OpenAI: general-purpose AI, years before the chat window
Before chat windows and household names, a lab aimed squarely at general-purpose AI looked premature to most of the field.
The Bet
When OpenAI was founded in 2015, most useful AI was narrow: systems that recognised images, translated text or mastered a single game. Talk of general-purpose AI was, for many researchers and investors, a distraction from practical work.
Moonshot’s bet was that progress in large-scale learning was compounding faster than the consensus allowed, and that a team organised around the most ambitious version of the problem would attract the talent to pursue it.
The mainstream was pricing AI as a feature. We thought it could become a platform.
Moonshot’s View
We believed the mainstream was pricing AI as a feature, something added to existing products. We thought it could become a platform: a general capability that many products would be built on.
We also believed the people mattered more than the plan. At the very beginning, the most important asset was a research culture willing to scale ideas that others thought had hit their limits.
Room for Disagreement
Skeptics argued that general-purpose AI was decades away, that research labs rarely become durable businesses, and that the compute required would be ruinously expensive. Each of those concerns was reasonable at the time.
We took them seriously. Our disagreement was with the timeline, and with the assumption that a research-first organisation could not find its way to products people use every day.
Notable
OpenAI, in brief.
- Sector
- AI research
- Founded
- 2015
- Founders
- Sam Altman, Greg Brockman, Ilya Sutskever and others
- Moonshot backed
- Early
- The contrarian call
- General-purpose AI was closer, and bigger, than the field believed
- Money raised
- $192.6B to date
- Company value
- $1.25T
- Moonshot 25
- No. 1 of 25 overall
Source: money raised and company value from Caplight, as of Nov 17, 2026. View source
Pitch Moonshot
Building something the crowd doesn’t believe in yet?
Tell us in about two minutes. No warm intro needed, and every pitch is read by the team.