Variant Perception

Where We Disagree With the Market

The market is treating Treasury's pending FEOC final rules as a symmetric binary risk on First Solar's $1.6B Section 45X benefit; on the evidence, FEOC is asymmetric upside for FSLR and symmetric downside for almost every other US-eligible competitor, which is the opposite of how sell-side has been adjusting price targets since February. Consensus has cut targets from north of $300 to roughly $240, the stock sits 14% below its 200-day, and 11 NEOs sold together at $190 — the entire stack has repriced as if the credit-driven margin profile is the bull's last leg. Our reading is that backlog erosion is concentration-driven (one distressed BP affiliate accounted for 6.6 of the 8.3 GW debooked), the FY26 EBITDA-only guide is a Pillar Two/credibility reset rather than a new ceiling, and the policy direction (OBBBA preserving 45X, India 126% tariff, China-export-curb reports) has been confirmatory in a single direction the bear case still ignores. The disagreement resolves on a known calendar: tonight's Q1 print, the proposed FEOC regulations, and whether new bookings re-engage by Q2.

This is not a contrarian "market is too pessimistic" call. The bear's 45X-as-illusion math is real on a static reading. What we think the market is mispricing is the direction of the policy lever — every observable signal in the last six months has pushed FEOC toward a regime where FSLR is the only large-scale eligible operator, and the Street is pricing that as if it is a coin flip.

Variant Perception Scorecard

A 64 on variant strength reflects a real disagreement that is materially mispriced (FEOC is a $7-10/share lever) but where the bear's near-term setup is also genuine — backlog is rolling, insiders sold, technicals broke. The disagreement is high-conviction on the policy-mechanics direction and lower-conviction on the timing and on whether one more quarterly miss intervenes before resolution. Consensus is unusually legible here (15+ PT cuts, Zacks #5, death cross), which makes the disagreement easy to define but does not by itself make us right.

Consensus Map

The five high/medium-confidence issues all run in the same direction: the market is pricing FY25 economics as a peak, the next print as a miss, and FEOC as a symmetric risk. Where consensus is genuinely ambiguous is on litigation — the press cycle is loud but no complaint has been filed in six months, and the underlying allegations (guidance disappointment, not manufacturing defect concealment) are different enough from Smilovits that the comparison is doing more work than it should.

The Disagreement Ledger

Parser Error: syntax error at or near "32"

LINE 25: ... — the textbook Form 4 cadence after an earnings event. CFO Bradley's 32% reduction was a single 10b5-1 trigger at $200...
                                                                          ^

LINE 25: ... — the textbook Form 4 cadence after an earnings event. CFO Bradley's 32% reduction was a single 10b5-1 trigger at $200...
                                                                          ^

1. FEOC asymmetry. A consensus analyst would say the OBBBA preserved 45X but introduced FEOC restrictions whose final form is unknown, and that until Treasury speaks, the prudent stance is symmetric haircut. The evidence does not actually support symmetry: First Solar produces CdTe modules with zero Chinese polysilicon, has its own US-domestic glass and tellurium supply, and is the only one of seven large-scale US-eligible module makers without a Chinese cell or wafer supplier. Every restrictive FEOC outcome that reduces FSLR's eligibility reduces every other US module maker's eligibility by more, not less. If we are right, consensus has to move EPS power toward $16-18 and the multiple toward the 5-year median of 19x — a roughly $60-70 spread to current consensus. The disconfirming signal is the cleanest available: published Federal Register proposed regulations, plus the next 45X credit-sale price, both will print on a known calendar.

2. BP-concentration vs customer cycle. A consensus analyst would point to the 80→50 GW backlog drop and the post-print cluster of customer-default disclosures and conclude the demand book is fraying. The named litigation and the disclosed termination ledger show 6.6 of the 8.3 GW of debookings were one BP affiliate; the remaining customer base is concentrated in NextEra and Silicon Ranch — both investment-grade names that took 10%+ of FY25 sales. If we are right, the 50.1 GW backlog is a real contracted floor at premium ASPs, not a way station to lower numbers. The cleanest disconfirmation is a second multi-gigawatt termination from a non-BP customer in 2H26 reporting; the cleanest confirmation is one or more new multi-GW bookings disclosed on tonight's Q1 call.

