S&P 500 / AI P&L

30 days ending Aug 17, 2026

Download data

What 30 days of earnings disclosures actually show

AI talk is common.
The financial story is thin.

Filings often mention AI. Far fewer companies explain what they spend, what changed financially, or whether the investment pays back.

Inspect a company Show me the evidence
How S&P 500 AI financial evidence branches Of 500 frozen issuers, 433 reported in the window and 67 did not. The 433 reporters split only into 411 non-provider adopters and 22 AI providers. All four evidence branches use the 411-adopter base. The 0 of 411 comparable-pair result means the comparison is missing; it is not a return rate. Frozen universe 500 / 500 · 100% Reported in window 433 / 500 · 86.6% No report in window 67 / 500 · 13.4% Non-provider adopters 411 / 433 · 94.9% AI providers 22 / 433 · 5.1% Named a specific internal AI use 47 / 411 · 11.4% SPENDING Said they spend on AI 18 / 411 · 4.4% Gave a dollar number 3 / 411 · 0.7% Isolated AI dollar total 1 / 411 · 0.2% MEASURED RESULT Measured a result 13 / 411 · 3.2% Issuer cost effect 2 / 411 · 0.5% Cigna: customer medical-cost result excluded from issuer P&L 2027 FUTURE AI-only 2027 dollar guide 1 / 411 · 0.2% RECONCILIATION Comparable pairs 0 / 411 · not a % Not a downstream rung Missing comparison ≠ zero return The 13 measured results are not a subset of the 3 spending numbers.

  1. How common is AI talk versus a financial story?

    253 / 428 · 59.1% of unique periodic filers mentioned AI or machine learning; 47 / 411 · 11.4% of adopters named an internal financial or operating path.

  2. Who says they are spending?

    18 / 411 named a spending path, 3 / 411 gave a number, and only Incyte gave a ring-fenced total.

  3. Who measured a result?

    13 / 411 measured something, usually operations; 2 / 411 tied AI to a current issuer cost effect.

  4. What is actually guided for 2027?

    1 / 411: Broadridge’s $25M FY2027 AI-driven productivity outlook. It is forward, not realized.

  5. Did anyone show spend pays for itself?

    0 / 411 showed a comparable pair. The 10% return calculation is NOT COMPUTED, not zero.

  6. What about AI sellers?

    22 providers sit outside the 411 and are shown separately so their much larger dollars do not distort adopter percentages.

  7. How complete?

    All 500 have a disposition, all 433 reporters were adjudicated, and 0 remain unresolved.

500 issuer universe

433 reporters adjudicated

411 adopters

22 providers

0 unresolved

S&P 500 universe at the August 17, 2026 close · Disclosure window: July 19–August 17, 2026, America/New_York · A 30-day earnings window, not a year of AI activity

How to read each claim

Confidence in the wording, not the company.

Recorded

The audited record supports exactly the narrow wording shown.

Constrained

The source is real, but mixed attribution or scope makes the everyday interpretation too broad.

Not shown

No qualifying disclosure was found. This never means zero activity or return.

Coverage
All 433 reporters were searched and all 500 issuers have a window disposition.
Citation
The highlighted sentences link to retained primary documents.
Meaning
Counts follow separate gates for spending, operating results, issuer financial effects, guidance, and reconciliation.

This record cannot reveal undisclosed AI, the blended-program AI share, net profits, or annual S&P prevalence.

Show me the evidence

Read the sentence. Then read its limit.

03 / Company evidence

Explore every company.

Search all 433 in-window reporter records. Open any row for the disclosure, the limit on its meaning, and the retained source.

Loading company evidence…

Open a row to read the disclosure, its limit, and the source.

04 / Providers

Provider economics are a separate story.

Twenty-two AI technology providers are excluded from the 411-company percentages. Their larger spending and revenue evidence generally still lacks an isolated bottom-line return.

22provider-side issuers

Nine core providers and thirteen AI technology vendors are retained for comparison, not blended into the 411-company adopter denominator.

05 / Methodology

Strict by design.

The study grades what companies actually disclosed during one fixed window. It does not fill silence with estimates.

Which denominator applies?

500 frozen issuers
The complete coverage frame; every issuer has a window disposition.
433 in-window reporters
Every reporter was adjudicated, including five with limited source access.
428 unique periodic filers
This denominator applies only to the 253 mention rate: 59.1% explicitly named AI or machine learning.
411 non-provider adopters
This denominator applies to spending, measured results, 2027 guidance, and comparable pairs.
22 providers excluded
Provider economics remain visible but do not enter the 411-company percentages.

Five access-limited source sets stay in the denominator.

  1. 01

    Absence is not zero.

    No disclosure means no qualifying evidence was found in the retained sources. It does not mean zero activity, spend, or benefit.

  2. 02

    Providers stay separate.

    External AI product and infrastructure economics do not enter the adopter numerator. Incidental vendors remain only when they fail the provider materiality test.

  3. 03

    Operational benefit is not P&L.

    Time saved, usage, seats, model quality, and anecdotes need an economic bridge before they can qualify as realized financial gain.

  4. 04

    The denominator and window are frozen.

    The observation window is July 19–August 17, 2026. Each percentage names its applicable denominator above.

  5. 05

    Blended totals are not AI-only.

    PayPal’s $1.5B target covers a broader restructuring, AI, and automation program, so it cannot become an AI-only savings amount.

  6. 06

    0 / 411 comparable pairs means return is not computed.

    A missing spend-to-return comparison is not a 0% return. The 10% return test is disabled until a valid dated cash-flow method exists.

Research codes used in the downloadable ledger

These compact codes are preserved for auditability. The page leads with their plain-English meaning.

D0–D2
From no qualifying internal discussion to a named internal AI path.
S0–S4
From no spending path shown to increasingly specific quantified spending evidence.
B0–B4
From no result shown through operating measures, forward benefits, and current financial effects.
G0–G4
From no dated financial outlook to increasingly specific or blended guidance.
J0–J2
Whether disclosed spending and return can be reconciled on a comparable basis.

06 / Downloads

Audit the work.

The site is a browsing layer. These retained artifacts remain the source of truth.

Final reportMarkdown · complete narrative Company evidenceCSV · 433 company records Source coverageCSV · denominator and source-access status Statistics ledgerCSV · machine-derived rates Issuer dispositionsCSV · all 500 issuers Final auditJSON · counts, hashes, and PASS status