Free list of US companies to download, by state, industry and size

Sign up with your work email: the 10M+ U.S. dataset is on your dashboard, one zip with the companies and the people who work there. From the API, GET /files/ lists it as us_10m. Then keep only what you need:

import duckdb
duckdb.sql("""
  select NAME, URL, HQ_CITY, EMPLOYEE_COUNT_RANGE
  from 'us_smb_mid_market_companies.parquet'
  where HQ_STATE_CODE = 'TX' and LINKEDIN_INDUSTRY = 'Software Development'
""").write_csv("texas_software_companies.csv")

The whole file, people included: the free US B2B leads dataset. The owners of those companies, with an email or a phone: build a free B2B leads list with SQL.

All 6,700,867 companies of the 50M+ dataset, by the employee range on their LinkedIn page. The free list holds 1,748,139 companies of 11 to 500 employees; all 6,700,867 unlock once 3 people you invited sign up.

The 4,429,454 companies whose headquarters state we know (66.1%), all sizes and 11 to 500 employees. Both columns count the 50M+ dataset's companies (2,113,537 of 11 to 500 employees); the free list holds 1,748,139 of that size.

A state's page has its companies by city, industry and size, and 20 of them picked at random. The state and industry pages leave out the 13,083 companies abroad whose LinkedIn address gives a region code that is also a state's, Perth in Western Australia as WA the largest, which this page's tables count.

LinkedIn's industry, filled for 5,892,571 companies across 494 industries. The largest, Construction, holds 270,515.

One page each, with the companies by state and size and 20 of them picked at random: construction companies, real estate companies, medical practices, business consulting companies, hospitals and health care companies, IT services and consulting companies, advertising companies, software companies, financial services companies, retail companies, technology and internet companies, restaurants.

The free list's columns, with their fill rate in the October 2026 release. A field is filled only where we have it: nothing is guessed.

Every field of both files: Dataset fields in the docs. A niche already cut, as a CSV: the free lists.