Renowned investor Elad Gil on how the great AI race is likely to erupt

Elad Gil, a successful founder and prolific investor, has already been called Silicon Valley’s largest solo venture capitalist, given the massive amounts of capital he’s invested in recent years, including on behalf of institutions that reportedly include Harvard’s endowment.

His track record largely explains his quiet rise. For example, Gil invested in the Series A round of highly rated payment software company Stripe 11 years ago and has invested in many of the subsequent rounds. He also took stakes in note-taking app Notion, cloud collaboration platform Airtable, military tech contractor Anduril, and design tool Figma, which agreed to sell to Adobe for a whopping $20 billion last September — though Adobe is still working to get authorities of the Justice Department based on the merits of the deal.

During a conversation late last week, Gil — who occasionally blogs but maintains a bare-bones site — declined to answer specific questions about how much he manages or some of the amounts he has invested in companies. But quantitative VC outfit TRAC calls him a “superforecaster” who has funded at least 155 companies, and whose “batting average” is .671, meaning 67% of his early-stage investments have yielded at least follow-on rounds, according to TRAC data. (It says at least 30 startups in Gil’s portfolio have become “unicorn” companies, though, as Gil himself points out, many valuations will shift over the next 18 months. “The really tough times are coming,” he says.)

When we spoke with Gil, we asked what founders should do when things go from bad to worse. We also talked about his continued fascination with AI and some of the early checks he wrote to startups that are now raising serious venture dollars, including Character. AI, backed this year by Andreessen Horowitz, Perplexity.AI, backed by NEA, and Harvey, backed by Sequoia Capital.

Last but not least, Gil shared how he uses AI to scale his own work. You can listen to our full interview; in the meantime, fragments of that conversation follow, edited in length.

TC: Years ago you wrote a book called High Growth Handbook, about scaling startups from 10 to 10,000 people. Do you now think there was too much focus on growing so fast?

EG: The focus of my book was on what you do next when you reach that magical moment of product-market fit. . . I think this mantra of growth for growth’s sake really came up especially during the COVID period. When capital became really cheap and available, people started scaling when they weren’t really suited to the product market. They started scaling up before they had a lot of customers, or before it was clear they had a moat that would create some sort of defense for their business. I think where things derailed was people started raising money years earlier than where they were. And then they started taking against that money they raised instead of taking against the company they had.

We hear many stories from employees eager to talk about mismanagement within their company as things go south. Do you have any advice for businesses on how to scale back without blowing up completely in the process?

A lot of the assessments that people are actively doing right now are asking, where do I think this will be in one or three or five years? And if it doesn’t work, what should I do? Those are really hard choices to face. People have to make decisions between downsizing the team and maybe changing direction, or trying to sell the company because it’s clearly not going to work. Do they shut down and give money back? If you look at when people have raised a lot of money in recent years, most of that happened in 2021. And if people raise money for three to four years and when [they] have nine months to go, that means a lot of people have to start fundraising by the end of this year. So I feel like the really tough times are coming. I think this is still a bit of a warm-up period, or an anticipation period.

There’s just a huge backlog of companies that are about to go under that should have closed years ago, but just kept going.

Regarding your own investments, can you indicate how much you have raised in recent years and how many companies are in your portfolio now? You raised $620 million per SEC filing in 2021. . .

I haven’t really talked much about it [and] I don’t really know the exact number [of portfolio companies] straight away. Traditionally I just invested my own money. Then things started to get bigger in terms of the allocations I could invest in, so in some cases I did what are known as SPVs, or single purpose vehicles or investments.

Right now my model is a bit of a hybrid where anything small, or if people just want me to be an angel, I can personally do it. If anything gets bigger, I could use a fund. If something gets really big, I can use a mix of personal money, fund money, and SPVs. I’ve tried to maintain a flexible approach so that as I work with different companies at different stages I can tailor my work to what they really want and need. I want to avoid the situation where I have a huge fund and feel the need to start investing a bunch of money and force it on people and start acting more or less badly.

You paid attention to Generative AI before others. Are you completely surprised by what has been unleashed into the world? [on the generative AI front] in the past six to twelve months?

For me, in a way, the big moment was seeing things like very early generative art based on GANs [which is a class of AI and machine learning algorithms]; it was just striking what non-artists could do. Then a little bit after that, when GPT-2 and then GPT-3 came out, that was obviously a moment where there was such a big move between them that it was obviously a big change.

Are you an investor in OpenAI?

I’m not involved in most things that happen at the foundation level, but I don’t want to talk about specific companies or anything like that.

You sat on a panel in LA earlier this month with Ashton Kutcher, whose Sound Ventures just raised a growth fund to expressly support just six or so fundamental model companies — three of which it’s already invested in: OpenAI, Anthropic, and Stability.AI. What do you think of that strategy?

There are a handful of companies that are really ahead of the game in developing these basic models. And I think there will be some scale and capital effects for them, at least for the very latest models. So you know, GPT-4 still feels a cut above everyone else, and Google clearly has the capabilities to build something against it. Anthropic repeats its cloud model. There are a few other players. There are Cohere and A121 [Labs] and such. But right now it seems that proprietary models are one or two generations ahead of open source, and if you assume that each model is going to be a lot more expensive than the previous generation of models, then you can bet that this trend can continue for at least another exist for a few years.

That means two things. One is that when there’s GPT-7 or whatever, open source might be the equivalent of GPT-6 or GPT-5.5. And GPT-6 will probably be incredibly performative. It’s probably going to do all sorts of amazing things. So that leads to the question of what are the really advanced things that you need the most advanced models for and that’s where I think there’s going to be a lot of the value in the industry — but I think a lot of it will also just going to the things that are a generation or two behind. And I think open source will also play a role there.

So I think of it as a world where we’re going to have a handful of very large, closed, proprietary models and an oligopoly market similar to the cloud world where we have Azure, AWS and Google Cloud as the big three players. I think the models can of course converge there as well. But then we have a bunch of open source that people will use for all sorts of things in parallel.

Again, for much more with Gil, including why he thinks more founders should consider quitting and returning capital while they still can, listen to our longer talk here.

Leave a Comment