• The share of US businesses using AI rose from 17% to 25% in 10 months, climbing in 22 of the nation’s 24 big metropolitan areas.

• Denver, Phoenix and DC lead the country. Dallas-Fort Worth, Baltimore and Philadelphia are higher than the national average. Only St. Louis and Riverside, California, went backward.

• The data also show that using AI has decoupled from building it. San Francisco is only mid-pack, and New York and Los Angeles sit near the bottom.

The fastest-growing use of artificial intelligence in American business isn’t happening where the technology gets built.

Denver and Phoenix lead the country’s largest metros in the share of businesses using AI at work, each around 36%, according to fresh US Census Bureau data. San Francisco, the industry’s capital, sits mid-pack. New York and Los Angeles rank near the bottom.

Nationally, the share of businesses that said they used AI in any function in the past two weeks climbed from 17.2% to 24.6% in just 10 months — a steep rise, and a broad one. Of the 24 big metros with a year-ago baseline, 22 grew. Only St. Louis and Riverside, California, slipped.

The pattern confirms something Technical.ly has tracked at the state level: AI use at work is spreading well beyond the coastal tech hubs.

Using AI and building AI are two different economies. The map of adoption looks nothing like the map of computing power and capital, which is still concentrated on a handful of corridors — as a recent Brookings analysis of the East Coast’s AI footprint laid out.

Among Technical.ly’s core ecosystems, DC ranks third in the entire country at 34.8%. Others top the national average: Dallas-Fort Worth (33.3%), Baltimore (30.6%) and Philadelphia (26.9%).

Two caveats:

  • First, these are survey estimates with real error bars; standard deviation runs between 1.2 and 4.5 points. Treat any gap under about five points as a tie.
  • Second, and more important: The wording — “used AI in any business function in the past two weeks” — is the lowest bar there is. It counts a marketer who tried a chatbot once the same as a company that genuinely re-tooled its operations.

Adoption is the cheap metric. It’s easy to measure, and it’s climbing fast. The expensive, unanswered question is whether any of this has changed how the work actually gets done — or who gets paid for it.

That question sits underneath the biggest bet in the economy right now: the hundreds of billions flowing into data centers and the physical plant to run AI. Whether that buildout becomes essential infrastructure or stranded assets depends on firms turning casual use into real production. One lesson from the 20th century deployment of electricity and 21st century broadband: The dispersion of a new technology predicts how widespread its benefits will be.

AI at work is spreading fast, and nearly everywhere. Who will benefit is the open question.