First Solar

Refresh delta (since 2026-06-20)

What changed — six material moves

1. A federal module price floor landed, and it is above First Solar's own ASP. On 2026-08-06 the President signed a Section 232 proclamation imposing a minimum import price of $0.38/W on solar modules ($0.22/W cells, $100/kg ingots and wafers, $21/kg polysilicon) plus a 15% polysilicon tariff, effective for goods entered 2026-12-04 . First Solar's US gross bookings since the April call were struck at **~$0.36/W** . This is the single most consequential change in the file: the bear case's strongest leg — "US oversupply crushes ASPs" — now runs into a statutory floor set above where First Solar is currently contracting.

2. The backlog shrank 5 GW in six months. Contracted backlog fell to 45.1 GW / $13.6B at 2026-06-30 from 50.1 GW / $15.0B at 2025-12-31 . The technology-adjuster slice fell to **21.6 GW with up to $0.5B of upside** from 23.2 GW / up to $0.6B. Q2 gross bookings were only 1.9 GW US + 1.1 GW India (H1) against 3.7 GW sold — **First Solar is shipping the book faster than it is refilling it**, by choice ("we continue to prioritize pricing, contract quality... over short-term booking volume" — Widmar ).

3. Revenue went negative year-on-year for the first time in the file. Q2 net sales $1,056.2M, −3.7% YoY (vs $1,097.2M) — the decline is a clean comp effect (prior-year contract-termination revenue), but the *headline* is a shrinking top line, and H1 tells the ASP story: sales +8.2% on volume **+16.8%**, i.e. **realised price per watt down ~7% YoY**.

4. The 45X credit stopped converting to cash. Government grants receivable ballooned to $1,263.3M from $625.2M in six months — a −$649.3M drag on operating cash flow . Cause: **no 2026 credit-sale agreement has been signed**; the last was October 2025. H1 cash from the credit was $95.2M (residual on the Oct-2025 sale) + $117.6M direct from Treasury (on *2024* credits) = **$212.8M against $862.5M recognised in P&L**. FY26 guidance explicitly assumes **"No 2026 Section 45X tax credit sales"** . First Solar also stopped factoring receivables ($492.8M factored in FY2025, zero in H1 2026) ``. Both are cash-quality changes the P&L hides.

5. A securities class action and a derivative suit now name the company and its officers. Filed EDNY 2026-06-23 (Claire Day v. First Solar; class period 2025-02-26 → 2026-02-24) alleging §10(b)/§20(a) and Rule 10b-5 violations — that management "understated the extent to which our responses to U.S. tariff policy were likely to negatively impact projected performance in 2026." A shareholder derivative suit followed 2026-07-28 (Manh Ho) on substantially similar allegations, adding breach of fiduciary duty, gross mismanagement and waste ``. The prior dossier stated "no disclosed material securities or accounting litigation against the company." That is no longer true.

6. The international fleet is a stranded-asset problem management now names out loud. ~1.8 GW of fully-finished international capacity (Malaysia/Vietnam) is idle pending Section 232 / FEOC clarity, burning ~$30M per quarter in underutilization costs . FY26 underutilization is guided at **$115–135M** and the 10-Q added new impairment language: geopolitical developments "may result in future decisions to **reduce, pause, or cease operations** at these facilities," with PP&E "sold or otherwise disposed of before the end of their previously estimated useful lives" . The class action alleges these plants ran at ~20% utilization ``.

Two structural positives worth naming alongside: First Solar is now effectively debt-free — long-term debt $0 (from $282.6M), total debt $37.6M, debt-to-equity 0.00x, after fully prepaying the India DFC facility ($328.2M) in May and repaying $672.4M in H1; the revolver was refinanced in February into a $1.5B senior unsecured five-year facility, undrawn . And the data-center demand thesis got a name: **Steel River Energy Center** (Cypress Creek / Google, Mississippi County, Arkansas) broke ground 2026-07-15 — 1.6 GWdc + 1.9 GWh in phases 1–2, 2.5 GWdc / 2.9 GWh at full build by 2029, on First Solar modules already inside the backlog .

What held — does the structural thesis stand?

Yes, and it hardened in both directions. The prior dossier's central claim — "you are not buying a module company, you are buying a leveraged, policy-dated bet" — is no longer an inference from a 10-KForm 10-KA company’s audited annual report to the US regulator. The most complete thing it publishes. footnote. It is now the company's own guidance line item:

The Q2 actuals tie out to the same place: gross profit $605.0M less the $444.5M of income-related government grants booked into Cost of Sales less the $88.6M IEEPA tariff-refund benefit = $71.9M, a 6.8% pre-subsidy gross margin ``. The 45X credit again exceeded net income — $444.5M vs $422.6M in the quarter, $862.5M vs $769.2M in H1. The 57.3% headline gross margin is a fiscal transfer with a factory attached.

