Elad Gil

Discussion w Arthur Mensch, CEO of Mistral AI

2024-03-22 20:19:30

Topics covered:

• Mistral

• Open and closed source AI

• Future tech (small models, context windows, etc)

• EU AI & startup scene

• Enterprise AI needs

• Building fast moving teams

Video link:

Transcript:

DYLAN FIELD

Hi everybody. Welcome. Thank you so much for being here. I am so glad that we're able to host this at Figma. My name is Dylan Field. I'm the CEO and co founder of Figma. And a big welcome to everybody here, but also everyone who's joining us via live stream as well. And I'm really excited for tonight. I think this is going to be a pretty credible conversation and I'm proud to be able to introduce the two folks who'll be having it.

So first, Elad Gil. Elad is not only a dear friend and mentor of mine, but also to many in Silicon Valley and the startup community globally, and also Arthur Mench. Arthur is a former academic who has turned CEO and co founder of Mistral. And mistral for the one or two people in the room that do not know, is breaking incredible ground in open source models and I would dare say changing quite a lot about the future of AI.

And with that, I'll pass it off for their fireside. Welcome.

ELAD GIL

Oh, thanks. Thanks so much to Figma for hosting us, and thanks everybody for making it today. And of course to Arthur. Arthur made a heroic effort to join us where he literally had to jump out into traffic, grab a bike and bike over here. So thank you so much for coming.

ARTHUR MENSCH

Discovering the US, I guess.

ELAD GIL

So from a background perspective, you got your PhD in machine learning, you were a staff research scientist at DeepMind, and then you started Mistrall, and you started it, I believe, with both some folks from Google, such as yourself, and then some folks from meta and the llama project there. You folks have taken an open core approach, which I think is super interesting and we can talk about in a little bit. But I was just curious to start off, what was the impetus for starting? Mistral All, how did you decide to do it? What were the motivations and the initial formation of the company?

ARTHUR MENSCH

Yeah, so I think this has always been on the mind of me, Guillome and Timothe. So I was at DeepMind, they were at meta, and I guess we were waiting for the hour, and the hour came with GPT to some extent, so that we realized we had an opportunity to create a company pretty quickly with a good team that we could hire from day one and go and try to do speedrun a bit because we weren't starting the first. So that's how we got started.

ELAD GIL

I guess the people who are probably watching the live stream, I think the people in the audience are probably well versed with what Mistral does. Can you explain a little bit about the set of products you have the platform, all the various components now.

ARTHUR MENSCH

Yeah, for sure. So Mistral is actually a company building foundational models. We are the leading in open source models. So we have started the company by creating text to text generation models, which are really the foundational block for creating today's generative VA applications. I know we're at Figma, so we're not focusing on images yet, but this is obviously coming at some point. And so, yeah, the differentiation we have is that we took this open core approach to release Mistral 7B mixed hole 87 B in December and build a platform on top of these open source models with the addition of commercial models that we introduced in December and then in February. So we're building an open source models and we're building a portable platform for enterprises with focusing on the developers and building tools for developers.

ELAD GIL

How long did it take from when you founded the company to when you.

ARTHUR MENSCH

Launched 7B took four months, approximately.

ELAD GIL

Yeah. That's amazing. So I think one of the things that's really noticeable is the immense speed in terms of how rapidly Mistral actually launched its very first product and then the rapid adoption of that right as 7B came out suddenly, I think people realized that you could have these small performant models that were very fast. Inference time and time to first token were very cheap, which made a big difference if you were doing things at high throughput. How did you build something so rapidly? Or how did you focus a team on such a singular goal so quickly?

ARTHUR MENSCH

Well, I guess we thought about what was missing in the field and we realized that small models were actually quite compelling for people. We saw a community building on top of llama at the time, on top of llama 7B. But llama 7B wasn't good enough. And so we realized that we could make it much better. We could make a 7B model much better. And so that's the sweet spot we targeted for our introduction to the world. And basically we had to build the entire stack from scratch. So getting the data, building the training code, getting the compute, which was a bit of a challenge because in these four months we were ramping up. So we started at zero GPUs and we actually trained on like 500 GPUs, 7B, I guess we went fast because the team was very motivated and so not a lot of holidays during these four months. And generally speaking, AI teams that succeed and go are typically like four to five people. And AI teams that invent things have always been this size. So we are trying to have an organization where we have squads of five people working on data, working on pretraining, and so far this has worked out quite well.

ELAD GIL

Is there anything you can share in terms of what's coming next in your roadmap?

ARTHUR MENSCH

Yeah, so we have new open source models, both generalist and focused on specific verticals. So this is coming soon. We are introducing some new fine tuning features to the platform and we have introduced a chat based assistant called the Shah that is currently just using the model. So it's pretty raw. It's a bit like chat GBT V zero, and we're actively building on building data connectors and ways to enrich it to make it a compelling solution for enterprises.

ELAD GIL

What kind of verticals do you plan to focus on, or can you share that yet?

ARTHUR MENSCH

Well, I guess we started with financial services because that's where most of the maturity was. Basically we have two go to markets. So enterprises starting with financial services because they are mature enough, and the digital native. So talking to developers like building AI companies or introducing AI to formerly non AI companies, and so that's the two, I guess, go to market pools that we're talking to. The first one through some partnerships with cloud because as it turns out, they're a bit controlling the market in that respect. And then through our platform, we're talking directly to developers.

ELAD GIL

I guess on the cloud side, one of the relationships you recently announced was with Microsoft and Azure. Is there anything you can say there about that relationship or that access that it's providing you to the enterprise?

ARTHUR MENSCH

Yes, this opened up new customers. A lot of enterprises can't really use third party SaaS providers easily because you need to go through procurement, risk assessment, et cetera. But if you go as a third party provider through the cloud, you actually get an accelerator. And so when we believed Mistral Large on Azure, we got like 1000 customers pretty right away. The truth is you need to adapt to the fact that enterprises are using cloud and they don't want to introduce new platforms easily. And so you need to go through that, at least at the beginning.

ELAD GIL

And then one of the things that a lot of the industries focus on right now is scaling up models and ever larger, ever more performant versions. How do you think about the scale that you all are shooting for in the next six months or year? Or is the plan to have very large models over time? Or how do you think about the mix of things that you want to offer?

ARTHUR MENSCH

Yeah, so we first focused on efficiency to be able to train models more efficiently than what was currently done. And then once we had achieved this efficiency, we started to scale so that's why we did another fundraising, and that's why we started to increase the amount of compute we had. And so we can expect new models that will be more powerful because we are pouring more computes in it, models that might be a bit larger, because when you grow the compute, you need to increase the capacity of models. But something that remains very important for us is to be super efficient at inference and to have models that are very compressed. And so that's the kind of model that will continue shipping, especially to the open source world.

ELAD GIL

One of the things that was pointed out to me that I'd love to get your views on is that as you reach certain capabilities within a model, you can start to accelerate the pace at which you can build the next model, because you can use, say, a GPT four level model to do rlaif, or to generate synthetic data, or to do other things that really accelerate what you're doing. Data labeling, all sorts of things, in some cases superhuman performance. How do you think about using models to bootstrap each other up? And does that actually accelerate the timeline for each subsequent release?

ARTHUR MENSCH

Yeah, I guess. Generally speaking, two years ago, RLHF was very important. Today it's actually less important because the models have become better, and they're actually sometimes good enough to self supervise themselves. And what we have noticed is that we scale as we scale. This is definitely improving. So that means that the costly part of going through human annotations is actually reducing. And so this is also lowering the barrier of entrance.

ELAD GIL

I guess another sort of adjacent area is reasoning. And a lot of people feel that as you scale up models, they'll naturally acquire reasoning. And then there's other approaches and entire companies that have recently been founded around just focusing on the reasoning aspect of some of these models. How do you think about that? Are you going to be training sub models for reasoning, or do you think it's just going to come out of scaling the existing models? Is it a mix of the two?

ARTHUR MENSCH

Well, I guess reasoning comes from. At this point, the only validated way of improving reasoning is to train models on larger data and make them bigger. There's obviously some possibilities that you have by building an auto loop, adding new function, calling, adding data as well for the model to reason about grounded aspects instead of trying to imagine stuff. So I guess we don't pretend to have like a secret recipe for reasoning, but we've made models that are pretty good at reasoning by focusing on the data. We're pretty good at using mathematics in our data. And so that's a good way of improving reasoning. There's many ways in which to improve it. Code has helped as well, and so there's no magic recipe, but just focusing on the little things makes it work.

ELAD GIL

Yeah, I guess one of the reasons I ask is, I feel like if you look at sort of the world of AI, there's a few different approaches that have been done in the past. One is sort of the transformer based models and scaling them. The other is a little bit more in the lines of like, Alphago and poker and some of the gaming related approaches where you're doing self play as a way to bootstrap new strategies or new capabilities. And those are in some sense forms of reasoning. And I know that there are certain areas where that may be very natural to do in the context of model training. Code would be an example. There's a few others where you can sort of test things against real rubric. And so I don't know if you folks are considering things like that, or if that's important or not in your mind.

ARTHUR MENSCH

So Guillaume and Timote were doing theorem proving with LLMs back in the day at meta. So that's very linked to, well, using LLM as the reasoning brick and then building an auto loop that involves sampling, that involve Monte Carlo research, all these kind of things. I think the one thing that was standing in the way for this is the fact that models have very high latency, and if you want to sample heavily, you need to make them smaller. And so it's very much tied to efficiency. So as we grow efficiency, as hardware increases in capacity as well, you become more able to explore more and to sample more. And that's a good way effectively to increase reasoning through the autoloup development.

