The Exchanges

Q For people who have not seen your lectures just yet, highly recommend it. Who have you spoken with? I know you brought on different guests and you bring them for the students. What was the plan there?

A Oh, you know, we had a, we had a great lineup. It's a, it's a labor of love. And, um, I was really excited to bring back some of those conversations back to school. Um, you know, if you're a student and you're making a big life decision, I wanted to make sure that you saw the whole AI stack from chips to data centers, to models, Um, and the, ultimately the AI applications. Uh, and so we had a great set of speakers across all of the parts of the stack. We had folks from, uh, NVIDIA and Grok. Um, we had Ali Gozi from Databricks. We had Tuhan from BaseStand. We had Sachin from OpenAI, the folks from Anthropic. Uh, so it was a great lineup and we're very lucky to have them.

AI assessment note: “We had folks from, uh, NVIDIA and Grok. Um, we had Ali Gozi”

Q So Brad Lightcap, he captured some of the momentum from their customers with GPT-Five. I want to go through some of these and then get something that you've heard from your portfolio. So he heard some really good momentum and I'll show the tweet from Cursor, Lovable, Harvey, Amgen, Uber, and Notion. I know you have a particular instance. Do you want to share this?

A Absolutely. Um, this is another example of what you said, which is The Twitter narrative was so volatile, but if you actually call the experts, the actual customers, and so we called a series of customers, we called all of the coding AI companies we know from, from, you know, a bunch of their customers, customers of the, the two or three largest labs to customer service and cybersecurity. And, you know, Early, right? Two weeks in, but good so far. Working with the team, as we said, there, it's a product in motion. But one particular one stands out is, uh, our company Expo. Their benchmark, on their benchmarks, they saw a significant step up with GPT-V, much better than any other coding or design or math startup saw in their ability to find cyber exploits. And, you know, summarizing, basically, the jump was from just under 60% to 81%, switching from, uh, Anthropik's latest model to OpenAI's latest model, GPT-V. And the jump was so significant. I think it was higher than what the team had expected, but certainly the Expo team had expected by quite a bit, but maybe even the OpenAI team. Uh, and so they've published the results, and it's becoming pretty clear that AI is much better than humans at this thing called hacking.

AI assessment note: “one particular one stands out is, uh, our company Expo.”

Q Okay. Well, we need to talk about where is spend today in AI and where will this be in the future? I know you have some great charts about this. They're not pyramids.

A It is probably the biggest question in AI right now. Where will the value accrue? And we've been studying this now for, for a couple of years, and you know, I, you know, I'm sure the audience sees the large CapEx announcements that the hyperscalers are making. Um, They add up to hundreds of billions of dollars that are going into the ground, building data centers, um, acquiring chips and the infrastructure to, to, to enable all that. And so we decided to study the same with the build of prior super cycles, with the internet, cloud, mobile, AI. And we put it together to see where does value accrue over time. And, you know, I'll summarize it for, for the audience here, which is in the cloud super cycle, you know, you have about. Four hundred billion dollars of revenue that is being earned by the applications. Let's call it the layer one. You've got about two hundred billion spent on infrastructure. This is AWS and GCP and Azure, et cetera. And you've got about fifty billion on semis. This is Intel, AMD, et cetera, the CPU layer. The shape of this is 402 150. Enter AI. Nvidia earned, Nvidia alone at the chip player earned roughly forty billion dollars in data center revenue in Q one last quarter. So analyze that's about a 160. Let's say the total industry is 170, 180, maybe let's say 200. The, uh, infra layer is, um, 20, thirty billion. This is inference revenue that's being earne…

AI assessment note: “Nvidia alone at the chip player earned roughly forty billion dollars in data center revenue”

Q have a cult. They have a cult brand. They've achieved this. It's actually insane the rate they did this. I think you published a chart on this, and then also the other charts from, um, East meets West. Those are great. We'll add all these in. I'm curious from your Standpoint as an investor and a user and understanding their commercial efforts. How does the consumer adoption bleed into enterprise?