What changed about the thesis is the second leg. Before 2026-08-06 the policy support was one instrument (a manufacturing credit that phases down from 2030) and the bear case had a clean kill shot. After 2026-08-06 there are two instruments pointing the same way — a credit on First Solar's cost line and a price floor under its competitors' revenue line, and the second has no announced sunset. Balanced against that: the backlog is contracting, the cash conversion has broken down, and there is now a live securities action alleging the company mis-told this exact story.

Competitive Advantages — REFRESHED (moat structure changed)

The prior dossier called the moat "policy, not product," and warned it was durable "only as long as the policy persists." Two things happened.

(a) The policy moat gained a price floor. Section 232's MIP schedule ($0.38/W modules, $0.22/W cells, effective 2026-12-04) does something 45X never did: it constrains what *imported* competition can charge. 45X subsidised First Solar's cost; the MIP raises rivals' floor. On the call, an analyst framed the question as "the potential for floor prices in the $0.40 per watt or higher range" — management did not push back on the premise. Adjacent: the FCC issued a ruling on Chinese solar inverters days before the print, which Widmar read as "our U.S. government trying to ensure that we don't have any overreliance on adversarial countries" ``.

(b) The IP moat became an offensive weapon at scale — and it is now a dated catalyst. First Solar has escalated its TOPCon patent campaign into a USITC Section 337 investigation (No. 337-TA-1494), instituted 2026-03-25, naming Axitec, Canadian Solar, JA Solar, JinkoSolar, Mundra, Philadelphia Solar, Hanwha QCells, Runergy, Trina Solar and VSUN, with BYD America, Tesla and IC Star Solar intervening as respondents. First Solar seeks a general exclusion order barring importation of infringing TOPCon products by any foreign entity. Hearing 2027-02-16; preliminary ruling May 2027; final ruling September 2027. The parallel Delaware suits against Canadian Solar and Trina are stayed pending the ITC outcome ``. A general exclusion order on TOPCon would be a larger structural event for US module pricing than 45X. It is also the reason the stock trades on legal-docket dates.

What did not change: the efficiency deficit is real and unaddressed on the current product; CuRe is now in high-volume manufacturing at Perrysburg and "exceeding expectations" in field data, but the perovskite Series 6 pilot line only reaches operational readiness in H1 2027 `` — i.e. still a 2027 story, still behind tandem timelines.

Newly visible weakness in the moat: the domestic-content advantage now strands the international fleet. Approximately 41 of the 45.1 GW backlog carries some form of domestic-content requirement ``, and management's stated allocation order is US integrated fleet first, South Carolina finishing second, international last. The overseas plants that were a cost advantage in 2019 are a liability in 2026.

Segments / mix — REFRESHED

Still a single operating segment (CdTe modules); the CODM measures gross profit and consolidated net income ``. The meaningful disaggregation is now geography of production and price realisation, and it has bifurcated sharply:

That is a ~1.8× price gap between the two books, and it is the mechanical explanation for the H1 ASP decline: "a lower sales price per watt associated with the higher volume of modules sold in India" ``. India is a volume-filler at roughly half the US price. Any mix shift toward India compresses realised ASP regardless of what US pricing does.

Utilization, which is now the segment story: global fleet 86%, US 98%, India 86% . Malaysia and Vietnam are conspicuously **not broken out** — and the call explains why: ~1.8 GW of finished international capacity sits idle pending Section 232 / FEOC outcomes. Underutilization cost was **$29M in Q2** (vs $15M) and **$55M in H1** (vs $35M), guided to **$115–135M** for FY26 .

Revenue trajectory:

• FY2023 $3,318.6M → FY2024 $4,206.3M → FY2025 $5,219.4M ``
• Q1-2026 $1,044.2M (+23.6%) → Q2-2026 $1,056.2M (−3.7%) → FY2026 guided $4.9–5.2B, i.e. roughly flat to down 6% on FY2025 ``. The growth company stopped growing this year, on guidance.

---

Earnings Result — Q2 FY2026 (reported 2026-07-30) — REFRESHED

A large EPS beat on a revenue miss, with a one-off doing a lot of the work.

``

vs consensus: EPS $3.92 vs $3.01 estimate — a ~30% beat; revenue "in line to a narrow miss" depending on the compiler ``.

