0:00
I I'm very dismissive of this idea that like people are going to vibe code their own [ __ ] Like whatever.
Search this video
0:04
Like it that that isn't the issue.
0:06
The issue to SAS companies [music] is every single day that you are hitting your plan, >> [music] >> you are destroying equity value. Like think about that.
0:16
Our whole careers we learned [music] you lay out a plan, you execute against it like relentlessly and violently, you hit your plan or you exceed your plan, and you keep on building that and that's how you [music] build equity value.
0:28
And now you're going to hear me hear me just like every day you hit that plan, you're [ __ ] up. You're [ __ ] up. You're destroying value.
0:47
>> I love asking [music] you and all your partners this every time we hang out, which is, okay, you've got these singular investments.
0:52
You don't do that many of investments each per year.
0:56
And then the ones that go on to work like so far Sierra and and Firework certainly are, you get to learn so much about the world through the lens of the company.
1:04
So I'd actually love to do both of those, maybe starting with Firework.
1:07
What what do you know or what have you learned about the world and how it's reordering itself based on watching the world through the lens of Firework that would maybe be surprising or interesting.
1:17
>> These are two trillion, three trillion, four trillion parameter models.
1:22
Well, it turns out running those models [snorts] is damn hard.
1:28
And running them efficiently is like super [snorts] hard.
1:33
And the way to see this and everybody can see this, which is everybody from AWS to Azure to GCP to the neo clouds to the Firework space hunt together of the world, like all of them.
1:47
They all run these stock open source models that are available in their developer pools and everything else.
1:55
And and like that's what it is.
1:58
And the performance difference for a fireworks versus like a cloud provider is like 5x.
2:11
And that is just the speed performance.
2:14
Then you add on top of that the like throughput, which is not visible externally.
2:18
It's only visible if you look at if you know the economics of these businesses.
2:23
And you're like, wait a minute.
2:26
This is the same open-source model with the same Nvidia hardware.
2:33
And there's a 5x performance difference and a multiple x throughput difference and you know, and and the way to just simply understand that is like these companies are paying the margins of the cloud providers and running on top and making money. >> Yeah.
2:47
>> And so like how how can that be?
2:47
And so what I think my big takeaway on it was like, wow, this stuff is actually really hard to run.
2:53
Like it's just really hard to run.
2:55
And um and and there's a lot of expertise involved in doing that and it's kind of like this very specific expertise that that exists.
3:04
And it it is the kind of thing that when you look at it as an investor from the outside and we should talk about like early days of AWS, but like when you look at it from the outside, it's like this is a commodity.
3:15
This this this is just as like pass-through resale. >> Scale game.
3:19
>> Yeah, scale game like whatever.
3:19
Like this doesn't And you're like, oh wait a minute.
3:22
No, it turns out it isn't. It isn't at all.
3:24
>> Do you think that's just a moment in time thing and I'd love to just hear you I'd love to really riff on like cloud.
3:29
You watch cloud very carefully and closely.
3:30
You know a lot about it and the adoption curve there versus how people use these things and the the nature of those two businesses in comparison.
3:36
It's tempting to say like there will be one or two scale winners like there typically have been in a commodity market >> Totally.
3:43
>> where cost to serve is everything and scale drives cost to serve down and like that's the whole story.
3:48
>> I I just I think the AWS example's so good.
3:50
Okay, so 2006 you have S3 and EC2 launch, right?
3:53
Their compute platform and their storage platform is the first two AWS offerings in 2006.
3:58
And you kind of start talking about it in late 2006 or whatever.
4:02
2006, 2007, um I think the 2007 annual letter like Bezos talks a lot about about um about AWS and what's important and why it's interesting and everything else.
4:12
And the investor reaction is just not good.
4:16
And I think if you put 30 of the smartest investors at that time in a room and ask them like what's the probability that this AWS business is a good business with durable long-term margins and like super interesting and everything not commodity.
4:31
I think you would have gone zero for 30 with really smart people that you and I know who were who were around at that time in '07.
4:39
Okay, fast forward from '07 to 2014.
4:39
I joined the venture business in 2014 and a really common narrative in 2014 was oh my god, AWS is going to eat everything.
4:50
There's no enterprise opportunity left.
4:52
It's going to eat databases and infrastructure, but it's going to eat the apps, too, and they're going to offer at the cheapest and best and you know, and you got to remember like we all have these like amazing SaaS and software businesses that we were involved with and investors and part of the reason people loved them was they were annuities and they ran at like 85% gross margins and like everything else.
5:13
And so it's just like, oh my god, AWS Amazon can offer the exact like 8% gross margin and like they're just going to crush this whole thing.
5:19
Yeah, your marks my opportunity like the whole thing, whole narrative.
5:23
And and think about from like 2014 to to now in enterprise. Okay?
5:29
Like of course, you have Snowflake, direct competitor to Amazon Redshift ran on Amazon.
5:41
You're out-Amazoning Amazon on Amazon. >> [laughter] >> Okay?
5:44
But it's not just Snowflake.
5:44
You had Confluent, and Elastic, and [ __ ] Databricks. >> Mhm. >> Amazing.
5:53
That's the infrastructure layer.
5:53
Then you have the whole app layer.
5:56
In the app layer, like think about offerings that they offered at the beginning, Datadog, $100 billion company today.
6:04
Like they had a competitive offer, right? And and they did it.
6:06
And and of course there was tons and tons of roadkill.
6:11
There was tons of roadkill.
6:13
They did run over a bunch of stuff.
6:15
But even then in 2014, the thesis was the the view that AWS was going to eat everything was massively wrong, not because of all the examples I just mentioned.
6:25
It was massively wrong because Azure and GCP were irrelevant then, and fast forward to 2026 and they're unbelievable businesses.
6:33
So now, is AWS the biggest?
6:33
I think it's like a 40 30 20 split.
6:36
You ended up with an oligopoly of of that.
6:38
And even outside of those big three, you have Cloudflare, which is a cloud provider in a different sort, which is another $100 billion company.
6:49
So you have these like smaller players that emerged as $100 billion companies outside of it. So what's the takeaway?
6:56
The takeaway to me is like oh, there's just a bunch of like kind of zero-sum thinking and not realizing like how big >> What if it all works? >> What if it all works? And so it all worked.
7:09
And of course like getting the relative winner right matters, and there was roadkill, and there were companies.
7:14
So it's all of those things still matters.
7:15
I'm not saying like it's just like, you know, like spray and pray.
7:16
I'm not saying that at all.
7:19
But I'm just saying like the market was so big that one vendor could not scale and consume it all.
7:28
And they just couldn't consume the whole industry.
7:31
Now, I just like again, like it's different now and like all these things.
7:35
But if I just like just that whole notion right now with what you and I are seeing in AI, sure feels like it rhymes, right?
7:41
It's like >> It's like Dropbox is going to do everything. >> Right.
7:45
It's like it can't like really?
7:47
Like is that is that really right?
7:47
To me that view that it's just like oh this one company is going to eat it all doesn't hold and I would tell you that scaling while cloud scaled very quickly cloud did not scale anywhere close to as quick as what's happening right now.
8:01
And so just for that company to actually scale and deliver it, scaling in this case requires like a ton of infrastructure build up, right?
8:12
From like energy power shell you know, chips, memory um obviously the algorithms and everything else on top.
8:22
So it just it feels it feels to me like we're going to end up with an oligopoly um of of winners and there will be I I really believe they will be these like hundred billion dollar crazy smaller winners.
8:37
Um and so like that's you know, that's like that is just feels like that the same thing is kind of happening.
8:45
>> Can you talk about it also from the demand side and compare it to how you watched cloud get adopted in enterprise versus how you're seeing AI get adopted now?
8:53
>> So if you kind of go back to like 2010 2011 like even you know, three now you're like three four years into to AWS being an offering enterprises were super skeptical.
9:05
Like super skeptical like traditional enterprise blue blue chip enterprise.
9:08
Um you had digital natives out here, you had new companies forming that were like using the cloud, you know, I think kind of famously like Snapchat uh was built on GCP and I think at one point in this era maybe 2012 2013 2014 like Snapchat was like 40% of GCP.
9:26
Like it was just like those kinds of things were happening.
9:30
>> That's like Cursor being 30% of all these things.
9:32
>> 100% 100% same exact thing happened.
9:35
Same exact thing happened.
9:35
Enterprises were very skeptical, I think, of cloud until like they really, you know, by 2014, 2015, 2016 is like, oh, yeah, yeah, yeah.
9:43
You know, then I think the big banks and financial services and insurance companies and more conservative blue chip enterprises were like, oh, wait a minute.
9:50
We This is actually, you know, different and we're going to have to pay attention and you know, it became an issue in recruiting for them because they couldn't get the best developers cuz the best developers wanted to work on the easiest platforms and all of these kinds of things happened.
