Phase Space AI

Financial Model Notes

CrowdStrike [CRWD]

CrowdStrike Holdings, Inc. [CRWD] — Financial Model Notes

As of 2026-07-29. Provenance, construction, adjustments, and the weaknesses in my own numbers. Reproduction scripts are committed under work/.

1. Data provenance

Item Source Retrieval
Prices — raw and split-adjusted Alpaca v2/stocks/bars work/fetch_prices.py, work/split_check.py, work/gap_check.py
Fundamentals SEC XBRL companyfacts CIK 0001535527 work/edgar_pull.py, work/extract.py
Balance sheet, income statement, cash flow, Note 1 10-Q Q1 FY27 crwd-20260430.htm, read directly work/fetch_docs.py, work/q.py
DBNRR history FY23, FY24, FY25, FY26 10-Ks, full text work/fetch_hist.py, work/dbnrr_hist.py
Split announcement, ARR, guidance 8-K Ex-99.1 crwd-20260603xex991.htm, crwd-20260303xex991.htm work/fetch_pr.py
Multiple history own P/S with an explicit 4.0 split factor work/multiple_hist3.py
Implied path assets/reverse_dcf.py work/run_paths.py, work/run_exit.py, work/tm_sens.py
Peer anchor XBRL + balance sheets, 11 comparators work/anchor.py
Option chain Alpaca options contracts + snapshots work/liquidity.py

Recency asserted. Period 2026-04-30, filed 2026-06-04 — 55 days old. Not stale.

2. The split — the correction, and the rule it establishes

This is the largest numerical error found tonight and it deserves the full workings.

2a. Detection

The split was found by comparing raw and split-adjusted Alpaca series rather than by trusting either one:

Date Raw close Split-adjusted close Volume
2026-06-30 $763.14 $190.79 3,695,172
2026-07-01 $772.74 $193.19 2,693,103
2026-07-02 $193.98 $193.98 10,439,037
2026-07-06 $199.38 $199.38 10,199,550

A −74.9% raw session with a ~4x volume step and a continuous adjusted series is a 4-for-1 split. Confirmed against the company's own words in the Q1 FY27 release: "a four-for-one split of the company's outstanding shares of Class A common stock in the form of a stock dividend ... record date June 25, 2026 ... trading is expected to begin on a split-adjusted basis on July 2, 2026."

The same scan was run on all five names. Only CRWD split in the window. PANW's 2-for-1 (2024-12-16, −48.5% raw) and 3-for-1 (2022-09-14) fall outside it and are already reflected in its 815m count.

2b. The correction

Screen Correct
Share count 254,564,820 1,018,148,000
Basis dei cover, 10-Q filed 2026-06-04 — pre-split 254,537k Class A × 4
Price $181.82 (7/28) — post-split $179.50 (7/29) — post-split
Market cap $45,694m $182,758m
Error — 4.0x, or $137.1bn
EV / TTM sales 8.33x 35.13x

The screen's share count was correct as filed. The cover page of a 10-Q filed on 2026-06-04 reports a pre-split company, and it reported it accurately. The error was pairing it with a post-split price.

2c. Why the mandated cross-check does not catch this, and what does

The net income ÷ shares ≈ filed EPS check PASSES on the screen's own numbers:

27.774 ÷ 257.881 (pre-split diluted) = $0.108 against filed diluted EPS of $0.11 ✓

Both the share count and the EPS are pre-split, so they are internally consistent. The EPS check validates a share count against an income statement — it cannot detect a mismatch between a share count and a market price.

The check that does catch it: does the implied market capitalisation reconcile against the last corporate action in the price series? A −74.9% single-session move 27 days before the memo date, with no corresponding revenue or balance-sheet event, is a split by elimination.

Generalisable rule this establishes, and it is new:

Every share count must be date-checked against the last corporate action in the price series, not merely taken from the latest filing. A cover-page count is a point-in-time fact about a pre-split company; a price is a fact about a post-split one. The EPS cross-check is necessary and NOT sufficient.

This is the same family as the other two data defects in this cluster — ADBE's net-cash sign error (a well-formed LongTermDebtNoncurrent value last filed in 2015) and WDAY's share count (a well-formed dei value last filed in 2018). In all three cases the number was correctly sourced, correctly formed, non-null, and used against a mismatched basis. Calibration item D1 says "existence is not validity." These three say the stronger thing: correctness in one basis is not validity in another.