Read the beat honestly. Of the +11.7 pp gross-margin expansion, management attributes the largest single piece to "an $88.6 million net benefit related to expected IEEPA tariff refunds less estimated amounts payable to customers" . That is ~$0.82/share pre-tax — i.e. roughly the entire beat versus consensus. The CFO flagged it is an estimate: it "reflects our current estimate of expected recoveries... and remains subject to refinement" . Origin: the **US Supreme Court ruled the IEEPA tariffs unlawful on 2026-02-20**; the Court of International Trade ordered CBP to refund; First Solar filed claims in Q2 and began receiving payment . Non-recurring, estimated, and reversible.

Guidance — reaffirmed, unchanged, and already once-cut. FY2026 (reaffirmed 2026-07-30, first given 2026-02-24) ``:

Guidance also states, under "INCLUDES": Section 122 tariffs at 10% through 2026-07-24 (150 days); Section 301 tariffs starting 2026-07-24 — Malaysia and India 10%, Vietnam 12.5%; estimated IEEPA tariff recoveries and commercial obligations; and "No 2026 Section 45X tax credit sales."

Q3'26 preview: volume sold 3.9–4.5 GW (US 3.2–3.7 GW), adjusted EBITDA $625–775M.

Context the prior dossier missed: this guidance was not a neutral bar. When first issued on 2026-02-24 it sent the stock down ~15%, because $4.9–5.2B was far below the ~$6.16B consensus at the time ``. "Reaffirmed" means holding a number that already disappointed, not defending an ambitious one.

Balance-sheet flags — three, in order of importance.

1. Government grants receivable $1,263.3M (current $284.5M + noncurrent $978.8M) vs $625.2M at 2025-12-31 ``. The noncurrent tranche carries at $978.8M against a disclosed fair value of $953.4M — the company's own fair-value note prices a discount on collection timing. The 45X credit is being earned far faster than it is being received.
2. Cash fell $1,125M in six months — cash + marketable + restricted $1,741M vs $2,866M . Uses: $672.4M debt repayment, $279.8M capex, and the working-capital build. Net cash of **$1.7B is already at the bottom of the $1.7–2.3B year-end guide**, and management describes its long-term target range as **$1.5–2.0B** . There is not much slack left.
3. Working capital outran sales. AR $1,476.2M (+$182.2M) on falling quarterly revenue; inventories (current + noncurrent) $1,273.6M vs $974.2M (+$299.4M) — while factoring was stopped entirely (zero receivables factored in H1 2026 vs $492.8M in FY2025) . H1 operating cash flow **−$359.8M**; H1 free cash flow **−$639.6M** .

Offsetting, and genuinely good: long-term debt $0, total debt $37.6M, equity $10,320.3M, debt-to-equity 0.00x, $1.5B undrawn unsecured revolver, effective tax rate still only 5.5% for H1 (up from 3.1% as Pillar Two global-minimum-tax bites) ``.

Market reaction / what is priced in. Shares rose ~2.5% on 2026-07-31 and rallied further into early August, then gave back — $238.82 on 2026-08-10 (−4.49% intraday), against $250.84 on 2026-08-07 and $257.70 at the prior dive on 2026-06-20 . Market cap **$25.67B**, trailing P/E **15.4x**, forward P/E **11.97x**, 52-week range **$176.47–$320.95** . Net of a 30% EPS beat and a favourable tariff proclamation, the stock is down 7.3% since the last dossier. That is the tell: the market is discounting the quality of the print, not the size of it.

Unusual vs own history: 57.3% gross margin is an all-time peak and is composed of a $444.5M subsidy plus an $88.6M estimated tariff refund on $1.06B of sales. Do not carry it forward.

Earnings Calls — sentiment shift — REFRESHED

Transcript now on disk (transcripts/2026-q2.md, ingested 2026-08-10 from MarketBeat, full presentation + Q&A). Caveat for anyone quoting it: the scrape's speech-to-text renders gigawatts as "MW" in several places ("45.1 MW of contract backlog", "1.8 MW", "3.5 MW", "41 MW"). The 10-Q and the deck confirm all of these are GW. Use the filing, not the transcript, for units.

The arc across four calls:

• Mid-2025 (post-OBBBA): relief. 45X retained through 2032, tail risk removed, tone pivoted to capacity execution ``.
• Q4-2025 / 2026-02-24: the break. Guidance well below consensus, stock −15%, tariff impact guided at $125–135M, permitting and policy uncertainty cited ``.
• Q1-2026 (2026-04-30): "record start," margin-expansion narrative, R&D stepped up for CuRe/perovskite; guidance held but explicitly not raised ``.
• Q2-2026 (2026-07-30): confident on execution, openly hedged on policy — the two halves no longer match.