ELAD GIL

And then I guess the other thing a lot more people are talking about or thinking about is memory and some ability to maintain a longer view of state in different ways across actions or chaining things for agents. Do you expect to go down any sort of agentic routes anytime soon? Or is the focus much more on sort of core APIs that are enabling in all sorts of ways?

ARTHUR MENSCH

So that's what we started to enable with function calling, which is a good way to handle, to start creating agent that store states. So memory, when we talk about memory, like of conversation, the way you make it happen is that you basically introduce some crude functions on your middleware part that you give to the model, and so it can actually use that to update its memory and its representation. And so function calling is the one multipurpose tool that you can use to create complex agent. It's hard to make it work, it's hard to evaluate as well. So I think this is going to be one of the biggest challenge. How do you make agent that work, evaluate them and make them work better for feedback? And this is one of the challenge that we'd like to tackle on the product side.

ELAD GIL

And then I guess the one other thing that a lot of people have been talking about recently is just context window. And for example, I know that there's some recent results around, for example, biology models, where if you increase the context window, you can end up with better protein folding and things like that. So the context and the context really matters. I think Gemini launched a million, up to a few million sort of context window, and then magic, I think, has had 5 million for a while. How important do you think that is? Do you think that displaces other things like rag or fine tuning? Are all these things going to work coincident with each other?

ARTHUR MENSCH

So it doesn't displace fine tuning because fine tuning has a very different purpose of pouring your preferences and basically demonstrating the task. On the other hand, it simplifies rag approaches because you can pour more knowledge into the context. And so what we hear from users is that it's like a drug. So once you start to use models with a large context, you don't want to go back. And so that's effectively something we want to try to improve and extend. There's a few techniques for making it happen. On the infrastructure side, it's actually quite a challenge because you need to handle very large attention matrices, but there are ways around it.

ELAD GIL

I see what you're saying. So basically, like on the RAM, on the GPU, you basically ran out of space for something as you're building a bigger and bigger context window. Or is it something else?

ARTHUR MENSCH

Yeah, there's a variety of techniques you need to rethink for sharding and communication to handle the big matrices. And then you do pay a cost because it basically becomes slower because of the quality cost.

ELAD GIL

When do you think we hit a moment where these models are better than humans at most white collar tasks? Do you think that's two years away, five years away, ten years away?

ARTHUR MENSCH

I guess it depends on the task. There's already a few tasks on which the model are actually better. And so I expect this to unfold pretty quickly, actually. So hard to say a date, but I would say in three years this is going to look very different, especially if we find a way to deploy agent and to evaluate them and make them robust and reliable.

ELAD GIL

What about displacing the CEO of Figma? No, I'm just kidding. Just kidding. Dylan, please keep us. So I guess there's a lot of different foundation models that people are starting to work on, right? There's obviously a lot of attention on the LLMs, and there have been diffusion models for image gen, although it seems like people are moving more and more towards image or transformer based approaches for image and video and other things. Are there big holes in terms of where you think there are gaps where people aren't building foundation models, but they should be?

ARTHUR MENSCH

I would say we've seen some things happening on the robotic side, but I think it's still at the very early stage on the audio. This is covered on video. This is starting to be covered, essentially, like models that can take actions and become very good at taking actions. I don't think this is very well covered. There's some progress to be made there, but, yeah, overall, I expect all of this to converge towards similar architectures and at the end of the day, like a joint training as we move forward in time.

ELAD GIL

So do you think eventually everything is a transformer based model?

ARTHUR MENSCH

Well, transformer are a very good way of representing associations between tokens or between information, so it really does not really matter, but it seems to be enough, it seems to be a sufficient representation to capture most of the thing we want to capture, and we know how to train them well so we can transfer information between what we learn from text on images, et cetera. And so that's why I think this is going to be quite hard to displace.

ELAD GIL

Do you think that'll also apply to the hard sciences? If you're trying to do, like, physics simulation, material sciences, pure math.

ARTHUR MENSCH

I don't expect just next token prediction to solve that. And so you do need to move to the outer loop, and you need to figure out also a way to make models interact with simulators, potentially, because at some point, you need the model to learn the physics, and so you need to bootstrap that with the simulator. But I'm not an expert, to be honest.

ELAD GIL

And then all these models, of course, need a lot of GPU, and people have very publicly talked about how there's a GPU crunch right now, and there's shortages of different sorts. When do you think that goes away, or do you think that goes away?

ARTHUR MENSCH

So I think that probably eases as the H and rit comes, and we'll start to see some competition on the hardware space, which is going to improve cost, I think. I expect that also as we move to foundational models that are multimodal, et cetera, we can actually train on more flops. And so I don't think we haven't hit the wall there in scaling. And so I expect this is probably going to continue on the training part and on the inference part as we move on to production and we have models running agent on the background. So really removing this bottleneck that we had at the beginning, which was the speed at which we could read information, I expect that inference capacity will spread pretty significantly.

ELAD GIL

Do you think that will be done through traditional GPU based approaches, or do you think we'll start having more and more custom asics, either for specific transformer models where you burn the weights on the silicon, or more generally for transformers in general, where you can just load a set of weights or something.

ARTHUR MENSCH

So the good thing about the fact that everybody is using transformer is that you can specialize hardware to this architecture and you can make a lot of gains there. There's a few unfortunate bottleneck on Nvidia chips, for instance, the memory bandwidth is a problem. And so by moving on to more custom chips, you can reduce significantly the cost of inference. It's not really ready yet, so we're not betting on it right now, but I really expect that this is going to improve cost pretty significantly.

ELAD GIL

So mistral really started off as a developer centric product, right? You launched to something that was very open source. Now you're starting to serve a variety of enterprises. Is there any commonality in terms of the types of use cases that people are coming with or the areas that enterprises are most quickly adopting these sorts of technologies or approaches?

ARTHUR MENSCH

Yeah. So enterprises adopts the technology for mostly three use cases. So the first one is developer productivity. And usually they kind of struggle with the off the shelf approach because it's not feed to their way of making, of developing. They also use knowledge management tools, and usually they've built their own assistant connected to their database. And the last thing is customer service. So the most mature company have made a large progress toward reducing their human engagement with customers and just making it much more efficient. And so these are really the free use cases we see with enterprises. And with AI companies, it's much more diverse because they are a bit more creative. But yeah, overall, enterprises have these free use cases. It's also the reason why we are starting to think of moving a bit on the value chain and offer things that are a bit more turnkey, because sometimes they need a little bit of help.

ELAD GIL

Yeah, that makes sense. I'm guessing many people here saw the tweet from the CEO of Klarna where he's talking about customer success and how they added a series of tools based on top of OpenAI that basically reduced the number of people they needed by 700. In terms of customer support. They launched it in a month and they had 2.3 million responses in that single month. So it seems like there's this really big wave coming that I think is almost under discussed in terms of impact of productivity, impact of jobs and things like that.

ARTHUR MENSCH

Yeah, so we saw even more diverse use cases. One of them was having a platform that engaged with temporary workers to try and find a job for them. So through texting, and so the customer in question went from 150 people activating this well, engaging directly with customers to seven, and they were actually able to scale the platform much more and to enable temporary workers to work more easily. And generally speaking, this approach of automating more of the customer service is a way to improve the customer service. And so that's, I think, what is exciting about this technology.

ELAD GIL

What do you think is missing right now, or what is preventing enterprise adoption from accelerating further?

ARTHUR MENSCH

So our bet is that they still struggle a bit to evaluate and to figure out how to verify that the model can actually be put in production. What's missing are a bunch of tools to do continuous integration, also tools to automatically improve whatever use case the LLM is used for. And so I think this is what is missing for developer adoption within enterprises. Now for user adoption within enterprises, I think we're still pretty far away from creating assistant that follows instruction, well that can be customized easily by users. And so yeah, on the user side, I think this is what is missing.

ELAD GIL

One thing that I think you've been very thoughtful about is how to approach AI regulation. And I know that you've been involved with some of the conversations in terms of EU regulation and other regulation of AI. Could you explain your viewpoint in terms of what's important to focus on today versus in the future and how to think about it more generally?

ARTHUR MENSCH

Yeah, so we had to speak up because at the time, in October, there was a big movement against open source AI. And so we had to explain that this was actually the right way to today, make the technology secure and well evaluated. And overall we've been continuously saying that we're merging very different conversations about existential risk, which is ill defined and that has little scientific evidence for this is merged with a discussion about, I guess, national security and AIA and LLMs being used to generate bioweapons. But again, this is something that is lacking evidence. And then there's a bunch of very important problems that we should be focusing on, which is how do you actually deploy models and control what they are saying? How do you handle biases? How do you set the editorial tone of a model in a way that you can evaluate and control? And I think this is the most important part. How do you build safe products that you can control well and that you can evaluate well? And this is the one thing we should be focusing on. That's what we've been saying for a couple of months because we were a bit forced to speak up.

ELAD GIL

Yeah, it seems like one of the areas that people are kind of worried about in the short term on AI is things like deepfakes or people spoofing voices or other things like that, either for financial attacks, for political purposes, et cetera. Do you all have plans to go down the voice and sort of multimodality side?

ARTHUR MENSCH

So generating things that are not text is effectively a bit more of a trap on the safety side, and that we've avoided it effectively. Imitating voices and deepfakes are very concerning. And this is not something that we pretend to be able to sort text. It's much easier because there's never this kind of problem. So you can generate text is generating text is never an enabler of very harmful behavior. Misinformation has been mentioned, but usually misinformation is bottlenecked by diffusion and not by creation. So by focusing on text, we kind of circumvent these issues, which are very real.

