This Creator Investor On The Truth About The AI Gold Rush | Term Sheet

0:00 Are we in an AI bubble?

0:02 Ai literacy is like sex ed.

0:04 You have to know about it.

0:05 AI is devouring venture capital in 2026

0:08 over $240 billion according to AI startups

0:11 in q1 up 150% and AI is taking 80 cents of every venture dollar,

0:16 though the Gold Rush is here, most Americans are still playing catch up.

0:20 Claire Zhao saw that gap and started

0:23 posting videos to educate the masses this week.

0:25 Light Speed, managing more than $40 billion in assets,

0:28 created a role that has never existed in venture capital,

0:31 an investor who's also a creator, maybe I've just been,

0:34 you know, bought into the cult, and that investor is Claire.

0:38 Welcome to term sheet.

0:40 I'm Ali Garfunkel.

0:42 Now, before we get to our interview with Claire, a little bit of news.

0:47 This is a little bit of a back

0:49 to the future moment in the private markets this week,

0:52 Cerberus is, at long last, going public.

0:54 Now, for the uninitiated, cerebris is an AI chip startup that initially filed

0:59 to go public back in 2024 and we're here.

1:02 They filed again.

1:03 They are now about to hit the public markets this week.

1:07 Now if you're sitting here wondering a chip startup,

1:09 the answer is yes, and do they compete with Nvidia?

1:12 The answer is actually yes.

1:13 Now cerebrus is looking to go out at about a $35 billion valuation,

1:18 which is very, very, very infinitesimal compared to Nvidia's 5 trillion.

1:22 But for a long time, the company has billed itself as an NVIDIA challenger.

1:27 Now, which one are you the bull or the bear?

1:29 If you're the bull, what you probably believe is that cerebrus, over time,

1:33 is going to build on its existing partnerships with open AI and AWS

1:37 and ride that inference wave to a trillion dollar valuation or more.

1:41 Now if you're the bear,

1:42 what you probably believe is that this is ultimately hype driven.

1:44 The company has customer concentration problems,

1:46 and that eventually the public markets will not

1:49 be able to take so many chip companies.

1:52 Which one are you let me know in the comments or send me an email,

1:55 and now for my interview with Claire.

1:57 So if you are into tech and you are on social media.

2:01 You have probably seen Claire Zhou explaining SpaceX, explaining the AI bubble,

2:07 explaining funding rounds, Claire, you have found a lot of success as a creator.

2:11 You have hundreds of 1000s of followers.

2:13 You are also, for a time, the youngest investor ever at GSV ventures.

2:18 And now you have a new title.

2:20 You are going to be joining Lightspeed as what

2:22 is the first of its kind ever in venture.

2:24 You are going to be a creator who's also an investor.

2:28 Yep.

2:28 Can you tell us more about that?

2:29 So I've been in the venture space for about six years.

2:33 I became partner at GSV, joined as an associate,

2:35 grew through the ranks, and have been thinking deeply about space.

2:38 And about a year ago, decided to start creating content.

2:41 What was the first day?

2:42 What was day one of creating content?

2:44 March 2025, and I had actually been thinking about it for a long time.

2:48 So it did take some courage to start

2:49 putting my face out there on short form media,

2:52 because I didn't think any other venture folks were out there.

2:55 So, you know, I always thought that short form media, at least for me,

2:59 I saw so many eyeballs there, but serious content never extended there.

3:03 And so yeah, march 2025, I was like,

3:06 why don't I just challenge myself post every single day, see what happens?

3:10 And it just so you've been thinking about it,

3:12 and then you got to sort of critical mass in your head, yeah.

3:15 And then it came out into the world.

3:17 What was the first set of videos?

3:18 Like?

3:18 I mean, as of any creator,

3:19 they will tell you that their first, honestly, first 50,

3:22 is all throwing spaghetti at the wall and figuring out what sticks,

3:25 figuring out how your audiences react to it.

3:28 And the great thing about content, compared to something like your venture role,

3:31 where you need five to 10 years to have any sort of feedback loop,

3:34 is you can find out in literally the first three minutes.

3:37 I can tell when and how a video will perform in the first three minutes,

3:41 you learn with different platforms, but how the algorithm spreads?

3:45 I feel like I've taken the most like productized,

3:48 techie approach to building content.