What is new in the language. The upbeat register survives on manufacturing and technology: "record second quarter and first-half sales volume," "surpassed 100 GW of cumulative module sales," CuRe "exceeding expectations," "our fully integrated domestic manufacturing fleet remains substantially committed through 2028" ``. But the commercial register turned conditional in a way it was not in April:

• Bookings are now explicitly deprioritised: "we continue to prioritize pricing, contract quality, appropriate risk allocation, and long-term value over short-term booking volume." That is a defensible strategy and a pre-emptive explanation for a shrinking backlog.
• The international fleet is now openly in limbo: "production planning and utilization levels in Malaysia and Vietnam continued to be influenced by U.S. market demand drivers and economics, including the pending Section 232 polysilicon and derivative investigation... We expect greater policy clarity will help inform the long-term operating profile for the approximately 1.8 [GW] of fully finished international capacity." Nobody says "long-term operating profile" about a plant they intend to run.
• A schedule slipped: South Carolina Phase II moved to mid-2027, framed as enabling earlier CuRe incorporation. Phase I remains on track for H2 2026 ``.
• A quiet write-off inside R&D: the R&D increase reflects perovskite investment "and the impairment of certain R&D equipment that is no longer expected to be used as part of our technology roadmap" — Alex Bradley ``. Unquantified.
• Cost pressure that did not exist in April: "in some cases, domestic freight costs are now approaching international shipping economics," plus "non-recoverable domestic trade expenses above our previously assumed forecast."

What they started saying: Section 232, FEOC, "policy clarity," "long-term operating profile," hyperscalers by name. What they stopped saying: anything resembling an ASP-upside narrative for 2026 — management now explicitly flags "limited ASP upside from Q.O.R sales in 2026," pushing the technology-adjuster benefit into 2027–2028.

Net sentiment read: down a notch versus Q1, despite better numbers. The confidence is concentrated in the things First Solar controls (fabs, technology, backlog quality) and has visibly drained from the things it does not (tariffs, FEOC, the fate of two overseas plants, when the credit turns into cash).

Comps — REFRESHED (lightly; unsourced multiples still refused)

EV/Sales, EV/EBIT and 5-yr ROE remain not sourced to a citable figure and are left blank rather than invented — unchanged from the prior dive.

The comp problem got worse, not better. At $238.82 and 11.97x forward, FSLR screens cheap. But the forward "E" now carries the company's own disclosure that $2.10–2.19B of $2.4–2.6B guided gross profit is a tax credit. Applying an 12x multiple to that earnings stream is not a valuation, it is a bet on a statute. On a pre-credit basis the same guidance implies ~7% gross margins and roughly break-even operating income before the credit — a number no module multiple applies to.

Catalysts and ±5% moves — REFRESHED (since 2026-06-20)

The prior dossier's finding — "it is a policy stock, not an earnings stock" — was reconfirmed, twice, inside seven weeks.

Earlier in 2026, pre-boundary but load-bearing for the pattern: −10% on 2026-01-07 (Jefferies downgrade citing de-bookings and margin compression) and −15% on 2026-02-26 (FY26 guidance ~20% below consensus) ``.

Dated catalysts now on the calendar — the most concrete set this name has had:

1. 2026-12-04 — Section 232 minimum import prices take effect ($0.38/W modules). The single largest ASP variable.
2. Q4 2026 / FY26 close — whether First Solar signs any 2026 45X credit sale, or takes the ~$1.3B receivable into 2027. Tests the net-cash guide directly.
3. Pending — final FEOC regulations.
4. 2026-08-24 — lead-plaintiff deadline in the securities class action; consolidation and an amended complaint follow.
5. 2027-02-16 / May 2027 / Sept 2027 — USITC 337-TA-1494 hearing, preliminary ruling, final ruling on a general exclusion order against TOPCon imports.
6. H2 2026 / mid-2027 — South Carolina Phase I production start; Phase II completion (slipped).
7. H1 2027 — perovskite Series 6 pilot line operational readiness.
8. Ongoing — a decision on the ~1.8 GW of idle Malaysian/Vietnamese capacity, and whether it produces an impairment charge.

---

Management — REFRESHED (governance signals, no leadership change)

No management change. Mark Widmar remains CEO, Alex Bradley CFO, Jason Dymbort General Counsel ``. The prior dossier's read on archetype and capital-allocation discipline carries.

Three new governance data points, all from inside the delta window:

1. The 2026 annual meeting (2026-05-13) showed real dissent ``. Of 92.1M shares represented:

• Paul H. Stebbins: 17,745,590 against — ~22% of votes cast against, by far the highest.
• Michael Sweeney: 11,038,235 against (~14%). Michael J. Ahearn (founder/chair): 6,066,472 against.
• Say-on-pay: 10,350,148 against (~13%) — not a revolt, but not a rubber stamp.
• PwC ratification: 12,029,611 against (~13%) — unusually high for an auditor vote.
• A shareholder proposal to lower the threshold to call a special meeting failed 34.3M for vs 45.1M against — i.e. ~43% of votes cast supported it. That is a meaningful minority pushing for shareholder power at a company that returns no capital.