10:02
And so, the difference now feels really profound to me because while it is 100% the case that blue chip enterprise AI is not like well absorbed and well adopted and not not the same thing as going to cursor and, you know, walking the halls of cursor versus um walking the halls of, you know, a big New York financial services firm in terms of their AI use. They want it.
10:31
They want to figure it out.
10:33
They're running experiments.
10:33
They are spending against it.
10:35
They're trying to figure it out.
10:36
They're talking about it.
10:38
They're not being dismissive about it.
10:40
And I think they view it probably as more of an opportunity um than they did the cloud in terms of the potential impact on their business.
10:47
I think they probably view it more as a threat than they did the cloud and maybe they just also learned lessons from the cloud in terms of like what what's possible here.
10:57
And so, I I feel that they will continue to to try to figure it out and absorb it.
11:04
Having said all that, I do think one of the more interesting things, if you go from Silicon Valley out into the world and talk to these companies, these enterprise companies, and you realize like what the pace of adoption is and what the barriers for adoption are and everything else is, I think that like there's a ton of opportunity to help the enterprises get there.
11:25
To be one of the things that I I tell the companies I work on is like, hey, let's be their AI sherpa.
11:32
Let's be their AI sherpa."
11:34
If we're in that position to be their AI sherpa, where we're crossing both worlds, that's very valuable.
11:40
And I think that will continue to work.
11:44
>> Ramp is the only platform built to make your finance team leaner, faster, and better, saving businesses 5% annually on average, so you can stay [music] focused on growth.
11:51
Ramp customers grew revenue 3.
11:53
2 times faster than the average American business.
11:55
Visa, Versel, Cursor, Stripe, Notion, 11 Labs, Shopify, and [music] 70,000 other businesses all run on Ramp.
12:02
Mine does, too, and so should yours.
12:04
Learn more [music] at ramp. com/invest.
12:08
>> The best AI and software companies from OpenAI to Cursor to Perplexity use WorkOS to become enterprise-ready overnight, not in months. Visit workos.
12:15
com [music] to skip the unglamorous infrastructure work and focus on your product.
12:21
>> Felix by Rogo is a personal finance agent [music] that turns a single prompt into finished, client-ready work using your firm's own templates, context, [music] and standards.
12:28
Send Felix an email like, "Take these comments and turn them for me."
12:31
or "Update my tracker with the context of these emails."
12:33
or "Run the ability to pay math on this buyer."
12:37
>> [music] >> and Felix sends back finished PowerPoint decks, Excel models, and source research.
12:41
Felix works the way your team already does, delivering work quickly and accurately around the clock.
12:45
[music] Learn more at rogo. ai/felix.
12:50
>> What is What is Sierra teaching you about the adoption of this stuff?
12:53
And it's kind of actually it's really interesting contrast to Fireworks, where Fireworks is the sort of infrastructure provider.
13:00
Sierra is feeling its way through what are really cool things that we can do for end consumers, enable companies to do for end consumers.
13:06
Starting with customer service, but I know with Horizon now, you know, going beyond that.
13:11
Again, same question as for Fireworks, like what what do you What do you know that the world doesn't fully appreciate because of how you've seen that business evolve?
13:18
>> I think this is a good It's a great example.
13:19
So, first of all, like my partner Peter and Brett, I think this is their third company working together.
13:25
So, it's like a 20-year relationship, which is just like an amazing and great place to start.
13:28
And one of the things that I think makes, you know, Brett so special is, um and the team is they're technologists, but they actually have like lived in enterprise world for a long time and like really understand it and everything else.
13:42
And I think actually he's personified this whole idea of, "Hey, let's be their AI sherpa.
13:47
Like, let's start with customer service where very automatable and like be their AI sherpa."
13:50
And now we have these long-running agents I can do with Horizon that can do more and more stuff.
13:55
And one of the things that I think is most interesting about this to me is, yes, it's an application company on the surface, but they're doing real AI work.
14:04
They're experimenting with the models and they're building agents and they're very close to the metal of the models and the capability and the harnesses and they're understanding the jagged edge of AI capabilities, which is very different than the human smooth arc that we understand intuitively.
14:23
They understand that jagged edge of capability and they build around it.
14:28
You saw the evolution of Cursor from like IDE to like tab auto complete to agentic work like over and over and over again.
14:35
They were obsolating their work from like >> Yeah.
14:40
>> 6 months ago, you know?
14:40
And it's what Brett Taylor calls like sand castles, right?
14:44
We We used to be building castles.
14:45
Now we're building sand castles that are going to get washed away.
14:47
And that's If you don't think of software that way, >> Yeah.
14:50
>> Yeah, you have to embrace it.
14:50
Because if you're an artisan and you're like, "Hey, I built these like perfect foundation of castle and the bricks and it's like perfect and I really care about this and it's going to be here for 100 years," you're just not going to make it.
15:01
So, as you get emergent new properties in the models, which is, you know, you get a new every 4 weeks, you get new capabilities, they understand that jagged edge, they understand that application of that jagged edge or that valley to their customer base.
15:16
And they're they're filling in those gaps and translating it.
15:20
You can't just be superficially applying things.
15:22
I think it's actually a complete inversion of how product development used to work to how product development happens now.
15:30
And product development, you know, it always used to be say you'd also say like, "Hey, product managers you're trained all product managers shouldn't you know, should understand the technology but like really shouldn't be thinking about implementation and shouldn't be specifying implementation and shouldn't be doing this and doing that."
15:42
And the product manager's job is to really understand the customer and translate that problem to the engineer so that the engineers build a solution like that was traditional product management.
15:50
Like good luck doing that now.
15:52
That's a horrible way to do it.
15:52
You can't do it that way.
15:56
You actually really need to understand the nuances of model capabilities, what they're great at and where they fail and you need to understand the customer problem. And put those together.
16:04
And bridge those gaps to build valuable solutions.
16:07
I think that's one of the things that Michael and team at cursor did so well.
16:13
From the very beginning is they really understood the jagged capability and built a product that allowed that translation.
16:19
From developer to that jagged capability and kept on iterating on that as the edge changed.
16:25
>> It's kind of ironic that like all that sounds like the returns to being technical are going up even as models supposedly are taking away technical edge.
16:34
>> [laughter] >> Well, I I I yeah, yeah, I think there's two things I yeah, it's 100% and and actually [clears throat] I think there's this weird thing where there was this all discussion about like the traditional roles of product manager and designer and engineer and like whatever it is.
16:44
And I think what they really are are there are people who understand customer problems. People who have taste.
16:56
And people who understand the jagged edge of AI capabilities and are curious about it.
17:00
And like those are the three things.
17:04
If you're going to be if you don't have those three things, you're going to do great.
17:06
And it doesn't really matter if you're an engineer or a product manager or designer.
17:10
But if you have taste and understanding of customer problems and understanding the jagged edge, you're going to do well.
17:18
>> When we first did this so many years ago now, it's just crazy.
17:19
It'd be fun to revisit some of the ideas that we talked about the first time with SaaS.
17:22
But, you used this term that I've used ever since, which you called it the competitive frontier, meaning like the things that will determine the winners and the losers.
17:30
My partner Bree has this great idea that stuck in my head, which is that everything is a jump ball right now. >> Yeah.
17:36
>> And I'm really curious, in addition to this like idea of sand castles versus real castles, what else you're seeing amongst the people that are becoming competitive winners?
17:44
Like, is it different Are the traits different for winners now, personality-wise or business strategy-wise or business model-wise, versus what you learned in the SaaS era?
17:55
>> I'm very dismissive of this idea that like people are going to vibe code their own [ __ ] Like, whatever.
17:59
Like, it That That isn't the issue.
18:01
That isn't the issue at all with SaaS companies.
18:04
The issue with SaaS companies is the competitive frontier completely shifted.
18:08
And everything they thought they were building against and that would make them win are not what's going to make them win. Let's say databases.
18:14
Like, database Databases have been a phenomenal area for software for a long, long time. Great margins. Why?
18:21
You had app developers that would build against these specific database interfaces that existed for that database.
18:29
You would have more and more data over time.
18:30
And so, migrating an app from one database to another database was a giant project that was like very, very difficult to do.
18:37
And so, these were unbelievably sticky businesses that you could generate a ton of margin, right?
18:45
Of course, you got Oracle and SQL Server and like and a whole slew of smaller players like that did really, really well in databases.
18:52
Well, let's think about that in the context of AI.
18:55
So, like now, you don't have a developer building against a database interface.
18:58
You have Claude or or Codex building against a a database interface. One.
19:03
Two, the beauty of database interfaces is they're very, very well specified. >> Mhm.
19:10
>> Well, it turns out AI is very good >> [laughter] >> good at things that are very very well specified.
19:15
Three, agents don't get tired of monotonous work of translating one specification to another.