3:50 You basically have all these different variables,

3:52 like how shareable a piece of content is,

3:55 watch time, whether it drives engagement and commenting.

3:58 So you might have things that are quite interesting,

4:01 but people don't stay for long enough.

4:03 Or you might have things that are people don't actually like the subject,

4:06 but it's highly shareable,

4:08 like anytime I talk about what the future labor market looks like,

4:12 highly shareable because people are concerned

4:14 and they want to share with their friends.

4:15 Oh, this is what this person has said, or this is what this report has shown.

4:19 I think I've now built this feedback loop in my head,

4:22 and kind of have gotten a very good sense of what will stick

4:26 and what won't with but that leads to the audience that I built.

4:28 What topics Did you find resonated with your audience early?

4:31 Like, what was some spaghetti that stuck on the wall and you said,

4:34 Oh, okay, that's it?

4:35 Yeah, I've seen a lot of content try

4:37 to, for better or for worse, dumb things down,

4:39 and I didn't want to take all the things I

4:41 was seeing in tech and dumb it down for the audience,

4:43 because I think a lot of people are able

4:45 to actually understand the stories just haven't been translated for them.

4:48 Especially as a woman, I've seen a lot of finance content.

4:51 It's like for the girlies.

4:52 And I'm like, No, I think, you know, as women, we can understand this.

4:55 No, I actually do get Yeah, we can.

4:57 We can.

4:57 I would just like to hear as you know, and it's.

5:00 Regular form.

5:00 And so, you know, I wanted to say first,

5:02 authentic to the content that I was interested

5:04 in, for bias to what the ecosystem didn't necessarily have,

5:07 which was an investor's perspective, or kind of that macro,

5:10 30,000 foot view perspective on where the world was heading,

5:13 and trying to play a translator.

5:14 I think there were a lot of headlines.

5:16 And I do think for your regular audience.

5:19 They're being flooded with all these headlines,

5:22 but they don't know what it means.

5:24 Does it mean that, you know, this space is going to be this space

5:27 is being targeted by anthropic and therefore I

5:30 need to start developing my own skill set

5:32 in design now that they're developing cloud design.

5:34 You know, I've spoken extensively about the AI supply chain.

5:37 So what does that mean as someone who you know,

5:40 might not necessarily have access into understanding

5:43 the world of Nvidia and memory and chips.

5:46 You know, you were the youngest partner at GSV.

5:49 Did you always want to be in venture?

5:51 Was that always the plan?

5:52 I knew.

5:52 I always wanted to do something in the intersection of tech,

5:55 business and education.

5:57 I've always played translator.

5:59 My mom doesn't speak great English when we were moving

6:01 clay translator for her navigated living in both Asia and us,

6:05 while I was at grad school, had the opportunity to join GC,

6:08 where I could merge these three worlds.

6:10 And so took up the opportunity, and it was one of the most amazing experiences,

6:14 just being able to dive deep into an industry

6:17 as someone who's always been passionate about future technologies,

6:21 being able to translate all this future technologies into what

6:23 that would look like in education and workforce settings.

6:26 Is there any part of investing that is a translation task,

6:29 especially in the AI era, I've always thought of venture as a narrative job.

6:35 Some of the best investors that you know I

6:39 can point two or look up to our storytellers.

6:41 They're able to sell you on blog.

6:43 Who are you thinking of when you think of that Fred Wilson's blog,

6:46 looking at the content that Bill Gurley and Brad Gerstner

6:49 have been able to put out in their BG two pod,

6:51 I've always looked at those as great models

6:54 of investors who also navigate storytelling very well.

6:58 I've always seen those jobs is going hand in hand,

7:01 because when you are building yourself as an investor,

7:05 you are selling why you should be a partner, right?

7:08 You're selling why this vision makes sense.

7:10 You're selling why your vision for the future of the world

7:13 is supposed to align with this specific company and venture specifically,

7:17 because you are existing in multiple spaces,

7:20 you're existing in the same rooms as AI researchers

7:23 who haven't been exposed to the financial metrics, right?

7:25 And so you have all these founders who

7:27 might currently exist in the AI research world.

7:29 They're deep in founder mode, but then eventually,

7:32 as a potential IPO or they have an amazing exit,

7:35 they have to learn how to navigate the finance world, right?

7:37 And then, at the same time,

7:38 through all of this, a company is also needing to tell its story to the masses,

7:42 like, what do you represent?

