Share of AI Papers on arXiv with an Industry Author (Monthly)

## What this measures

Of AI papers on arXiv in OpenAlex (primary topic in the Artificial Intelligence or Computer Vision subfield) with a publication date in the month, out of those where OpenAlex identified at least one author institution, the share with at least one author at an institution OpenAlex types as a company.

## Why it matters

Industry's share of AI research shows how much of the frontier is set inside companies rather than universities and labs.

## How to read it

The series starts in February 2016 at 11.4%. The latest reading, for September 2026, is 13.3%. The high is 27.9% (June 2022) and the low 6.3% (July 2016).

## What it does not say

OpenAlex identifies at least one institution for between 46% and 89% of these papers a month (lowest in July 2025); a paper without one is left out of the base, so a month with weak coverage reads noisier. Single months move several points; read the quarterly average. Affiliations for recent months keep being added, so the last six months are rewritten weekly. A paper with one company author counts the same as an all-company paper. OpenAlex dates a paper by its publication, and for a preprint later merged with its conference version that can be the proceedings date, so months with a large conference whose papers come heavily from industry labs (June 2022, October 2021) jump; read the quarterly average.

## Source

OpenAlex works with a location on arXiv whose primary topic is in the Artificial Intelligence (1702) or Computer Vision and Pattern Recognition (1707) subfield, counted by Palanor with grouped queries; windows skip 1 January, the date OpenAlex gives works known only by year. CC0. Source: OpenAlex.