A Look, I think, um, ultimately there's three legs to the stool on the usage numbers that you highlighted. It's obviously the number of users. It's how much time they spend per day. And what is the retention of those users? I would say those are The three legs of the stool that we measure to, to have a sense of the health of a, uh, a consumer app. And on the first one, as I mentioned, I think they are far and above larger than any standalone AI app today. On, in terms of time spent, you know, we've published this analysis. They are far and above time spent per user per day compared to any of the other apps, uh, AI apps today. Uh, actually not just AI apps. I, you know, one of the analysis we did is they are now larger than some of the mainstream consumer apps like X. Yeah. And, um, retention, you know, the, it's, it's, It's pretty wild, actually, so what we call smile curves are starting to show up in retention, meaning that, you know, in a chart where you see the number of users that stay on an app after they started the journey, typically you would expect an exponential decay of the users on this app. For very few apps, um, the, the, the curve actually goes up, so it looks like a smile. And ChatGPT exhibits a smile curve. The other apps that exhibit smile curves are Instagram, And TikTok. And the implication is users are finding so much value on this, on this, on this app that …

AI assessment note: “I have this beautiful thing at my personal phone... and I want that at work.”

Q Some people are extreme doomers, and David Sachs tweeted not too long after this, exposing the doomers, talking about the doomers. Guess what? We're not there yet. And what we've seen with these model releases over and over again, that the world is not Ending. We're not gonna get destroyed quite yet. It's gonna take longer to build. Where do you think we are in the AI adoption curve?

A Yeah. Uh, Jeffrey Moore, uh, wrote this book called, uh, Crossing the Chasm. I'll, I'll refer to some of the concepts that he describes in that book, but I think David's not wrong. Look, there's with any sufficiently advanced technology, you've got the tech enthusiasts or, or, or, or call it the developers or the super users who will adopt it first. And so I think of it as like technology that raises the ceiling and technology that raises the floor and on the ceiling, you've got super users. Tech enthusiasts, typically 10 to 15% of the population. And then at the ceiling, you've got the vast majority of the population, the majority, even the skeptics. And so I think the way adoption plays out for really any technology is I evaluated on, hey, is this new product release raising the ceiling or is it raising the floor? So for example, let's talk about raising the floor. Um, What people who are, who the vast majority in laggards need is remove friction. Look, they don't wake up to use AI. They wake up to do their life, live their lives, and AI's got to find a job, find a way to be a part of their lives. Advanced voice mode. Reduces the friction. I've got a action button dedicated to voice mode. Incredibly useful. Um, one-eight-hundred-chat-gpt. WhatsApp with chat-gpt or meta-ai, you know, those are, I would say, um, ceiling feature releases that will accelerate the adoption, uh, ov…

AI assessment note: “with any sufficiently advanced technology, you've got the tech enthusiasts”

Q Well, speaking of performance, Sorcery is sponsored by Brex. They're all about spending smarter, moving faster for over 30,000 companies, 200 public ones. Um, and I'm curious from your perspective, as we think about companies today in this super cycle, moving super fast, becoming really competitive, what are the key characteristics that you look for in founder performance?

A Ultimately at, you know, we, we invest across venture growth, um, uh, pre IPO and then public markets. I would say there's really two questions that I'm trying to assess when, when, when looking at businesses, is this a great business? Is this a generational business that will go create a large company, maybe a public company? Question number one. Question number two, is this a great price? Is this a great structure? Uh, does the math pencil out, et cetera? And I spend 99% of my time answering the first question. Is this a generational business led by incredible leaders, uh, inspired by a great mission? And, um, While every business is unique, it's, it's hard for me to give you a template for, for, for all of them. I think the thing that I obsess about the most is asking the right questions. You know, every time, you know, we, we study a business, there'll be two or three or four questions that'll define the, the, the, the business at that given time. And, um, Knowing the right questions to ask, actually, actually that's probably the process that takes a little bit of time. The answers will typically either be knowable or not, and we can get to them. Uh, but that's, that's where, that's where I spend a lot of time.

AI assessment note: “While every business is unique, it's, it's hard for me to give you a template”