7:43 What is your brand?

7:44 What are you doing to change the world?

7:46 And so I think of venture as existing in this interesting space where

7:50 you are helping a founder navigate all of these spaces, and in doing so,

7:54 you have to play translator quite well,

7:56 because one you have to master all these spaces,

7:58 but also able to tell the story across all these spaces,

8:01 because those stories look very different.

8:03 Depending on whether you're talking to a general

8:05 consumer or if you're talking to a banker,

8:07 or if you're talking to another founder.

8:09 Where is AI awareness for your audience right now?

8:12 I mean, definitely mixed.

8:13 I operate by day in this space that is very techno optimistic.

8:17 And I think by default, I am pretty techno optimistic.

8:20 I grew up with a dad who would bring us to every single sci fi movie.

8:24 He actually collects sci fi movie props.

8:26 So what is the best sci fi movie prop he has?

8:29 Maybe.

8:29 What is your favorite?

8:30 Well, shout out to Claire's dad.

8:32 Wherever you are, having the Terminator head in your basement is

8:35 probably a good reminder of what happens when AI goes Skynet route.

8:38 I've always been a big believer in what

8:40 AI and technology can unlock for society.

8:43 I think, you know, humanity is always innately wanted to build and invent.

8:48 It's kind of like core to who we are as a species.

8:51 So I exist in that space on a daily basis,

8:54 and it's so fun being able to be at a place like Lightspeed or at GSB,

8:59 having peers that feel that excitement or existing in Silicon Valley,

9:03 where so many people believe in this vision at the same time.

9:08 I think creating content has and building this audience

9:11 with the general masses has actually been a humbling experience.

9:14 In many ways, it's pushed me to go beyond this kind of techno

9:18 optimistic bubble and recognize that a lot

9:21 of people aren't necessarily anti tech.

9:24 They just feel like they've been left out of the journey and therefore,

9:29 and because nobody's playing translator for them,

9:32 their exposure to this story is only doom and gloom oriented.

9:36 It's only about how tech is going to take their jobs.

9:38 You look at like the Maslow's hierarchy of needs.

9:40 AI has done a pretty bad job of addressing every part of the stack.

9:44 It's like, literally, your water need, right?

9:47 Like, data centers taking your water, your purpose, your job.

9:50 And so I've never heard that, yeah, actually, yes, yeah.

9:54 And so, like, of course there's panic, of course there's anxiety, yeah.

9:57 And I don't, I don't fault people for having anxiety.

10:00 I sit in my day job and I see how much excitement there is for AI,

10:04 for humanity, for science, for healthcare,

10:08 for just scientific progress, the fact that we might have orbital data centers,

10:13 I don't know, and that was always like a movie concept for me.

10:16 But I think for your regular audience,

10:19 they're only hearing the negative headlines,

10:21 the most extreme headlines, and they're not necessarily seeing the incremental,

10:25 exciting change that we are seeing,

10:26 that I feel I have privileged access to through my work adventure.

10:29 What do you think is the most important thing any

10:33 general audience needs to understand about AI that really this is,

10:36 I mean, everybody says, an unprecedented time, but two,

10:39 it really is kind of a jagged frontier, right?

10:42 It's jagged intelligence.

10:43 I think it's interesting, because when I post anything about AI,

10:47 I will see either a reaction that is, oh, no, this doesn't work.

10:51 I've tried it, and terrible output,

10:53 like these investors and these founders must be lying.

10:56 But then I'll also see reactions that are like, this thing's gonna take my job.

11:02 What do we do?

11:03 We shut it down.

11:04 And so it's almost this tension where we

11:06 need to kind of center where AI is today.

11:09 There's a lot of promising where AI will go.

11:11 But I also think there's maybe a discrepancy

11:14 between how people are using it today

11:17 and what the technology is enabling at kind

11:20 of the tech bro Silicon Valley level.

11:23 I think, for example, I'm seeing a lot of people only exposed

11:26 to AI in the form of a better search engine.

11:29 And I think that's therefore,

11:30 when you tell them that this, you know, something like an alpha fold,

11:32 is going to help us discover entirely new protein

11:35 structures and therefore unlock personalized vaccines like that is

11:38 a there's just a massive gap between better search

11:41 engine and personalized medicine that could eventually extend lifestyle,

11:46 find cures to cancer, auto immune diseases.