19:19
So, it turns out that now all of a sudden database migration which used to be like the like number one thing you would not do in software is like kind of trivial.
19:34
Yeah, it's just like put some money against it, move it. So, what changed?
19:37
Well, what changes is the criteria to be an amazing database company changed.
19:43
It isn't that we don't need databases or that everyone's going to build their own database or whatever.
19:48
Like that isn't what's going to happen.
19:49
What's going to happen is the criteria changed.
19:51
So, now you're going to have a ton more applications that start obviously as we're seeing everywhere.
19:57
Like people are going to experiment a lot more cuz it's much cheaper to experiment.
20:00
It's much cheaper to start a new application.
20:04
So, now you need databases that scale from like basically zero usage to if it works all the way through.
20:09
Like that matters a lot more.
20:12
Your cost matters a lot more.
20:12
You want to be able to spin these things up, spin them down, tear them apart over and over again.
20:18
So, the iteration speed goes up in what you need on a database.
20:22
Ultimately, I think the cost becomes the arbiter of this.
20:23
So, the cost and then this like kind of zero to infinity scaling and transportability and everything else around that become like the arbiters of who wins and who doesn't.
20:34
That's really different than like hey, I specked this database for our user thing and I you know, procured a license and I ran it on this kind of hardware and like everything else.
20:42
And and so >> [snorts] >> there've been elements of course of these things over time.
20:47
But, I think that just criteria changed.
20:50
And I think one of the big messages to these SaaS companies you know, a few years ago was like you have a choice.
21:03
Get to AI or be worth three times revenue.
21:06
And I think a lot of a lot of these public SaaS companies are trading at six times or whatever.
21:12
But keep in mind in '21 they were going for 30 times. Right? Like everybody, right?
21:16
And so you're like three times, wow.
21:19
I've grown 4x since then, right?
21:19
This is the whole multiple compression is a [ __ ] I've grown 4x, the multiple has gone down by a factor of six, so I'm worth less even though I've grown 4x in 3 years.
21:30
And gotten to break even and like all these things.
21:32
So like that was the first message.
21:33
But I think that even it became really visceral to me where every single day that you are hitting your plan, you are destroying equity value. Like think about that.
21:48
Our whole careers we learned like you you lay out a plan, you execute against it like relentlessly and violently, you hit your plan or you exceed your plan, and you keep on building that, and that's how you build equity value.
22:01
That's [clears throat] what the whole management team's learned.
22:03
That's what these CEOs learned. Like pretty out.
22:05
This is like everything learned.
22:06
And now you're going to hear just like every day you hit that plan, you're [ __ ] up. You're [ __ ] up. You're destroying value. You're destroying value.
22:14
And And the point of like you know, saying that to them was to set them free.
22:19
Another articulation of this from my my friend Anne Lee Skates it was was just like the the CEOs who were going through this transitory period and had a business that was at a hundreds hundreds of millions and they thought they had you know, they're already they were were working on their business from 8:00 a. m. to 5:00 p. m.
22:36
and then trying to do AI from 5:00 to 8:00 in the evenings.
22:42
And what they needed to be doing >> is the inverse. >> is the inverse.
22:44
And it's so hard to do that because of all of the training and all of the muscle memory and all the inertia and everything that we learned about we're all hill climbing in a way and like, you know, the success model was like set the plan, execute the plan, build value, compound value. and it's like, oh no. No, no. Stop that.
23:03
You're you're you got to completely invert.
23:06
And I think there's a very very long way to get back to your question of like what the profile or what the mentality of the the kind of winners are right now.
23:18
But if we look at, you know, Brendan from Recore or Lynn from Fireworks or or Max or Brett, any of these people, they're so nimble about what the eval is.
23:33
Like, what is what are they optimizing against?
23:35
They're so nimble on all of it.
23:38
And if you look at every one of those companies, the evolution of the business, it's just the business is constantly evolving.
23:43
And they've done such an excellent job at that.
23:46
Um and I think that is very very different than certainly the the way I was like taught.
23:57
>> What's your sense of the disorienting nature of model progress?
24:02
And like how to We were talking You said earlier, for every four weeks like I got a a new jagged edge I got to figure out.
24:09
You're one step removed from that as an investor versus as the technical founder, you know, with your with your hands on the metal.
24:13
How are you behaving differently uh than you would have, you know, 3 years ago or something because of the pace?
24:20
>> Anytime that I'm talking to a a founder about, you know, a problem in their company or what they're doing or move they're making or anything else, which is what I spend 80% of my day doing.
24:31
I'm very This is how we used to do it. >> Mhm.
24:35
>> This is what we would typically do.
24:35
This would be the typical readout on this like this would be the old school readout of why this candidate is better than this candidate.
24:43
Now like, let's reevaluate that in the context of today.
24:45
Let's reevaluate that in the context of an unstable technology substrate.
24:50
Let's reevaluate that on you know, in the context of like a business model that's growing this way versus that way.
24:57
And and and and so I've really started to question like every assumption and every lesson that I learned before like which of it translates and which of it doesn't.
25:07
And and so like that's a huge difference.
25:11
I'll give you a really concrete example.
25:14
It which is which has been very um disruptive inside of um these like scaling AI companies.
25:19
So these scaling AI companies, we've had a lot of leaders come from the prior generation with great experience and everything else come to these scaling AI companies and completely flame out.
25:30
Like you and you see it across the industry.
25:34
And so the question is like why?
25:34
And it's these are these are some of the best leaders from four or five years ago.
25:40
They had all the lessons. They learned it all.
25:43
You know, they're excellent.
25:45
But somehow it's not translating.
25:45
And and there's just like some impedance mismatch between like the AI founders potentially, the needs of the business, and what these people are bringing.
25:57
And I saw it like really abruptly with with a particular sales leader who we hired who's like, "Hey, we can't hit any of these things cuz the way that software sales has been taught forever is a quota capacity model.
26:09
You have a quota capacity model. Each rep does this.
26:12
In the early days of a company, the quotas are, I don't know, 1. 2, 1. 5 million.
26:17
You know, maybe the ISRs are at 750 or 850.
26:19
And then over time, it scales up and enterprise gets to 2. 5 million.
26:23
And that's the whole thing. >> it works. That's how it works.
26:25
That's how all these financial models are built.
26:28
>> You start with the quota capacity model.
26:29
You take a discount on attainment.
26:29
This is This is what we can do. Boom boom boom.
26:32
Now, like fundamentally, without like realizing it, everybody was implementing something that was based on pushing demand, not pulling demand.
26:47
And for so many of these companies, they're operating here, these customers, and there's a new AI-enabled product that comes along, and it's just [ __ ] magic.
26:56
So, these companies are selling magic. Like, magic.
27:00
Well, it turns out if you're selling magic and you're the first one there >> So, a lot more than 2 >> you're going to sell a lot more than 2 million.
27:06
And so, the whole notion of a quota capacity model and that being working like, it it's not that it doesn't matter.
27:13
It matters kind of, but it's definitely not the first-order thing or constraint.
27:16
And so, you have these like execs come over with this just like, here's our math and here's the territory assignment and here's what we would do and, you know, first you do West Coast and then you do East Coast and then you do Central, all of these things.
27:26
And it's like, oh wait, no no, it doesn't work like that at all.
27:29
And so, one of my big things that I started to realize is as I'm interviewing these folks and and and talking to them is just like, hey, you need to check everything at the door. Just like, check it all.
27:38
Which is probably good practice anyway, but check all the baggage.
27:43
Check everything that you learned and just learn this from first principles. Like, how's it working?
27:47
What are really the bottlenecks on delivery?
27:50
What are the bottlenecks on demand?
27:51
Because it turns out that in a lot of these companies, you have reps doing 10 or 20, 30 million. >> 50 recently.
28:01
>> I mean, like, and so it's like, okay. >> Yeah.
28:04
>> It turns out that's different.
28:06
>> What is like the best sales person that you've seen that's doing it in a de novo way doing?
28:10
>> Honestly, the best sales person in any of these companies is the founder.
28:12
And what they're doing is bridging the jagged edge to what the customer's capability is, and that's it.
28:16
And it's like, I mean, it sounds so simple, but it's not.
28:19
But But that's what they're doing.
28:21
And um and then the you know, the market is just so big.
28:25
Like, just like we were talking about with cloud, I think the biggest mistake everybody made was the market, they just undersized the market.
28:32
And it turns out the market's just really really big. And this market is big.
28:38
>> It feels like right in this moment in time, actually just this morning you and I are in this great this great group um chat together where the discussion is the the kind of demand for intelligence like, I don't know.
28:46
It seems kind of unlimited and it seems like the smarter the thing gets the more demand there is and that maybe the bottleneck is just capital.
28:53
Like the the world just feels like it needs to take a breath. >> Yeah.
28:56
>> Like the RSI kind of concept if you apply it across technology it doesn't need to breathe.
