Phase Space AI

Financial Model Notes

Kodiak Gas Services [KGS]

Kodiak Gas Services [KGS] — Financial Model Notes & Defect Register

Screen record validation

Screen: reports/scan_universe/KGS_analysis.json, as-of 2026-07-30, Quality PASS, Valuation PASS at +13.8pp.

screen field screen value verified value source result
shares 88,771,027 88,771,027 10-Q cover page 2026-05-06 exact
revenue_ttm 1,324,217,000 1,324,217,000 four filed quarters exact
net_cash −2,692,640,000 −2,692,640,000 cash $94,363k − debt $2,787,003k exact
ev 7,921,253,490 7,921,253,490 exact, and LEASE-EXCLUSIVE
op_margin_pct 26.0 25.99 (FY2025 filed) EDGAR OperatingIncomeLoss correct — screen used EDGAR, not AV
revenue_cagr_demonstrated 22.7 4.9 organic Q1'26 vs Q1'25 invalid — acquisition artefact (screen's own note flags it)
exit_multiple 24.2, GROWTH_MATCHED, peer_n: 223 7.5x ROIC identity invalid — 223-name "peer" set is a universe median
terminal_margin 0.26, "own trailing operating margin" 0.270 own TTM filed correct in method, one quarter stale
required_cagr_pct 8.9 28.2 reverse_dcf, identity exit multiple wrong by 19.3pp
valuation_margin_pp +13.8 −23.3 sign flip, 37.1pp swing

The screen's inputs are almost all correct. Its verdict is not. Every balance-sheet and revenue figure ties to the filing exactly — this is the cleanest screen record validated in this cluster. The failure is entirely in two derived parameters: an acquisition-contaminated demonstrated CAGR and a 223-name exit-multiple anchor.

Defects found — Alpha Vantage

1. NEW DEFECT — AV operatingIncome overstates KGS filed income from operations

Not on the DATA_DEFECTS.md list. Measured:

period AV operatingIncome EDGAR filed OperatingIncomeLoss error
FY2023 $244.1m $244,110k 0.0% — clean
FY2024 $249.4m $249,450k 0.0% — clean
Q1 2025 $100.6m $89,186k +12.8%
Q2 2025 $106.3m $99,730k +6.6%
Q3 2025 $104.0m $64,200k +62.0% (+$39.8m)
Q4 2025 $103.7m $86,894k +19.3%
Q1 2026 $116.5m $106,812k +9.1%
FY2025 $414.6m $340,010k +21.9% (+$74.6m), 5.7pp of margin

The excluded lines are Kodiak's loss on sale of assets ($61.6m FY2025) and long-lived asset impairment ($6.3m). AV was clean on FY2023 and FY2024, when those lines were smaller — so this is not "AV is always wrong on this field," it is AV drops certain operating charges and the error scales with them. Per the defect-list discipline: this is a hypothesis with measured instances, not a universal rule.

Consequence if unchecked: a terminal margin set from AV would be 31.7% instead of 27.0%, and steady_state_check.py — which consumes AV — reported KGS's "latest margin" as 31.7% for exactly this reason.

2. av_vs_edgar.py returned VERIFIED while the above was true

sym   cmp  dis  worst  inv  cal  outcome
KGS     4    0     --    0    0  VERIFIED

Four periods compared, zero disagreements. The tool compares revenue; it does not compare operating income. A VERIFIED result from it is evidence about the top line only. Recorded so the next reader does not treat VERIFIED as a clean bill of health for the income statement.

3. Fields checked and found CLEAN on KGS — reported per instruction

field result
SPLITS queried explicitly, returned {"symbol":"KGS","data":[]}no split, no reverse split
cashAndShortTermInvestments not applicableshortTermInvestments and longTermInvestments are both absent from the balance sheet, so there is nothing for the composite to silently exclude. Cash $94,363k ties to the 10-Q
commonStockSharesOutstanding not used; the 10-Q cover-page figure was used instead and it matches the screen exactly
revenue, all quarters ties to EDGAR to the dollar
EARNINGS_ESTIMATES returned 16 records with full 7/30/60/90-day revision history — not the empty-array failure seen on KLAC/ZS/NOW
EARNINGS_CALL_TRANSCRIPT 9 of 12 quarters, speaker/title/sentiment populated
ebit, ebitda, sellingGeneralAndAdministrative not used — all opex lines taken from the FY2025 10-K

4. Defect in our own tooling — normalized_fcf.py

Documented in full in KGS_Research.md §6. Two offsetting errors on this name: D&A-as-maintenance-capex overstates sustaining spend 2.0x (21.1% vs a disclosed 10.6%), and revenue-per-dollar-of-growth-capex overstates capital efficiency ~10x ($1.44 vs a physical $0.14) by crediting the CSI Compressco revenue step to organic capex. They cancel to within 1pp here. They will not cancel on a name that is acquisitive without being capital-intensive.

Model construction notes