11:48 Yeah, and I think this that already existed,

11:51 and that's actually a harder story to tell than,

11:54 Oh, it's going to take your job.

11:55 Yes, I think that there's enough messaging

11:58 and storytelling around the superintelligence and singularity story,

12:01 but what about all the cool advances that we are making for humanity?

12:05 And what about logging the incremental step

12:07 changes in both positive and negative directions?

12:09 We've only heard the very broad headline, and so there's not that nuance,

12:14 and therefore it's just this ominous thing that you have to be afraid of.

12:18 And so hopefully the work that we're able to do is really play translator.

12:21 That this is exciting.

12:23 It's less scary than you might think.

12:26 But there are also things that we as an industry

12:28 want to invite the public to be involved in.

12:32 What I've been really surprised by is actually just seeing that these AI

12:36 labs actually do really care about how people use these tools,

12:39 how they're exposed to it.

12:42 They're working to invite people to the table.

12:44 Obviously, you know, we're not in a perfect world.

12:46 No, we're not, yeah, but I think there's, but the effort matters.

12:50 Yeah, as I think about maybe where the public sits

12:53 with broader AI sentiment and where the tech world is building,

12:56 there is definitely silos,

12:58 but at least I'm seeing in the public sphere an interest in tech,

13:01 there's a an innate desire to learn and participate in the future.

13:05 And I think for the big tech world, there's a desire for people to feel

13:09 as excited as we are about about these technologies.

13:13 You've talked before about how it's important

13:16 for people to get educated about AI, particularly students.

13:19 What would you compare that importance to?

13:22 Ai literacy is almost like sex ed.

13:24 You don't have to love this technology.

13:26 You don't have to promote it, but you have to know about it, and if you don't,

13:31 the ways that people interact with this technology is

13:33 going to be kind of the worst form of it.

13:36 When we talk about diffusion of this technology to young people,

13:38 the ways that they're going to learn about it is through Snapchat or meta AI,

13:42 as opposed to learning about how it can be used

13:46 to accelerate scientific discovery or how

13:49 it might enable their own entrepreneurship.

13:51 Having knowledge and awareness is critical,

13:54 and if you don't even give them access to that awareness,

13:58 you're doing the service to those people,

14:01 education has actually been one of the most

14:04 high friction industries of introducing any search technology.

14:06 And so I've actually thought a lot

14:08 about the diffusion of different technologies,

14:10 specifically the diffusion of like AI technology.

14:13 I think what's interesting is you have all

14:16 these different factors at play in the education space,

14:21 you simultaneously have kids who are using this as a cheating

14:24 tool institutions who are trying to ban it,

14:26 but then also want to use it, because if you don't use it,

14:30 then you're not preparing people for the future.

14:32 But then simultaneously, Gen Z actually having a new found angst against AI.

14:37 I think we recently did a survey with Gallup, and it found that for Gen Z,

14:44 the of all the different emotions, anxiety and angst, grew the most around AI.

14:50 So Alex, I think, yeah,

14:52 I've been thinking a lot about how technology spreads and education it.

14:56 You know, education and healthcare, I think are two spaces where, uh.

15:00 So you often see different velocities of adoption

15:04 at the institution level versus at the user level.

15:07 And I think education is one of those interesting case studies where, you know,

15:10 you simultaneously have your end user as a student,

15:13 adopting it faster than ever, but then having mixed feelings about it,

15:18 and then institutions scrambling to figure out, you know,

15:20 they're like, We need to pilot all these things,

15:22 but then often isn't actually putting the work to implement it correctly.

15:25 This technology is inevitable.

15:26 It is exciting.

15:27 It's the future, and we can't pretend it's not there, yeah?

15:31 And if so, we need to talk about it, yeah.

15:32 And if you don't talk about it,

15:34 it's a disservice to people that these institutions are serving.

15:36 I do think there is something really interesting.

15:38 If you're in school, for example,

15:40 it's probably easier to go to chat GPT for certain things.

15:42 Certain things.

15:43 So I imagine there's a way to sort of triage best practices and be like,

15:47 Okay, this is a way that AI can help you learn.

15:49 This is a way that it maybe would hinder you from learning.

15:52 When I have seen the concern, very valid concerns around AI usage,

15:57 I often ask, like, how do you use AI?

