Coopetition in AI
Coopetition, cooperation between competitors, appears on 31 recent papers split between the management and operations research journals, with 7 in the last quarter, in line with earlier quarters. AI firms compete fiercely while sharing open models, benchmarks and safety research, and renting computing from each other. When does sharing help a rival more than oneself?
What prompted this
Since October 2025, "coopetition" appears on 31 papers, 14 in management journals and 14 in operations research, mostly on supply chains and platform ecosystems. Open-source models, shared benchmarks and cloud partnerships among AI firms are among the clearest current cases, and none of the 31 papers names them.
Questions
- Do firms release open models to set standards, to recruit developers or to commoditize a rival's product, and do the stated and revealed motives match?
- When a firm both sells computing to a rival and competes with it in models, how does it price, and does the rival hedge by building its own?
- Does participation in shared safety and benchmark efforts correlate with market position, and does it change when a firm leads?
Data that could help
Model release histories and licences, cloud partnership announcements and contract terms, and benchmark and safety consortium memberships.
Next step
Build a timeline of open model releases and cloud partnerships among the main AI firms and relate release decisions to each firm's position in the market at the time.