OpenAI is also a startup investor—here’s its investment thesis

OpenAI, the company behind influential AI models like ChatGPT and DALL-E 2, is looking to take a broader stake in the AI game. And it’s doling out millions to AI startups in an effort to make that happen.

The SF-based AI R&D company, founded in 2015 by Elon Musk, Y Combinator’s Sam Altman, Stripe’s Greg Brockman, and others, is the same organization that helped kick off the current generative AI boom. The company owns and operates large-scale AI models that a number of startups are built atop, including, most recently, ChatGPT. OpenAI announced a $100 million startup fund in May, with support from Microsoft and other investors, and in November, it debuted the fund’s first accelerator program.

The program, called Converge, will fund chosen startups with $1 million each in equity investment as well as “early access to OpenAI models” and a five-week program of workshops and office hours from OpenAI. The news comes on the heels of OpenAI dishing out even more cash to its chosen startups in recent months: In October, it reportedly agreed to lead recent funding for Descript, and in November, the company led a $23.5 million round for Mem and a $5 million round for Harvey.

“[We’re] thinking about, ‘What are the application areas that we really don’t understand super well today?’” Brad Lightcap, COO of OpenAI and manager of the OpenAI Startup Fund, told us in an interview. “I think there’s going to be a class of things that are new, like any platform shift. Who could have predicted Airbnb in 2005?”

We chatted with Lightcap about the new program’s development, investment trends, ideal startup characteristics, red flags, and more.

This interview has been edited for length and clarity.

What made OpenAI decide to introduce the accelerator program, and how long did the process take from ideation to debut?

We really started investing seriously about a year ago…We launched the API, I think, just over two years ago, and really, the conversation started to snowball after [that]. We wanted to pull the fund together fairly quickly thereafter, just because we felt the opportunity was real.

We spend a lot of time with founders and builders, and… we saw a high level of intellectual curiosity about the types of things that our models could bring to those companies. Speak [AI] is a great example of a company that we had engaged with and known for a very long time—their aspiration is to build commonly accessible language learning for everyone…So we had had the conversation with them for a long time about, “How do we have a GPT-3-style model that might work in that context to supercharge language learning?”…Ultimately, we wanted to put a fund in place to really support those types of initiatives and others like them.

Which characteristics of a startup make it a “prime investment” for OpenAI’s program?

Number one is the ambition of the vision. We believe strongly that the type of technology that we build is going to have broad-reaching effects that are broadly positive, broadly beneficial, to the world. At the end of the day, it’s on founders to envision what that is and bring it to life. What motivates us, first and foremost, is founders who have that vision and the scale of that vision. It’s one thing to say, “I want to build an application to make language learning better,”—it’s another thing to say, “I actually want to build a tutor that lives in everyone’s pocket that’s personalized, it’s real-time, it’s on 24/7, and it’s affordable.”

And then there’s an appreciation for the technology—and that could be both in its potential, but also in the fact that we have to use this responsibly. It’s a new technology, there are guardrails that we have to understand and safeguards that we have to put in place and respect. That’s a big part of OpenAI’s mission as well. So we look for founders who are respectful of the magnitude of the technology and want to understand the impact of it socially

What’s a red flag that could make OpenAI lose interest in a potential investment?

Folks whose vision for what they want to build is not substantial…We want to back ambitious founders; we want to back founders who see no limit on the potential of what they're building and how the technology can empower it…We really like…founders who are building for really, really big markets, and really building for broad use. The more we can see a path to “Why is this something that ultimately is going to matter for a lot of people in the world?” the more excited we get.

On that note, are there any investment ethos that OpenAI is following, declining investment for certain areas or use cases—like military use, to name one potential example?

We take more of a kind of first principles-based approach to that, and so we think through the criteria for what makes a great investment for us. It's premature to write categories off wholesale—I think those lines also get quite blurry.

Speaking of safety standards and guardrails, who’s on the board deciding on these investments? Any ethicists or external advisors?

We really take OpenAI’s principles here, and so day-to-day, most of the decision-making is really me and Sam [Altman]. And then our core team, which includes a broad group of folks inside OpenAI.