Phase Space AI

Valuation

Exzeo Group [XZO]

Exzeo Group [XZO] — Valuation (RE-UNDERWRITTEN 2026-07-29, v2)

Supersedes archive/XZO_Valuation_2026-07-29_superseded.md. Spot $16.45 (2026-07-29 close, Alpaca SIP). Framework: MEMO_BRIEF v2. No position verdict is issued.

---

0. Why this name was re-underwritten, and what changed

A terminal-margin audit over 84 covered names found XZO carrying the single largest gap in the corpus: a terminal EBIT margin of 22.7% against a trailing actual operating margin of 48.7% — −26.0pp.

The prior memo identified that gap in prose ("a terminal margin of 22.7% that sits 26.0pp below the 48.7% margin the company earns today") and then held the parameter anyway. The FAIL verdict it published rested entirely on the number it had just described as wrong. That is the defect: the diagnosis was correct and the correction was never applied.

The coordinator has since traced the origin. coverage_scan.py keyed its "sector peer median" on a field that does not exist in SEC companyfacts (there is no sic key), so every company landed in one bucket and the "sector median" became a universe-wide median of 14.35%; 70 of 111 records carry 14.4% while claiming max(own, sector peer median). The 22.7% was an industry-p75 cap stacked on that artifact. It was not an estimate of anything about Exzeo. It has not been reconciled against — it has been discarded.

Note the shape of the error. The 12-month target barely moves, because the terminal margin enters only the five-year instrument. That is why a −26.0pp parameter error survived a full memo: nothing the reader looks at first was visibly wrong. The correction cuts the required CAGR by 4.2pp and takes the margin from −9.9pp to between −5.7pp and 0.0pp. The conclusion is still FAIL, but it is now a marginal FAIL rather than a decisive one, and it fails for a reason that is actually about Exzeo.

---

1. Company state — declared before the instrument is chosen

Declared: STATE C for instrument purposes, STATE A economics. This is a genuine gap in the taxonomy and is reported as a framework finding, not papered over.

The three breaks, each documented: 1. Carve-out IPO 2025-11-06 (424B4, 8,000,000 shares at $21.00). Pre-IPO periods are carve-out presentations of a division of HCI Group, not standalone accounts. This is an accounting-basis break. 2. Two new related-party contracts effective the first day of FY2025, which produced the entire FY2025 step-up (+62.0% reported; +11.7% excluding them). 3. Corporate identity change — CIK 1873951 filed as TypTap Insurance Group, Inc. from 2021-08-03 to 2023-04-07. A 2021 S-1 was filed and the IPO did not proceed.

State B is inapplicable: no exogenous commodity or utilisation cycle governs revenue (Florida catastrophe exposure sits at the customer, not here). State D is inapplicable: $220m of revenue at a 61.3% gross margin and $107m of operating income is not venture-like.

Consequence: the terminal margin is BUILT through the opex bridge (the State-C requirement) rather than read off a trailing average (the State-A method), because only five quarters of the current economic structure and three quarters of standalone public reporting exist. evidence_grade: C. Per valuation.md rule 5, that constrains position size, not the operating assumption.

---

2. Inputs, rebuilt from primary sources

Every figure is from EDGAR XBRL (CIK 1873951) or the filed 10-Q/10-K. Discrepancies against both the screen and Alpha Vantage are in §6.

Quarterly operating margin, five consecutive quarters at the current structure (10-K + 10-Q, primary):

Revenue has sat in a $52–56m band for five straight quarters. The screen's FY23→FY25 CAGR of 56.7% is a step function measured across itself.

---

3. Terminal EBIT margin — 44.5%, built and justified in prose

3.1 The bridge

m_EBIT,T = m_gross,T − R&D − S&M/G&A − other. Reconciled against the TTM actual first, to prove the bridge closes on real numbers before it is projected:

Hard constraint: 44.5% ≤ 60.0% ✓. The bridge closes on the actuals too: 61.3 − 4.05 − 8.3 − 0.32 = 48.7% = the reported TTM operating margin, to the decimal. The projection is the same arithmetic with three named cost increases and no revenue-side credit.