ELAD GIL

I think one of the things that's very striking about Mistral is, and I should say in Europe in general right now, is there's a very robust startup scene. And if I look at the two biggest pockets of AI right now in terms of startup formation, it's basically here in Silicon Valley, and then it's like the Paris London corridor, and you have eleven labs and you have Mastrall and you have all these great companies forming. What do you think is driving that?

ARTHUR MENSCH

I think there's a couple of historical reasons. In London there was, and there still is DeepMind, which was a very strong attractor of talents across the world. And in Paris in 2018, both DeepMind and Google opened offices, research offices, and it went and augmented the existing research scene that was already pretty strong, because as it turns out, France and also a couple of other countries in the European Union have very good education pipeline. And so junior machine learning engineers and junior machine learning scientists are quite good. And so that's one of the reason why today we have a pretty strong ecosystem of companies on both the foundational layer, but also on the application layer.

ELAD GIL

Yeah, the French seem a lot smarter than the British. So. No, I'm just kidding.

ARTHUR MENSCH

I'm not the one saying that.

ELAD GIL

The other thing that I think is kind of striking is you start to see a lot of different AI based companies focused on regional differences. So, for example, when you launched, you included a variety of different european languages. I know there's models being built right now for Japan, for India, for a variety of different geos. And one could argue that either you have large global platform companies that serve everywhere, except for maybe China, because China is likely to be firewalled in some ways, just like it has been for the Internet more generally. Or you could imagine a world where you have regional champions emerge. And in particular, you could almost view it like Boeing versus Airbus, where the governments of specific regions decide that they really want to fund or become customers to local players. What do you view as sort of the future world, and how does that evolve in terms of global versus regional platforms?

ARTHUR MENSCH

So we've taken a global approach to distribution. I guess there was another path that we could have taken, which was to focus on just the european market, pretending that there was any form of defensibility there. We don't think this is the case. Technology remains very fluid and so can circulate across countries. On the other hand, the technology we're building is effectively very linked to language, and language is, well, English is only one language across many. And as it turns out, LLMs are much better at English than other languages. So by also focusing more on different languages, we managed to make models that are very good at european languages in particular, versus the american models. And so there's a big market for that. And similarly, there's a big market in Asia for models that can speak asian languages. And there's a variety of scientific problems to be sorted and solved to address these markets, but those are huge, and those haven't been the focus of us companies. So it's effectively an opportunity for us as a european company to focus a bit more on the world globally.

ELAD GIL

Okay, great. I think we can open up to a few questions from the audience, and if people want to just ask, I can always just repeat it in the back there, please. Yeah, right there. If you want to speak loudly, I can repeat what you say. The question is, do you plan to release closed source versions of your model or always be open source?

ARTHUR MENSCH

So we have commercial models already. So to an extent we haven't been open sourcing everything. We are a very young company, but our purpose is to release the best open source models. And then we are basically coming up with an enterprise surrounding and some premium features that we can sell to sustain the business. And so our strategy today, and that might evolve with time, is to have both very strong open source models, but also models that are much stronger actually at that point in time as closed source APIs. The one thing that we focus on also for our commercial models is to make deployment of these models very portable and very flexible. So we have customers to whom we ship the weights and allow them to modify the model, do client side fine tuning the same way they would do it with open source models. And so in that sense, we have some coherence across the commercial family and the open source family.

[AUDIENCE QUESTION ON MAIN USES CASES]

ARTHUR MENSCH

Knowledge management, developer productivity. So coding basically.

[AUDIENCE QUESTION – PLANS TO DO CODING SPECIFIC MODELS?]

ARTHUR MENSCH

Yeah, we have plans. Not doing any announcement today, but we do have plans.

[AUDIENCE QUESTION – NEW ARCHITECTURES AND RESEARCH]

ARTHUR MENSCH

We've been mostly into production at that point because the team was pretty lean, but we're not dedicating a couple of full time employees like finding new architectures, finding well, making research. And I think this is super important to remain relevant. So as we scale, we will be able to afford more exploration. It's also very linked to the compute capacity you have. So if you want to make some discoveries and make some progress, you need to have enough compute. And we're a bit compute bound because of the shortage on H 100, but this is going to improve favorably. So we expect to be doing more research and more exploratory research, I guess because we've been doing research from the.

ELAD GIL

Starter, I guess related to that, it seems like in general, your team has a very strong bias for action, and you move very quickly. How do you select for that in people that you hire? Are there specific things you look for, interview questions you ask?

ARTHUR MENSCH

So we look for AI scientists that can do everything from going down the infrastructure stack to making, extract, transform and load pipelines to thinking about mathematics. So we've been trying to find full stack AI engineers, and they tend to have a strong bias for action. Really, the focus we had is to find low ego people willing to get their hands dirty with jobs that are considered boring by some AI scientists because it's a bit boring. But this has been actually quite productive. And because we focused on the right things and the back. I guess the team is now quite big, so there's a bunch of challenges associated to that. I was surprised by the amount of inbound that we had and the amount of representation that I had to do, especially as we got drawn into political stuff, which we would rather have avoided, but we kind of didn't have a choice. So this was definitely a surprise for me, generally speaking. I was also surprised by the speed we managed to have, because it actually exceeded our expectations. But, yeah, I had pretty little idea of what the job of a funder would be when we started. It's quite fun, but it's effectively surprising. I was imagining myself as still coding after a year, and it's actually no longer the case, unfortunately. But, yeah, that's the price of trying to scale up pretty quickly.

ELAD GIL

You get to do HR coding now, which is even better.

ARTHUR MENSCH

Yeah.

ARTHUR MENSCH

So the reason why we started the company is to have a production arm that creates fission value, to have a research arm. And to be honest, there isn't much demonstration of existence of such organs, because you do have a few research labs that are tied to cloud companies that have a very big top line and using it to sustain research. We think that with AI and with the value that the technology brings, there is a way for doing it. But I guess this still remains to be shown. And that's the experiment we are making with mistral.

ELAD GIL

Probably. One last question. I know Arthur has a hard stop, maybe way in the back there.

[AUDIENCE QUESTION – HOW MUCH PERFORMANCE CAN A SMALL MODEL REALLY HAVE]

ARTHUR MENSCH

Yes, I think you can squeeze it to that point. The question is whether you can have a 7B model that beats Mistral Large. This starts to be a bit tricky, but there might be way. I also expect the hardware to improve, like the local hardware to improve. And so that will also give a little more space and a little more memory. And yeah, I see more potential there, because effectively you're a bit constrained by scaling loads. That tells you that at some point you do saturate the capacity of models of a certain size.

ELAD GIL

What is the main constraint? Or what do you think is the thing that it asymptotes against for scaling loads?

ARTHUR MENSCH

You can make 7B models very strong if you focus on a specific task. But if you want to pour all of the knowledge of the world onto 7GB, well, it's actually quite ambitious. So one thing is, for instance, multilingual models at this size are not a great idea. So you do need to focus on a specific part of the human knowledge. You want to compress I guess one.

ELAD GIL

Last question for me and then we can wrap up is a friend of mine pointed this out to me, which basically, if you think about what you do when you're training a model is you spin up a giant data center or supercomputer and then you run it for n weeks or months or however long you decide to train for, and then the output is a file.

ARTHUR MENSCH

You're basically zipping the world knowledge. It's not much more than that, actually.

ELAD GIL

Yeah. How do you think about either forms of continuous training or retraining over time or sort of longer training runs that get tacked on? I know some people are basically training longer or longer and then dropping a model and then they keep training and then they drop a model. And so I don't know how you think about where the world heads.

ARTHUR MENSCH

Yeah, this is an efficient way of training, so that's definitely interesting for us.

ELAD GIL

Okay, great. Well, please join me in thanking Arthur.

Other Firesides & Podcasts

• Arthur Mensch: Mistral.AI

• Apoorva Metha: Starting Instacart

• Satya Nadella: Building Microsoft

• Brad Gerstner: Altimeter and Macro

• Clem Delangue: Hugging Face, Open Source, AI

• Dylan Field: Figma, AI & Design, Education

• Reid Hoffman on AI, Big Tech, and Society

• Sam Altman, CEO OpenAI

• Emad Mostaque, Stability.AI

• NoPriors AI Podcast

My book: High Growth Handbook. Amazon. Online.

Markets:

• Things I Don’t Know About AI

• 2023 Unicorn Market Cap & Rise of AI

• AI Regulation

• Early Days of AI

• AI Safety: Technology vs Species Threats

• Startup Decoupling and Reckoning

• Defensibility and Competition

• AI Platforms, Markets, and Open Source

• Changing times (or, why is every layoff 10-15%?)

• AI Startup Vs Incumbent Value

• AI Revolution - Transformers and Large Language Models 

• Startup Markets Summer 2022

• False Narrative Around Theranos

• Index Companies

• Defense Tech

• Collaborative Enterprise

• Industry Towns: Where you start a company matters

• Markets are 10X Bigger

• Hot Markets 2019

• End of Cycle?

• Machine Learning Startups

• 3 Types Of Platform Companies

• Defensibility and Lock-In: Uber and Lyft

• Uber And Disruption

• Who Cares If Its Been Tried Before?

• The Road To $5 Billion Is A Long One

• How To Win As Second Mover

• End Of Silicon Valley

• Social Products

• Hot Markets For 2015

Startup life

• Capital Efficient Businesses

• Startups Are An Act of Desperation

• Back To The Office

• Hiring Executives and Bad Advice

• When executives break

• Fear of Sales

• A brief guide to startup pivots

• 4 Ways Startups Fail

• Founder Investors and Scout Programs

• Better Meetings

• Magic Startup Moments

• Founder Investors & Scout Programs

• Jobs, Wozniak, Cook

Co-Founders

• How To Choose A Co-Founder

• Unequal Cofounders

• How To Fire A Co-Founder

• Founders Should Divide and Conquer

Raising Money

• Preemptive rounds

• Don't Ask For Too Much Money

• Building VC Relationships

• Founders Should Divide And Conquer

• Lead VC Vetos

• What Is A Good VC?