29:00
The agents don't get tired but the world feels kind of like oh man it sure would be nice to have three months just to like digest this a little bit and for capital to form and and and evaluate its prospects and the scale is getting so big.
29:13
Like does it feel to you like this can just keep going or are we just going to get tapped out of money that can be invested in these things when it seems like we could consume like any amount of money to build any amount of stuff and serve any amount of inference like What do you feel about the market? >> I I I would say this.
29:31
I am worried about energy.
29:35
If you think about the models as translating compute into intelligence.
29:40
Like quite simply what do models do?
29:41
They they very effectively translate compute into intelligence.
29:46
>> [snorts] >> What's the demand for intelligence?
29:47
Well, it seems like a a lot a lot, right?
29:50
So then it follows that we will kind of continue to like have more and more demand on compute.
29:55
I'm saying compute broadly not not chips, not storage, not whatever.
29:58
But then like what do we need for compute?
30:02
We need energy like a lot of energy.
30:05
There's all this topic of distillation and and and Chinese open source and all these things but the bigger thing to me is that I think China's bringing on 10 times as much energy next year as we are in the US.
30:21
If energy is what you need for compute and is the bottleneck and there's unlimited demand for intelligence then it stands to reason that if we have a lot less energy then we will have a lot less intelligence or a lot less tokens or a lot more expensive tokens.
30:41
And if we have a lot more expensive tokens, then >> Supply demand.
30:44
>> You're going to end up with less. And that seems very bad.
30:46
And so, to me, I think the energy bottleneck, however that that is, and it'll manifest in 20 different ways, you know, gas turbines go up and down, natural gas get up and down, and solar, whatever, like all these different things, where it's like that to me is is probably more concerning.
31:06
And if I'm thinking about it from a regulatory perspective or government perspective, and I think the administration is doing some things around this, like you know, fostering really investment and development of all energy.
31:23
That and I I mean all, like solar, nuclear, gas, like whatever, do it all.
31:27
Like we should do it all.
31:29
And and it'll work itself out, you know, but again, this is one of these things where yes, one will be relatively better than the other, and I don't know, and I'm not smart enough to predict which one's which, but like it'll all work.
31:40
>> Speaking of compute, I would love to hear the Cerebras story.
31:42
I I haven't heard you tell the full version of this.
31:45
I think it's the reason I'm asking about it is I'm deeply interested in compute, I have big investments in compute, and I'm fascinated by it.
31:52
It's just like the most magical thing to watch happen.
31:55
It's mind-boggling when you get close to one of these things what humans have been able to do on these chips and in these systems.
32:01
I think you invested in 2016 or thereabouts.
32:02
Um so, I think it was your first first foray into like extremely difficult hardware type investment. >> It's so hard.
32:10
>> And and I'd love you Now, the world is full of opportunities like this, whereas back then it was like a, you know, really a one-off.
32:17
Teach me everything you've learned about hardware investing through Cerebras.
32:22
>> Mostly >> [laughter] >> Mostly, it's really hard.
32:25
It's an amazing example to me about like the naivete required.
32:28
So, you know, the the company came in in 2016, it was five founders in a deck.
32:35
Um I really did not want to go to the pitch but it was my job cuz I'm like why are we going to go to hardware investment like this crazy since we made a semi investment I think basically the team was excellent and then the kind of first slide was just like GPUs actually suck for deep learning they just happen to be a hundred times better than CPUs.
32:58
And as soon as you said it yeah you have to remember this is pre Transformer.
33:02
Like open AI is just like weird research lab at this time.
33:06
Like Nvidia was worth like 40 billion not 4 trillion.
33:10
The TPU hadn't been announced like none of that.
33:12
So this is like early early.
33:15
But the whole idea just like as soon as he said it I was like oh [ __ ] of course of course like why?
33:21
And I'm I had been spent the last 18 months trying to figure out like applications of deep learning and looking at the security thing and looking at this medical imaging thing and like all of this other stuff thinking like hey there must be something here that's going to be really transformed by this stuff.
33:32
And so anyways go through this like whole journey we end up investing which was amazing we just like we first met on Wednesday part of meeting on Monday and had a bunch of meetings in between and it was like got just built a lot of conviction that this was a great swing and I'll tell you what I understood and I really just didn't understood so little.
33:55
But basically there are three things that we know how to do to speed up deep learning in hardware still till this day.
34:01
Increase the number of cores.
34:04
Increase the communication between cores.
34:07
Bring the memory closer to the compute.
34:09
Those are the three things. That's it.
34:10
Those are the only three dimensions that we know. In hardware.
34:13
So my articulation of what they said to me and like honestly all I understood was if we let's just take all three of those things to their logical maximum.
34:22
You have a wafer scale chip at that time you would have 450,000 cores on it you'd have something like 20 gig of SRAM on the chip so you never have to go off chip to get to memory.
34:35
And because they were all in the same wafer, the communication between cores is maximized.
34:39
So, this is the best you could do on that process.
34:41
And I think first the first chip was 7 nm or something.
34:42
And so, like you're like, "Okay, like that's it.
34:46
Like that's what we we do."
34:50
And it turns out that in software, if you have like that kind of like logical block diagram of like why it works and everything else, you're kind of 80% of the way there.
34:59
And it's a matter of like go to market execution.
35:04
In hardware, um you're like 2% of the way there.
35:08
>> [laughter] >> Like there's things like physics and and, you know, and and this entire supply chain of vendors and that like, you know, like there's of course TSMC, which everyone knows, but it's not just TSMC.
35:18
There's 30 other vendors that like matter in like putting all this stuff together and everything else.
35:24
And so, like I didn't know any of that.
35:25
And like really didn't know any of this.
35:27
So, you know, fast forward from 2016 to like 2019, I think they they got their first parts back.
35:33
And and it's like you go through this like bring up, and then it's like bring up, "Oh, yes, we got a part back."
35:39
And then it's like bring up.
35:41
>> [laughter] >> Yeah, and then you got to go through like we bring up, and there's like 14 steps of bring up and everything else.
35:46
And then by [snorts] like 2020, we had our like first thing that like works.
35:52
And then you're like then it's just this march of like actually getting it to work.
35:57
And, you know, one of the lessons that I've learned on this stuff is basically you go through all these sims and everything else in hardware and and semis in particular, and that that basically is your roofline.
36:10
Like that's the maximal performance.
36:13
Best it's ever going to be is like what that is.
36:16
And then every bit of like software and reality and compilers and kernels takes away from that roofline.
36:21
You might start at 10% of the roofline.
36:23
Once you bring it up, and then you're like good to good to good to you know these guys are grinding for for months and years to like get closer and closer and closer to the roofline.
36:35
It's really different and it's really hard.
36:36
I'm I'm actually like astonishingly bullish if I kind of rewind part of the reason we made the investment was if you looked at the four prior generations of compute that like in my lifetime.
36:49
So you had CPUs, you had graphics, networking, mobile.
36:57
There there's a new workload each time.
37:00
So you had multi-purpose compute and led to the CPU.
37:04
You had massive parallelism which you know led to the graphics processor.
37:07
Graphics processor offered massive parallelism which led to graphics.
37:09
Then with networking you needed really low latency chips and so they had low latency chips.
37:13
And then with with mobile you needed really power efficient chips.
37:17
And so each case we ended up with a new hundred billion dollar company.
37:22
And the question the first question, you know, going back to 2016 was like is AI that big a new workload?
37:28
Because there'd been many many other attempts for specialized chips for other things that like really just didn't end up mattering.
37:36
Like there were there were some fine outcomes but they just didn't really end up mattering etc. Right, exactly.
37:40
And so it's just like well, okay, well you need something that's like a really really big workload. Okay, so that's one.
37:45
So and we we had a lot of conviction on that.
37:47
And then two was the nature of the workload did it introduce a new constraint or problem in it.
37:55
What I learned was like basically the AI workload benefited from the parallelism of GPUs massively.
38:03
But GPUs didn't solve the core to core communication, basically the layers of the network problem.
38:07
And so like you're like, okay, wait a minute.
38:09
Like there is a new constraint which is communications communication bound problem.
38:15
And so then you're like, okay, is AI a new giant workload that is going to have specialized chips?
38:20
You know, basically everything that I just said was the entire everything I knew at that time.
38:26
And I I do actually think so.
38:26
That that's obviously played out.
38:28
And you know, like in each prior generation, we obviously we got Intel, we got Nvidia, we got Broadcom Avago, we got Qualcomm and ARM like, you know, in each of these generations.
38:40
And like you're there will be like these giant winners, standalone winners.
38:44
Obviously the TPU itself is is a winner, Trainium is a winner.
38:48
Um you know, you've had Groq and Cerebras, you know, Etch.
38:50
Like it'll keep getting fought out.
38:52
Um and you know, but I think that will end up being big.
