Artificial intelligence is spreading quickly through parts of the U.S. economy and is reshaping how some companies work while barely touching others. Early adopters are already using AI to automate tasks and analyze data, but the basic ingredients for broader adoption, such as skilled workers, reliable broadband, and jobs that pay enough to attract and keep talent, remain unevenly distributed across states.
A new report by National University maps an emerging “AI gap belt” of states that are structurally more vulnerable as AI adoption accelerates. The study builds a state-level AI gap risk score that ranks where communities are most at risk of being left behind, combining four ingredients—business AI use, educational attainment, broadband quality, and median wages—into a single 0–100 number. Higher scores mark states on the AI gap belt with greater risk, while lower scores signal stronger conditions for adopting and benefiting from AI.
Key Takeaways
- West Virginia sits at the top of the AI gap risk scale with a score of 100, followed by New Mexico (89.1) and Louisiana (86.8), putting them among the most vulnerable states in the country.
- At the other end, the District of Columbia anchors the low‑risk side. It has the strongest education‑and‑wages foundation score at 100.
- North Dakota leads on broadband quality with a score of 100, whereas Alaska sits at the bottom with 0.0, and New Mexico and West Virginia also lag at 20.1 and 30.1.
Overall AI Gap Risk Score
The AI gap score summarizes each state’s overall risk of falling behind on AI, after combining AI usage, education, broadband, and wages into a single score. Higher values mean more risk relative to other states; lower values mean stronger overall conditions.
Highest risk
- West Virginia (WV) 100
- New Mexico (NM) 89.1
- Louisiana (LA) 86.8
- Mississippi (MS) 86.4
- Wisconsin (WI) 86.1
- Alabama (AL) 85.5
- Arkansas (AR) 84.1
- Alaska (AK) 83.1
- Iowa (IA) 82.0
- Oklahoma (OK) 81.9
Lowest risk
- Arizona (AZ) 58.8
- Connecticut (CT) 58.8
- Nevada (NV) 58.1
- Massachusetts (MA) 58.0
- Washington (WA) 55.7
- Delaware (DE) 52.1
- Utah (UT) 51.1
- Maryland (MD) 50.5
- Colorado (CO) 38.5
- District of Columbia (DC) 0
Education and Employment Conditions
This section captures the strength of each state’s talent and earnings base, combining the share of adults with at least a bachelor’s degree and the typical annual wage into a single standardized score. Higher values mean stronger “foundations” for AI adoption; lower values mean thinner pipelines of skilled workers and weaker wage structures.
Strongest conditions
- District of Columbia 100
- Massachusetts 51.9
- Colorado 47.5
- Maryland 42.9
- New Jersey 42
- Washington 40.7
- Connecticut 40.1
- Vermont 38.5
- New York 38.3
- Virginia 36.4
Weakest conditions
- West Virginia 0.0
- Mississippi 0.1
- Arkansas 2.0
- Louisiana 5.5
- Kentucky 7.4
- Oklahoma 7.8
- Alabama 8.5
- Nevada 8.9
- Indiana 12.4
- New Mexico 13.1
AI Adoption
This section measures how many businesses in each state say they are currently using AI, based on responses to a recurring national business survey. Higher values mean that a larger share of surveyed firms in that state report using AI tools or systems in their operations; lower values mean relatively fewer firms are using AI today.
Highest adoption
- District of Columbia 100.0
- Delaware 65.4
- Colorado 63.7
- Arizona 53.7
- Nevada 51.6
- Utah 50.6
- Florida 47.0
- Maryland 44.3
- Washington 43.6
- Wyoming 43
Lowest adoption
- West Virginia 0.0
- Wisconsin 6.6
- Iowa 7.1
- Puerto Rico 10.6
- Maine 11.9
- Alabama 12.2
- Mississippi 12.6
- New Hampshire 13.6
- New York 13.9
- Oklahoma 14.1
Infrastructure Broadband Quality
This section summarizes how strong and “future-ready” each state’s internet infrastructure is, using an index of broadband quality that reflects the share of locations able to access higher-speed service tiers. Higher values mean better broadband quality and capacity; lower values point to weaker digital infrastructure.
Best infrastructure
- North Dakota 100.0
- Nevada 94.0
- Connecticut 92.9
- Utah 90.1
- New Hampshire 87.2
- Tennessee 86.2
- Kansas 81.5
- Nebraska 81.1
- New Jersey 78.8
- District of Columbia 78.2
Weakest infrastructure)
- Alaska 0.0
- New Mexico 20.1
- West Virginia 30.1
- Montana 30.5
- Idaho 37.4
- Wisconsin 39.2
- Rhode Island 39.5
- Michigan 42.4
- Virginia 44.1
- Louisiana 46.9
This report shows that the AI gap is not an abstract, national-level issue but a concrete, geographic one, concentrated in an emerging AI gap belt that runs through states like West Virginia, New Mexico, and Louisiana. States in this belt face overlapping weaknesses in AI adoption, education, broadband, and wages, while places such as the District of Columbia, Colorado, and Massachusetts benefit from reinforcing strengths that keep their risk low.
