245–255m of revenue → implied ASP ~$0.299/W. The statutory §45X module credit is 7¢/W, so at 835 MW the module-level credit is roughly $58m per quarter — consistent in order of magnitude with the $41.4m of Q1-2026 receivable movement on lower volume, and it implies the credit is roughly 20–23% of revenue at current ASPs.

A credit worth ~20% of revenue, netted into cost of sales, unquantified in the filings, against a reported gross margin of 7.4% to 16.4%. That is the whole accounting problem in one sentence.

Once G2_Austin produces cells, the cell-level credit (4¢/W) is additive, taking the combined credit to 11¢/W — roughly 37% of the current ASP.

---

5. Balance sheet

Balance check: 1,027.871 + 72.505 + 236.702 = 1,337.078 = total assets. ✓

Cash fell 74.6% in one quarter, $182.450m → $46.367m, recovering to $79.1m only via the 2026-04-16 equity offering. Unrestricted cash of $79.1m is 7.9% of TTM revenue and 15.5% of the revised $510m G2_Austin Phase 1 capex.

5.1 Debt, by instrument (2026-03-31) — the table the screen read partially

Carrying value $378.406m, reconciling to the balance-sheet captions: convertible notes 153.381 + long-term debt 122.604 + long-term related-party debt 54.185 = 330.170 non-current carrying, plus 48.236 current. ✓

5.2 Net cash, all conventions

Screen error: net debt understated by $114.1m, from omitting the $48.236m current portion of debt and the $72.505m redeemable preferred (with a rounding reconciliation against the carrying-value basis).

A coincidence worth flagging so it is not mistaken for accuracy: the screen's EV of $1,444.2m is within 0.5% of the verified $1,436.8m — because the stock fell 10.5% between the screen date and the memo date, offsetting the $114m debt error. At the same spot the screen's EV would be $1,322.7m, understated by 8.6%.

---

6. Share count and dilution

+79.1% in twelve months.

Weighted average shares: FY2024 140,538k → FY2025 173,640k (+23.6%) → Q1-2026 basic 278,539k / diluted 285,252k. Q1-2025 basic was 155,933k, so Q1 year on year is +78.6%.

Revenue per weighted share, Q1: $0.343 (2025) → $0.638 (2026) = +86.1%. The operating ramp outran the issuance. Dilution did NOT invert the per-company view for T1 — the cluster hypothesis holds for FuelCell and fails here, and that should be recorded as a genuine positive.

Potential dilution ahead

Roughly 28% of potential dilution is already contracted, before any G2_Austin equity.

---

7. Discontinued operations — a live cash drain the screen ignores

The legacy FREYR businesses (European operations and Coweta County, Georgia) were classified as held for sale in 2024.

Two points for the record. The $2.0m per month is a contractual obligation under the Transaction Agreement ("Unless waived by the counterparty in writing, we expect to incur fees of $2.0 million per month until the business is disposed of"), and none of the $56.4m of liabilities appears in the screen's enterprise value.

And T1 has already had to repay a $20.0m government grant. In a cluster memo about subsidy dependence, that is the cleanest available demonstration that a government grant is a conditional liability, not income.

---

8. Related-party map

Because the counterparty is simultaneously customer, shareholder, lender, licensor and sales agent, the relationships are worth tabulating.

A module manufacturer's 25-to-30-year warranty is the product, and T1 has lost the counterparty that stood behind it while stating it "expects to establish a new warranty framework." That is a product-integrity risk, not just a legal one.

---

9. Integrity checks run

---

10. What this model does not contain

• No Section 45X point estimate. The company does not disclose it. Only a floor is derived, from the movement in the government-grants receivable, with the direction of the bound stated.
• No twelve-month price target. The own-multiple history is eight months long and is mechanically contaminated by a fifteen-fold TTM-revenue ramp. Declared UNIDENTIFIED. See TE_Valuation.md §3.
• No demonstrated revenue CAGR. Six quarters of module revenue from a partial base quarter, with a −50.5% sequential quarter inside it. The screen's null is the correct value and it is preserved.
• No semiconductor comparators. SIC 3674 is a misclassification; the comparator is PV module manufacturing.
• No backlog table. The company discloses none. Deferred revenue of $138.195m (current plus long-term) from the related party is the closest thing and is presented as such.
• No consensus estimates and no Street target. Alpha Vantage's shared quota was not spent; a Street target requires a paid feed.
• No full Q2-2026 financial statements. Only the 2026-07-28 preliminary release, labelled preliminary throughout.
• No KORE Power purchase price. Not disclosed in the preliminary release.
• No transcript mention counts. Substituted with filing-text counts, with the substitution declared, and the series is only two periods long because only two periodic filings cover the module business.