AI supply is expanding unevenly across research, code, and products
A 19-month longitudinal study compares scientific output, visible open-source implementation, and product launches to measure where AI technology supply is accelerating and where translation still stalls.
AI technology supply is broadening, with pronounced sector divergence
The cross-source supply index changed +34.1% from Q1 2025 to May–July 2026. Agents & automation expanded fastest (+143%), while developer tools moved most slowly (-46.9%). The market is expanding, but research, implementation, and productization are not moving in lockstep.
01
Market snapshot
Comparable measurements from independent market surfaces.
205,365
AI-category papers
Complete monthly arXiv counts across seven computer-science categories.
Generative media470 papers · 130 repos · 15 products · PH excluded from index: small sample
-34.6%
Developer tools234 papers · 88 repos · 15 products · PH excluded from index: small sample
-46.9%
04
What is changing beneath the headline growth
Structural findings that a single aggregate number would hide.
151.7 vs 121.5
Research supply is still translating into products with a lag
In the latest three complete months, the arXiv index averaged 151.7, compared with 129 for visible new repositories and 121.5 for ranked product launches. Productization is expanding, but it has not matched the amplitude of research output.
+143%
Agents are absorbing adjacent developer-tool supply
Agents and automation are gaining while generic developer tools lose relative share. This does not imply less software tooling: coding, workflow, and orchestration projects are increasingly described and distributed as agent systems rather than standalone assistants or utilities.
+78.2%
Trust and safety is becoming a supply category, not only a policy topic
Growth appears across papers, implementations, and launches rather than in research alone. Provenance, runtime security, model evaluation, privacy, and governance are becoming buildable layers around deployed AI systems.