1. Insider activity remains one-directional. Reported ~135 sales and zero purchases over the prior six months, ~$26.1M of stock sold ``. Consistent with vesting-driven comp; still no insider-buying signal to lean on — and no buyback either, so shareholders have received nothing while cash fell $1.1B.
2. Capital allocation shifted, quietly and defensibly. The company deleveraged to zero long-term debt, refinanced into a larger unsecured revolver, and cut capex nearly in half year-on-year (H1 $279.8M vs $494.1M) while holding an $0.8–1.0B FY guide it is running far behind ``. Read charitably: prudence into policy uncertainty. Read sceptically: the growth capex that justified the no-dividend, no-buyback stance is being deferred, and shareholders still get nothing.

The open governance question this refresh raises: the securities class action and derivative suit allege that management's communication about tariff exposure and Southeast Asian utilization was materially misleading during 2025-02-26 → 2026-02-24. This dossier cannot adjudicate that. What it can note is that the Q2-2026 disclosures — 1.8 GW idle, ~$30M/quarter underutilization, explicit "reduce, pause, or cease" impairment language — are markedly more forthcoming than what the prior dossier could find in the FY2025 10-K, which is itself consistent with either a genuine deterioration or a disclosure correction. Both readings are live.

Forensic Red Flags — REFRESHED

The prior dossier's verdict was "the concern is earnings composition, not integrity." That verdict now needs a qualifier: integrity is under litigation, and the composition got worse on the cash side.

1. THE central flag, now stated by the company itself. FY2026 guidance: gross profit $2.4–2.6B, Section 45X credits $2.10–2.19B . **~85% of guided gross profit is a tax credit** . Q2 actual: government grants recognised in Cost of Sales $444,480K against net income $422,569K — the credit exceeds net income, as it did in FY2025 and as it does for H1 ($862,513K vs $769,188K) ``. Every multiple in Lens 7 is applied to this.
2. NEW — the credit stopped converting to cash, and it is a guided choice. Government grants receivable $1,263.3M (from $625.2M), a −$649.3M operating-cash-flow drag in six months. No 2026 credit sale has been signed (last: October 2025); FY26 guidance assumes none. H1 cash from the credit: $95.2M residual + $117.6M Treasury = $212.8M against $862.5M recognised ``. The noncurrent tranche carries at $978.8M vs a $953.4M fair value — a self-disclosed haircut. Earnings quality did not deteriorate through accounting; it deteriorated through collection.
3. NEW — receivables factoring stopped entirely. $492.8M of trade receivables factored in FY2025 (accounted for as secured borrowings, with $6.5M of interest expense); zero in H1 2026, and no outstanding liabilities at either balance-sheet date ``. This makes the YoY operating-cash-flow comparison flatter than it looks and is part of why AR rose $182.2M on falling revenue.
4. NEW — securities class action and derivative suit. EDNY, filed 2026-06-23 and 2026-07-28 respectively; §10(b)/§20(a)/Rule 10b-5, breach of fiduciary duty, gross mismanagement, waste. Company "disputes the allegations and intends to defend the action vigorously"; "not in a position to assess the likelihood of any potential loss... or to estimate the amount or range of possible loss" ``. Unquantified contingency.
5. NEW — international PP&E impairment risk is now disclosed language, not inference. "It is reasonably possible that the operational status of one or more of our international facilities may be adversely affected by geopolitical developments... which may result in future decisions to reduce, pause, or cease operations... could result in a decrease in the value, and possible impairment, of this property, plant, and equipment" . Total PP&E net is $5,643.5M; the international share is not broken out, so the exposure cannot be sized from the filing. **~1.8 GW of finished international capacity idle at ~$30M/quarter** .
6. NEW — an unquantified R&D equipment impairment already ran through Q2 opex, disclosed only verbally on the call, not sized in the 10-Q ``.
7. Backlog quality — the erosion is now measurable, not anecdotal. 50.1 GW/$15.0B → 45.1 GW/$13.6B in six months; tech adjusters 23.2 GW/$0.6B → 21.6 GW/$0.5B. The BP/Lightsource matter is unresolved: $384.6M triggered, only $61.0M ever recognised as revenue, $323.6M still being litigated, and defendants have counterclaimed for $175M plus return of $15M credit support; First Solar's motion to dismiss the counterclaims was argued 2026-07-29 ``. Outcome not yet in the file.
8. Tax-credit indemnification, carried. First Solar indemnifies credit purchasers against legislative changes affecting previously transferred 45X credits. No accrued indemnification liabilities at either date `` — the exposure is entirely off balance sheet.
9. Variable consideration, carried and slightly larger. Six categories of price adjuster (technology, freight, commodity, wattage, domestic content, tariff) can move recognised revenue up or down; contracts "have in the past and may in the future be breached by our customers or be subject to renegotiation" ``.
10. The $88.6M IEEPA benefit is an estimate booked into gross profit. Management says it "remains subject to refinement as additional information becomes available" ``. It is ~15% of the quarter's gross profit and ~100% of the beat versus consensus.