3.2 Why 44.5% and not 48.7% — the explicit causal bridge

valuation.md rule 2 permits a terminal margin below the current one but requires "an explicit causal bridge, not prohibition." Here it is, and it is 4.2pp of it:

• +2.6pp of permanent SG&A. SG&A went from $2,706k in Q1-2025 to $5,216k in Q1-2026 — a 92.8% increase on a 6.0% revenue increase. That is the public-company cost load (board, audit, D&O, SOX readiness, IR) and it is permanent. FY2025 as reported carries only a partial quarter of it; the run-rate is 9.4% of revenue against 7.2% for FY2025 as a whole.
• +1.5pp of incremental selling cost. Third-party revenue was $2,211k in Q1-2026, 4.0% of revenue, from a base of exactly zero a year earlier. Winning non-affiliated carriers requires a sales motion Exzeo has never had to fund, because HCI was delivered to it. Growth away from the affiliate is not free.
• −1.3pp of gross margin, conservatively. Cost of revenue fell 3.4% YoY while revenue rose 6.0%, and the company guides improvement. Terminal gross margin is set at 60.0%, below both FY25 and TTM, taking none of it.

What is deliberately NOT in the bridge, and why. HCI Group owns 82.5% of Exzeo and sets the take rate (8.5–25.0% of premium by counterparty), so it captures 17.5% of any value transferred out of Exzeo. That is a real and named risk. It is not in the terminal margin, because per valuation.md rule 1 ("assert, never clamp") and rule 5 ("uncertainty reduces POSITION SIZE, never the operating assumption") a governance risk belongs in the Downside Criteria and in sizing. Burying it in the margin is precisely how the 22.7% got there, and it is the NTRA double-counting defect in a different costume.

44.5% sits 4.2pp below TTM (48.7%) and 0.7pp below the most recent quarter (45.2%). It is not below the company's demonstrated economics in any sense that requires a special defence — it is essentially the quarter just printed, with the public-company cost load taken as permanent.

---

4. The implied-path test — the Valuation Criteria

assets/reverse_dcf.py. Solved for: revenue CAGR. Held fixed: terminal EBIT margin 44.5%, exit multiple 11.0x EBIT, WACC 10.0%, horizon 5y, net cash $329.897m, shares 90.918m, TTM revenue $220.107m.

4.1 The exit multiple is derived, not chosen

valuation.md rule 4: the exit multiple is not an independent parameter.

EV_T / EBIT_T  =  (1−t)(1−g/ROIC) / (WACC−g)

• t = 26.0% — Q1-2026 income tax expense $7,181k on pre-tax income $27,587k.
• ROIC is very high, effectively unbounded. Invested capital is negative: equity $274.941m less net cash $329.897m = −$54.956m. Capex runs $0.3–1.2m per quarter on $55m of revenue. So g/ROIC → 0.
• Therefore EV_T/EBIT_T ≈ (1−t)/(WACC−g) = 0.74/(0.10−g).

Base set at 11.0x, the identity value at g = 3.25%. Two independent cross-checks agree:

The prior memo's 18.0x was also too high, for a subtler reason than the screen's 22.8x. It was built from AON / BRO / GSHD / RYAN / BWIN / AJG on market-cap/EBIT — a levered basis on net-debt companies — and then applied to a net-cash company on an EV/EBIT basis. The identity and the name's own traded history converge at 10.2–12.3x, and they are the two references that do not require a comparator set at all. Correcting the exit multiple downward partly offsets the terminal-margin correction upward, which is why the net change in the required CAGR is only −4.2pp and not −14pp.

4.2 THE RESULT

The market requires an 11.7% revenue CAGR for five years, holding the terminal EBIT margin at 44.5% and the exit multiple at 11.0x EBIT, at a 10.0% WACC.

