Technology

Benchmark, the storied venture firm, sees the AI race as still wide open

Yesterday in Helsinki, this editor interviewed four of the six general partners at Benchmark, the nearly 30-year-old Silicon Valley firm that’s known for some notable bets (Uber, Dropbox), paying each general partner exactly the same way, and for continuing to raise similar-size funds over its many years rather than balloon in size. 

We were speaking at Slush, a major yearly event for the European startup ecosystem, so I naturally asked why the firm was making such a big showing, given that it’s hard enough getting the Benchmark team to appear in Silicon Valley together.

Victor Lazarte, a gaming company entrepreneur who joined Benchmark five months ago as its newest GP, admitted that there was “no business reason” for Benchmark to come other than its interest in understanding all things “exceptional.” (Helsinki is truly gorgeous.) 

Larzarte was equally candid when the conversation turned to soaring valuations in recent years, and I asked about his own gaming company, Wildlife Studios. It raised a $60 million Series A round from Benchmark in 2019 at a $1.3 billion valuation and, less than a year later, was assigned a valuation of $3 billion when Vulcan Capital led a subsequent $120 million round. Larzarte said the company had really made “like, no progress” in between rounds, but that because Benchmark had funded the company, “everyone” subsequently wanted to invest in the company. (He said that, in retrospect, taking on too much money at too high a valuation so quickly was a “mistake.”)

Not last, we talked about how strange it is to be living through a general downturn and a boom in AI investing at the same time. On this front, the team was clear in its assessment that today’s high-flying but closed large language model companies aren’t going to be the breakaway winners that many expect them to be. (Worth noting: Benchmark is not an investor in such closed LLM companies, including OpenAI and Anthropic.) You can catch our conversation in this longer broadcast; meanwhile, you can find some excerpts below, lightly edited for clarity.

Regarding Benchmark’s views on the sweeping trend of AI in everything, partner Miles Grimshaw said we’ll be collectively astounded at how backwards our current use of software will look just a few years from now.

I think if we look back at ourselves in a few years — maybe even a year — it will feel like we were primates kind of mashing rocks together to make fire. In two years, it’ll be weird that you had to click all these buttons in Salesforce and navigate around and that it didn’t do more for you. User expectations of what’s possible are ratcheting up, and you’ve got tectonic forces at play for imaginative, creative founders to take advantage of. 

I think the question [ties to] the startup opportunity versus an incumbent opportunity. You can never tell founders where they should go — that’s not what we do. But one of the places to maybe avoid — the traps — is: don’t be Microsoft. Don’t be [part of] the Copilot game [meaning Microsoft’s AI-productivity tool that’s powered by OpenAI’s ChatGPT]. That’s what they’re doing. It serves their business model. It serves their product environment very well. But be more creative and ambitious than just Copilot. 

Peter Fenton, the most senior member of Benchmark’s team, weighed in to add that:

We didn’t invest in a large language model. Maybe this is unique to Benchmark, but our view is the capital intensive [companies are tricky]. We’ve been in some — we all took Ubers here [to the event] today [and that was a Benchmark portfolio company]. And capital-intensive businesses and venture backed companies have historically not been great partners. 

Our [belief] is that open source will end up having a profound effect on the ecosystem. We’re all, in a way, soldiers in the army of ‘tear down anything that’s getting capital intensive and overbuilt’ and then propagate a developer-driven world. And these experiences in AI are going to be built by developers who are imagining stuff that no one can fathom at a large language model, because they’re serving a different kind of platform-horizontal need. So yeah, we hope [the closed LLM companies] do well. We love the innovation. But I’ve been particularly drawn to the idea that there’s an open source founder who’s probably going to surpass almost everything that you can do with capital.

Other outtakes from our conversation include Fenton discussing a big miss by Benchmark that came up during the chat (by accident, candidly), which is Airbnb.

You mentioned Airbnb. That’s one of those on our long list of deep regrets. When I joined the industry, you could buy 20% to 25% of a company in a Series A investment for a number today that sounds like a seed round — $7 million to $10 million. Because we had an ownership threshold that was impossible to achieve [when Airbnb was first fundraising], we missed the opportunity. And we’ve sort of relaxed that as a constraint because it’s not a question of what can Benchmark own. It’s: What is the company’s potential?

We also talked about what makes a Benchmark company in 2023, with GP Sarah Tavel saying the focus very much remains on nascent teams:

Of the investments that we’ve made so far this year, some large percentage of those [were] actually at incorporation of the company. So more often than not, it is actually two people who see an opportunity, and we’re getting there before they even left their last job to start that company.

We really focus on, ideally, being that first board member, the first partner to a founder when they’re embarking on this journey, and a large percentage of the time, being the first money [that] two people raise for their idea.

Speaking of board seats, I asked about the latest trend in Silicon Valley, that of VCs who say board seats don’t matter because the real information between founders and investors is transferred between board meetings. Here, Fenton suggested that as a fiduciary, it would almost be negligent for a VC not to take a board seat where possible.

It’s an interesting hack, the venture business, where we codify a relationship typically with money. But then we join the board governance structure, and the person who takes our money, we have power over, in theory. With governance structures and boards, you can hire and fire the CEO. That’s the biggest job of the board. 

In my view, the really great businesses are built with boards that have a partnership with the CEO, that have a gaze to the horizon of what’s possible that’s bigger than any one person. And I think that the integrity of that structure has been tested throughout the entirety of our C corp business model. [When the industry] moved into crypto, we got rid of boards; we said, ‘Who needs boards? Who needs company building and all that stuff?’ And it created interesting token value, but I don’t think it built equity value . . .

My sense is we are moving through a period of time where the idea of governance — we just went through it at OpenAI — percolates up to the top of people’s consciousness. And we can see what happens when the governance structures are misaligned. And I have a personal view that my partnership with a great CEO is deeply enhanced by knowing that I’m carrying the fiduciary responsibilities that they carry with them close to their heart, and that if I’m not serving on the board, I can be effective, but it’s not the same. 

Finally, getting back to that valuation discussion with Lazarte, I wondered how Benchmark counsels its startups on valuations, given that the bigger their follow-on valuation, the better in some ways for earlier investors but, sometimes, the worse for the founders, who may have fewer options when their companies become overvalued. Here’s what he had to say:

When I partnered with Benchmark [as a founder in 2019], I really wanted to work with Peter because I felt that he was someone who could help me transform the company, and I was lucky that he wanted to work with me, right? And then, just being transparent, we were in a period where there was a lot of capital chasing deals, and there’s the fact that after Benchmark invests in a company, everyone wants to invest in the company. So this second round that we raised, we really made, like, no progress. But there were so many people who were interested and I [was thinking that] we’re a company from Brazil and we’re trying to move to Silicon Valley. And we were always very low profile. But like all of a sudden, it was, ‘Oh, Benchmark invested,’ and there were these people coming in. And then I made the decision of, okay there are these funds that [are being] invested [at] twice the valuation when really not much progress was made. And I made the decision of, okay, with more money, perhaps we can do more. 

But in retrospect, I think that myself and a lot of the founders . . .made the mistake of raising too much capital. The problem is when you raise too much capital, you start going in unnatural directions, you start deploying more capital than what is natural to that business. And then you grow your team, but bigger teams, lots of times they don’t produce more. In fact, they produce less. And once you do that, you have to go through the painful process of reducing the team. So the best founders are not trying to maximize for unnatural valuations, because that does distract from the core purpose of building the company.