Regulatory findings (required sub-section). Re-fetched regulatory/regulatory-findings.md on 2026-08-10 (period 2021-08-10 → 2026-08-10):

• SEC Litigation Releases: 0 naming First Solar ``.
• AAERs: 0 in the same window ``.
• Non-SEC enforcement: no FTC/DOJ/CFPB action, consent decree or penalty surfaced in this sweep ``.
• Private securities litigation: PRESENT and NEW — the EDNY class action and derivative suit above. Note the distinction the automated regulatory fetch cannot make: EDGAR LR/AAER searches only cover SEC enforcement, not private plaintiff suits. A clean regulatory-findings file is not evidence of a clean litigation docket, and on this name it now materially understates the position.
• Verdict: no regulator has acted. Private plaintiffs have. The forensic picture is no longer "composition, not integrity" — it is composition, deteriorating cash conversion, and an unresolved integrity allegation the company denies.

---

Forward Projection (FY2026 / FY2027 / FY2028 EPS) — REFRESHED

Model integrity — read this before any number below

model.xlsx was rebuilt this run (our model, MODEL_VERSION 2.0): 5 quarters of history (2025-Q1 → 2026-Q2), opening balance sheet sourced from our figures (yes). But our model reports:

"No computed values. The workbook has 51 formulas and no cached results — open it in Excel, save, and re-run. Nothing below is a real number yet."

So: value per share, enterprise value, WACC, the balance check and every FY+1/+3 output are BLANK. The share price and CAPM inputs on the Assumptions sheet are also unset. I am therefore citing no model output. Every figure below is `` with the arithmetic shown, anchored on filed actuals and company guidance. This is a real gap in the research layer, not a rounding issue — the workbook needs one manual Excel open-and-save before it can be quoted.

Anchors

• H1-2026 actual diluted EPS $7.14; H1 net income $769.2M; H1 adjusted EBITDA $1,163M ``
• FY2026 guided adjusted EBITDA $2.6–2.8B, with ~$225M of addbacks (SBC, 45X credit discounts, underutilization, start-up) ``
• H1 D&A $295.2M; diluted shares 107.7M; H1 effective tax rate 5.5% and rising on Pillar Two ``
• Consensus FY2026 EPS ~$18.04 (range $13.35–$24.73); consensus FY2027 net income ~$2.54B, ≈ $23.6 EPS on 107.7M shares ``

FY2026 base case — built from the company's own reaffirmed guidance

Adjusted EBITDA (guidance midpoint) $2,700M
less guided addbacks ($225M)
= EBITDA $2,475M
less D&A (H1 $295.2M x2, PP&E flat) ($600M)
= EBIT $1,875M
plus net interest & other (H1 +$18.0M,
 annualised, cash falling in H2) +$30M
= pre-tax income $1,905M
less tax @ ~6.5% (H1 5.5%, Pillar Two rising) ($124M)
= net income $1,781M
/ 107.7M diluted shares
= FY2026E EPS ~$16.5

``

Cross-check against the half already reported: H1 net income / H1 adjusted EBITDA = 66.1%. Applying that conversion to guided H2 adjusted EBITDA ($2.7B − $1.163B = $1.537B) gives H2 net income ~$1.015B, FY ~$1.784B, EPS ~$16.57. The two methods agree to within 0.5%.

This is a real disagreement with consensus. To reach $18.04, H2 net income must be $1.174B on ~$1.537B of H2 adjusted EBITDA — a 76% conversion, versus the 66% First Solar actually delivered in H1, in a half that included the $88.6M IEEPA windfall and excluded the back-loaded production start-up ($55–65M remaining) and underutilization ($60–80M remaining) that guidance explicitly pushes into H2. We read reaffirmed guidance as implying ~$16.5, roughly 9% below the $18.04 consensus.

Three-year scenarios

The number that matters more than EPS, restated: the 45X credit runs 2023–2032 with phase-down beginning in 2030 ``, and the backlog delivers through 2030. Base case is well supported through 2028; a DCF must still apply a steep terminal haircut at 2029–2032. What is genuinely new is that the terminal haircut is now smaller than it was in June, because the Section 232 MIP has no announced sunset and would survive a 45X expiry.

