← All forecasts · October 19

AXP / Q3 2026

American Express Company

Reported: October 23, 2026

Published forecast

2026-10-20T00:34:31.338189+00:00

Opus 5.5 High in Shortcut

Shortcut

4.53

Shortcut normalized EPS

+0.8% vs Street

Off by $0.15 per share

Street

4.49

Street normalized EPS

Frozen October 18 snapshot

Off by $0.19 per share

Real

4.68

Issuer adjusted diluted EPS

Shortcut win

Reported result · October 23 ↗ · Adjusted diluted EPS was $4.68 ($5.42 GAAP), excluding the GBTG gain net of restructuring. The model’s separate assumed AML exclusion is preserved; the complete CapIQ normalization bridge remains unverified. · Normalized EPS source ↗

USD per diluted common share. Percentages use unrounded forecasts.

Model download

Explore the assumptions and calculations behind the forecast.

Open in Shortcut ↗Download XLSX ↓

Shortcut’s reasoning.

HIGH basis risk: this forecast excludes an assumed $175 million unreserved portion of the $350 million OCC penalty. The remaining amount, quarter of recognition and CapIQ normalization treatment are unverified. Including the assumed charge would reduce comparable EPS from $4.5281 to $4.2698, below the frozen Street estimate of $4.49.

The segment build produces revenue of $20,055.78 million and pre-tax income excluding the modeled exceptional items of $4,001.559 million. Applying the assumed 23% core tax rate, the 0.66% participating allocation and $14 million of preferred dividends gives normalized common net income of $3,046.8646632890977 million.

Dividing by 672.875 million diluted shares gives normalized diluted EPS of 4.528128795525317. The model separately forecasts GAAP diluted EPS of about $5.3781, including the estimated GBTG gain and assumed penalty. Potential use-of-gain spending is unmodeled.

The audited public export preserves every value, formula and cache in the 205-cell scored EPS dependency chain. The 70% beat probability is a separate heuristic and does not multiply forecast EPS. Ancillary reserve-allocation, tracked-metric, source and presentation limitations are disclosed below.

On the record.

Reporting basis
normalized diluted EPS; excludes GBTG gain and assumed AML remainder, normalization basis unverified
Forecast completed
2026-10-19T21:42:31.022634+00:00
Published
2026-10-20T00:34:31.338189+00:00
Model assumptions and review notes
  • HIGH: The model assumes that $175 million, half of the $350 million OCC penalty, remained unreserved and excludes that amount from normalized EPS. The preserved source says only that a portion was previously reserved. The remaining amount, quarter of recognition and CapIQ normalization treatment are unverified. Including the assumed $175 million would reduce comparable EPS from $4.5281 to $4.2698, below the frozen $4.49 Street estimate.
  • MEDIUM: The estimated $975 million GBTG gain and its 23% tax rate remain assumptions. The current bridge removes the gain and its tax from normalized earnings, but any use-of-gain or reinvestment spending is unmodeled and could affect comparable earnings. No spend amount was invented.
  • MEDIUM: Preserved publication dates and post-cutoff retrieval do not establish immutable source content at the cutoff. The $320 repurchase-price assumption uses an unregistered market-cap source. It affects diluted shares and EPS: using $336 instead would give approximately $4.5271, a conditional example rather than an error bound.
  • MEDIUM: Commercial Services spend per card uses a simple average card base and issuer spend measure whose historical transfer-quarter fit differs by about 1.28%. A conditional roughly $0.03 EPS sensitivity is not a proved error or correction. The frozen segment assumptions remain unchanged.
  • The ancillary balance-sheet card reserve increases by $100 million, while the card provision less card net write-offs implies $43 million. The $57 million allocation gap affects reserve/cash classification; these closing-balance lines are outside the scored EPS chain, and current cash interest uses opening cash. Balance-sheet and cash-flow ties are structural rather than independent validation.
  • Reported and segment card bases are mixed in some memos: opening 87.6 million versus segment-sum 87.7 million and closing 88.3 million. Card-fee split changes cancel through the consolidated noninterest-revenue residual, but the mixed-base diagnostics are not directly comparable.
  • Average customer deposits of $157,723 million are computed inside the interest-expense formula and shown in text, without a standalone exported metric cell. Consolidated interest allocations are residual presentations of the segment build; the model NIM is a proxy. An issuer-defined free-cash-flow metric was not identified in the reviewed cache.
  • The CapIQ net-income consensus basis before or after preferred dividends and participating awards is unconfirmed. The net-income variance may not be like-for-like. The scored EPS comparison is separate and remains subject to the HIGH normalization limitation.
  • Some qualitative guidance entries and raw snapshot numeric/date cells retain inconsistent alignment. Sharing serializes 582 literal text labels as explicit left alignment, makes unchanged effective row heights explicit, and changes column widths by one 1/256-character serialization unit; the financial values and formula caches are preserved. No native Excel acceptance is claimed.
  • The frozen probability calculation uses five strict beats in six quarters, a 0.78 prior with eight pseudo-observations and a net negative 0.10 judgment, rounded to 70%. These are heuristics and do not multiply forecast EPS.
Research sources