38:56
And actually I think there's a new sixth one that's coming um generation, which is and I'm really excited we've made an investment that's unannounced in this, but I think that for the first time in a long time, there's actually room for a new CPU approach.
39:12
The thing that's happening right now and and you you see this reflected in all the semi stocks and everything else is the LLMs, which are running on you know, accelerators and GPUs, generate code. The code runs on CPUs.
39:29
And right now it's running on classic CPUs we've had around forever.
39:35
But there's a whole bunch of constraints on CPUs that have existed and CPUs have dragged all this baggage forward that you might not need to anymore.
39:43
And so I'm actually like really excited about about that possibility. >> category. >> The next category.
39:50
>> Does the experience with Cerebras make you want to do a lot more investing in companies >> [ __ ] no.
39:56
>> [laughter] >> But why not?
39:56
Like like >> The defining >> 2019, we're sitting in a board meeting and this thing is melting.
40:02
It's like [ __ ] melting, okay?
40:04
We've raised $500 million or something.
40:09
And it's like wait, what? Like what?
40:11
It's melting or it's burning or something and like we're looking at and I'm like holy [ __ ] Like you know, and I realize like $500 million isn't that much in today's era, but like I was just like, "We're going to lose all this money.
40:23
Like, this is not going to work."
40:24
And you know, and and I'm I didn't talk to Andrew about this, obviously, just what that team did. Insane.
40:31
Like, in insane in terms of the technical They're built differently.
40:35
They They're They're built so differently, and I have so much respect and thanks to them um for what they've done.
40:42
Look, I I I you know, joking aside, I think that there are efforts that make you really proud to be a venture capitalist and like to because you're funding something that like makes a difference and matters.
40:59
And for me, I'm an investor because like that's a means to work with companies, not because I like fundamentally like love investing or something like that.
41:08
I like the working with the companies.
41:11
That's my That's my favorite part of it.
41:15
And working with teams like that and companies like that is so special on these like giant ambitious efforts.
41:23
And um you know, and I said this before, like well before 2018 or whatever it was, like whether this Rubis worked or didn't, I think it was an effort that was worth venture capital.
41:34
Like, it Now, that's the kind of thing you should do.
41:38
You should try to build something that people have tried for 50 years and have been unable to, but now we think we could do and there's a reason and application for it and everything else. I love that.
41:47
And And so, I I do like those kinds of things, joking aside, and you know, we do have a robotics company.
41:51
We actually have a defense company that that I'm also really excited about.
41:54
And And then this like, you know, CPU project that we're talking about.
41:58
And so, like I I think these kinds of things are actually really fun um and interesting good use of venture capital.
42:05
But, they they definitely aren't easy.
42:07
>> the the sort of productive naivete that you described where it's probably a virtue that you didn't know more than you knew, otherwise you wouldn't have done it. >> Yeah. Yeah.
42:15
>> This is really my experience with etch that like when we first called around asking like if you asked experts, everyone says like in any of these fields, you ask experts they're going to tell you don't do it. Yeah.
42:22
Like it sucks, it's too hard, base rate is too low.
42:25
Young people can't do it or they're 47 reasons.
42:27
Um is there anywhere where that is it's just a bridge too far that could be like I'm thinking like bio or something like this where you just are unwilling to invest if you're naive?
42:38
>> Actually like most of the time all of these like stereotypical statements are correct.
42:44
Like they're not like correct like three times out of four, they're correct like 19 times out of 20.
42:48
Maybe 99 times out of 100.
42:51
The thing that Bruce, one of our founders, always said is just like what could go right?
42:57
That we have to ask ourselves is like what could go right and do we see that path? And it's like yeah.
43:02
You know, young people can't build chips or you shouldn't do a no show me company or you shouldn't do this or like whatever.
43:06
Like all of that stuff is actually totally right. Except when it isn't.
43:11
And so apparently there's a saying that someone someone said to one of my partners which was if it doesn't work it'll be for all of the reasons that your partner said.
43:24
If it does work, it will be because those reasons [music] didn't matter.
43:27
>> Vanta automates security and compliance for over 16,000 fast-moving companies like Ramp, Cursor, and Harvey keeping them audit ready around the clock.
43:34
It's the number one [music] agentic trust platform and it now helps companies like yours watch for the risks that show up between audits across [music] your vendors, your AI tools, and your whole environment.
43:45
Every new tool your team signs up for, [music] every vendor that turns on AI features is an opportunity for something to go wrong and most security programs weren't built for AI's pace of growth.
43:54
The Vanta agent works like a 24/7 GRC engineer in the background finding issues, drafting fixes for you, and cutting vendor assessment time by up to 50%.
44:05
Whether you're a fast-growing startup or a global enterprise, [music] Vanta helps you earn and prove trust.
44:10
Invest Like the Best listeners get a special offer for $1,000 off at vanta. com/invest.
44:17
Ridgeline is the first end-to-end system of [music] record with embedded AI for investment management firms, running portfolio accounting, reconciliation, reporting, trading, and compliance on one unified platform.
44:29
Firms are moving off legacy technology and onto Ridgeline because of how far ahead Ridgeline's AI features are compared to anything else in [music] investment management software.
44:36
I've been hearing from a lot of investment managers about AI, and they fall [music] roughly into two camps, with some unsure of where to even start and others convinced they can build their own order management system over a weekend.
44:46
The reality is that running an investment firm will always require governance controls [music] and a single source of truth for your data, and no amount of AI enthusiasm changes that requirement.
44:57
Ridgeline is built on exactly that foundation, which is why I believe that firms that come out ahead in the AI era will be the ones running on Ridgeline's unified platform.
45:05
If you're serious about your firm's AI strategy, [music] Ridgeline should be part of that conversation.
45:09
You can request a demo at ridgeline. ai.
45:13
>> You and I think are both interested in >> [music] >> robotics. >> Yes.
45:17
>> Uh it's not controversial that if robotics works, it's going to it might dwarf what we're currently living through.
45:25
What do you think has to be true for it to work?
45:27
Like obviously it's exciting.
45:27
I want a robot in my house folding my laundry.
45:30
It sounds great, but it it's kind of one of these classics, it's always 10 years away and it's been that way for a long time.
45:35
What do you see happening?
45:36
What has What has to happen for this to actually be a thing in the near to medium term?
45:41
>> We've had classic robotics forever and they're all over the place and they're on assembly lines and manufacturing lines and all these things.
45:47
It's where you're doing repetitive tasks in controlled environments.
45:49
Repetitive task in controlled environments is like more or less solved and like that you know, that'll continue to happen.
45:57
But having tweaked tasks in real-world environments.
46:03
That's where you need the AI.
46:05
Like that's where you need the AI plus the plus the robots.
46:09
And the trick with it all is you need a model that can do that.
46:14
And of course, the problem the first problem is there is no internet scale data to bootstrap the whole thing, right?
46:22
LLMs are all bootstrapped on the internet.
46:24
Um which is a ton of human knowledge.
46:27
And the equivalent for that for robots doesn't exist.
46:31
Um and people are trying different things with videos and simulations and um and teleoperation.
46:37
So like there's a lot of different ways, but if you kind of think about take teleop as an example, like how much you need to teleop robots to get to an internet scale data.
46:51
This first step is like getting a good set of data to bootstrap the model.
46:59
And one of the I think insights that you can have is with the internet, there's like a lot of slop data.
47:04
Like even before AI generated all the stuff, there was also a bunch of junk data.
47:09
And some data was more valuable than others.
47:11
Like you may value, okay, certain things on Reddit more valuable than other things.
47:14
You might value Wikipedia more than like other forums.
47:17
You might value GitHub more than other things.
47:19
And and all of the model companies did that, right?
47:22
They prioritized data um that was more valuable and less valuable and and ran that through the model.
47:27
So, I think one of the most interesting things that these AI robotics companies are doing are saying, "Okay, well, let's just go after the high value data to start."
47:36
And if we go after the high value data, then we can kind of bootstrap the the this model.
47:40
Now, once you do that, then can you through the pre-training process get to a place where you can have very small auxiliary examples of data that you add in post-training and all of the sudden it works for that.
47:58
That's the magic we have with LLMs is you have this giant pre-trained base and then you add a little magic on it um in RL and post training and and you teach it.
48:10
You teach it a new thing that it wasn't in the pre-training set and and you kind of go from there.
48:15
And I think the exact same thing is happening in robotics.
48:16
We're investors in in Sunday Robotics which is going after this exact pipeline.
48:22
It's a cool company and and they you know they're doing household robots but the the key part before you get to household and all of that stuff matters much less actually than like can you get this like training pipeline to work and how do you do that?
48:36
And one of the lessons I learned and I look back I try to learn from history cuz you know it doesn't repeat but it rhymes and and if I look at autonomous vehicles as an example which is they're robots basically they're they're AI plus robots you look at Waymo and you look at Tesla and they were both designed vertically integrated in their own way.
