Phase Space AI

Valuation

T1 Energy [TE]

T1 Energy [TE] — Valuation

Spot $3.72 (2026-07-29, −10.5% on the day after the preliminary Q2-2026 release) EV $1,436.8m · TTM revenue $996.7m (to 2026-06-30E) · EV/Sales 1.44x

Valued as energy manufacturing, not semiconductors. SEC SIC 3674 is a misclassification and the terminal margin is the parameter it corrupts. That correction is the substance of this document.

---

1. Fixed inputs, named

---

2. Implied-path test (Valuation Criteria) — FAIL

2.1 There is no demonstrated growth rate, so there is no demonstrated − required margin

criteria.md requires "the margin in percentage points: demonstrated − required. This is the number the strategy ranks on." For T1 Energy that number does not exist, and asserting one would be the defect rather than the output.

The reason: module revenue began after the 2024-12-23 Trina acquisition. The filed series is six quarters long and its first quarter is a partial period.

A series that runs +148%, +59%, +70%, −51%, +41% across six quarters, whose largest quarter was an inventory dump to beat a tax deadline and whose base quarter was a partial period, has no demonstrated CAGR. The screen recorded this correctly: cagr_base_revenue: null, revenue_cagr_demonstrated: null, growth: null, note "quality inputs missing: growth." That is the right answer and it is preserved here.

So the test is run in the other direction: what does the price require, and is that anywhere near what the business has shown?

2.2 What the price requires, at terminal margins a PV module assembler can actually earn

assets/reverse_dcf.py, --solve cagr, all §1 parameters fixed.

First, the comparator that matters. First Solar is the most cost-advantaged US module manufacturer in existence — vertically integrated, proprietary CdTe thin film, a diversified third-party customer base, and the largest 45X recipient in the industry. Its product operating margin, once §45X is removed, is −0.05% on a trailing-twelve-month basis. That is the honest ceiling for what commodity US module manufacturing earns without a subsidy. T1 Energy buys its cells, assembles PERC and TOPCon modules, and sells 99.86% of them to one related party.

Read the top row. At a terminal operating margin of zero — which is where the best US module manufacturer's product economics actually sit — the reverse DCF has no solution at any exit multiple in the solver's range. Terminal EBIT is zero, so no multiple produces value, and the price cannot be justified anywhere.

At a terminal margin of 2%, which is generous for a purchased-cell assembler with one related-party customer, the price requires a +50.6% compounded five-year revenue CAGR — revenue rising from $996.7m to $7.7bn by 2031. T1's own 2026 production target is 3.1–4.2 GW; at the Q2-2026 realised ASP of $0.299/W, $7.7bn of revenue implies roughly 26 GW of annual module shipments, more than five times the 5 GW nameplate capacity of G1_Dallas and G2_Austin Phase 1 combined.

Result: FAIL. Stated in the form the SMR precedent established:

At any terminal operating margin consistent with what commodity US PV module manufacturing actually earns without a subsidy, there is no solution in range — the price cannot be justified anywhere in the plausible range. Solutions appear only at terminal margins of 2% and above, and each requires a compounded five-year revenue CAGR of +35% to +99% from a company with no demonstrated growth rate at all and a product operating margin of at worst −36.0%.

2.3 The inversion: what terminal margin the price requires

--solve terminal_margin, all else fixed:

At flat revenue and a 15x exit, the price requires a 15.48% terminal operating margin. Set against what has been observed:

A 51.5-percentage-point margin swing, from the only product margin ever observed to the one the price requires. Even granting +30% compounded revenue growth for five years, the required 4.17% terminal operating margin sits 40 points above the observed product margin and above First Solar's product margin.

2.4 The multiple, on an EV/Sales basis

Today's EV/Sales is 1.44x. The EV/Sales formulation makes the price look far more defensible — at 1.5x exit the required CAGR is only +9.1% — and that is precisely why it must not be the primary output here. An EV/Sales exit multiple silently embeds a margin assumption. A commodity assembler with a negative product margin does not deserve 1.5x sales in perpetuity; First Solar trades at 3.6x on a positive reported margin and a −0.05% product margin. The EV/Sales route is reported for completeness and is not the test, because the question in dispute is the margin, and this instrument hides it.