The margin — demonstrated − required:

Verdict on the Valuation Criteria: FAIL, marginal. The honest band is −5.7pp to 0.0pp, centred near −3pp. It is not PASS WITH ARGUMENT: that grade requires "a specific, evidenced reason (a named product cycle, mix shift, pricing action)", and the only candidate is third-party diversification whose entire evidence is $2,211k — 4.0% of one quarter's revenue. Real, named and evidenced; roughly two orders of magnitude too small to underwrite 11.7% for five years. It is logged for monitoring, not scored as an argument.

Implied compression, as a number: trading at 10.87x EV/TTM EBIT against an 11.0x exit — implied expansion of +1.2%, i.e. essentially none. This is the largest single change from the prior memo, which required +65.6% multiple expansion on top of 15.9% growth and a halving of the margin. The corrected path asks the business to grow at 11.7% and nothing else. That is a coherent question. The prior one was not.

Terminal value share of EV: 100% by construction (reverse_dcf.py discounts a terminal EV only). Above the 60% threshold, so the reverse DCF is the mandatory primary long-horizon output and no forward DCF competes.

4.3 Sensitivity — on the exit multiple, never on scenario probabilities

A. Required CAGR vs exit multiple (terminal margin 44.5%)

The flip point against the live run-rate sits at ~14x, above the identity's warranted range and above every session XZO has ever traded at (own max 12.8x). Against the organic 11.7%, the flip point is 11.0x — exactly where the identity lands. The name is priced at fair value for its organic growth rate and expensive for its current run-rate. That is a far more precise statement than the prior memo could make.

B. Required CAGR vs terminal margin (exit 11.0x)

*at 11.0x. At the prior memo's 18.0x the old parameterisation gave 15.9%.

C. Required terminal margin at observed growth (exit 11.0x)

At the live run-rate the price requires a 57.9% permanent EBIT margin — above anything Exzeo has printed in any quarter (max 50.3%) and above the 60.0% terminal gross margin less any opex at all. That is the honest bear statement, and it is arithmetically much stronger than the prior memo's 35.4%.

D. Required exit multiple at observed growth (terminal margin 44.5%)

---

5. The 12-month target

5.1 Near-term consensus

Alpha Vantage EARNINGS_ESTIMATES, retrieved 2026-07-29 — returned data (3 analysts on revenue, 1 on EPS). Recorded explicitly because an empty response from this endpoint does not establish absence of coverage (confirmed empty on KLAC, a $222bn company).

Revisions positive. NTM revenue ≈ $255.3m.

One Street assumption the filings do not support. FY2026E of $235.6m implies Q2–Q4 at $60.0m/quarter against Q1-26's actual $55.5m — an 8% sequential step-up beginning in a quarter that has already closed and is not yet filed. Nothing in the Q1 10-Q flags it. Q2 earnings tests this directly and it is the single most falsifiable input in the bull case.

5.2 The multiple anchor is UNIDENTIFIED — declared, not substituted

The anchor is declared UNIDENTIFIED. The prior memo called it "LOW-CONFIDENCE, bordering UNIDENTIFIED"; that was too generous and is corrected here. valuation.md requires the declaration where history is too short or spans a regime change. Both hold, three times over: 107 observations is under six months and cannot establish a 12-month mean-reversion band; the window contains one earnings cycle; and it spans the 180-day IPO lock-up expiry (~2026-05-05), a structural change in float. No peer median has been substituted anywhere — that is the failure mode the rule exists to prevent. The number below is a range with a midpoint, not a point estimate, and it should not be traded as one.

5.3 The target

NTM EBIT = consensus NTM revenue $255.3m × 45.0% — the margin actually delivered in Q1-2026, not the 48.7% TTM, because the SG&A step-up is permanent. NTM EBIT ≈ $114.9m.

12-month target: $16.50 — +0.3% to spot. Range $14.50 – $18.30 (−11.9% to +11.2%). Anchor UNIDENTIFIED.

Cross-check. Forward P/E on NTM EPS ~$1.075 = 15.3x; ex-net-cash 11.9x. Neither is demanding. The target is flat because the stock sits at the ~70th percentile of its own short band while NTM EBIT is roughly flat against TTM on the SG&A step-up — not because a house view is imposed.