16:00 And it is pretty telling that they, again,

16:03 use it only as kind of a better search engine,

16:06 or they're using it as, like a one shot output,

16:09 and they're just taking it as a completely written essay,

16:11 and they're like, send it in.

16:13 When people exhibit agency over this technology,

16:16 it is a massive unlock for those people.

16:20 At least for me, I feel like the floor has risen on my ability to design things,

16:24 ability to code, ability to write.

16:26 Do I think it's replacing me?

16:27 And do I ever use it?

16:29 You know, do I take the first output that I ever get from chat,

16:32 GPT or Claude, and submit it?

16:34 That's probably not good practice.

16:35 And I think in the same way,

16:37 so fair that you probably wouldn't ideally take, you know,

16:40 someone else doing your you know the cheating problem has existed

16:42 for a very long time in the same way that you would,

16:44 since the beginning of school.

16:45 Yes, yeah, if in the same way that you won't

16:47 learn if you ask your friend to do your homework.

16:49 It's the same thing with chat GPT doing your homework.

16:51 I think it ultimately comes down to how you use this technology.

16:54 I'm curious about your take on where

16:57 the conversation around data centers is in 2024

17:00 GPUs were selling faster than Taylor Swift

17:02 tickets or something like that, which is crazy.

17:05 What analogy would you use for the conversation around AI hardware,

17:10 the conversation around data centers and where we're at right now?

17:14 I don't think everybody who has entered Tech has a negative,

17:18 villainous vision of the world, and they also want to improve it.

17:22 It's just that there's kind of a difference

17:25 in how we get there among the tech industry.

17:28 The prevailing mindset around things like data centers

17:31 or the harms that the AI industry is making right now is that these are problems

17:36 that will be solved by AI improving itself, this idea that, yes,

17:41 this massive upfront physical build out will have physical harms today.

17:47 However, we're already seeing advances where closed loop data centers

17:51 or advances in ship efficiency or even orbital data centers,

17:56 there is a genuine desire to improve the world,

17:59 and the physical build out issue is one where

18:03 it's almost a difference in how you approach a problem.

18:06 It's like the vision is that we get to a place

18:09 where the technology is able to help us solve climate issues,

18:12 is able to help us solve the data center build out issue.

18:15 Obviously, I think if you're not a believer

18:18 in the iterative intelligence that AI promises,

18:21 it's hard to see, and you just see

18:23 a bunch of reckless decision making by a couple stakeholders.

18:26 So I understand where all sides are coming from.

18:30 There's an interesting opportunity for, again, tech to tell that story.

18:34 Because I think right now,

18:36 there is just kind of almost a we're just going to do this.

18:38 You've seen the headlines where just we're going

18:40 to build a Manhattan sized data center, and not really,

18:43 not really explain the why or what this unlocks,

18:45 or where it fits into the supply chain, or how it enables other technologies.

18:52 So I think again, always comes down to a difference in how we

18:58 get there and who has visibility into the entire roadmap, if you will.

19:04 You've invested in AI startups, right?

19:06 You've served as an advisor for Arizona

19:08 State University's external AI thought leadership group,

19:11 which launched chatgpt edu, but you've also warned about the dangers of AI.

19:16 What is your personal relationship to that tension?

19:19 I think this technology is kind of like Promethean fire, right?

19:23 There's, it's hard to assign it this very binary good or bad,

19:29 good or evil status, because it's really how the technology is used.

19:34 It's like, you know, I think even electricity, right?

19:36 Like electricity, can enable amazing things for society,

19:40 but it also did technically enable put

19:45 the lamp lighters out of business eventually, yes, over a long period of time,

19:49 but it is what happened exactly, and so you can there's always an argument to be

19:53 made about why a technology can be positive or negative.

19:56 You know, this is the first time ever you.

20:00 It is not necessarily just a tool.

20:02 It is also able to replicate certain functions that humans have served.

20:08 I think a lot about the ways that AI

20:11 is being used in place of certain relationships, right?

20:15 Or, you know, my fear that AI is reducing our metacognition as as people,

20:24 our ability to learn, how to learn,

20:27 because we just kind of take AI's answer as as the final answer,

20:30 and then you're set with it.

20:32 There's no inquiry, there's no critical thinking,

20:34 and I can seem so decisive and be so confidently wrong, yeah.

