About

An economic intelligence report, built to be checked.

The NLT143 AI Economy Index asks where AI is creating economic value in America, who receives it, what it costs, and what would make the gains broadly shared. Report 001 covers data through November 2026.

Who made this

Researched, designed and built by David T Phung for NLT143 RESEARCH. Corrections and new sources are welcome: message @davidtphung on X.

Follow @davidtphung

Why “Index”

Like other economic index reports, this is a series of measures, not one number. No composite “AI prosperity score” is published, because adoption, output, wages and community outcomes are measured in different units by different sources. Each is shown in its own units first.

Independence

No sponsor. Several sources are AI or software companies with a commercial interest in AI adoption (Anthropic, Ramp, OpenAI, McKinsey). That is disclosed next to each one; it is not treated as proof that their numbers are wrong.

How to read the labels

Observed
Measured and published by the source for that period.
Modeled estimate
A number produced by a model or estimate (e.g., power from satellite imagery).
Estimated effect, one study
An effect estimated in one study's sample; not a national total.
Forecast
A projection about the future.
Scenario, not a prediction
Assumption-driven paths you can change; not predictions.
Claim, unverified
Stated by a company or advocate; not independently verified.
Insufficient evidence
Not enough evidence to say.

Every chart also states its unit, geography, period, source and retrieval date, and has an “Inspect the evidence” drawer. Missing is never shown as zero.

Sources

Business adoption and outcomes

Who uses AI, how much they pay, and what changed afterwards.

Anthropicdownloaded and validated

Anthropic Economic Index release 2026-06-26 ('Cadences')

Used for. Which tasks and occupational groups people bring to Claude, by state and country; work vs personal use.

Type
vendor telemetry
Covers
Global; 121 countries published (May 2026); U.S. states + D.C. + Puerto Rico (overall metrics only), 2026-04-01 to 2026-05-31
Counts
Sampled Claude.ai chat and Cowork conversations (Free, Pro, Max) and 1P API traffic excluding Claude Code
License
CC BY
Interest
Anthropic (model developer; commercial interest in AI adoption)

Limits. One platform. Task labels are not users' occupations. Consumer plans and 1P API only; Claude Code reported separately. Suppressed cells mean shares sum to under 100%.

Ramp Economics Labpublished aggregate only

Ramp AI Index

Used for. Monthly paid AI adoption by vendor, sector, size and state; spend per employee; token spend and volume shares.

Type
vendor telemetry
Covers
United States, businesses on Ramp; state, NAICS sector, size class, vendor breakdowns, 2023-07 to 2026-10
Counts
85,000+ U.S. businesses using Ramp corporate cards and bill pay (per Ramp)
License
Ramp website terms; reuse terms not verified
Interest
Ramp (spend-management vendor)

Limits. Not a probability sample. Misses free tools and personal accounts (Ramp says likely undercount). Paying is not productive use.

U.S. Census Bureaudownloaded and validated

Business Trends and Outlook Survey (BTOS): AI / 'Super Intelligence' core questions

Used for. The national, state, sector and firm-size measures of business AI use; the only representative adoption series here.

Type
representative survey
Covers
U.S., 50 states + D.C., NAICS sectors, size classes, top 25 MSAs, 2023-09-11 to 2026-11-01
Counts
Employer businesses (excluding farms) in the BTOS sample frame
License
Public domain

Limits. Low response rates; questionnaire breaks; no collection Oct 6 to Nov 16, 2025 (funding lapse).

OpenAInot yet verified

The State of Enterprise AI 2025

Used for. Not used: the page refused automated retrieval.

Type
customer survey
Covers
Enterprises (global)
Counts
OpenAI enterprise customers + worker survey
License
Proprietary
Interest
OpenAI (model developer)

Limits. Returned HTTP 403 to automated retrieval on 2026-11-18; not used for any figure.

Sources

People and places

Jobs, wages, output and communities.