• How To Choose The Right VC For You

• Signs a VC Just Isn't That Into You

• Series A Crunch

• Questions VCs Will Ask You

• How To Raise A Successful VC Round

• Differences Between Funding Rounds: Series Seed, A, B, C...

• Financing Approaches Most Likely To Kill Your Company

• Party Rounds: How to Get A High Valuation For Your Seed Startup

• 20 Questions To Ask Yourself Before Raising Money

• The 7 Types Of Angel Investors

• Fundraising Will Take You 3 Months

• How To Sell Secondary Stock

Things I Don't Know About AI

2024-02-21 21:37:01

In most markets, the more time passes the clearer things become. In generative AI (“AI”), it has been the opposite. The more time passes, the less I think I actually understand.

For each level of the AI stack, I have open questions. I list these out below to stimulate dialog and feedback.

LLM Questions

There are in some sense two types of LLMs - frontier models - at the cutting edge of performance (think GPT-4 vs other models until recently), and everything else. In 2021 I wrote that I thought the frontier models market would collapse over time into an oligopoly market due to the scale of capital needed. In parallel, non-frontier models would more commodity / pricing driven and have a stronger opensource presence (note this was pre-Llama and pre-Mistral launches).

Things seem to be evolving towards the above:

Frontier LLMs are likely to be an oligopoly market. Current contenders include closed source models like OpenAI, Google, Anthropic, and perhaps Grok/X.ai, and Llama (Meta) and Mistral on the open source side. This list may of course change in the coming year or two. Frontier models keep getting more and more expensive to train, while commodity models drop in price each year as performance goes up (for example, it is probably ~5X cheaper to train GPT-3.5 equivalent now than 2 years ago)

As model scale has gotten larger, funding increasingly has been primarily coming from the cloud providers / big tech. For example, Microsoft invested $10B+ in OpenAI, while Anthropic raised $7B between Amazon and Google. NVIDIA is also a big investor in foundation model companies of many types. The venture funding for these companies in contrast is a tiny drop in the ocean in comparison. As frontier model training booms in cost, the emerging funders are largely concentrated amongst big tech companies (typically with strong incentives to fund the area for their own revenue - ie cloud providers or NVIDIA), or nation states wanting to back local champions (see eg UAE and Falcon). This is impacting the market and driving selection of potential winners early.

It is important to note that the scale of investments being made by these cloud providers is dwarfed by actual cloud revenue. For example, Azure from Microsoft generates $25B in revenue a quarter. The ~$10B OpenAI investment by Microsoft is roughly 6 weeks of Azure revenue. AI is having a big impact on Azure revenue revently. Indeed Azure grew 6 percentage points in Q2 2024 from AI - which would put it at an annualized increase of $5-6B (or 50% of its investment in OpenAI! Per year!). Obviously revenue is not net income but this is striking nonetheless, and suggests the big clouds have an economic reason to fund more large scale models over time.

In parallel, Meta has done outstanding work with Llama models and recently announced $20B compute budget, in part to fund massive model training. I posited 18 months ago that an open source sponsor for AI models should emerge, but assumed it would be Amazon or NVIDIA with a lower chance of it being Meta. (Zuckerberg & Yann Lecunn have been visionary here).

Questions on LLMs:

• Are cloud providers king-making a handful of players at the frontier and locking in the oligopoly market via the sheer scale of compute/capital they provide? When do cloud providers stop funding new LLM foundation companies versus continuing to fund existing? Cloud providers are easily the biggest funders of foundation models, not venture capitalists. Given they are constrained in M&A due to FTC actions, and the revenue that comes from cloud usage, it is rational for them to do so. This may lead / has led to some distortion of market dynamics. How does this impact the long term economics and market structure for LLMs? Does this mean we will see the end of new frontier LLM companies soon due to a lack of enough capital and talent for new entrants? Or do they keep funding large models hoping some will convert on their clouds to revenue?

• Does OSS models flip some of the economics in AI from foundation models to clouds? Does Meta continue to fund OS models? If so, does eg Llama-N catch up to the very frontier? A fully open source model performing at the very frontier of AI has the potential to flip a subportion the economic share of AI infra from LLMs towards cloud and inference providers and decreases revenue away from the other LLM foundation model companies. Again, this is likely an oligopoly market with no singular winner (barring AGI), but has implications on how to think about the relative importance of cloud and infrastructure companies in this market (and of course both can be very important!).

  • One of the most brilliant things in the Llama2 terms of use is the open commercial use of the license if you have fewer then 700 million users[1]. This obviously prevents some large competitors from using their models. But it also means if you are a big cloud provider you need to pay a license to Meta for Llama, which Microsoft has already done. This creates an interesting long term way for Meta to control (& monetize) Llama despite being open source.

    

• How do we think about speed and price vs performance for models? One could imagine extremely slow incredibly performant models may be quite valuable if compared to normal human speed to do things. The latest largest Gemini models seem to be heading in this direction with large 1 million+ token context windows a la Magic, which announced a 5 million token window in June 2023. Large context windows and depth of understanding can really change how we think about AI uses and engineering. On the other side of the spectrum, Mistral has shown the value of small, fast and cheap to inference performant models. The 2x2 below suggests a potential segmentation of where models will matter most.

• How do architectures for foundation models evolve? Do agentic models with different architectures subsume some of the future potential of LLMs? When do other forms of memory and reasoning come into play?

• Do governments back (or direct their purchasing to) regional AI champions? Will national governments differentially spend on local models a la Boeing vs Airbus in aerospace? Do governments want to support models that reflect their local values, languages, etc? Besides cloud providers and global big tech (think also e.g. Alibaba, Rakuten etc) the other big sources of potential capital are countries. There are now great model companies in Europe (e.g. Mistral), Japan, India, UAE, China and other countries. If so, there may be a few multi-billion AI foundation model regional companies created just off of government revenue.

• What happens in China? One could anticipate Chinese LLMs to be backed by Tencent, Alibaba, Xiaomi, ByteDance and others investing in big ways into local LLMs companies. China’s government has long used regulatory and literal firewalls to prevent competition from non-Chinese companies and to build local, government supported and censored champions. One interesting thing to note is the trend of Chinese OSS models. Qwen from Alibaba for example has moved higher on the broader LMSYS leaderboards.

• What happens with X.ai? Seems like a wild card.

• How good does Google get? Google has the compute, scale, talent to make amazing things and is organized and moving fast. Google was always the worlds first AI-first company. Seems like a wild card.

Infra companies

There are a few types of infrastructure companies with very different uses. For example, Braintrust provides eval, prompt playgrounds, logging and proxies to help companies move from “vibe based” analysis of AI to data driven. Scale.ai and others play a key role in data labeling, fine tuning, and other areas. A number of these have open but less existential questions (for example how much of RLHF turns into RLAIF).

The biggest uncertainties and questions in AI infra have to do with the AI Cloud Stack and how it evolves. It seems like there are very different needs between startups and enterprises for AI cloud services. For startups, the new cloud providers and tooling (think Anyscale, Baseten, Modal, Replicate, Together, etc) seem to be taking a useful path resulting in fast adoption and revenue growth.

For enterprises, who tend to have specialized needs, there are some open questions. For example:

• Does the current AI cloud companies need to build an on-premise/BYOC/VPN version of their offerings for larger enterprises? It seems like enterprises will optimize for (a) using their existing cloud marketplace credits which they already have budget for, to buy services (b) will be hesitant to round trip out from where their webapp / data is hosted (ie AWS, Azure, GCP) due to latency & performance and (c) will care about security, compliance (FedRAMP, HIPAA etc). The short term startup market for AI cloud may differ from long term enterprise needs.

• How much of AI cloud adoption is due to constrained GPU / GPU arb? In the absence of GPU on the main cloud providers companies are scrambling to find sufficient GPU for their needs, accelerating adoption of new startups with their own GPU clouds. One potential strategy NVIDIA could be doing is preferentially allocating GPU to these new providers to decrease bargaining power of hyperscalers and to fragment the market, as well as to accelerate the industry via startups. When does the GPU bottleneck end and how does that impact new AI cloud providers? It seems like an end to GPU shortages on the main clouds would be negative for companies whose only business is GPU cloud, while those with more tools and services should have an easier transition if this were to happen.

• How do new AI ASICS like Groq impact AI clouds?

• What else gets consolidated into AI clouds? Do they cross sell embeddings & RAG? Continuous updates? Fine tuning? Other services? How does that impact data labelers or others with overlapping offerings? What gets consolidated directly into model providers vs via the clouds?

• Which companies in the AI cloud will pursue which business model?

  • It is important to note there are really 2 market segments in the AI cloud world (a) startups (b) mid-market and enterprise. It seems likely that “GPU only” business model default works with the startup segment(who have fewer cloud needs), but for large enterprises adoption may be more driven by GPU cloud constraints on major platforms. Do companies providing developer tooling, API endpoints, and/or specialized hardware, or other aspects morph into two other analogous models - (a) “Snowflake/Databricks for AI” model or (b) “Cloudflare for AI”? If so, which ones adopt which model?

• How big do the new AI clouds become? As large as Heroku, Digital Ocean, Snowflake, or AWS? What is the size of outcome and utilization scale for this class of company?

• How does the AI stack evolve with very long context window models? How do we think about the interplay of context window & prompt engineering, fine tuning, RAG, and inference costs?