3. Quarterly series construction

CRWD's fiscal Q4 is not separately tagged in XBRL; only the full year is. Q4 is derived as the annual value less the three filed quarters, with contiguity enforced before any TTM is summed.

Quarter ended Revenue ($m) Derivation
2025-04-30 1,103.434 filed 10-Q
2025-07-31 1,168.952 filed 10-Q
2025-10-31 1,234.244 filed 10-Q
2026-01-31 1,305.375 derived: FY26 4,812.005 − (1,103.434 + 1,168.952 + 1,234.244)
2026-04-30 1,385.629 filed 10-Q

TTM revenue = 1,168.952 + 1,234.244 + 1,305.375 + 1,385.629 = $5,094.2m. Matches the screen exactly.

Revenue tag note: CRWD reports under RevenueFromContractWithCustomerIncludingAssessedTax — the Including variant — where CRM and WDAY use Excluding. A pipeline keyed on one variant only finds nothing for CRWD.

4. Net cash — the one the screen got right

Component 2026-04-30 ($m) Screen Mine
Cash and cash equivalents 4,552.8 ✓ ✓
Long-term debt (745.8) ✓ ✓
Restricted cash, noncurrent 123.7 included excluded
Strategic investments 66.3 excluded disclosed, excluded
Operating lease liabilities (current + noncurrent) (19.9) + (55.6) noncurrent as debt disclosed, excluded
Net cash ex-lease +3,807 +3,875 HEADLINE

A $68m difference on a $183bn company — immaterial, and CRWD is the one name of the five whose net cash the screen got essentially right. That deserves recording alongside the criticisms.

Tags checked per the brief: LongTermDebtNoncurrent 745.8 @ 2026-04-30 ✓ live · LongTermDebt absent · ConvertibleDebtNoncurrent absent · ShortTermInvestments reports 0 from 2024 onward and MarketableSecuritiesCurrent was last filed 2020-01-31 — CRWD holds its liquidity in cash equivalents, so there is no securities line to miss. Had a pipeline used the 2020 MarketableSecuritiesCurrent value it would have invented $647m of assets; the screen did not.

5. The commission-amortisation change — the accounting-quality finding

From the Q1 FY27 10-Q, Note 1, verbatim:

"In February 2026, the Company completed an assessment of the estimated period of benefit of commissions earned upon the initial acquisition of a contract, or subsequent upsell, and determined that it should increase from four to five years. This change in estimate was effective beginning in fiscal year 2027. Based on the carrying value of the related deferred contract acquisition costs as of January 31, 2026, the effect of this change in estimate for the three months ended April 30, 2026 was a reduction in sales commission expense of $27.9 million."

($m) As reported Ex the change
GAAP loss from operations (30.6) (58.5)
GAAP net income attributable to CrowdStrike +27.8 ~(0.1) or worse

32% of the operating-loss improvement from −$118.7m to −$30.6m is a change in accounting estimate. The maiden GAAP quarterly profit of $27.8m is smaller than the $27.9m the change contributed.

Extending the amortisation period of a capitalised cost lowers current-period expense and raises the deferred asset. Corroborating balance-sheet movement: deferred contract acquisition costs, noncurrent, rose $655.7m → $743.2m (+13.4%) while amortisation of deferred contract acquisition costs in the cash-flow statement fell $102.9m → $98.9m despite revenue growing 25.6%. The direction is consistent with the disclosure.

It is properly disclosed in Note 1. It is nowhere in the press release, which presents the quarter as a profitability milestone.

Three further items in the same $27.8m (Research §5b): +$36.4m of realized gains on strategic investments (cash-flow statement), a $6.9m tax benefit against a $21.1m expense a year earlier, and $18.2m of pre-NCI income attributed to non-controlling interests versus −$0.8m. Strip either the accounting change or the investment gain and the profit becomes a loss.