Tracked forecast (recorded here only; our model create NOT run — unattended pass):

"FSLR FY2026 GAAP diluted EPS ≤ $17.25 (i.e. below the $18.04 consensus), p ≈ 0.65, resolves 2027-02-28."

Bull vs Bear — REFRESHED

Bull case (updated). First Solar now sits behind two policy instruments, not one. The 45X credit subsidises its cost base ($2.10–2.19B in 2026 alone), and from 2026-12-04 a Section 232 minimum import price of $0.38/W puts a statutory floor under imported module pricing — above the ~$0.36/W at which First Solar is currently booking US volume ``. On top: a debt-free balance sheet (0.00x debt-to-equity, $1.5B undrawn unsecured revolver), a 45.1 GW / $13.6B backlog through 2030 of which ~41 GW carries domestic-content requirements only a US-integrated fleet can serve, a US fleet running at 98% utilization and substantially committed through 2028, an offensive §337 case that could exclude TOPCon imports outright from September 2027, and demand that is now named rather than theoretical — Steel River (Cypress Creek / Google), the largest US solar project, on First Solar modules. At 11.97x forward the market is not paying for any of this.

Bear case (updated — three legs strengthened, one weakened).

1. The earnings are ~85% subsidy, on management's own guidance, with a clock on it. $2.10–2.19B of credit inside $2.4–2.6B of guided gross profit. Strip it and FY26 is a ~7% gross-margin business. Phase-down begins 2030.
2. STRENGTHENED — the subsidy is no longer converting to cash. $1,263.3M of grants receivable, no 2026 credit sale signed, factoring stopped, H1 FCF −$639.6M, net cash already at the bottom of the year-end guide. A subsidy you have booked but not banked is a receivable from a government, priced by the company's own fair-value note at a discount.
3. STRENGTHENED — the backlog is shrinking and the counterparty record is bad. −5.0 GW / −$1.4B in six months; BP/Lightsource still owes $323.6M and is counterclaiming $175M; Pine Gate went Chapter 11 in November 2025 ``.
4. NEW — an unquantified triple contingency: a securities class action, a derivative suit, and a possible impairment on ~1.8 GW of idle Southeast Asian capacity, none of which the company will size.
5. WEAKENED — "US oversupply crushes ASPs." This was the prior dossier's second-strongest bear leg. The Section 232 MIP directly targets it. It is not dead — the floor applies to imports, not to domestic c-Si capacity built with the same 45X money — but it is materially less potent than it was on 2026-06-20.
6. Unchanged — technology. CdTe at ~19.7% module efficiency still loses to TOPCon and to eventual perovskite-silicon tandems; First Solar's own perovskite Series 6 pilot only reaches operational readiness in H1 2027 ``, and a piece of the R&D equipment roadmap was just impaired.

Pre-mortem (18 months out, thesis broke). It is 2028-02. The Section 232 MIP was enjoined or narrowed in litigation during 2027 (the IEEPA precedent — an entire tariff regime voided mid-stream by the Supreme Court in February 2026 — made challenging trade actions look winnable). The 45X receivable, never sold, was collected slowly and at a discount, and net cash breached the $1.5B floor, forcing First Solar to draw the revolver for the first time. Malaysia and Vietnam were closed with a nine-figure impairment. The securities class action survived a motion to dismiss and settled. The backlog, run down from 50 GW to the high-30s because management "prioritised contract quality," met an ASP environment it could not reprice into. The stock re-rated from a subsidised-growth multiple to a commodity-manufacturer multiple on a pre-credit earnings base — and the company never had a bad operating quarter.

Are multiples too high? At 11.97x forward the headline is undemanding, and cheaper than at the last dive. The honest answer is unchanged and now better evidenced: the multiple is roughly fair against the policy-supported base case and far too high against pre-subsidy earnings power, because ~85% of the "E" is a fiscal transfer. What changed is that a second, non-expiring instrument (the MIP) now supports part of that E — which is why the structural bear case has narrowed rather than broken.

Devil's Advocate (short-seller) — REFRESHED

Dismantling the bull case, updated for what actually happened.