• How does FTC (and other regulator) prevention of M&A impact this market? There are at least a dozen credible companies building AI cloud related products and services - too many for all of them to be stand alone. How does one think about exits under an administration that is aggressively against tech M&A? Should the AI clouds themselves consolidate amongst themselves to consolidate share and services offered?

Apps questions

ChatGPT was the starting gun for many AI founders. Prior to ChatGPT (and right before that Midjourney and Stable Diffusion) most people in tech were not paying close attention to the Transformer/Diffusion model revolution and dislocation we are now experiencing.

This means that people closest to the model and technology - ie AI researchers and infra engineers - were the first people to leave to start new companies based on this technology. The people farther away from the core model world - many product engineers, designers, and PMs, did not become aware of how important AI is until now.

ChatGPT launched ~15 months ago. If it takes 9-12 months to decide to quit your job, a few months to do it, and a few months to brainstorm an initial idea with a cofounder, we should start to see a wave of app builders showing up now / shortly.

• B2B apps. What will be the important companies and markets in the emerging wave of B2B apps? Where will incumbents gain value versus startups? I have a long post on this coming shortly.

• Consumer. Arguably a number of the earliest AI products are consumer or “prosumer” - ie used in both personal and business use cases. Apps like ChatGPT, Midjourney, Perplexity and Pika are examples of this. That said, why are there so few consumer builders in the AI ecosystem? Is it purely the time delay mentioned above? It seems like the 2007-2012 social product cohort has aged out. New blood is needed to build the next great wave of AI consumer.

• Agents. Lots and lots of things can happen with agents. What will be strong focused product areas versus startups looking for a use case?

This is one of the most exciting and fast-changing moments in technology in my lifetime. It will be fun to see what everyone builds. Looking forward to thoughts on the questions above.

Thanks to Amjad Masad and Vipul Prakash for comments on a draft of this post.

NOTES

[1] Yes I occasionally read terms of use for fun.

MY BOOK
You can order the High Growth Handbook here. Or read it online for free.

OTHER POSTS

Firesides & Podcasts

• Apoorva Metha: Starting Instacart

• Satya Nadella: Building Microsoft

• Brad Gerstner: Altimeter and Macro

• Clem Delangue: Hugging Face, Open Source, AI

• Dylan Field: Figma, AI & Design, Education

• Reid Hoffman on AI, Big Tech, and Society

• Sam Altman, CEO OpenAI

• Emad Mostaque, Stability.AI

• NoPriors AI Podcast

Markets:

• 2023 Unicorn Market Cap & Rise of AI

• AI Regulation

• Early Days of AI

• AI Safety: Technology vs Species Threats

• Startup Decoupling and Reckoning

• Defensibility and Competition

• AI Platforms, Markets, and Open Source

• Changing times (or, why is every layoff 10-15%?)

• AI Startup Vs Incumbent Value

• AI Revolution - Transformers and Large Language Models 

• Startup Markets Summer 2022

• False Narrative Around Theranos

• Index Companies

• Defense Tech

• Collaborative Enterprise

• Industry Towns: Where you start a company matters

• Markets are 10X Bigger

• Hot Markets 2019

• End of Cycle?

• Machine Learning Startups

• 3 Types Of Platform Companies

• Defensibility and Lock-In: Uber and Lyft

• Uber And Disruption

• Who Cares If Its Been Tried Before?

• The Road To $5 Billion Is A Long One

• How To Win As Second Mover

• End Of Silicon Valley

• Social Products

• Hot Markets For 2015

Startup life

• Capital Efficient Businesses

• Startups Are An Act of Desperation

• Back To The Office

• Hiring Executives and Bad Advice

• When executives break

• Fear of Sales

• A brief guide to startup pivots

• 4 Ways Startups Fail

• Founder Investors and Scout Programs

• Better Meetings

• Magic Startup Moments

• Founder Investors & Scout Programs

• Jobs, Wozniak, Cook

Co-Founders

• How To Choose A Co-Founder

• Unequal Cofounders

• How To Fire A Co-Founder

• Founders Should Divide and Conquer

Raising Money

• Preemptive rounds

• Don't Ask For Too Much Money

• Building VC Relationships

• Founders Should Divide And Conquer

• Lead VC Vetos

• What Is A Good VC?

• How To Choose The Right VC For You

• Signs a VC Just Isn't That Into You

• Series A Crunch

• Questions VCs Will Ask You

• How To Raise A Successful VC Round

• Differences Between Funding Rounds: Series Seed, A, B, C...

• Financing Approaches Most Likely To Kill Your Company

• Party Rounds: How to Get A High Valuation For Your Seed Startup

• 20 Questions To Ask Yourself Before Raising Money

• The 7 Types Of Angel Investors

• Fundraising Will Take You 3 Months

• How To Sell Secondary Stock

A return to company core mission & values

2024-01-31 18:32:43

This post is not about whether a company (or other organizations) should, or should not have, viewpoints on every political and current event. That is ultimately up to the organization’s governance.

However, if you run an organization that is being distracted from its core mission and fatigued on politics at work- now is a good window to let employees know your organization is refocusing on its core mission. Broader politics is no longer part of the day to day at the company, while other performance & cultural values have returned back into focus.

You could potentially tie this to a bigger view point on the nature of work at the company or organization, and a re-affirmation of key cultural values. For example, that you are a performance-based culture, focused on your customers as primary stakeholders of your business, etc. Shopify did this extremely well and gracefully back in 2021.

Why now?

There is a grace period right now between the Israel Oct 7 terror attack and the Biden / Trump election. It is possible 2024 will be a bumpy year, so acting now in relative calm might be the easiest time to make a change. In this window of relatively few, new politically charged events it is a good time to make a statement, that your company will no longer be making statements on every news event. :)

What to let people know?

• Your organization’s focus is on your core mission or business. It may be useful to enumerate your mission, some of the core cultural values, and a reminder of which customers you serve, etc.

• Your organization is not the government or political adjudicator. Your organization will no longer make political statements on broad based societal topics. It may occasionally comment on areas unique to its specific business. For example, it may be important for a fintech company to engage in financial services policy and regulation, or for a healthcare company to work on healthcare policy.

• Your organization will not allow for internal political conversation on official channels (e.g. slack) for political discourse. Not everyone wants to participate and many find this distracting. If you want to have political conversations you are free to do so on your own time with your own personal chats, platform etc.

What else?

• People tend to self select into organizations that they will thrive in. If you are explicit in terms of company values through the interview process, all hands, website etc you will select out people who view work as a primary political venue.

• A small subset of people may resist change of any sort (be it this one, or any other), and when change happens act out or disrupt work. This may lead to the voluntary, or involuntary, departure of a handful of people.

• You may get tested in the next political event. Staying firm and unapologetic tends to work well and prevents slippage of focus and mission. You have told people about the change and now you are sticking with it.

• You do not necessarily need to potentially tie this to a bigger view point on the nature of work at the company or organization, and a re-affirmation of key cultural values. If you want you can just send a short note about the shift and mention it at the next all hands. However, it is an opportunity to remind your fellow employees of what mission you serve, or that for example, that you are a performance-based culture, focused on your customers as primary stakeholders, etc. Shopify did this extremely well back in 2021.

• For some organizations refocusing back to organizational core mission is a smooth and easy transition. For others there are a few bumpy weeks and then things resolve and life moves on.

MY BOOK
You can order the High Growth Handbook here. Or read it online for free.

OTHER POSTS

Firesides & Podcasts

• Apoorva Metha: Starting Instacart

• Satya Nadella: Building Microsoft

• Clem Delangue: Hugging Face, Open Source, AI

• Dylan Field: Figma, AI & Design, Education

• Reid Hoffman on AI, Big Tech, and Society

• Brad Gerstner on Macro & Startups

• Sam Altman, CEO OpenAI

• Emad Mostaque, Stability.AI

• NoPriors AI Podcast

Markets:

• Capital Efficient Businesses

• 2023 Unicorn Market Cap & Rise of AI

• AI Regulation

• Early Days of AI

• AI Safety: Technology vs Species Threats

• Startup Decoupling and Reckoning

• Defensibility and Competition

• AI Platforms, Markets, and Open Source

• Changing times (or, why is every layoff 10-15%?)

• AI Startup Vs Incumbent Value

• AI Revolution - Transformers and Large Language Models 

• Startup Markets Summer 2022

• False Narrative Around Theranos

• Index Companies

• Defense Tech

• Collaborative Enterprise

• Industry Towns: Where you start a company matters

• Markets are 10X Bigger

• Hot Markets 2019

• End of Cycle?

• Machine Learning Startups

• 3 Types Of Platform Companies

• Defensibility and Lock-In: Uber and Lyft

• Uber And Disruption

• Who Cares If Its Been Tried Before?

• The Road To $5 Billion Is A Long One

• How To Win As Second Mover

• End Of Silicon Valley

• Social Products

• Hot Markets For 2015

Startup life

• Startups Are An Act of Desperation

• Back To The Office

• Hiring Executives and Bad Advice

• When executives break

• Fear of Sales

• A brief guide to startup pivots

• 4 Ways Startups Fail

• Founder Investors and Scout Programs

• Better Meetings

• Magic Startup Moments

• Founder Investors & Scout Programs

• Jobs, Wozniak, Cook

Co-Founders

• How To Choose A Co-Founder

• Unequal Cofounders

• How To Fire A Co-Founder

• Founders Should Divide and Conquer

Raising Money

• Preemptive rounds

• Don't Ask For Too Much Money

• Building VC Relationships

• Founders Should Divide And Conquer

• Lead VC Vetos

• What Is A Good VC?