6. Adjustments applied, with reasons

# Adjustment Effect Why
1 Share count × 4 for the split market cap +300% §2. The defining correction.
2 Apply a 4.0 split factor to all P/S observations before 2026-07-02 percentile series internally consistent Otherwise the history is discontinuous at the split.
3 Add back the $27.9m commission-amortisation benefit GAAP operating loss (30.6) → (58.5) §5. A change in accounting estimate is not operating leverage.
4 Note the $36.4m investment gain, $28.0m tax swing and $19.0m NCI swing the $27.8m profit is non-operating §5. Disclosed rather than netted, because each is separately identifiable.
5 Anchor CRWD separately from the other four names exit multiple from a 20–34% growth cohort, not 7–20.6% Research §11. A single shared anchor across an 8.8%–25.6% spread is the diagnostics-set defect.
6 Declare the EBIT-basis exit multiple UNIDENTIFIED; run the primary test on sales primary output is sales-basis Valuation §2b: three of six growth-matched comparators have negative GAAP EBIT and the three positive ones span 49.9x–177.8x.
7 Use the five-year P/S percentile (90th), not the full history (72nd) percentile 90th The pre-2022 tail is the 2020–21 bubble at 52–76x sales. Using it would be a regime-change error.
8 Step shares up to ~1,033m for forward work −1.5% to forward per-share values Buyback covers 0.59x of SBC; the count rose 0.46% in Q1 FY27. One of only two names in the cluster where the forward count rises.
9 Score Quality PASS, correcting a screen FAIL INFLECTION margin-CHANGE test met on the run-rate Research §10. The screen measured FY2026 vs FY2025 — the July-19 damage year — not the run-rate.

No SBC adjustment. SBC is 22.8% of revenue — the heaviest in this cluster — and the −4.0% GAAP operating margin carries it. Ex-SBC operating income of +$804m (16.7%) is reported for context and is never used as a terminal margin. The guided 24.7% non-GAAP margin is run only as a sensitivity, and the name fails at 35% too.

7. What the model deliberately does not contain

8. Known weaknesses in my own numbers

Stated so a reviewer can attack them.

  1. I could not verify quoted spreads for five of the six deepest option strikes. The Alpaca snapshot returned a live quote for only CRWD261218C00165000. Open interest of 5,883 on the Jan-27 $175 strike establishes that positions exist; it does not establish today's tradeable spread, and I decline to infer it. The HCA precedent is that assuming depth without pulling it is how an uninvestable vehicle gets proposed.
  2. I have not established what the non-controlling interest is. $18.2m of Q1 FY27 pre-NCI income went to minority holders against −$0.8m a year earlier — a $19.0m swing on a $27.8m reported result — with $41.5m of NCI equity on the balance sheet. It is disproportionately large relative to earnings and the sections of the 10-Q I read do not identify the entity. This is a genuine gap in my understanding of this income statement and it is material to any GAAP-EPS forecast.
  3. The four tuck-in acquisitions' revenue contribution is assumed immaterial, not verified. $881.4m of cash plus $22.9m of replacement equity awards is not trivial, and goodwill rose $904m. If Seraphic and SGNL.AI contributed even $15m in the quarter, organic growth is ~24.5% rather than 25.6%. The company discloses no split and I did not locate a pro-forma revenue disclosure in the sections I read.
  4. The 8.7x sales-basis anchor rests on six observations spanning 3.39x–19.29x — a 5.7x range. That is very wide, and it is why the full range is carried through every sensitivity. The verdict turns positive only at the top observation (SNOW, −26.1% operating margin), which I regard as not a credible year-5 anchor for a name that should be profitable by then — but that is a judgement, and a reader who accepts SNOW's multiple as the anchor gets +2.2pp instead of −19.2pp.
  5. The year-5 deceleration cross-check assumes CRWD decelerates to 15–18% by 2031. If it holds above 20%, the bracketing comparators shift toward FTNT's 15.52x and the anchor rises materially. I checked this and reported that the median moves to ~8.8x, within 1% of the base — but that check assumes the current comparator multiples persist.
  6. RPO went backwards ($9.0bn → $8.8bn) and I have treated it as Q1 seasonality. The year-on-year comparison ($7.9bn at Q3 FY26) supports that reading, but I have not verified CRWD's RPO seasonality across multiple years.
  7. The P/S history uses a 45-day post-period-end price proxy, not actual filing dates.
  8. The 30% probability on the downside case is a judgement, not a computed figure. It is anchored on the fact that the scenario has already occurred once (July 2024) with a measured effect on DBNRR (119% → 112%), which is better grounding than most, but it remains a judgement and will be Brier-scored as one.