3. EBITDA guide reread. A consensus analyst would frame the EPS-to-EBITDA switch as "management is hiding something" and follow the EPS estimate cuts. The transcript explicitly cites Pillar Two corporate-minimum-tax volatility. At the $2.7B EBITDA midpoint and a normalized D&A and tax structure, the implied EPS is $15-17 — flat to FY25's $14.21, not a step down. If we are right, the FY26 net-sales guide below FY25 framing that has dominated commentary is doing all the work, and the actual earnings trajectory is flat at peak. The disconfirming signal is the Q1 EBITDA print versus the $400-500M floor, plus the cash tax rate the company discloses on Q1.

4. Cluster sale mechanics. A consensus analyst would say that 11 simultaneous NEO sales right after a guide cut is the kind of pattern courts and ISS care about. The People tab and the Form 4 timestamps show every trade printed at the same price ($190.36) in the same blackout-reopen window — the signature of a 10b5-1 plan triggered by the post-print release, not eleven independent decisions. CFO Bradley sold up into $200, which is information about the floor, not the top. The cleanest disconfirmation is a second cluster sale within 30 days of the Q1 print without any open-market buys; the cleanest confirmation is any open-market purchase by Widmar or Bradley in the post-Q1 window.

Evidence That Changes the Odds

The five strongest pieces of evidence (rows 1-4 plus 6) all bend in the same direction: every observed move on the policy lever has favored a US-domestic, China-free producer, and the disclosed economics of FY26 are flatter than the headlines suggest. The two pieces of evidence we treat as least helpful are the revolver (lender behavior is a coincident indicator, not a vote) and the 45X-cash-conversion math (real but already widely modeled).

How This Gets Resolved

The two highest-information signals here are rows 1 and 4 — Treasury rules and post-Q1 insider activity. Both have observable filings on a known calendar and both directly map to the disagreement. Rows 2, 3, 5, 6, 7 are calibration signals; they shift the probability but do not by themselves decide it. A clean read needs Treasury to print proposed rules and at least one open-market insider buy in the same six-month window.

What Would Make Us Wrong

The strongest case against the FEOC-asymmetry view is that we are projecting a clean policy outcome on a politically discretionary process. Treasury has wide rule-writing latitude under OBBBA, and the narrowest plausible reading of "foreign entity of concern" — one that drags in upstream IP licensing, foreign R&D linkages, or pre-OBBBA wafer-step tooling supply — could fold FSLR's CdTe production into the restricted bucket alongside Chinese-tied competitors. We do not know how Treasury will define material assistance, which is the single rule-writing decision that determines whether FSLR's existing glass and tellurium supply chain crosses the threshold. If the final rules include foreign-IP-origin tests or tighten material-assistance to a point that even a fully US-built fleet cannot clear, the asymmetry inverts: FSLR loses the $1.6B credit base and the multiple compresses to KeyBanc's $135 path. This is not the base case in the upstream evidence, but it is a non-trivial tail.

The second thing that would make us wrong is a second multi-gigawatt customer default that is not BP. The remaining 50.1 GW backlog includes meaningful weight from two top customers; if either NextEra or Silicon Ranch has to invoke a termination clause for project economics that have nothing to do with FSLR, the "concentration noise" framing collapses and the bear's customer-cycle read becomes correct mid-flight. The 10-Q AR aging line and the SG&A "expected credit losses" disclosure are the early warnings; another $50M-plus expected-credit-loss provision would force us to reconsider.

The third thing — and this is where the bear's near-term setup is genuinely strongest — is a Q1 EBITDA print below the $400M floor combined with a FY26 guide cut. Two consecutive misses below management's own publicly committed anchors close the credibility window, and at that point the FEOC-asymmetry argument has to compete with a tape that has lost the right to be optimistic. We can be analytically right on FEOC and still be wrong on the stock for two more quarters, which is itself a problem for any thesis that has to be defended in front of a committee. We are also openly skeptical of our own read on the cluster insider sale: 11 NEOs at the same price in the same window is the dictionary definition of 10b5-1 mechanics, but it is also possible those plans were adopted close enough to the Q4 print to make the unwinding economically informed. The Pomerantz investigation is parked on this exact question.

The first thing to watch is whether Treasury publishes proposed FEOC regulations in the Federal Register before Q3 2026 — that single document moves the disagreement from theoretical to settled, and it arrives on a calendar the market is already pricing.