Phase Space AI

Financial Model Notes

AtriCure [ATRC]

AtriCure [ATRC] — Financial Model Notes

Data provenance

Item Source Value
Statements AV INCOME_STATEMENT / BALANCE_SHEET / CASH_FLOW (normalized; 81 quarters, 20 years)
AV-vs-EDGAR audit av_vs_edgar.py --quarters 8: 0 disagreements of 6 compared; trust flag NO solely because of the D&A sign (below) revenue/GP/EBIT accepted
EDGAR revenue Q2 2026 RevenueFromContractWithCustomerExcludingAssessedTax @2026-06-30 $153.6m vs AV $153.6m — ties
EDGAR gross profit Q2 2026 us-gaap:GrossProfit $119m vs AV $118.6m — ties
EDGAR operating income Q2 2026 us-gaap:OperatingIncomeLoss $10m vs AV $9.7m — ties
Cash @2026-06-30 EDGAR CashAndCashEquivalentsAtCarryingValue $168m
Debt @2026-06-30 EDGAR LongTermDebt / LongTermDebtNoncurrent $61m
Net cash EDGAR-derived $107m
Price Alpaca 1-min bar 2026-07-29 19:59Z $40.89
Splits AV SPLITS: none, ever no split-basis risk
Transcripts AV EARNINGS_CALL_TRANSCRIPT 2023Q1–2026Q2 (14 quarters) 34–67 utterances, 33–53k chars. ATRC is the only name of the three with a Q2 2026 transcript.

TTM to 2026-06-30

$m % of revenue
Revenue 569.6
Gross profit 433.7 76.16%
R&D 96.2 16.89%
SG&A 323.6 56.81%
GAAP operating income 13.9 2.44%
D&A 21 3.69%
EBITDA (EBIT + |D&A|) 35 6.10%

Bridge ties exactly: 76.16 − 16.89 − 56.81 = 2.46% vs 2.44% reported (0.02pp rounding).

The §6 regression check — run deliberately

known-silent-failures.md §6 records that a prior run double-counted $174m of ATRC net cash on an equity multiple, worth 5.5pp of expected return, and that peer_multiples.py was rewritten to own the equity↔enterprise bridge and to raise on mixed-standard comparisons. That specific failure was checked for in this run:

Defects found

1. AV depreciationAndAmortization is signed NEGATIVE for ATRC (−$5m for the quarter ending 2026-06-30, −$1m for 2025-03-31), flagged by av_vs_edgar.py as DNA_SIGN. AV's ebitda field was never read. EBITDA computed as operating income + |D&A|. This is the brief's named AV defect #1, confirmed on 2 of 3 names in this run (ATRC and BSX; DXCM clean).

2. AV OVERVIEW.OperatingMarginTTM = 0.0629 is NOT a TTM figure — it is exactly ATRC's Q2 2026 single-quarter GAAP operating margin (6.29%). The true TTM figure is 2.44%, so the field overstates the TTM operating margin by 3.85pp, i.e. by 2.6x. This is the most damaging instance of this mislabel in the run, because 6.29% and 2.44% support quite different Quality readings. Confirmed systematic: BSX's 0.206 = its Q1 2026 quarter, DXCM's 0.214 = its Q1 2026 quarter. 3 of 3 names. Do not use this field.

3. AV OVERVIEW.ProfitMargin = 0.0185 and PERatio = 194.71, both computed off a single quarter's annualised earnings on a business that crossed into profitability four quarters ago. Read as levels by a Quality screen with no archetype gate, these two fields alone are sufficient to produce the FAIL diagnosed in ATRC_Research.md §1.

4. Share-count discrepancy, 3.7%, reported not silently adopted. - AV BALANCE_SHEET.commonStockSharesOutstanding: 49.1m - AV OVERVIEW.SharesOutstanding: 50.87m

50.9m was used (the higher figure, producing the higher EV of $1,974m vs $1,901m and therefore the higher required CAGR — the conservative direction). Effect on the required CAGR: ~+0.7pp. At 49.1m shares the required CAGR would be ~18.2% and the margin −0.8pp rather than −1.51pp. The verdict does not change, but the record should show the number was not adopted silently.

5. AV EARNINGS_ESTIMATES has no coverage for ATRC — returns {"Information": …} on a premium key with paced requests, retried 6 times with backoff, while the same call succeeded for BSX. Consensus Criteria is BLANK; per criteria.md this blocks nothing. Company guidance ($600–610m revenue, $85–89m adjusted EBITDA) used as the near-term base instead.

6. A minor debt disagreement, resolved toward EDGAR. AV shortLongTermDebtTotal = $71.0m; EDGAR LongTermDebt = $61m. The $10m gap is most likely lease obligations included by AV. EDGAR's $61m was used, giving net cash of $107m rather than $96.7m — the less conservative of the two, so it is flagged: using AV's figure would raise the required CAGR by ~0.1pp. Immaterial either way.

7. FY2021 operating margin of +20.12% is an outlier and was quarantined. It sits inside a FY2019–FY2025 range of −21.4% to −0.6%. It is almost certainly a one-off credit (CARES-era or settlement). It is not used as evidence of demonstrated operating margin anywhere — the terminal-margin build uses TTM actuals and the guided forward differential only. Had it been included in a max(own, …) or multi-year-average terminal margin rule, it would have corrupted the build.

8. The prepared/Q&A splitter defect. Under a naive split on the operator's "question-and-answer" phrase, 7 of 14 ATRC quarters returned n_prepared = 0 — attributing all prepared remarks to Q&A. This silently inverted the single most important reading in the research document (that "profitability" is a prepared-remarks word and the trial-revenue concession is a Q&A word). Fixed by splitting at the first utterance whose title contains "Analyst". The first version produced a complete-looking table that was wrong for half the periods and raised nothing — the canonical shape of every entry in known-silent-failures.md.

Consistency checks