20:38 And I think about like, okay, then AI psychosis.

20:40 And I think about the AI girlfriends and boyfriends and, you know,

20:43 the loneliness pandemic.

20:44 And so I there's a lot of things on my mind, but I think ultimately,

20:49 I believe in the strength, the collective strength of humanity and our ability

20:55 to steer it in the right way.

20:57 I genuinely do believe that people in a lot of these companies

21:02 have the right motives that they want to build a better future.

21:06 And I do think you know whether or not

21:08 you agree with the ways that we get there.

21:11 I think it's hard to just say that have

21:14 this very kind of one dimensional view that, like everyone,

21:16 is an AI villain and trying to build an AI super villain, super intelligence.

21:20 I genuinely do believe that people, there wouldn't be so many smart,

21:26 intelligent people in this ecosystem thinking

21:29 about what's possible with this technology,

21:32 if there wasn't a promise of a better future.

21:35 And maybe, or I don't know, who knows,

21:38 I sometimes also kind of challenge my own assumptions.

21:40 I'm like, maybe I've just been, you know, bought into the cult,

21:43 and therefore that's what they want us to believe.

21:45 But for now, I think that's kind of where I sit.

21:48 I think anyone who says they know where the world is going is,

21:52 yeah, they're messing sometimes, the best that you can do is actually just

21:57 help explain what is happening to your best ability,

22:00 and make sure that people feel like they're part of it and feel informed.

22:04 And sometimes, you know, there are days where I create content and you know,

22:09 I really do say like, I don't know,

22:11 but here's what the tech is saying, and here's how I might explain it,

22:15 and here's how I might explain the storyline.

22:17 And one thing I've, I've always tried to hold true of my content

22:20 is that I never really want to tell people what to believe,

22:24 but rather, I will just explain it for them.

22:26 And you can decide whether or not you love or hate the vague AI labs.

22:30 You can decide whether or not you love or hate the data centers.

22:33 I've never wanted to, you know,

22:35 steer or sway anyone in one very strong direction.

22:39 Obviously, just my biases, that I am techno optimistic,

22:43 and I'm wary of that, and that genuinely,

22:45 generally I will cover probably things that lean more techno optimist,

22:49 but at the same time,

22:51 I think I've tried intentionally not to make extreme statements,

22:57 like I don't think there has ever been a video

23:00 where I've said AI is going to take this specific job.

23:02 I've talked about reports and research that has happened in that space,

23:07 but I feel like I've tried to be

23:10 quite intentional about not being overly rah rah tech,

23:13 but also not being overly anti tech either,

23:16 and really just trying to present the facts in a digestible way.

23:20 You've clearly developed a really nuanced view

23:22 of the relationship between Silicon Valley and Main Street.

23:26 Should we say?

23:27 How does it inform how you invest to your point?

23:31 Sometimes in venture, we tend to talk to each other.

23:33 And you know, if another fund is doing a lot in this space,

23:37 it must mean that it must matter, and that's where the world is going.

23:41 And I think that leads to some of the level of behavior where it

23:45 is just kind of this incestuous loop of orbital data centers must be the future,

23:50 so we must write it check into a server in the space.

23:53 And I think there's grounds for, you know,

23:56 the investments that you know Lightspeed as a firm is making.

24:00 There's, like, clearly, a lot of research and rigor behind it.

24:03 In venture in general, there's a lot of noise,

24:07 and it's hard to tell what's actually reaching audiences.

24:10 I think one good example is, as an investor, all I hear about is,

24:17 or at least, I am a massive user of Claude.

24:20 It feels like there's been a lot of kind of Claude maxing,

24:25 if you will, among the tech industry, like almost often a jump.

24:29 But then I think there's recent data

24:32 that shows that for most your general masses,

24:35 what they know is AI is chat, right?

24:37 So I think there's helpful.

24:39 It's helpful to have this real time sounding board,

24:43 where anytime I talk about a specific topic,

24:45 I can get a sense of what the public is feeling about that topic,

24:50 or if they actually if it's resonating, if they've even heard of it.

24:53 I felt like I was already seeing

24:55 this difference in usage between the general public.

24:57 There's a lot of Gemini usage.

24:59 Message, chat, GPT usage,

25:01 and Claude was kind of this, this, you know, archaic thing.

25:05 They're like, what's Claude?