On calibration item B16 (16 of 16 house targets below spot, median 46.1% below Street): this target is not below spot. It is +0.3%, and it lands there mechanically. Flagged so the record can separate it from B16.

No external professional target was retrievable from the entitled sources. Reported as missing, not invented.

5.4 Why the two horizons disagree, and why that is correct

Over 12 months XZO is fairly priced: $330m of net cash (22.1% of market cap), 11.9x ex-cash forward earnings, a $12m buyback authorised 2026-05-26, and a mid-band multiple. Over five years the price requires 11.7% growth against a business that has printed 6.0% for five quarters and whose organic rate is 11.7% only if you credit the FY2025 ex-contract figure as repeatable.

The name is not expensive. It is uncollateralised — 98.9% of revenue is set by a counterparty that owns 82.5% of the equity and captures 17.5% of anything it takes back. That is a sizing constraint and a downside case, not a margin haircut.

---

6. Defects found — every one, with magnitude

6.1 Alpha Vantage ebit is income before tax, not operating income (NEW — independently found here)

AV's ebit equals its own incomeBeforeTax in every XZO period — it includes interest income on a $330m cash pile. Corroborates the coordinator's AMAT finding (9.7pp). Use operatingIncome / totalRevenue.

6.2 Alpha Vantage ebitda overstated by $6.416m (+5.8%)

AV: $116.639m. Computed as operating income + cash-flow-statement D&A: $110.223m. AV's figure is derived from the wrong ebit above.

6.3 Alpha Vantage RevenueTTM overstated by exactly the same $6.416m (+2.9%)

AV OVERVIEW.RevenueTTM = $226.523m. Sum of AV's own four quarterly reports = $220.107m. The identical $6,416k appears in both the RevenueTTM and EBITDA errors, which points at one spurious period in AV's TTM aggregator. AV's summary fields disagree with AV's own statements — an internal inconsistency the vendor's own checks cannot see. AV OperatingMarginTTM 47.5% vs computed 48.7%.

6.4 Alpha Vantage cost of revenue / gross profit wrong in every period tested

Revenue and operating income match exactly; the error is confined to the split. This directly attacks the m_EBIT,T ≤ m_gross,T ceiling test, which would have been run against a gross margin 1.6pp too low.

6.5 Alpha Vantage statement depth for XZO is 14 quarters and 6 annuals, with holes

The brief promised "~81 quarterly and 20 annual periods." For XZO: 6 annual (FY2021 missing entirely) and 14 quarterly (all of FY2022 and FY2023 absent, Q1–Q2 2024 absent, Q4 2021 absent). A recent IPO carries only what its S-1 disclosed. AV's operatingExpenses is also internally inconsistent across periods (Q1-25 $28.610m vs Q4-24 $3.123m vs Q1-26 $6.806m), and its R&D + SG&A components double-count D&A relative to its own operatingExpenses subtotal. OVERVIEW.CIK returns "None".

6.6 XZO's OWN XBRL carries a 1,000× share-count scale error (NEW — filer-side defect)

WeightedAverageNumberOfDilutedSharesOutstanding is tagged 78,709 for Q1-2025 and 88,385 for Q1-2026 — against 76,102,000 / 77,494,000 / 80,171,000 in the annual periods. Both quarterly figures are off by 1,000×. Worse, basic and diluted are identical in every single period on file, which is implausible for a company recognising $0.7m of quarterly SBC and post-IPO RSUs; and 88,385(,000) does not reproduce the reported $0.22 EPS ($20,406/88,385 = $0.231; $20,406/90,918 = $0.224 → $0.22).

Resolution: EV is built on point-in-time shares outstanding, verified at 90,918,430 from two independent tags. The diluted count is recorded INDETERMINATE. Direction of the residual: diluted ≥ basic, so EV is understated and the 11.7% required CAGR is a floor. At a 3% dilution the required CAGR rises to ~12.2%, which does not change the verdict.