2.5 What the screen's version produced, and why

terminal_margin: 0.091, basis "industry median of mature profitable peers (pre-profit subject)" — drawn from the TECH universe that sic: 3674 assigned. At 9.1% and a 15x exit the required CAGR is +11.2%, which against a null demonstrated CAGR produces quality: INDETERMINATE and no valuation verdict.

The chain of consequence is worth stating exactly, because it is a template for the whole class of error:

1. EDGAR assigns SIC 3674 to a solar module assembler.
2. The screen maps SIC 36 → sector: TECH.
3. The terminal margin is set to the TECH median of mature profitable peers: 9.1%.
4. That margin is 9.15 percentage points above First Solar's product operating margin and 51.5 points above T1's own.
5. At 9.1% the required CAGR is +11.2% — a number that looks achievable for a ramping manufacturer.
6. At any margin the industry actually earns, there is no solution at all.

One wrong classification field moved the answer from "unjustifiable at any multiple" to "requires 11% growth."

---

3. Twelve-month target — UNIDENTIFIED

valuation.md: "If the history is too short or spans a regime change, declare it UNIDENTIFIED rather than substituting a peer median." Both conditions hold, and the second is mechanical.

3.1 The history is eight months, and it is contaminated by construction

A point-in-time EV/Sales series requires four filed quarters of revenue. T1's fourth quarter of module revenue was filed on 2025-11-14, so the series begins there and runs to 2026-07-29: 175 trading days, roughly eight months.

This distribution is worthless as an anchor, and the reason is arithmetic rather than judgement. The multiple "collapsed" from a 5.47x median to 2.39x today because TTM revenue was still filling in from a near-zero base — revenue went from $53m in Q1-2025 to $996.7m on a trailing basis. The denominator grew roughly fifteen-fold in five quarters. A multiple series whose movement is dominated by denominator construction rather than by price measures nothing about valuation. Reading "1st percentile of its own history" as cheap would be the exact error the framework exists to prevent.

The twelve-month target is UNIDENTIFIED. No number is substituted, and no peer median is imported.

3.2 Why not First Solar's multiple

FSLR is the obvious comparator — same product, same 45X-in-COGS accounting, same OBBBA exposure. It is also vertically integrated with proprietary CdTe technology, has a diversified third-party customer base, and trades on a positive reported operating margin. T1 buys its cells, assembles commodity PERC/TOPCon, and sells 99.86% to one related party. Importing FSLR's 3.6x EV/Sales would embed FSLR's cost position and customer diversification into T1's valuation. That is the growth-matched-anchoring rule applied to competitive position, and the violation would be the same: extrapolating beyond the support of the sample.

3.3 What would make it identifiable

Two years of post-ramp trading history with a stable TTM revenue denominator, plus a positive product gross margin that gives EV/EBIT a defined sign. Neither exists within twelve months. The scenario set in §5 replaces the target.

---

4. Peer spread — INDETERMINATE, declared

Named peer: First Solar [FSLR]. Not tradeable as a spread — the two differ in vertical integration, technology ownership, customer diversification and the sign of the product margin. Declared UNIDENTIFIED rather than defaulted to a sector median. Recorded for the RV fork.

---

5. Downside Criteria — with the going-concern case argued

Type: MEASURED. Logged; blocks nothing.

Named cause: a single related-party offtake relationship is renegotiated, reduced or lost, and/or a PFE material-assistance determination disqualifies the 2026 §45X vintage — while an unfunded $510m plant and a $66.0m preferred redemption consume a balance sheet holding $79.1m of unrestricted cash.

5.1 The going-concern case — argued and flagged

criteria.md: "A going-concern bear case must be argued explicitly and flagged." Flagged.

The company's own assertion is narrow: "We believe that we have sufficient liquidity to meet our contractual obligations and commitments for at least the next 12 months." It then lists the conditions, and they are the case against:

• Unrestricted cash $79.1m against total debt principal of $404.508m, and cash fell 74.6% in the quarter to 2026-03-31 (from $182.450m to $46.367m) before an April equity raise.
• G2_Austin Phase 1 is unfunded at $510m, revised up 20% on 2026-07-28, with first cells pushed to Q1 2027. The company "continues to target a comprehensive financing solution, which includes a significant debt component." Targeting is not securing.
• $66.0m of preferred is mandatorily redeemable 2027-12-23 if unconverted. At $3.72 both series are deeply in the money at $1.70/$1.90 conversion prices and should convert, which removes the cash call and adds ~35.7m shares. If the price falls below the conversion prices, the cash call returns.
• The legacy European business is still unsold, accruing $2.0m per month in fees, with $35.2m already accrued and $56.397m of current liabilities of discontinued operations. Q1-2026 loss from discontinued operations was $(24.321)m — larger than the loss from continuing operations.
• $135m committed to the Evervolt IP purchase on 2026-07-28, payable in part in shares at a 15% discount to VWAP, alongside the July 2026 KORE Power closing. Two acquisitions in the month the plant's capex rose 20%.
• The 45X credit is the difference between a −6.9% and a +16.4% gross margin, and its 2026 eligibility rests on cell sourcing where "a portion" is certified non-PFE and "the remainder" is under diligence.
• A government grant has already been repaid — $20.0m on the Coweta County disposal (2025-02-15), with $7.0m more of repayable grants outstanding. Grants are conditional liabilities here, demonstrably.
• Accumulated deficit $1.113bn against $236.7m of total equity.

Conclusion: the going-concern risk is real and it is a financing risk, not a demand risk. The company must raise a large debt package into a plant that is late and over budget, while its only customer is a related party and its reported margin depends on a credit whose eligibility is under diligence. T1 Energy equity is a levered option on completing G2_Austin's financing.

5.2 Scenario set

No twelve-month target is issued (§3), so these are scenario anchors, not a target range.

The bear case is severe because the capital structure is severe: at 0.6x sales the enterprise value is roughly $540m against $404.5m of debt principal, $72.5m of preferred and $56.4m of discontinued-operations liabilities. The equity is thin under the debt, which is what makes the financing outcome binary rather than gradual.

5.3 Dilution — the one place T1 screens better than FuelCell

Revenue per weighted share, Q1: $0.343 (2025) → $0.638 (2026) = +86.1%, against weighted shares +78.6%. The operating ramp outran the issuance. Dilution did not invert the per-company view here, and the cluster hypothesis therefore holds for FCEL and fails for TE. It should be recorded as a genuine positive.

Ahead is different: ~35.7m shares from preferred conversion (+12.8%), up to ~42.7m from the Evervolt consideration at a 15% VWAP discount (+15.3%), plus listed warrants (TE WS), private warrants and the Anti-Dilution Right. Roughly 28% of potential dilution is already contracted, before any G2_Austin equity.

---

6. Model integrity notes

• net loss attributable to common ÷ weighted shares ≈ filed EPS: Q1-2026 $(21,409)k ÷ 278,539k = $(0.0769) vs filed $(0.08). ✓
• FY2025: $(380,789)k ÷ 173,640k = $(2.193) vs filed $(2.19). ✓
• Quarterly revenue sums to annual: FY2025 53.452 + 132.767 + 210.522 + 358.554 = 755.295, where Q4-2025 is derived as FY less the three filed quarters. ✓
• Quarterly gross profit sums to annual: 17.781 + 32.761 + 21.148 + (16.109) = 55.581. ✓ The derived Q4-2025 gross loss is the finding in §2.1 of TE_Research.md, and it falls out of an identity rather than an estimate.
• TTM to 2026-03-31 = 132.767 + 210.522 + 358.554 + 177.647 = 879.490, matching the screen exactly. ✓
• Balance sheet ties: total liabilities $1,027.871m + redeemable preferred $72.505m + total equity $236.702m = $1,337.078m = total assets. ✓
• Debt note ties to the balance sheet: total principal $404.508m less current $48.236m = non-current $330.170m, against balance-sheet captions of convertible notes $153.381m + long-term debt $122.604m + long-term related-party debt $54.185m = $330.170m carrying. ✓
• 45X is a derived floor, not an estimate. The lower bound is the cash-flow-statement movement in Government grants receivable, net — recognised-minus-collected. Wherever any credit was collected in cash, the true recognised amount is higher and the product margin is worse. The direction of the bound is therefore safe. The company does not disclose the amount and none is invented.
• Q2-2026 figures are preliminary, from the 2026-07-28 8-K, and are labelled as such throughout. The 10-Q has not been filed.
• The Form 25-NSE filed 2026-07-10 relates to warrants only. The common continues to trade on the NYSE and is priced daily. Not treated as a delisting.