Phase Space AI

Financial Model Notes

Catalyst Pharmaceuticals [CPRX]

Catalyst Pharmaceuticals [CPRX] — Financial Model Notes

Provenance, derivations and reproducibility for every figure used in this memo. Framework: Criteria, 2026-07-29. Analysis date 2026-07-29.

No Excel model was built. The security was extinguished for cash on 2026-07-15 and no forward projection has a referent. Building a three-statement model for a company that no longer exists would be work product without information content. What follows is the full derivation of every number that is used, so that each is independently reproducible.

---

1. Data sources

Recency assertion (per valuation.md data hygiene): latest 10-Q filed 2026-05-11 for period ending 2026-03-31; latest 10-K filed 2026-02-25. At the analysis date the newest financial data is 119 days old, and it is the last data that will ever exist — Form 15-12G suspended reporting on 2026-07-24. Not stale by error; terminal by fact.

---

2. Revenue — TTM build and verification

Quarterly us-gaap:RevenueFromContractWithCustomerExcludingAssessedTax, four quarters to 2026-03-31:

Screen value 596,958,000 — exact match. The Q4 derivation is stated because Catalyst does not tag a discrete Q4 period; the "TTM that skipped Q4 entirely" failure mode noted in the brief is explicitly avoided here.

Annual series (Total revenue incl. licence/other): 2021 140,833 · 2022 214,203 · 2023 398,204 · 2024 491,734 · 2025 588,989.

---

3. Product-level revenue — the organic/acquired decomposition

Source: Note 2, "net product revenue disaggregated by product", FY2023 / FY2024 / FY2025 10-K, and 10-Q Q1-26. These are audited, company-published tables. Nothing here is reconstructed or estimated.

FYCOMPA 2023 covers 2023-01-24 (acquisition) → 2023-12-31. AGAMREE 2024 covers 2024-03-13 (launch) → 2024-12-31. Both partial periods are footnoted by the company.

Q1: FIRDAPSE 83,731 → 98,859 · AGAMREE 22,042 → 36,713 · FYCOMPA 35,627 → 13,771 · total 141,421 → 149,390.

Derivations:

Total CAGR FY22→FY25       = (588,989 / 214,203)^(1/3) − 1 = 40.1%   [reproduces screen exactly]
FIRDAPSE CAGR FY22→FY25    = (358,380 / 213,938)^(1/3) − 1 = 18.8%   [same-product / organic]

Revenue added FY22→FY25    = 588,989 − 214,203 = 374,786
  FIRDAPSE  358,380 − 213,938 = 144,442  →  38.5%
  FYCOMPA   113,341 −       0 = 113,341  →  30.2%   (acquired, Eisai, 2023-01-24)
  AGAMREE   117,086 −       0 = 117,086  →  31.2%   (in-licensed, Santhera, 2023-07-11)
  ACQUIRED / IN-LICENSED      = 230,427  →  61.5%

---

4. Margins

TTM EBIT = Q2-25 66,298 + Q3-25 66,267 + Q4-25 (257,778 − 195,932 = 61,846) + Q1-26 73,232 = 267,643.

No us-gaap:GrossProfit tag exists for this issuer. The screen correctly returned gross_margin_pct: null rather than evaluating nan > threshold to FAIL — a correct D1 handling, worth recording positively.

Q1 comparison used for the royalty mechanism: Q1-25 cost of sales 17,911 (= 9M-25 61,179 − ... derived as H1-25 38,525 − Q2-25 20,614), giving gross margin 87.33% → 90.31%, +2.98pp.

---

5. Balance sheet and share count

Net cash = $755,859k, cash-only basis. Screen value — exact match.

Scale cross-check (net income ÷ shares ≈ filed EPS):

214,326 / 122,290.866 = $1.7526  vs filed basic EPS   $1.75  ✓
214,326 / 127,257.929 = $1.6842  vs filed diluted EPS $1.68  ✓

Single share class; no dual-class dimensional-tagging problem. Discrepancy reported, not silently adopted: the screen's 122.381m basic shares understate the ~130.2m fully-diluted count implied by the transaction's stated $4.1bn equity value by 7.8m shares (6.4%), and therefore understate EV by $246m (7.9%).

---

6. Working capital — DSO derivation

DSO = accounts receivable ÷ quarterly revenue × days in quarter.

FY2025 cash-flow statement: accounts receivable −61,001, inventory −18,089, accounts payable −5,391, accrued expenses +24,164 → operating cash flow 208,670 against net income 214,326 (FY2024: 239,808 against 163,889).

Company's stated cause (FY2025 10-K MD&A): distributor contract renegotiation — reduced fees paid to the customer recorded as a reduction in gross-to-net expenses, in exchange for monthly rather than semi-monthly settlement; a ~$27.0m December-2025 receipt landed 2026-01-02. Adjusting Q4-25 AR for that $27.0m gives 99,477 → DSO 60.0 days — still up 19 days on the H1-25 average of 43.2. The revenue benefit of the fee reduction is not quantified by the company and is not estimated here.

---

7. Own-multiple history — construction

Purpose: the 12-month target's multiple anchor, per valuation.md step 3.

Method. Daily closes 2020-07-27 → 2026-07-14 (Alpaca, IEX feed, adjustment=raw, n = 1,498 sessions). For each session, enterprise value is computed point-in-time as close × shares outstanding − cash & equivalents, using the most recent quarterly XBRL values lagged 45 days to approximate filing availability. TTM revenue and TTM EBIT are rolling four-quarter sums with Q4 derived as FY − 9M. Quarters with a negative TTM EBIT are excluded from the EV/EBIT distribution.

Reference points:

Limitation, stated: IEX-feed closes may differ marginally from consolidated closes; the SIP trades/latest and snapshot endpoints corroborate the final price to the cent ($31.49). SIP historical bars returned HTTP 403 on this entitlement.

---

8. Implied-path test — reproduction commands

R=~/.claude/skills/investment-memo/assets/reverse_dcf.py
python3 $R --spot 31.49 --shares 122.381 --net-cash 755.859 --revenue 596.958 \
  --years 5 --wacc 0.10 --terminal-margin <TM> --exit-multiple <EM> \
  --solve cagr --hist-cagr 0.401

EV implied by price = 31.49 × 122.381 − 755.859 = $3,097.9m (5.19x TTM revenue). Confirms the screen's ev: 3,097,918,438.

---

9. Volatility

Log-return population standard deviation, annualised by √252, on the IEX daily close series:

---

10. Mention-frequency corpus

Intended source per references/mention-frequency.md: Alpha Vantage EARNINGS_CALL_TRANSCRIPT, minimum 8 quarters. Actual: 2024Q1 retrieved; the second request returned the shared 25/day rate-limit Information payload. Transcript metric = INDETERMINATE. One quarter cannot support a time-series claim and none is made.

Substitute corpus, clearly labelled and not represented as transcripts: 15 quarterly earnings releases filed as 8-K Exhibit 99.1, 2023-02-07 → 2026-05-11, pulled from the EDGAR filing index (51 Ex-99.1 documents total; non-earnings releases excluded by title match). Word counts 1,269–3,420. All counts normalised per 1,000 words — the ISRG transcript-length artifact is controlled for. No prepared-remarks / Q&A split is available in this corpus and none is claimed. Full series in CPRX_Research.md §5.

---

11. Figures deliberately NOT produced