• How To Choose The Right VC For You

• Signs a VC Just Isn't That Into You

• Series A Crunch

• Questions VCs Will Ask You

• How To Raise A Successful VC Round

• Differences Between Funding Rounds: Series Seed, A, B, C...

• Financing Approaches Most Likely To Kill Your Company

• Party Rounds: How to Get A High Valuation For Your Seed Startup

• 20 Questions To Ask Yourself Before Raising Money

• The 7 Types Of Angel Investors

• Fundraising Will Take You 3 Months

• How To Sell Secondary Stock

Altimeter's Brad Gerstner on Macro, Tech and Startups

2023-12-06 00:52:07

Brad Gerstner, Founder of Altimeter, and I discuss:

• What can we learn from tech and internet history?

• Technology cycles

• AI, its impact, and how to invest in it

• Macro shifts

• Tech startup world

Video

Transcript

Brad Gerstner

Thanks for having me. I started four companies on the back of a napkin, like, day one. And so I get both excited and know the feeling of sitting where you all are sitting right now. Unfortunately, this guy gets to have all the day-one fun these days because Altimeter tends to invest a little bit later. And I'm sure we'll get into that today. 

But, my heart really still lies and the heart rate gets going talking about building new things. Because ultimately what we all do only matters if we're making life better for some consumer or we're making life better for some business, right? Like we're pushing humanity forward or we're not. Everything else is kind of bullshit and financial engineering and so being in the room with the people who are actually doing the work and building the things that keep this engine of innovation going, I couldn't wait to come and spend time with you.

Elad Gil

Yeah, thanks so much for coming. Maybe we can start with your background because you've done a variety of things. You were trained, I think, as a lawyer. You've started multiple companies. You were very early on at venture capital firms like General Catalyst, which have now really taken on a major role in the ecosystem. And then, you started Altimeter in, I believe, 2008.

Brad Gerstner

2008, yeah. So the quick and dirty. I went to law school in 1996. I was always passionate about technology, kind of an early adopter. But I see Andreessen on the cover of, I think, Time magazine in 1996 for the Netscape browser. I gathered all my friends in law school around one of the few computers we had that was connected to the Internet. I pulled up the Netscape browser, and I said, this changes everything. They're like, what are you talking about?

But in that moment, I realized I went to law school. I was trained as an analyst, but what I really wanted to do, where my passion was, was to get to this place, Silicon Valley. I grew up in Indiana. I hadn't traveled a lot. I had never been to California. So I bought a ticket and I flew to Silicon Valley. And that really just started me on this journey. I ended up practicing law, doing this stint in politics, and realized I wanted to go back to business school to really pivot into business. 1999, the world's blowing up. Internet 1.0. I feel like I was already late. I must have missed it. There are already big outcomes.

But of course, we were just getting started. I thought I was going to move to Silicon Valley. In fact, I came out here and interviewed with some small, little companies. I really wanted a small company with a ping pong table and getting that experience. So I interviewed with a company called TellMe - you know, which Mike McCue ran, which was kind of voice automation that Microsoft ended up buying, and then a little search company called Google that I thought was already too big with a few hundred employees, but loved it as a user.

And ultimately, my classmate, who became my wife, said, we're staying in Boston. And I knew these two guys who were trying to start a venture capital firm in Boston, David Fialkow and Joel Cutler, and they said, hey, come work with us. We need to put together a launch deal to launch the fund. And so let's incubate a launch deal, and we need to partner with the best firm and the hottest firm in the world to put ourselves on the map. So the hottest firm at the time was SoftBank. Before they weren't. Before they were. Before they weren't. They've had quite the up and down over the years. But SoftBank in 1999 was flying. They were hosting every party out here.

Everybody was starting an Internet company. We had an idea, David and Joel, really, for an online travel business that basically like Shopify for online travel. Go and build the booking engines, because Expedia was already doing air tickets, but they hadn't done hotels or vacation packages or cruises. So let's go build that booking infrastructure and then we'll license it to the brands. And that's what we did. We pulled the deal together, and within a couple of years, we had a business doing over a billion in gross bookings, and over 20 million in EBITDA, and we would go on to sell that business to Barry Diller in 2001.

I started two additional businesses before getting back to the investing world. And so the red thread for me was definitely a passionate analyst trying to understand and make sense out of the world, but really, as a founder, an entrepreneur, a venture capitalist, walked up and down Sand Hill raising money and had those feels of a first-time founder. All of the exhilaration, all of the terrifying feelings of that. I was running a company with 1,200 people when September 11 happened, and I had to gather them in a room and let 700 people go because the world of travel had stopped in a day.

And so unless you've done that and experienced those things, then I think it's very difficult to have the empathy to sit in the chairs that we now occupy and talk to founders who are on those journeys and so many people along the way. But David and Joel were really instrumental in helping me see what was possible.

Elad Gil

Yeah, it's kind of interesting. You mentioned the feeling of showing up too late. I remember Marc Andreessen talking about this, where he said he showed up in the 90s. The world of shrink wrap software was sort of the prior wave before the Internet. And everybody felt like they were showing up too late to Silicon Valley. Was tech still a thing?

And so I feel like it's almost a generational thing where every generation feels like it's showing up too late and then it turns out in hindsight, there's so much to do and so many opportunities.

What made you decide to start Altimeter in 2008? So to your point, the financial crisis has happened.

Brad Gerstner

I had started these three businesses and to be fair, they were all like base hits, doubles. But for a poor kid from Indiana, they felt like smoking home runs to me. Right. Because you just get economic freedom and it changes how you think. In fact, I remember Joel Cutler saying to me one time, it must have been 99 or 2000, I was fretting about something. He was like, we just need to get money in your bank account so you stop worrying about all the little shit. And I said, I'm happy to send you the wiring instructions. Like, we could solve this right now.

But I knew I wanted by the time I'd started that third business, I knew that I really wanted to run my own thing. And I had some observations about venture capital and about public market investing that were different than what was currently being practiced. So in venture capital, I saw a lot of firms that were very generalist firms. They were organized as traditional partnerships. They moved pretty slow. They were not very diverse. A lot of them at that time were in Boston. I consider myself a founder. Wanted to be in Silicon Valley, wanted to be with my people, wearing my T-shirts. I didn't want to wear suits. And I thought there was a huge advantage from also doing public markets.

So in my early first company, I was lucky enough to have two hedge fund investors invest in that company. Paul Reeder, who ran PAR Capital, and Seth Klarman, who ran Baupost. True legends in the hedge fund industry. Nobody talked about it as like, crossover investing, right? This predated Tiger and Coatue that both got started around that time. But if you go all the way back to Warren Buffett, in 1950, he had been investing in private companies and public companies, right? He commingled them. It really was the emergence of LPs who said, no, you're not allowed to do both. We want you in one of these buckets. That caused firms to start specializing and calling themselves kind of one thing or the other.

I had a theory. Technology companies are going to scale faster because of this thing called the Internet. The private markets and private capital will mature and be able to provide the capital they need to scale faster. And that there was a huge informational advantage from being in both. Right. That I could bring the value of that venture network to the public markets and the public markets to the venture investors. Because if you're bold enough in 2008 to think you're a venture brand that can break in against the Sequoias of the world.

You better have a wedge, like what's your wedge? And you better have a point of competitive differentiation. And so our wedge was really that we bring the knowledge, insights, research of the public markets, the relationship of the public markets to these earlier stage investors, and that we can be with them for a longer period of time. Go and do that Series A with Bill Gurley because he's excellent at helping you build that first two or three years, and find product market fit. Then we'll come in, take the baton, help you scale your capital, still with the sensibility of a founder who wears a T-shirt and lives here and lives and breathes it, but also has a foot in Wall Street, knows how to navigate, kind of scaling that capital. So I thought there was an opportunity.

I was lucky enough to be trained in the hedge fund business by Paul Reader and some of these greats so I understood the public market business. 2008, I certainly didn't think the world was going to be as bad as it was. I had some endowments who had promised me some money at the end of 2007. But this is a good lesson, starting any business, right? So 2007, the world's pretty good. Some people say, hey, we'll give you a couple of hundred million dollars. Go start a fund. We like this idea you have. Public and venture. Move to Silicon Valley. I get married at the end of seven. I have my first child in June of eight. And I had made the plan. I'm going to launch this firm.

And of course, the world starts melting down. And I don't know, entrepreneurs, they're pretty damn determined. And I said the horse has left the stable. We're doing this. I'll get whatever money I get. I had mentors who started with just a little. Seth Klarman had started 1981 with less than $20 million. My mentor, Paul Reader, had started with less than 5 million. So I was just like, the key is you're either committed or you're not. Are you starting or you're not?

And so, November 1, 2008, yesterday was our 15th anniversary. The first trade I ever placed right in my own fund, I bought Priceline was the first stock we bought. The world was melting down. It was hard to even get a technology platform at that time to do our business because everybody thought all the investment banks were going broke. And the guy who gave me my shot at opening up worked at UBS, and he said, we'll be your prime broker. And he said, Listen, here's the truth. I don't know if we're going to be in business in six months, but you're a good guy. Let's give it a shot. And if it works, it'll work together. And he sent me a text yesterday, and it worked beyond all of our wildest dreams.

Elad Gil

Yeah, well, congratulations on 15 years. That's pretty amazing.

Brad Gerstner

Thanks, it’s been a run.

Elad Gil

You folks have also done a variety of things. You've done privates like one of the early runs in Snowflake, you help anchor IPOs, you do the public market investing.

How do you view the range of things that you do today and how's that broken out by relative focus?

Brad Gerstner

Firms are organized in all sorts of different ways. You guys know the classic benchmark model. Six partners, they're each full stack. So they go source their own deals, they do their own deals, they don't have analysts. And the portfolio is just the natural byproduct of the work that they do. And each partner tends to do two or three deals a year.