48:59
You have a Tesla you have a fleet of Teslas that everybody owned who was collecting data on the Teslas and it was used to train the models to drive the Teslas.
49:08
And same thing with Waymo.
49:10
And I think part of the lessons is they gather very very high quality data they did their pre-training and and then they worked off of that and that was kind of that was a simplification to allow you to get to complete product or complete solution which of course is going to continue to improve and ultimately will generalize I'm sure it'll generalize in some some ways so you'll be able to strap it on any car and everything else.
49:34
So I think the same thing exact thing is happening in robotics where um you have companies like Sunday and others who are using techniques where they're vertically integrating the robot um and the model the data collection around the robot, Sunday uses gloves that are designed just with the robot hands, so they're they're they're perfect.
49:59
So, you get very good data transferability from one to another.
50:03
You do this pre-training and you have a great pre-training data set, you have a good model, and then you start adding these examples in URL and posturing on top, and you get cool emergent behavior.
50:14
>> Do you have a most visceral moment?
50:16
>> When we invested in Sunday the first time, um we saw we saw them, my partner Peter arranged a demo, we went down into the basement at the Stanford lab, and they had like this totally janky cardboard glove thing.
50:32
And you know, you see a few evolutions of it, and um the last time we we saw demo, we like went down in the basement um [snorts] of their now office building, and and you see and there was like a dozen robots just folding arbitrary laundry.
50:46
It wasn't It wasn't a demo, it was just trial and error, and then they have people like taking taking the clothes and then measuring them to make sure that they were folded properly, and like and they were like creating a rigorous baseline and evaluation criteria.
51:02
And I was like, "Oh my god, this is like this is happening."
51:05
>> Do you ever worry about uh how to pick the right customer for these companies?
51:09
Like every one of these things folds laundry, which like I don't think anyone likes folding laundry.
51:13
It like seems like a good a good use case, but it feels like we don't actually understand the demand for what these things will be used to do.
51:22
You ever worry about that they were sort of building solutions that will then be in searches of problems?
51:27
[snorts] >> I don't, and I'll explain why, and it's it's it's like certainly informed by watching the LLM evolution, which is like you know, I think some of the people who were involved in the early LLM at at OpenAI like really understood that like code was going to be important, and and obviously, you know, the topic team had this perspective that you could get to RSI if you got code generation going and automating AI research and whatnot.
51:55
But if you think about it, like the first use cases were very much language oriented.
52:00
They were very much like essay writing and editing and and um and like marketing like I think the first application that really took off was was Jasper which was just like writing marketing copy and you know, I think it's going to evolve a lot basically.
52:14
I think like laundry is kind of a good task because it's arbitrary. It's complex.
52:20
It requires dexterous manipulation. It's done on time. It isn't time sensitive.
52:24
So you can like if it takes three times longer, so be it. Who cares? It doesn't matter. Just let it run all day.
52:29
So I think it has like some of those properties, but I don't I don't think the task is actually that important.
52:35
Um what I think is much more important is are you pre-training an amazing model and then being able to post-train on top of it and get that flywheel going.
52:44
If you get that flywheel going, then you know, the task capability will just keep it'll keep multiplying.
52:52
>> Maybe this question will be annoying or slightly uncomfortable for you, but if you ask basically every founder and critically other investors of your type, almost everyone if I ask like who's the best board partner, will say you.
53:04
Like it you come up way more often than than anyone else that I've come across.
53:11
And I'm curious why you think that is.
53:13
Like what what it is that you're doing that other people the incentives are there like to to do a great job as a board partner.
53:20
What do you think that you're doing on the boards of these companies, partner with the founders, that's actually different from other really talented investors who are also nominally doing the same job, but don't come up nearly as often it when when asked that question.
53:34
>> One of the things that I've realized is we each are attracted to like different types of entrepreneurs and like where we have chemistry.
53:43
And I think one of the differences when I say I'm like I'm really like I'm an investor second and I try to be a partner first.
53:50
And and here's what I mean.
53:52
We'll We'll see companies come in.
53:52
We We have one coming yesterday and it's an what I would call an investment grade opportunity.
53:58
It's kind of like you can invest, it probably works, you make money, it like it's it's good.
54:02
And an investor would do that. And a partner wouldn't.
54:09
Because that's not sufficient for a partner.
54:14
Because like if unless you have real chemistry with that person where you feel like you're going to be able to work together really effectively and and I'm going to learn a ton from them and they're going to learn something from me.
54:27
And and together we're going to just like feed each other's loops, right?
54:31
Unless you feel that way, then like you can't be a partner.
54:36
And and so you pass on that.
54:36
But like that's that's a really important like fit element to me.
54:39
And so it kind of starts with this like mutual selection actually weirdly of like they want to partner with us and and and I want to partner with them and like I'm really like looking forward to working with them together.
54:52
And and I'll give you a really good example of where this like kind of comes into play for me.
54:55
If I take like Sagi and Benchling, Benchling is, you know, life sciences SaaS company is absolutely crushed, done really really well.
55:08
And then of course, you know, you had this like biotech crash and the company it became hard.
55:12
Like it became grindy and I and I did the company had never had any churn for the longest time.
55:21
Such that even on their like reports, you know, for every SaaS company you have this like, okay, gross ARR added, churn line, net ARR added, like everybody does the same thing.
55:29
They never had a churn line.
55:31
Never reported it for the first of like 6 years that I worked with the company.
55:37
So then they got 7 years of churn in like 12 months.
55:40
Turns out like life sucks when you get 7 years of churn in 12 months.
55:45
And like through that grind and through it all and related to the whole like wait a minute, the goalpost moved, we have to do something different, everything else.
55:57
They kept like thinking about how to apply AI for these biotech and pharma customers, which they're very close to.
56:03
How can we make it better for them?
56:03
How can we apply these models in their world in a way that they're excited about and and continue to iterate.
56:11
And like and it was grindy. And I was there for it.
56:15
And I I just like and that's one of these cases where it's like you you you know, you're excited to work with that person cuz like one, of course you think it's a really special opportunity and there's a way out, there's a path and we can find it.
56:28
But two, because of the joy of the game and like the relationship and like everything else, like that's part of it.
56:36
And I don't I don't know what it is more than that.
56:38
I it's it's very different.
56:40
I've seen different models and and lots of different models of venture capital work, right?
56:44
Like Moritz was a writer, Door was a sales guy, like, you know, Gurley was an engineer, like Peter's a career venture capitalist, like they're all different, right?
56:52
Um and but for me I think like that working with them, it's like when I call these people, I learn something and they push back on me and then I ask them questions.
57:04
And what I've realized is like so much of my job is they know the answer.
57:11
They know what they want to do, they know the answer and it's maybe asking questions of them to maybe help solidify their conviction or solidify their articulation of what they want to do and why.
57:25
And and and through that process, hopefully we get like, you know, 1% better a few times a year, we make a 1% better decision, 2% better decision a few times a year.
57:38
And if you do that over a decade, that compounds to real results.
57:41
One of the questions that I ask myself before making an investment is there are all these people that I care about through my life, like you care in yours.
57:53
Could I talk one of them into going to this company and honestly, intellectually honestly to myself, explain to them why this could be their life's work.
58:07
And if I can't do that, I should not invest.
58:08
Because it's just not like it just means that the project is not for me.
58:14
It It just doesn't line up in that in that way.
58:16
And so, as long as we have one of those things, it doesn't matter that much what it is to me.
58:21
It's just It's important.
58:24
It could make a big dent.
58:26
And if it can make a big dent and it's a special person, I'd love to work on >> Are there any other questions like that that you ask yourself before investing?
58:33
>> That's a particularly good one. >> Um yes.
58:35
Um the other question, if this person calls me at 9:00 p. m.
58:38
on a Saturday night, will I pick up the phone?
58:41
>> I call that the green button test.
58:42
>> Yeah, yeah, yeah, right.
58:42
Like it's just like it's like it's one of these things which is like if you don't, then then >> You just know if it just doesn't >> You just know.
58:47
Like for whatever reason, it's just It's a chemistry thing.
58:49
And they have to feel the same way, obviously.
58:52
And then, you know, one of the other ones is like if if it's right, doesn't matter.
59:00
Which is different than this like first one, but it's like you know, there's so many things that we could be right on as a business.
59:05
I I I looked at one last week and I told the entrepreneur, I was like, I I don't I really think that you can run build an amazing company here and you just shouldn't raise venture capital.
59:16
But there's like so many things that you can be right on that but they ultimately just don't matter. Like nobody cares.
59:21
That's a better way to say it.
59:23
If we're right, >> will anybody he >> Yeah.
59:26
>> And it's like if they won't care, then you're just not going to build enough equity value.
59:31
Um and I think that's a that's another useful one.
