AI unicorns publish almost nothing - someone finally counted
Stanford metascientist John Ioannidis went through every AI company ever valued above $1B and checked what they actually put into the scientific record. It is thinner than the field's reputation suggests.
- 317 AI unicorns from 1998 to 2025 examined, more than half never led a single paper or preprint
- 2,077 publications total where a company researcher was first or last author
- together they account for 1 in every 1,000 AI papers published in 2025
Ioannidis is the person who publicly questioned Theranos in 2015 for the same reason, so treat this as a diligence heuristic: no published record is not proof of fraud, but it is a question worth asking.
https://www.science.org/content/article/ai-s-top-startups-are-barely-publishing-their-research
📎 Read also:
→ Only 11% of companies run AI agents in production
→ WEF - the tech winners are integrators
→ $6T heading into AI infrastructure by 2030
Stanford metascientist John Ioannidis went through every AI company ever valued above $1B and checked what they actually put into the scientific record. It is thinner than the field's reputation suggests.
- 317 AI unicorns from 1998 to 2025 examined, more than half never led a single paper or preprint
- 2,077 publications total where a company researcher was first or last author
- together they account for 1 in every 1,000 AI papers published in 2025
Ioannidis is the person who publicly questioned Theranos in 2015 for the same reason, so treat this as a diligence heuristic: no published record is not proof of fraud, but it is a question worth asking.
https://www.science.org/content/article/ai-s-top-startups-are-barely-publishing-their-research
📎 Read also:
→ Only 11% of companies run AI agents in production
→ WEF - the tech winners are integrators
→ $6T heading into AI infrastructure by 2030