We have a public fund. So just to give you a sense of the path, we started with less than 5 million. In 2021. We peaked at over 21 billion in assets. We sent 6 billion and change of profits back to our investors - more than all the money we had ever raised from all of our funds combined. And today we're over 10 billion, roughly split equally between our public funds and our venture funds.

On the public side, that capital tends to be what they call evergreen. So somebody will give us their money, we'll go invest it in the public market. We're judged at the end of every year. We have a team of five people and their major is public markets and their minor is the venture deals that we do. So everybody on our team has a major and a minor. We all work in an open setting like this. So we're sharing ideas because again, we think that free flow of information is how we can really understand how these venture insights from the relationships that we have inform whether or not we should be buying or selling Nvidia, right? And obviously, you guys get why that works.

On the venture side, we're now investing out of our 6th venture fund, which is about $1.5 billion. We just had our first close on our 7th. That will be a billion dollars. And there we really think of our sweet spot as being the occasional Series A, but really the sweet spot is that Series B and Series C. So those are in the case of Snowflake, that Series B was $170,000,000. The Series C was 500 million. The Series B, it was still pre-revenue. They had been working on the product for three years. They had ten beta customers. So we had a sense, right, because we could talk to the customers, we could use the product ourselves. We understood why there was going to be product market fit and why rearchitecting the database native for the cloud, separating storage and compute could be so big. So I would say about 90% of the profits we've generated as a firm in venture were in companies that had less than $5 million in revenue at the time we invested.

Elad Gil

When you talk about $170,000,000 for Snowflake, you mean market cap right?

Brad Gerstner

Yeah, that was the total enterprise value of Snowflake when we got involved, and we're still one of their largest shareholders today, and it's over $50 billion in enterprise value. And I think there's a company that will most certainly double and triple from here in the fullness of time.

Let's say you compete for the Series B in Snowflake, and let's say you don't win it. If you are a stage-specific firm, then all the work you've done on that goes on the shelf, and you move on to your next deal. But at Altimeter we may lose the series B. We may choose not to do the series B, that we don't think the risk-reward is a good setup in the series B, but we can do the C. If we don't do the C, we can do the D or the E or we've bought in over 100 IPOs. 

We're one of the larger buyers in tech IPOs, so we can buy it in the IPO. And in the case of Snowflake, we bought it along every step along the way because the risk-reward kept getting better. Right. Because although the price was going up, the risk was going down. And so we always think about the distribution of those probabilities, the relative risk and reward when we're making the investment.

So if you ask me, like, Brad, where does your heart really lie? It lies in this room. Right. I'm still the guy who started those companies on the back of a napkin. But the truth of the matter is, we don't have the scale in our organization to be the best partner, who's going to sit there and hire your first engineer and spend that two to three years trying to find product market fit. So I partner with world-class guys like this guy and other world-class investors like Benchmark or Mike Speiser at Sutter Hill or Martin Casado and Andreessen and go through the list, and they're doing those A's, and then we come in and we really want to be the best partner possible to hand the baton to. And then when I can't satiate my own appetite for that early-stage stuff, I come in as an angel investor off my personal balance sheet. And I have, I don't know, far too many angel investments, but I can't resist the excitement in partnering with really world-class people at those early stages.

Elad Gil

So a lot has changed since you started Altimeter in terms of the venture markets, public markets, what do you think are the biggest shifts that have happened in terms of the entrepreneurial ecosystem from 2008 until today?

And what do you think will snap back? Because I feel like there are a lot of changes that happened during ZIRP (Zero interest-rate policy) that happened over the last couple of years. Some of those will keep going and some will shift. I'm curious how you think about this sort of macro side of this.

Brad Gerstner

And I might even rewind a little bit further because I think the pattern recognition of 1999 and then the period 2001, it's super important. So 99, everybody said, I can start a company. This is easy, it's get rich quick. Everything was going to the moon. You can take it public, it has no real revenues, no real profits. That was all make-believe, right? 

And by 2001, the only people who hung around Silicon Valley were the real believers, right? The people who wanted to grind, who knew that this was a decade-long journey to build something that was durable and important and that made the world a better place. And it was a pleasure when the tourists left and the people who really believe stayed, right? And that period of investing, 2001 to 2005, sure, we had a lot of challenges. Interest rates were higher, we had a recession we lived through. But I will tell you, we all knew that the Internet, those of us who stayed knew that the Internet was going to be as big as we thought it was going to be. And now it was just about much more sober investing to get you there.

By 2007, of course, like humans, it takes about ten years for us to forget the sins of our prior period. 2007, we saw some silliness start reemerging. All of a sudden, public market investors thought this venture thing was easy and they started throwing big checks around. In 2007, I was on the board of Zillow, and I remember when this happened at Zillow. But then 2008 hit and again, it wrung everybody out of the system, because if you don't have deep conviction at a moment of peril and distress, you are gone. Because the only people who have the stomach to stick around are the people who really believed in the first place.

And so what came out of 2008? I launched, there are always a lot of distress-era funds, by the way, Andreessen started at the exact same time that we started Altimeter. I remember being at the Allen and Company conference with him in Arizona, and we're both passing around our decks to raise money. Andreessen and Altimeter, and by the way, he's an incredible human and built an amazing business. And to even be in the same conversation, I feel inspired by.

But then we all lost our minds again, right? And this started to build from 2009 to 2013. 2014 was really again, the golden age. We had these two super cycles taking off. We had these mobile devices, we had cloud taking off, but we still had the post-traumatic stress of 2008 that kept pricing in check. So like, risk reward was fairly distributed.

But by 2015, 2016, markets going up, everybody's like, this is easy. Again, all the tourists show back up. And man, I thought by 2000, Bill Gurley thought by 18,19, man, this thing is so overcooked and we hadn't even gotten started. I mean, then we had this thing in March of 2020, we have COVID really hit. We think it's the end of the world. But no, it wasn't the end of the world. The fed went all in. Congress went all in. We jacked ourselves up on this super Red Bull high, right? Interest rates go to zero, and the only assets people wanted to own were risk assets. So the more unprofitable you were, but faster growing, the higher the multiple, right?

And those of us who had been through that period in 99, 2000, 2008, 2009 we're like, this does not make sense. This is not sustainable. The market is being manipulated by the Fed. And by the way, I think for good reason. If you remember, in March of 2020, we didn't know if COVID was ebola. We didn't know if it was going to wipe out humanity. And so what they were worried about is that we all just stop. We stay home, the economy literally stops, right? And they had to force everybody to take risk. They had to force everybody into the equity pool, if you will, to keep the economy going.

So at the start, I think totally justified. By the middle of 2021, I think the Fed lost their mind, stayed the course way too long, caused hyperinflation that we then had to come back and battle. But we knew this was upside down on the All in Pod. I was talking about this a lot throughout 2021, by the end of 21.

And then we had the reset in 22, a tech recession, massive reset in pricing. But here's the interesting thing, because I had to set all that up just to answer the simple question like, where are we now? We never got back to the level of despondency that we had in 2001 and 2009. I mean, there, it was washed out. People were so beaten up. They had looked into the abyss, they had witnessed death, and they wanted nothing to do with it. It was a scary feeling, right? I think 2022, yeah, people started to say, oh, we need to tighten our belts a little bit.

But I would say we only ratcheted back to what I would call maybe 2013, 2014 levels of fear and pricing. And now we're entering another super cycle in AI that obviously has some dynamics unto itself that we'll get into. So listen, I think we formed a lot of bad habits during low interest rate environment, and during ZIRP, I think it led to excess.

You know, I wrote this letter to Meta, time to get fit. And I said to Mark at the time, this is an open letter to all of Silicon Valley, right? And I encourage you to read the letter. And then I encourage you to read his response letter that he wrote in March called Year of Efficiency, where he said, flatter is faster, leaner is better. And man just his leadership, Elon's leadership, and others in demonstrating the need to wring out the excess out of these businesses and all the people that they were hoovering up.

That made it impossible to build a startup because the talent was getting overpaid in all these places. And so why would you ever leave? Made it more difficult for the entire ecosystem. And now that whole cycle is reversing and I see the right heartbeat coming back. But listen.

Set a clock, set a timer in your calendar for seven or eight years from now, because we'll forget all of this again and we'll all lose our minds again. And there'll be some event in the world and we'll all get jacked up on Red Bull again and make sure that you say, god, I remember that conversation about 2022 because that is the time you need to be dialing it back.

Elad Gil

That's when you sell and retire. I think it's kind of interesting because if I look at private markets versus public, because public markets have adjusted by 50% to 80%, depending on the company, private markets haven't really shifted much. A, because they're illiquid, and B, because people raise so much money.

And so I think one dynamic that we haven't quite seen is any real carnage happen in private tech, and so we can come back to that in a second. And then secondly, to your point on AI, I feel like the tech economy was almost like an oil economy and we ran out of oil and we discovered shale. It's sort of the thing that kept everybody going and said okay, now we still have energy. It sort of reinflated the entire ecosystem and environment because it's a fundamental technology shift.

Do you think the next year or two are going to see pain in private markets?

What do you think is going to happen to all the companies that raised a couple of years ago that may not have strong business models?

Do you think any reckoning has happened yet? What sort of reckoning do you think will come, if any?

Brad Gerstner

Well, I think we really have to bucket this. We have to cohort it because there's a lot of stuff in there. First, let's talk about all those companies that raised Series A's at inflated prices over the course of the last three years. You know, they're all in market, and unless they have AI in their name, there's no chance they're getting an up-round done, right?