• The kill switch got a second lock — but the first one is still the one that matters. Yes, the MIP is a floor. But a floor on import prices does nothing for First Solar if the 45X credit is trimmed: remove ~$2.1B and 2026 gross margin collapses from ~50% to ~7%. Roth Capital reporting suggests a 45X extension draft has been pushed past the midterms `` — i.e. the credit's future is an election variable, again, exactly as it was in November 2024 when this stock fell 29%.
• The MIP itself is not safe, and this company just proved it. On 2026-02-20 the Supreme Court struck down the IEEPA tariffs as unlawful, and First Solar booked $88.6M of refunds as a benefit — while simultaneously losing the tariff wall those duties provided. Section 232 is a sturdier statute than IEEPA, but the lesson of the last six months is that a US tariff regime can be voided in a single ruling, and First Solar's earnings sit on both sides of that trade.
• "Record backlog" is now a shrinking backlog, and management pre-announced why. 50.1 → 45.1 GW in six months while selling 3.7 GW/quarter and booking 1.9 GW/quarter of US volume. "Prioritising contract quality over booking volume" is what every company says while its book runs off. At H1's booking rate the backlog empties.
• The cash story is the short. Net income $769M in H1; operating cash flow −$360M; free cash flow −$640M; the gap is a $649M government receivable and a $300M inventory build with factoring switched off. Ask the only question that matters: if the 45X credit is worth $2.1B this year, why has the company not sold a dollar of it since October 2025? Either the transfer market has repriced (credit-sale discounts were ~4–5% in 2024–25) or First Solar prefers to wait — and either answer costs shareholders liquidity while they receive no dividend and no buyback.
• Two plants nobody will discuss. ~1.8 GW of finished international capacity, ~$30M/quarter of underutilization, "long-term operating profile" pending, and brand-new 10-Q language about "reduce, pause, or cease." The class action alleges those plants ran near 20% utilization while investors were told US module prices were stable. Management has neither sized the impairment nor ruled it out.
• The one-off did the beating. EPS $3.92 vs $3.01 consensus. The IEEPA estimate contributed ~$0.82/share pre-tax. Strip it and the "30% beat" is a single-digit beat on a −3.7% revenue quarter.
• The competitor bulls underestimate is unchanged and getting bigger: the US c-Si fleet being subsidised into existence by the same 45X credit — which the MIP protects too, since domestic production is not an import.
• Assumptions that must hold for today's $238.82: (1) 45X survives at full rate through 2029; (2) the Section 232 MIP takes effect on 2026-12-04 and is not enjoined; (3) the ~$1.3B credit receivable collects at or near carrying value; (4) no material impairment on Malaysia/Vietnam; (5) the securities litigation resolves immaterially; (6) the backlog stops shrinking; (7) no perovskite/tandem leapfrog before ~2029.
• If it breaks: haircut to a pre-credit margin (~7% gross, roughly break-even operating) and First Solar's normalised EPS is $3–6, not $16.5. Even a generous 15x on a $6 normalised number is a $90 stock. The low analyst target of $150 `` remains the market's mildest version of this scenario; it does not go anywhere near the honest one.
• Single scenario that permanently impairs: 45X trimmed after the midterms and the MIP narrowed in litigation — product moat never mattered, both policy moats gone, and a 7%-gross-margin manufacturer with a shrinking book and two stranded plants is repriced from a growth multiple to a cyclical one. Plausibility: low in the next 12 months, non-trivial by 2029–2031.

---

Company Overview — unchanged, see the previous dossier

(Only correction: the manufacturing footprint is described in the Q2 10-Q as "the United States, India, Malaysia, and Vietnam"; the fifth US facility has commenced operations and South Carolina is the sixth, Phase I H2 2026 / Phase II mid-2027, ~$0.3B investment, up to 3.5 GW of finishing capacity. Cumulative module sales passed 100 GW in Q2 2026, from 93 GW at the prior dive.) ``

Supply Chain — unchanged, see the previous dossier

(Tellurium, the wafer-free/polysilicon-free position, and the chokepoint analysis all stand. One reinforcement worth flagging: the Section 232 action is on polysilicon and its derivatives — a chokepoint First Solar does not touch. It taxes competitors' input and First Solar's output floor simultaneously. That asymmetry is the cleanest expression of the CdTe supply-chain thesis to date.)

Management Questions — carried, see the previous dossier

(The 15 questions stand. Six additions this refresh, ordered by information value:)

1. You recognised $862.5M of 45X credit in H1 but collected $212.8M, and have signed no 2026 credit sale. What discount is the transfer market currently bidding, and at what discount do you become a seller?
2. If no 2026 credit sale is signed, walk us to the $1.7–2.3B year-end net cash number line by line. You ended Q2 at $1.7B.
3. What is the net book value of the Malaysian and Vietnamese PP&E, and under what specific Section 232 / FEOC outcome do you take an impairment?
4. Your US bookings ASP is $0.36/W and the Section 232 module MIP is $0.38/W. Do you reprice the existing backlog, or does the floor only benefit contracts written after 2026-12-04?
5. Gross bookings ran 1.9 GW US in a quarter against 3.7 GW sold. At what backlog level does "disciplined contracting" become "we cannot replace the book at acceptable prices"?
6. The class action alleges Southeast Asian utilization near 20% during a period when you described US module pricing as stable. What was that utilization, quarter by quarter, across the class period?

---