Ramp and Revelio Labsdownloaded and validated

A New Look at AI's Impact on Jobs: Firm-Level AI Spending and Workforce Adjustment (Kharazian, Simon, Stevens)

Used for. The one firm-level study of AI adoption and hiring, reviewed on the Job accounting page.

Type
observational study
Covers
United States, linked Ramp/Revelio firms, 2021-01 to 2026-02
Counts
21,559 firms linked by web domain; 15,926 never, 3,969 low, 1,664 high intensity
License
Public working paper
Interest
Ramp and Revelio Labs (vendors)

Limits. Selected sample of Ramp customers; high-intensity firms small (mean 26.7 employees) and tech-adjacent (63.8%); 3 of 11 pre-periods flagged for the headline; significant only in Information.

U.S. Bureau of Labor Statisticsdownloaded and validated

Current Employment Statistics (CES): employment and average hourly earnings by industry

Used for. Payroll jobs and hourly pay for every sector and for AI-supply industries (data processing, electrical equipment, trades).

Type
administrative record
Covers
United States, 2019-01 to 2026-10
Counts
Nonfarm payroll jobs
License
Public domain

Limits. Jobs, not people; excludes farms and self-employed. Benchmark revisions each February.

U.S. Bureau of Labor Statisticsdownloaded and validated

Occupational Employment and Wage Statistics (OEWS), May 2025, national

Used for. Employment and wages for occupations in the news: radiologists, nurses, teachers, truck drivers, electricians, developers.

Type
administrative record
Covers
United States, 2025-05 to 2025-05
Counts
Wage and salary workers in nonfarm establishments
License
Public domain

Limits. Only latest year retrieved; trend analysis (e.g., radiologists 2016-2025) needs flat files.

U.S. Bureau of Economic Analysisdownloaded and validated

GDP by Industry: value added, real value added indexes, contributions to growth

Used for. Real output and each industry's contribution to GDP growth.

Type
administrative record
Covers
United States, 2005Q1 to 2026Q2
Counts
All industries
License
Public domain

Limits. Revised annually; quarterly industry detail lags GDP.

Federal Reserve Bank of St. Louis (FRED)downloaded and validated

Business applications (Census BFS), output per hour (BLS), real GDP (BEA)

Used for. Business applications, output per hour and real GDP context.

Type
administrative record
Covers
United States
Counts
Various
License
Public domain source series

Limits. Republication; vintages follow the source agencies.

Shelby County Health Department; TVA; MDEQ; Memphis City Councilnot yet verified

Colossus permits, utility approvals and council actions (as reported)

Used for. Permits, utility approvals, lawsuits and claims for the Colossus case study.

Type
journalism
Covers
Memphis and Shelby County, TN; Southaven, DeSoto County, MS
Counts
xAI / SpaceXAI facilities
License
Public records (accessed via reporting)

Limits. Tax bills, PILOT files, MLGW contracts and permit PDFs not yet obtained.

Sources

AI supply

Models, compute, chips, data centers, construction and supply chains.

U.S. Census Bureaudownloaded and validated

Value of Private Construction Put in Place: Data center

Used for. Monthly construction spending on data centers since 2014.

Type
administrative record
Covers
United States, 2014-01 to 2026-09
Counts
Private construction projects
License
Public domain

Limits. Excludes chips and IT equipment; preliminary values revise.

U.S. Census Bureaupublished aggregate only

Employment in Data Centers Increased by More Than 60% From 2016 to 2023

Used for. The widely cited 306,000 to 501,000 data-center jobs figure, audited for its industry boundary.

Type
administrative record
Covers
United States, states, counties, 2016 to 2023
Counts
Workers in the data-center industry (QWI; BLS for national counts)
License
Public domain

Limits. Article does not print the NAICS code; its national figures (306k to 501k) match the scale of NAICS 518 (data processing, hosting), which includes more than physical data-center operations.

Stanford Institute for Economic Policy Research (SIEPR)downloaded and validated

AI data centers: A hidden driver of U.S. jobs risks becoming a missed opportunity

Used for. Modeled manufacturing jobs supported by data-center construction.

