Benchmark’s take on AI competition: A conversation with Miles Grimshaw

At the end of 2020, Miles Grimshaw became the latest recruit to the storied venture firm Benchmark, and still is, Benchmark is a fundamentally has stubbornly refused to change the way it operates. Especially the size of the team and assets under management.

With just five general partners, few principals in its 28-year history in Silicon Valley, and one after another with less than $500 million in funds, the company competes with rivals with increasingly large checkbooks. It is said that he often experienced setbacks because he had to. Today, of course, many of the 2021-era deals and funds look like bad bets to those who backed them, but the benchmarks (companies that have past bets on Uber, Snap, WeWork, Sorare, etc.) traditionally seems wise in sticking to the early stages of to invest.

We were curious to hear how Grimshaw, a Yale graduate and former general partner of Joshua Kushner’s Thrive Capital, thinks about how to run the benchmark. . We were also interested in understanding his Benchmark thoughts on generative AI, another potential bubble.

Indeed, Grimshaw’s past and current companies seem to take very different approaches. Thrive jumped into OpenAI’s recent $300 million round at his $29 billion valuation. Meanwhile, in early spring, Grimshaw led his $10 million seed round at LangChain, which helps developers build more complex applications. above Large language models like those built by OpenAI.

We discussed these and many other things with Grimshaw last week in a conversation excerpted below and lightly edited for length and clarity.

TC: This is our first meeting. What did you study at Yale? What steps did you take from college to joining Thrive Capital in New York?

MG: Well, my trip to work with the Thrive team in 2010 was very accidental. I was probably a very classical liberal arts student in many ways, but strictly an economics major.

The question at the time was, “Can we make a silicon array?” People on the East Coast found each other, bonded, and said, ‘Yes, we can, let’s do it, let’s make it’. So when I was at Yale, I met Will Gabrik. He was actually in Yale Law School at the time. He later became one of Thrive’s partners and then its CFO. [and now a unit president] A Stripe employee, he and I became natural friends over our interest in products and application software. He happened to know Josh from his undergraduate days. And I started spending weekends with them in New York, and then I said, ‘Let’s team up. So I officially joined the group in 2013 when I graduated.

I was the 4th or 5th. I think we had a total of 8 people at that time, including the head of finance and the EA. So it felt like we were now in the corner of a very large office and outsiders were proving that it was possible, and it was actually a really good time.

May I ask: Are you Australian? Are you from New Zealand?

i’m british [Laughs.] I grew up in England until I was 12, then moved to Boston, so I have a very confused accent. I am the eldest of 7 siblings. Feel free to move on to the next eldest son (no accent). So I was on the verge of keeping a few words.

I had just read about it yesterday, and a person was said to have a “non-accented accent.” How are Thrive and Benchmark different from an operational perspective?

Benchmark keeps things simple and consistent. Only 5 of us. It is always a partnership of equals, [traditionally] 5-6 partners, each focused on 1-2 commitments per year. Much of it is in many ways not capital, but a service effort to help founders increase their potential for success and scale.

Has Thrive grown a lot? They have increased their assets under management.

I don’t know the exact numbers these days.But Josh and Karen [Zaki] and Vince [Hankes] And to the team over there, I have tremendous respect and admiration for their ambition to serve people in so many ways. I know there was a $2 billion fund after I left, but I think there was another $3 billion fund. I don’t know exactly how big the team is these days, but they continue to build out various features and support for founders, so the last I heard is a few times what they were when I left. I think it is.

In the AI ​​space, he led the Langchain seed round. What kind of business model is there? Is this a SaaS business?

Yes, that’s right. Between the application experience being built and the model at the bottom of the stack, there is a wide range of needs that great software can fill. [Thanks to the large language models we’re seeing like GPT-4], standards have been raised, expectations of software have also been raised, and no company wants to be a dinosaur walking around in awe after a meteorite, thinking everything is bliss. What LangChain supports is that application developers coming to the market can say, “Well, we have a language model that spit out tokens. is saying.

Where did the developers find LangChain?

It has a good reputation in many ways. We certainly don’t do marketing. We started earlier this year and now we have a few engineers, but there is no one on the team other than engineers. The team is obviously very active in the open source community. I think there are about 20,000 members on Discord right now.team is [also] Many events are flooded with application engineers and developers looking to learn, imagine, and create. And they are in the wake of all that energy.

I was curious that you led a $10 million seed round at this company and Sequoia led a $20 million round like a week later. is that so?

I forget the exact timing, but yes.

Has Benchmark invested in any of these foundational model companies?

We partner with GPU chip company Cerebrus [off which AI models run] But we are not foundation model level partners.

Have you had the opportunity to invest in these companies? Clearly, these outfits have raised a lot of money. I wonder if that was the factor.

We don’t start making decisions limited to $10 million, $5 million, and so on. We aim to be the first and most influential partner to the founders and start by aiming to serve on the board for 10 years. If this is a real body of work and the starting funding amount may be higher than usual, we will work it out together. .

But I haven’t yet found any conviction to imagine the possibility of one of them holding a huge market share in perpetuity. You’d see this in open source, but now open source is coming up and catching up fast.I can imagine the input to some of these [large language models] Whether it’s the amount of computing available on a chip or the cost of a chip, the cost will definitely come down over time. Knowledge is clearly pervasive, more people know how to do it, and it doesn’t require a lot of money just to try to figure out how to do it. And I’ve even seen depreciation rates for people like OpenAI’s model. For example, think about how quickly you got rid of all the money you spent on GPT-2 or GPT-3.

We certainly appreciate all the effort and we are sure they could be good business, but instead we focused on the developer layer above it. . And very focused and committed to meeting and partnering with people on that journey.