59:33
>> One of the coolest things that's happening right now is all of that that you just described has higher stakes and more leverage attached to it. >> Yes.
59:39
>> Which is manifested most simply in more dollars. >> Yes. >> Um and higher prices.
59:42
You and I have talked about this notion of um what high multiple meaning like high multiple on invested capital investing is like and what it has been like and what it's moving into. >> Yeah.
59:54
>> You did something recently which was you raised a growth fund for the first time in a long time.
59:59
And I think that is related to this concept of like these companies need more capital, the prices are higher, the outcomes are bigger.
1:00:05
Maybe we can earn the same multiple on a billion-dollar entry price that we could on a $50 entry price, you know, 10 years ago or whatever.
1:00:12
Can you talk through that evolution, talk through the partnership's like way of thinking about it and talking about it?
1:00:17
Like what you believe to be true that that results in this decision to do this?
1:00:22
>> Just go back to you like why did LPs, you know, starting with with Swensen and everyone else, like why did they start investing in venture capital?
1:00:32
And fundamentally it wasn't because they thought they could beat the Nasdaq or the index by like three percentage points a year or five or whatever stupidity.
1:00:38
Like it's because there were there situations where venture capital could drive these insane multiples on invested capital.
1:00:46
Like that's from a financial perspective, that's what they're seeking.
1:00:52
And for the longest time for most of the industry's history two things were synonymous.
1:00:59
Early-stage investing and high cash-on-cash multiples.
1:01:02
Like the way to get high cash-on-cash multiples was to do early-stage. And like that's it.
1:01:08
Like those two circles in the Venn diagram like almost perfectly overlap.
1:01:14
And I think the thing that's changed is recently, relatively recently in the last few years um because outcomes have gotten so much bigger and these markets are bigger and everything else, the circle of high cash on cash multiple opportunities is bigger than just early stage.
1:01:32
And it's not so so big that like there's like a gazillion new companies in there that you can generate 100 X's on. That's not true.
1:01:42
But, um, there are there are certainly many outside of early stage where you can generate really high high returns.
1:01:52
And so that's what That's it.
1:01:52
Like that's what we want to go after.
1:01:54
You could argue we're a few years late.
1:01:55
I And I I think I'd take that criticism.
1:01:57
I think that, um, but I think that opportunity exists for uh, on a go-forward basis.
1:02:04
Um, and so like we should go do it.
1:02:10
And you know, everything else that that we represent, which is the high conviction, high commitment, you know, partnership, everything else like that, that has to still be there.
1:02:21
>> As part of that discussion, what were what were like the other sides of the debate such as like maybe we would have said the same thing in '99 and 2020.
1:02:26
As markets get exciting, >> Yeah.
1:02:32
>> it's seem like the possibilities we all ex- do this extrapolation error. >> Yeah.
1:02:36
>> Um, what were the counter arguments to like let's despite all that still not do it?
1:02:40
>> mean, there's all the counter arguments that you would expect.
1:02:41
I I think the biggest counter argument that really made this time right versus 2 years ago or whatever, um, was you need a team that can do it.
1:02:51
Like it's just a different mentality.
1:02:52
Like there there are differences in how you evaluate and think about things.
1:02:54
All all the other stuff are there's there's, you know, why not change and stay within your circle of competency and all those things are true.
1:03:02
Um, but but to me that that was like the biggest one.
1:03:05
>> [clears throat] >> And we had a several examples over the last couple years where we had, I think, the right intuition on a company or an opportunity and we didn't do it >> Cuz it was outside the box.
1:03:21
>> Because it was outside the box.
1:03:23
And that's that's obviously dumb.
1:03:23
And I think it is quite different than a lot of um you know, the industry does.
1:03:30
We're we're we are really chasing these like very rare special companies that have like very high cash on cash opportunities um where we think there are just can be runaway successes and and we can invest in them.
1:03:43
>> Is there any lesson to be pulled from the many let's call it 20 to 100 X's that you personally have observed?
1:03:48
I mean, it's like such a crazy amount of return that like obviously it doesn't pencil in the beginning.
1:03:54
Like you can't make something pencil if it was that clear, like the price would be different. >> can't.
1:03:59
>> Um what is what have the 20 to 100 X's taught you in aggregate if anything?
1:04:05
>> Work with really special people.
1:04:05
You know, it's kind of like you want to work really hard. You want to work smart. And and get lucky.
1:04:13
Like it really all has to come together.
1:04:17
And there's a lot of things that are like timing dependent, you know, that you have no control over as a company.
1:04:26
Like you take the Cerebras example is a good one, which is like we took it public this time in May, but like we tried to take it public in 2024.
1:04:33
And it would have been taken public at a much much lower valuation and and like, you know, it would have been um and rough and it didn't it didn't work out because of Civis and all this stuff. So, timing matters.
1:04:42
Like and and you know, just the advancement of that 18 months like made all the difference in the world for a bunch of things that were honestly outside of our control.
1:04:51
There were some things that were in our control like getting inference running and everything else, but like there was a lot of stuff that was outside of our control.
1:04:56
These are all the classic things which is like you got to focus on what you can control.
1:04:58
That's one of the things that's really different in software companies.
1:05:01
With software companies you know, aside from like building on AWS or whatever, you pretty much own your whole stack.
1:05:07
And so, you're really fully in control of your your destiny in that way.
1:05:12
With hardware companies, you don't.
1:05:14
Like, there's an entire supply chain and all this other stuff.
1:05:17
HBM's a thing and DRAM's a thing and TSMC's a thing and and a lot of those cross geopolitical orders.
1:05:23
And so, geopolitics gets involved and and so, then that makes that complicated.
1:05:26
And so, that's a a really big difference.
1:05:29
You got to get lucky on the timing and macro and and and other stuff.
1:05:33
But, I think it really starts with like working with these like crazy people with unbounded opportunities.
1:05:39
And if you work with these crazy special people on unbounded opportunities, you know, then then you you get you get lucky uh from time to time. You're bound to.
1:05:47
I I This When I was 20 years old, I was working at an investment bank.
1:05:53
Ben Horowitz, Marc Andreessen um had started Loudcloud.
1:05:54
It was still in stealth and, you know, Ben gave me an offer to be his assistant.
1:05:58
Um And I was talking to this associate who like seemed like this like elder at times. He's probably 25.
1:06:06
Uh but like, you know, I was talking to this associate at the time.
1:06:09
And, you know, he he said this thing to me which I which which really stuck with me.
1:06:14
He was like, "Do you golf?"
1:06:15
And I was like, "Classic banking question. Do you golf?" No, I don't [ __ ] golf.
1:06:18
But, he's like he's like, [clears throat] "You know, with golf, you you keep on practicing and you like keep getting the ball on a three par like close close to the pin close to the pin close to the pin close to the pin."
1:06:32
And he's like, "You keep getting the ball close to the pin and you keep practicing."
1:06:35
And he's like, "That's hard work.
1:06:37
That's like working smart.
1:06:37
That's like That's what you want to keep doing."
1:06:41
He's like, "Getting the hole-in-one, that's luck."
1:06:43
I kind of love that framing and I I use it with my kids actually cuz what it says is like, "Yeah, there's luck involved."
1:06:50
And there really is luck involved.
1:06:52
But, there is actually a way to increase your luck.
1:06:55
And the way to increase your luck is get a lot of balls close to the pin.
1:06:56
And then like, yeah, eventually one will drop.
1:06:59
And and so, I like that that mental model.
1:07:03
So, I kind of go back to this for for companies which is in I think in you know, each of us invests in one or two companies a year.
1:07:11
I think in my 12 years I've invested in 18 companies total. >> Crazy.
1:07:15
>> Which is a relatively small number.
1:07:15
Um, so it's a very high conviction and very high commitment.
1:07:19
It's I've a lot of skin in the game.
1:07:21
I believe in these entrepreneurs.
1:07:22
I believe in these companies.
1:07:23
But if you keep on working with these like very special people and these opportunities magic can happen.
1:07:34
>> What have you learned about the best reasons and conditions for going public?
1:07:39
>> Benchmark does these Monday night dinners.
1:07:40
You've You've been once there like and so we had a a CEO last night of a uh, multi-hundred-billion-dollar private company and we had this whole conversation.
1:07:50
So it's kind of it's it's a little like fresh.
1:07:52
I think that ultimately when you go public you have a range of new opportunities in what you can do. You have public trust.
1:08:06
Actually, weirdly, cuz there's some transparency that comes with being public being a public company.
1:08:09
You obviously have a currency that you can do things with um, that ends up being there.
1:08:14
You have an unbelievable ability to raise capital, which I think is why the labs will ultimately go.
1:08:17
Although I think the trust thing is actually a really important element of why they should go and it's beneficial to the world and to America if they do go public is like It's like okay.
1:08:26
Yeah, see what's going on. Everyone can see it.
1:08:28
I think that's like a really beneficial setup.