Data by State
Methodology
Data Sources
This index is built entirely from publicly available federal data so that the results are transparent and reproducible.
- Business AI usage (BTOS)
Comes from the U.S. Census Bureau’s Business Trends and Outlook Survey, which regularly asks firms whether they are currently using AI in their operations. The AI adoption input is the average share of “yes” responses across recent survey waves. - Education levels (ACS)
Uses the American Community Survey from the U.S. Census Bureau to measure the share of adults age 25 and older with at least a bachelor’s degree in each state. - Broadband quality (BDC / FCC broadband data)
Draws on federal broadband data that report where higher-speed internet service is available. These data are summarized into a broadband quality index capturing access to faster, more “future‑proof” connections. - Earnings and wages (OEWS)
Uses median annual wage data from the Occupational Employment and Wage Statistics program to proxy local labor‑market strength and the ability to attract and retain talent.
Understanding the AI Gap Risk Score (0–100 scale)
The AI gap risk score is a combined “risk of being left behind” measure for each state. It is designed so that higher numbers mean more risk, bundling four underlying signals into one headline metric that is easy to compare across states.
The main change is presentation, not mechanics. The composite is still built from standardized inputs, but the result is now reported on a 0–100 scale, where:
- 0 = lowest-risk state in the dataset
- 100 = highest-risk state in the dataset
Every other state is placed in between based on where it falls relative to the best and worst performers. That makes the score immediately interpretable without needing to understand z-scores.
The Four Components
1. Low AI usage today (BTOS business survey)
This component is based on the average share of firms answering “Yes” to using AI in recent waves of the Business Trends and Outlook Survey. Because AI usage is treated as a positive:
- Lower AI usage compared with other states → higher AI gap risk
2. Low education pipeline (ACS BA or higher)
This component looks at what share of adults age 25 and older have at least a bachelor’s degree in each state. In other words, it asks: “Out of all adults 25+, how many have a four‑year degree or higher?”
Because having more college‑educated adults is a strength, states with a smaller share of adults with a bachelor’s degree or higher end up with a higher AI gap risk score.
Since higher educational attainment is a positive:
- Lower BA+ share → higher AI gap risk
3. Weak broadband capacity (broadband quality index)
This component uses a broadband quality index summarizing how many locations can access higher-speed internet tiers, as a proxy for whether a state’s digital backbone can support more data‑intensive work and tools.
Because better connectivity is a positive:
- Lower broadband quality → higher AI gap risk
4. Lower earnings power (OEWS wages)
This draws on median annual wage as a proxy for labor‑market strength and a state’s ability to attract and retain talent.
Because higher wages are treated as a positive:
- Lower median wages → higher AI gap risk
Because the four inputs are measured in very different ways (survey shares, percentages, index scores, and dollar wages), the first step is to put them on a common scale. In practice, that means converting each input into a standardized value that shows how far a state is from the national average, rather than using the raw units.
Once everything is on that common scale, each “good” factor is flipped so that being low on it always raises the risk score. In other words, less AI use, fewer adults with a bachelor’s degree, weaker broadband, and lower wages all move the index in the same direction: toward higher risk.
Those four pieces are then combined with weights to create a single “raw” risk score, where AI usage gets the largest weight, followed by education, broadband, and wages. This raw score is the statistical core of the index.
Finally, that raw score is stretched and shifted onto a 0–100 scale for readability. The lowest‑risk state in the data is set to 0, the highest‑risk state is set to 100, and every other state falls somewhere in between based on its position relative to those two endpoints. The published 0–100 number is therefore just a clearer, rescaled version of the same underlying composite.
How to Interpret the Score
The score is relative, not absolute. It is meant to show how a state compares to others in the dataset, not to provide a literal probability of harm.
- A state near 100
One of the most at‑risk states overall. It typically lags on AI usage and is also weaker on one or more of: bachelor’s degree attainment, broadband quality, and wages. - A state near 50
Roughly in the middle of the pack on the weighted combination of the four components. - A state near 0
Among the least at‑risk states, generally showing stronger AI adoption and/or stronger fundamentals: higher BA+ share, better broadband, and higher wages.
The weights reflect a practical view of what matters most for near‑term AI diffusion:
- 0.45 on AI usage because observed business adoption is the most direct near‑term signal of AI diffusion.
- 0.25 on BA+ attainment because talent is a slow‑moving constraint with long‑run implications.
- 0.20 on broadband quality because digital infrastructure is a core enabler, especially outside major metropolitan areas.
- 0.10 on wages because wage levels overlap with education and industry mix but still carry a useful signal about local labor‑market strength.