25:06 And then, you know, when the anthropic ad came out,

25:08 they're like, oh, it's Claude chats enemy.

25:11 And so it's things that I want chats.

25:14 And yeah, yeah, exactly, they're being answer isn't, no,

25:19 they're not not beefing, but it's just interesting, because you're like,

25:24 sitting in your you know, you're sitting in x,

25:26 where you're seeing all these investors talking about,

25:30 like, their armies of agents.

25:32 And then on the flip side, you're interacting with the masses who are still,

25:37 like, Claude is beefing with what's up with this?

25:40 Yeah, speaking of something that is a big topic on x,

25:43 and big topic sort of in venture land,

25:45 the sort of one person or the tiny team, Billion Dollar Startup.

25:50 Do you think we'll see that over the next year or two?

25:52 I have seen the ways that this technology provides tremendous leverage

25:58 for an individual that is able to take advantage of it.

26:01 I think, as always, with anything there is any sort of technological change is

26:08 oftentimes like two parts the tech

26:11 and then eight parts the actual change management.

26:14 And so do I think that, like we're going to see all

26:19 of these organizations size down and only run off of agents, managing agents.

26:25 I think it will take a long time to get there,

26:28 because there's still a lot of human context that's like the missing layer.

26:32 There's a lot of judgment, there's a lot of, you know,

26:35 figuring out which tasks are prioritized like I

26:38 think we have the actual units of power.

26:41 They're just not necessarily routed in the ways

26:44 that make sense in a human organization today.

26:47 Do you believe it is possible to have a moat in AI at this point?

26:52 I think so.

26:54 I think there is always, you know, a human moat.

26:58 There's always a judgment mode.

27:00 You know, I think if you think about all of these intelligent people,

27:06 there's you also have to figure out who is building in the right spaces,

27:10 and do they have the subject matter expertise

27:13 or the skill set to build in that specific space.

27:17 So I think while the floor has risen and what you're able to do,

27:21 and that there's never been more of an exciting time to build,

27:24 because you are now empowered by an army of agent coders,

27:27 and you can use Cloud design to help mock up your your brand design.

27:32 And I think generally, the floor has been lowered where anyone can

27:39 start building in a much more accessible way.

27:42 At the same time, there's still a need to figure out,

27:45 like, do you build in this specific industries?

27:48 Do you target this user?

27:49 Like those business modes still matter, because yes,

27:53 an AI can tell you and help you with making such founder decisions,

28:00 but at the same time, there is every single founder in a single

28:05 day goes through so many decisions that ultimately,

28:08 that's why I'm so excited about betting on founders

28:12 and investing pre seed and seed stage is because really

28:16 it does come down to people and their ability

28:19 to navigate this Tech and the ability to compounding over time,

28:22 make the right decisions.

28:23 I want to bring back something that you've said before,

28:26 that this is the first time we're able to produce intelligence,

28:29 and there's going to be a job for people who can speak to intelligence.

28:33 Well, how does that affect the kinds of startups

28:35 you'd be betting on for the next 10 years,

28:38 20 years, as with any technology, time and time again,

28:42 we've seen labor get abstracted to a higher level, right?

28:46 I think for a long time we had humans as manual computers.

28:52 They're just plugging wires in and out, human alarm clocks.

28:56 Human labor will always exist,

28:58 because you need that kind of judgment at the same time,

29:02 I do think we are seeing AI get to a place where AI is

29:05 able to produce a certain level of judgment to judge another set of AIS.

29:10 And that's kind of the vision of AGI as I

29:14 think about the teams that we're excited to back.

29:18 It's really the types of founders and teams

29:21 that know how to take advantage of this compounding value.

29:27 It's founders who are able to constantly adapt

29:32 to this ever evolving technology that's changing on a daily basis.

29:36 And so I think in that world,

29:38 the MO is not just whether or not you have an insane product,

29:44 it's also whether a team can quickly adapt that product when

29:48 a big lab or an existing incumbent tries to enter your territory, right?

29:53 I think so much of the bet is not just on what you're seeing today.

29:57 The bet is on that.

29:59 Civic team and their ability to navigate what is

30:02 arguably one of the most exciting times to build,

30:04 but also the most uncertain times to build.

30:07 We see these really high valuations, right?

30:09 We see companies that have billions in revenue sometimes

30:14 after just being around for a couple of years.

