Free range startups? Market scope, academic founders, and the role of general knowledge in AI

Abstract

Abstract Research Summary High‐tech startups develop technologies, the market applicability of which can vary widely, enabling startups to target a range of market segments. Using a question‐driven approach to contrast startups with and without academic founders, we investigate the difference in the market applicability between the two groups on a sample of 988 startups in the artificial intelligence (AI) field. Our findings reveal that academics' pursuit of basic research drives the creation of general knowledge, which in turn leads to wider market applicability. With fewer requirements for complementary downstream assets in the AI ecosystem, academics can more easily translate their general ideas to market applications and locate downstream in the value chain. Our findings highlight the role of problem‐formulation and ‐solving in startups and of academic startups within AI. Managerial Summary Using a sample of 988 startups in the Artificial Intelligence field, we find that startups with at least one academic on their founding team are associated with a higher number of verticals (potential market segments for the technology the startups developed) compared to startups without any academics. Teams with academic founders produce more general publications and patents than others, which drives the association with more verticals. Academics formulate and solve more general problems relative to non‐academics, leading to the creation of more general products that are applicable to a broader range of verticals. With fewer requirements for complementary downstream assets in the AI ecosystem, academics can more easily translate their general ideas to market applications and locate downstream in the value chain.

Cite this work

No public copy is stored for this record, so only the metadata is cited.

Stored only when both papers are in the corpus — 30 citations exist upstream.

Papers that cite the same references — the relational fallback while embeddings are still backfilling.