Type
modeled forecast
Covers
United States, 2026 to 2026
Counts
Domestic manufacturing industries supplying data-center components
License
Public policy brief

Limits. Not a count of hires; excludes substitution away from other uses of the same inputs (which would lower it); depends on a third-party capex forecast.

Epoch AIdownloaded and validated

Data on AI Models (all, notable, frontier)

Used for. Training compute of notable models and the release tracker (language, video, audio, image, science).

Type
compiled dataset
Covers
Global (organization country recorded), 1950 to 2026-11-17
Counts
AI models meeting Epoch inclusion criteria
License
CC BY 4.0

Limits. Recent months under-collected; compute often estimated.

Epoch AIdownloaded and validated

Frontier Data Centers

Used for. Power, chips and status of 77 tracked U.S. AI data centers, including Colossus 2 in Memphis.

Type
compiled dataset
Covers
Global (77 of 96 sites in the U.S.)
Counts
Large AI data centers tracked by Epoch
License
CC BY 4.0

Limits. Estimates, not metered power; capacity is not utilization.

Our World in Dataaccessible but not yet downloaded

Computation used to train notable AI systems

Used for. Cross-check only: it republishes Epoch, so it is not counted as a second confirmation.

Type
compiled dataset
Covers
Global
Counts
Notable AI systems (from Epoch)
License
CC BY

Limits. Same evidence family as Epoch: not independent confirmation.

U.S. Securities and Exchange Commissiondownloaded and validated

EDGAR XBRL company facts (19-issuer AI-economy cohort)

Used for. Revenue, capital spending and cash flow for 19 AI-economy companies, with accession numbers.

Type
corporate filing
Covers
U.S.-listed issuers (consolidated, global operations), FY2019 to FY2026
Counts
Defined cohort; not all public companies
License
Public domain

Limits. Universe file (company_tickers_exchange.json) returned 403; total capex is not AI capex; employees mostly untagged.

Sources

Forecasts and models

Projections, kept apart from observations.

Penn Wharton Budget Modelpublished aggregate only

Projected Impact of Generative AI on Future Productivity Growth

Used for. A U.S. GDP projection, compared in the Scenarios chapter.

Type
modeled forecast
Covers
United States
Counts
U.S. economy
License
Public

Limits. Assumes 23% of exposed tasks automated and near-full adoption within ~15 years.

NBER (Daron Acemoglu)published aggregate only

The Simple Macroeconomics of AI (w32487)

Used for. A productivity bound, compared in the Scenarios chapter.

Type
modeled forecast
Covers
United States
Counts
U.S. economy
License
Working paper

Limits. Built on early task-level studies; excludes new tasks/products.

McKinsey Global Institutenot yet verified

The economic potential of generative AI: The next productivity frontier

Used for. A global value-potential estimate, shown only as labelled and not re-verified.

Type
modeled forecast
Covers
Global
Counts
63 modeled use cases
License
Proprietary report
Interest
McKinsey (consulting; sells AI implementation services)

Limits. Page refused automated retrieval on 2026-11-18; figure not re-verified this pass. Global, potential, not realized U.S. GDP.

Refresh and reuse

Research cutoff 2026-11-18. Every source is fetched by a script that stores a dated, hashed copy, so any number can be traced back to the exact file it came from. Public-domain and Creative Commons data can be downloaded from the Methods page. Ramp series are shown with attribution and are not offered for download until their reuse terms are confirmed.

Credits

Anthropic Economic Index (CC BY). Epoch AI (CC BY 4.0). Ramp Economics Lab. U.S. Census Bureau, Bureau of Labor Statistics, Bureau of Economic Analysis, SEC EDGAR, Federal Reserve Bank of St. Louis (FRED). Stanford SIEPR. Penn Wharton Budget Model. NBER. Map geometry: U.S. Census Bureau cartographic boundaries via us-atlas.