30:16 Yes, that is absolutely impressive, but it doesn't mean it's a mature company,

30:20 and we shouldn't necessarily talk about them that way.

30:22 And how do you know if somebody can even manage that change?

30:25 Like, when you're looking at somebody to invest

30:26 in, how can you look at them and say, I think you could figure it out.

30:29 I think that's the hardest part about early stage investing,

30:33 is it really is betting on people.

30:35 And so there are ways, I think, to translate and use proxies,

30:39 right like I think you can look at how they've built previous products.

30:45 Have they built companies before?

30:47 How have they navigated those journeys?

30:49 You can talk to people that they've worked with better understand.

30:53 How do they approach change?

30:55 Are they, you know?

30:57 Are they quick to adapt in moment?

31:00 Are they kind of more stubborn in certain ways,

31:03 and maybe being stubborn sometimes is good, right?

31:05 Like, I think in many ways still TBD,

31:09 but question of whether it's stubborn or just

31:12 in the ability to implement the change, but, like,

31:15 maybe the fact that Apple hasn't played at the foundation model

31:18 layer might be to their advantage later on, when you know,

31:23 if the models do commoditize, they have always had the distribution layer,

31:27 and they get to benefit from all of that.

31:29 So I think there's all these examples

31:33 where it is really just pattern recognition,

31:36 and it's yes, gleaning data points through

31:39 the proxies that you can get through knowing people,

31:41 through seeing how they've worked,

31:44 what they've worked on, but it's also really betting

31:48 on that specific person and their vision for the world,

31:54 I think, also recognizing that this is

31:58 a partnership where For any venture fund, at least,

32:01 I've always been excited about taking

32:03 a hands on approach to partnering with people,

32:06 because there is every founding journey is different.

32:10 Even a founder has started a company previously and had a successful exit.

32:14 It is also really just who that founder is surrounded by as well,

32:18 and that includes not only their team,

32:20 but also the investors that are that are backing them.

32:22 You've previously said that the future of the AI boom is hardware.

32:25 What did you mean by that?

32:26 And why does it matter?

32:27 We've actually come to a place where we're investing both Bits and Atoms.

32:31 Again, I think this idea that in order to usher this next technological change,

32:38 there is a physical build out.

32:41 There are energy and compute and data bottlenecks.

32:43 Those are all very interesting problems

32:46 to solve for companies and for founders who

32:48 want to change the world and who want to build solutions for the space,

32:52 and there will always be the software layer that sits on top of that.

32:55 And I think for maybe the last two decades, at least,

32:59 the companies that I've seen you've probably indexed

33:02 a little bit more towards the software side of things,

33:05 just because for the primitives around hardware, never really changed, right?

33:10 It was always those things pretty much stayed the same.

33:15 And so where you had opportunity to change

33:17 the world was to build better software.

33:19 And that's why the kind of name

33:22 defining or category defining companies of those era

33:25 or of the last two decades have been things

33:29 like Airbnb or stripe or or Google even zoom,

33:35 all these companies that exist at software layer.

33:38 I think now there's an interesting opportunity where,

33:41 because AI is requiring both a physical build out and a software layer,

33:46 there's interesting opportunities in both.

33:48 And I'm excited to meet and continue

33:50 to back founders that are thinking deeply about,

33:53 you know, the priors that we've had,

33:55 updating those priors for this new world we're entering.

33:58 I also think, you know,

34:00 there's a lot of interesting things that AI unlocks for hardware.

34:05 You know, I see those paradigms is kind of going hand in hand,

34:10 where this idea that we have AI that can now unlock ambience, right?

34:14 This AI that can now take in computer vision, it can take in audio data.

34:18 What's the best way to capture all that data,

34:20 you need some physical manifestation of that.

34:22 So I think there is, yes, the physical part in the actual heavy infrastructure,

34:27 but I also think in the ways that, whether that's robotics or ambient wearables,

34:32 the ways that AI, in turn,

34:34 unlocks a new wave of even like consumer hardware as well.

34:39 And that was clear, knowledge really is the antidote to fear,

34:43 and she is someone absolutely worth keeping an eye

34:45 on as the public conversation around AI continues heating up,

34:49 that's it for term sheet.

34:50 I'm Allie Garfinkel, we'll see you soon.

34:52 You.