1:08:31
There's another element of it, which is What does a collegiate athlete want to do? >> Go pro. >> Mhm.
1:08:38
>> They want to play at a higher level. And is it harder? Yeah, it's harder.
1:08:40
Is the competition tougher?
1:08:42
Yeah, the competition's tougher. They move faster. They're tougher. They're bigger. They're stronger. The stakes are bigger. The stage is bigger. The scrutiny is bigger. All of that's true.
1:08:54
It's kind of the same thing with companies.
1:08:55
There are a handful, and it really is a handful, three, four, whatever, that can get to this like tremendous scale without going public because things have gone through their execution and excellence.
1:09:09
Dave, lots of free cash flow and they've done really well over a really long time and I think that's fantastic.
1:09:14
And you know, good for them, but you know, in general for everyone else, like get out there.
1:09:20
And then the other thing that I would tell you is there are windows for a particular type of company.
1:09:30
So, like yes, the SAS companies that went public in 2021 a whole boat of them have struggled and it's been tough in the public markets cuz their stocks ripped to this third this multiple compression issue.
1:09:40
They were trading at 30 times, they four accidents size, but now they're trading at six times.
1:09:45
It turns out you're under still, right?
1:09:47
Like that's that's a tough place to be.
1:09:49
Um Yeah, I will also tell you that there's 500 something probably SAS companies that are between 100 million and 500 million and that are private.
1:10:02
What what happens to them?
1:10:03
Those those employees never got a chance to sell.
1:10:05
Those employees don't have annual tenders.
1:10:07
Those employees don't have an opportunity to exit.
1:10:11
Like those investors don't have an opportunity to exit. They're stuck.
1:10:15
And you know, I don't think they're all going away and it as I said, I don't think they're all getting vibe coded and everything else, but ultimately, you know, the AI natives with their growth rate sucked all the oxygen out of the room and all the interest and the window was missed. And that's tough.
1:10:33
>> What are the biggest debates right now inside of the partnership?
1:10:35
I always love coming here and talking to you guys when there's something interesting going on because you know, you you debate. It's healthy.
1:10:41
It's going to be can be really fun to watch and I learn a lot from it.
1:10:46
What are those debates today?
1:10:48
>> There's a ton of debate around this AI infra apps, you know, full foundational models infra apps like ecosystem, where does value accrue and how does it accrue and like why where the motes and and how do we think about that?
1:11:02
Um but also like the business model innovation.
1:11:04
I think one of the things that's people don't understand about like why SaaS did so well versus traditional software was it wasn't just that it was a better delivery model and everything else.
1:11:17
There was actual business model innovation on it, right?
1:11:19
Like you you really did have this like subscription element that ended up being fantastic.
1:11:24
Um for both the company and the customers.
1:11:26
Like it was a win-win situation.
1:11:28
And so I think there's like that same thing actually exists in AI and you know, selling by outcome um and and that piece of it.
1:11:37
Um but then you know, wrapped up into that debate um and discussion is like how much value just accrues to the labs?
1:11:46
How much of the value just accrues to the semis?
1:11:51
I mean, that's that's a real discussion.
1:11:52
Like I am of the view that it all works.
1:11:57
As [laughter] I said, and I I really like it's a very weird thing.
1:12:00
It's like will the CSPs do well? >> Yes. >> Yes.
1:12:06
Will some of these neo clouds do well? Yes.
1:12:07
Will the fireworkses of the world do well? Yes. Will Nvidia do well? Yes.
1:12:11
Will these like chip startups do well or you know, some set of them? Yes.
1:12:19
Like are we going to have edge inference on our phones? Yes.
1:12:21
Are we going to have like near edge inference on pops? Yes.
1:12:27
Are we going to have big models in data centers? Yes.
1:12:28
There's so much zero-sum thinking, which is just like, okay, how do we cut up this pie and they're going to eat this much and like, oh no no no no, Anthropic or whomever is going to eat 98% of the value and they're going to do all the drug discovery and I was like, come on. >> Yeah. >> Like no.
1:12:42
That's not what's going to happen.
1:12:43
Like and and we have patterns like in the not that distant past where it's like this exactly manifested, right?
1:12:49
When I say like I think everything's going to work and I listed off all these other things, I it's really important to understand that that doesn't mean that every company that's doing every one of those things is going to work.
1:12:59
It actually means quite the opposite of that.
1:13:01
Like most companies in each of those areas are not going to work.
1:13:04
And it's actually more important than ever to have real differentiation and to to like really like take each of these thoughts to to their logical extreme and understand like, "Wait a minute.
1:13:15
You got to go all the way on these things and really be differentiated >> on anything else that you have your eye on, whether it's in the funding market, in the technology world, like anything at all that that you really are watching carefully?"
1:13:28
>> It's actually funny to me that some of the people are so so smart.
1:13:36
And yet they're in this like tech world where they just they're like kind of reaching these like deterministic almost conclusions, which they of like mass unemployment and like all of these different things.
1:13:50
And and I'm I'm going to give you a really concrete example, which I think is just so good.
1:13:54
Take Jeff Hinton and and radiology.
1:14:00
So, I think it was 2016 where he was like, "We should stop training radiologists.
1:14:02
Like AI's going to do it all better."
1:14:06
And like look, Jeff Hinton's like three orders of magnitude smarter than I am, but like could not have been more wrong.
1:14:16
But the actual thing that led him to make that statement and that conclusion was 100% correct.
1:14:21
Which is if you look at these radiology images, like we should be able to train AI to do a better job of like reading these things than humans.
1:14:29
And that's probably true.
1:14:33
Um and and and actually I think the studies and areas have shown that to be true.
1:14:38
And we have an investment in a company called New Lantern, which is approaching this.
1:14:41
But the big hurdle and the big thing that articulated was like, "Wait a minute.
1:14:46
First off, all of the aggregated training data set doesn't exist anywhere.
1:14:49
And so like what you see is like companies going after like chest CTs or like very specific elements, but your typical radiologist looks at a whole variety of things every single day from X-rays to CTs to MRIs of all parts of the body and everything else.
1:15:03
And so an AI climbing in specific areas like chest CTs is like very marginally helpful because it's only doing that one thing, which could be one of 20 things or or 40 scans that they read that day.
1:15:18
And so it's okay, so problem number one, you don't have the data just like we talked about in robotics and everything else to to to train the AI.
1:15:24
Problem number two, the whole healthcare industry is oriented around reimbursing doctors for making readouts.
1:15:29
And so like how is that going to work?
1:15:31
And like who's And there's liability associated with that.
1:15:35
And there's repercussions of getting something wrong or missing something and there's medical malpractice and everything else.
1:15:40
So like how are we going to avoid that?
1:15:41
And how are we going to get around that?
1:15:43
So that's That's like like problem number two, just like real world stickiness.
1:15:48
And [snorts] so I think in the end, we're going to end up with this application where AI really does help radiologists.
1:15:56
It helps radiologists get more and more higher higher throughput because the AI can do some parts and the radiologist does some parts and they're checking each other and everything else.
1:16:04
And you do kind of weirdly end up in this co-pilot situation um for for some time.
1:16:07
And then you're going to slowly have the AI read more and more of the scans and build up and build up and build up.
1:16:14
But the actual duration to get from here to there is going to take a long time and in the ensuing time, we need more radiologists, not less because oh by the way, everyone's getting more imaging than they used to get because the cost of imaging's going down in a Jevons paradox kind of way.
1:16:30
And so like my point on it is is like you have someone very very smart who really understands the capabilities, really understands what's happening, has the right data, but by not thinking about data in the real world application comes to the wrong conclusion.
1:16:47
And that's how I think of the unemployment thing.
1:16:49
I think it's just it's almost the exact same setup.
1:16:53
>> Eric, I love talking about markets and economics with you.
1:16:55
An absolute [music] blast. >> Thank you.
1:17:02
>> You know how small advantages compound over time?
1:17:03
[music] That's true in investing and just as true in how you run your company.
1:17:06
Your spending system is your capital allocation strategy.
1:17:10
[music] Ramp makes it smarter by default.
1:17:11
Better data, better decisions, better economics over time. See how at ramp. com/invest.
1:17:17
As your business grows, Vanta scales with you, automating compliance [music] and giving you a single source of truth for security and risk. Learn more at vanta. com/invest.
1:17:26
Ridgeline is redefining asset management technology as a true partner, [music] not just a software vendor.
1:17:29
They've helped firms 5x in scale, enabling faster growth, smarter operations, and a competitive edge. Visit ridgelineapps.
1:17:36
com to see what they can unlock for your [music] firm.
1:17:39
The best AI and software companies from OpenAI to Cursor to Perplexity use WorkOS to become enterprise ready overnight, not in months. [music] Visit workos.
1:17:47
com to skip the unglamorous infrastructure work and focus on your product.