6.7 Net cash: prior memo's $329.897m is CONFIRMED, with one omission

Cash $231.381m + AFS $98.516m; NotesPayable = 0; no debt tag in any period. The prior memo asserted "no borrowings" without addressing the $6.929m operating lease liability. On a lease-inclusive EV, net cash is $323.0m and the required CAGR rises ~0.2pp. Immaterial, but it was an unstated assumption. Contra the brief's expectation that net_cash was wrong on ~20 of 22 names — on XZO it was right.

Unresolved: whether the $98.5m AFS portfolio is unencumbered was not verified in the time available. Exzeo is a services company, not a carrier, so statutory deposits are unlikely, but this is an open input.

6.8 SIC misclassification — the root of the 22.8x exit multiple

XZO files under SIC 7372 "Services-Prepackaged Software." It is an outsourced insurance-operations business whose revenue is a take rate on affiliated gross written premium. The screen inherited 7372 and drew a 22.8x exit multiple from a 106-name software universe (exit_multiple_peer_n = 106), which alone accounts for most of the screen's PASS.

6.9 The first standalone 10-Q required amendment eight days later

10-Q/A (Amendment No. 1) filed 2026-05-15, eight days after the 2026-05-07 original. Per its Explanatory Note, the tabular disclosures for "Cost of Revenue," "Operating Expenses," "Investment Income," "Income Tax," "Adjusted EBITDA," "Adjusted Revenue" and "Adjusted EBITDA Margin" — seven tables — "inadvertently included incorrect information due to a technical issue." Financial statements were unchanged.

The prior memo cites neither the 10-Q/A nor its existence, and quotes "Adjusted Revenue $54.005m" and Q1-26 gross profit consistent with the AV figures rather than the amended filing. It therefore reported at least one Q1-2026 figure from a superseded table. The amendment itself is also internally inconsistent, stating cost of revenue as 41.0% of revenue in the table and 41.1% in the adjacent narrative.

For the record, the amended, authoritative Q1-2026 bridge: revenue $55,534 · cost of revenue $22,791 (41.0%) · gross profit $32,743 (59.0%) · SG&A $5,216 (9.4%) · R&D $2,306 (4.2%) · D&A $146 (0.3%) · total opex $7,668 (13.9%) · operating income $25,075 (45.2%). Adjusted EBITDA $26,531 vs $25,218 prior year.

6.10 SPLITS

SPLITS returns an empty array for XZO. Correct — the company listed 2025-11-06 and no split is possible. No split-basis risk. Verified rather than assumed, per the KLA 10-for-1 precedent (89% market-cap error).

---

7. Mention-frequency over time — prepared remarks vs Q&A

EARNINGS_CALL_TRANSCRIPT with speaker/title. Prepared remarks = utterances before the first analyst-titled speaker. Only two transcripts exist (2025Q4 and 2026Q1). 2026Q2 returns an empty array despite an 8-K item 2.02 earnings release on 2026-05-06 — recorded as a coverage gap, not as an absence of a call.

The single most important line is the third one. The entire bull case — customer diversification away from HCI — was raised three times in Q&A on the first call and not once, by anyone, on the second. Management stopped volunteering it and analysts stopped asking. Meanwhile mentions of "managed premium," which is HCI's premium, nearly doubled. The disclosure is drifting toward the affiliated metric and away from the diversification metric, and it is doing so on both sides of the call. "New customer" appears only ever in Q&A — never volunteered — which under the brief's own evidentiary rule makes it the weaker kind of claim to have to extract, and it is being extracted.

---

8. Peer Spread Criteria — MEASURED

Named peer: Erie Indemnity Company (ERIE) — a listed management company earning a contractual percentage of an affiliated insurance enterprise's premiums, controlled ownership, near-total related-party revenue. The closest public analogue that exists.

Spread: +31.2pp. Percentile UNIDENTIFIED — the comparison did not exist before November 2025.

The question this poses is the memo in one line: why does the captive fee conduit for a ~$900m Florida homeowners insurer earn a margin nearly three times the captive fee conduit for a $4bn reciprocal exchange — and what happens to the difference now that the parent that sets the price owns 82.5% instead of 100%?