All of those companies, if they raised at a billion dollars with no revenue, if they don't really have product market fit, they probably just disappear. Because the reset that's got to occur in those businesses is so demoralizing to the people in those businesses, so demoralizing to the investors. It's almost better to just shut it down, and start over. 

And if you look at the PitchBook data that I think JCal and I tweeted last week, but I think they talked about on the pod, I think this most recent quarter or month, we had 239 shutdowns. So if you look at the rate of shutdowns at Silicon Valley, it's going pretty parabolic. And these are the early-stage companies that were pre-product market fit, that were raising at Series B, Series C, and Series D valuations. So the best case scenario in those companies is you're going to do a big, and again leave the AI kind of deliciousness aside for the moment, I'm talking about the non-AI businesses. Those will all do big down rounds. The best of them may be a 50% down round if they want to raise new capital. Unfortunately, a lot of them, I don't like this pattern, but they're extending converts.

Listen, a convert is very dangerous. And the reason the convert is dangerous is that it comes home to roost at some point in time. And just expanding your convert, you're just denying reality. And delusion is not a strategy, right? So reprice the business, get the price right, and if that is not satisfying to you, if you don't own enough of the business at that point in time or whatever, call it quits and restart, right? Because you got to have the foundation right in these businesses. Otherwise, it's very difficult. Okay, so that was the early stage, guys.

Elad Gil

Over what period do you think that plays out, by the way?

Do you think that's the next year, the next two years, the next three years?

Brad Gerstner

I think I said last year it was a three-year journey, so now I think it's a two-year journey, right? It's when these companies run out of money and need to come back to market. And I see a ton of them in market. And just to put in perspective, for a firm like ours, our pacing over the last 24 months has been one 10th to one 20th, our normal pacing, because, listen, I can't do this. This thing in the public market corrected by 80%. And now you're coming to me and acting as though interest rates had not reset and that nothing happened in the public market.

I can't live that delusion. So until as buyers and sellers, we have an honest conversation, and by the way, I'm not looking to steal anything because I've been in your shoes, but we have to have an equitable distribution of risk and reward, right? Because I have a fiduciary obligation to the people who give me their money. I care about these institutions. They’re colleges, they’re foundations, they’re hospitals, they’re state pensions, teachers and whatnot. I'm going to fight like hell for these people. And so I can't just pretend nothing happened. So the early stage, I think, is going to have to weather this. And I expect the PitchBook data on shutdowns will continue to climb. By the way, it just started going up over the course of the last few months.

To your point, we're kind of just getting started now. Let's talk about the thousand unicorns. So these were the companies that were valued between a billion dollars and $50 billion that were further in their journey. Most of them had product market fit, had real revenues, maybe 50, maybe 100, maybe two or $300 million in revenue. But they were valued at 50, 70, 100 times ARR.

So I'm going to use software as kind of our analog here. Follow Jamin Ball at AltCap on Twitter. Jamin does probably the best analysis of software. He's a partner of Altimeter and his Substack is called Clouded Judgment. And it's just training in the art of thinking about this. And he has 1 foot in venture and 1 foot kind of in public markets.

But you'll see that for high-growth companies, profitable companies in the public markets, their revenue multiple is now about eight and a half times. Okay, so now just run the math. If you're raising money at 3 billion, 4 billion, 5 billion at 70 times, just run the forecast out. How long do you have to sustain that growth to grow into the public market multiple? That ultimately is your exit. And you just realize very quickly it's impossible. No software company has ever done that, right?

Worse yet, now we're entering an economic slowdown. And so these high-growth companies are now slowing down. So they have these multiples up here and they're now slowing down. So they're coming to you. And they're no longer growing at 70, 80, 90%. They're growing at 20 and 30 and 40%. You say, hold on a second here. Amazon AWS, which is on a $70 billion ARR run rate, is reaccelerating from 12% to 20%, right? You've got these massive businesses like a Snowflake, still growing at 35 or 40%. So you have 100 million in revenue. You're only growing at 20%. You're not profitable, and you want a premium multiple to these players. No chance.

So, again, I think that entire cohort, if you just wanted a rule of thumb, again, not any specific company, but you take Instacart, an amazing business built by amazing friends and people who killed it. But their last round in the private markets was $39 billion. Today in the public markets, they're worth six and a half. Why should any of these other companies have a better situation than that? That should be your case study. That should be your benchmark. If you're one of these unicorns and if you want a higher multiple or less adjustment down than they had, then show us why you've outperformed those expectations.

So that leaves us with only one thing, and that's a slice of the market called AI. And I love the shale analogy. So I think that what makes AI so difficult from a venture capital perspective today really is two things. Number one, we have this massive reset going on around interest rates, normalization of multiples, these drawdowns. But at the same time, we have excitement that only comes to us once a decade, that the Internet brought us, that cloud brought us, that mobile brought us, and we have it. And I think we're right that AI is that big.

Augmented intelligence is going to be what we probably invest against for the next two decades of our life. And so you can have that level of excitement just like we had about the Internet. 97, 98, 99. And you could be deadly wrong from an investing perspective. Think about this. I've used this analogy. 1998. Lots of us thought the Internet is going to be huge. And one of the number one winners or areas in which you can win will be search.

Okay, why? Because they're organizing the world's information. It's the point of entry to the Internet. You put up a little toll booth, you collect a little toll, whether it's per transaction or whether it's subscription like, great place to be. So gather all these people, get them on your platform, you be the toll booth. At the time, it was AOL and Yahoo that were really collecting the tolls, but we all knew there were a lot of challenger technologies, Alta Vista, Lycos, Infoseek, Go, AskJeeves, Google. And so every venture capital firm wanted a logo.

Okay? So what'd you have, you got those two things right, and they all had revenue. And all their revenue was going like this. Okay, why? Because it was easy to get revenue at that point in time because advertisers all wanted to be on the Internet, and these were the only games in town. So all of a sudden now for every venture capitalist, you had all three of those boxes checked. You were like, count me in. I have to have a logo, because otherwise how can I go back and talk to my LPs? And they say, well, who are you investing in search? And you're like, I don't have a search logo. They'd be like, you're an idiot. You're fired. So everybody thought they had to have a search logo.

The truth of the matter is, all those companies, you got all of that right, and all of those companies still went to zero. And you could have waited until 2004 and bought the Google IPO and captured 95% of all the profits ever generated by Internet search, okay? Now, that's one of the reasons I like to do both venture and public markets, because I did buy the Google IPO, right? I was like, man, this is way easier than that venture stuff. Just ride this pony for the next decade, which we did.

But the truth is, I kind of have this deja vu right now. Like, when I look at you and I have talked about every foundation model player in the world, and some people might say, oh, Brad, you're grumpy, you're over the hill. All you talk about is valuations and multiples. But no, I believe these are going to be really instrumental in changing the game. The only reason we have this level of excitement is because of the convergence of all this data set free, silicon that's come together, supercompute, and these transformational models that allow us now really to extract meaning out of an unlimited amount of data, right?

So trying to sort through that and say, okay, if that's true, what is going to have durable value capture? Am I getting head faked? Is this Alta Vista? Is this Lycos? Is this AskJeeves? Or is this Google? Right? And why did Google ultimately capture all the value? Because everybody went to Google, right? They owned the top of the funnel. They became the place. Not only did they own search, they were so crafty. They said, you know, this address bar that people used to actually type in URLs, we'll just turn that into a search box so that anything you type in there becomes a search too. I mean, the level of execution to build that natural monopoly, right? And they captured over a trillion dollars. I would argue it's over 110% of Google's value today is their search business.

And for the first time in 20 years, we have a challenge to that. So I was talking at the Javits Center in New York last week, thousand people with Jason and the audience, and we said, how many people have used ChatGPT in the last two days? I'll ask this room, how many in here have used ChatGPT or Claude or some equivalent in the last couple of days? Everybody's raising their hand. How many people used it as a replacement to what you might otherwise use Google for or search for? Okay, and about half of the people raised their hands, which is about what I would expect. Okay? And so I can tell you, for 20 years, if I asked that question, no hands would go up.

There was no alternative. Right? And so all I'm saying is, for the first time in 20 years, there's the possibility that something else may emerge as the top of the funnel for how we live our lives, how we plan our travel, how we get answers to questions that we have.

So I'm sitting here with my mobile device, with my team, and we're saying, okay, here's the challenge. I want to find the coolest, hippest business hotel on the West Side of New York near the Javits Center, and I want to know if it's available tomorrow night, and what it costs. You use Claude, you use ChatGPT. You use Meta AI. You use Google. Go. You use Booking.com. Okay?

And it's very clear to me that the promise that can now be fulfilled is the promise that Rich Barton and I started with 23 years ago. Rich was this brilliant engineer at Microsoft, Bill Gates. He was working on CDROMs on the computer. He's building travel CDROMs for the computer. He goes to Bill, he says, ”Hey, Bill, CDROMs, that's not the answer. We're going to put these things on Microsoft, they had an MSN service that competed with AOL. Let's put it on there, let's turn it into the internet.” That's what I feel in this moment. I said Rich for the first time in 23. Remember you wanted to put the world's smartest travel agent in everybody's pocket and instead all we got were ten blue links. Well, now we can fucking do it. Like, we can put the world's smartest travel agent in everybody's pocket, where we interact, we ask questions, right? They come to understand me. They know my calendar, my travel schedule, and guess what? They can book it, right? Because action bots are like months away, not years away. They're going to close the loop for us. So this gets me really excited because those are multi-trillion-dollar opportunities that are sitting out there waiting to be disrupted. And I think, again, we can go through a bunch of other things. So I would say valuations there are spicy and they're spicy because those of us who are looking at this saying, holy shit, this is shale, right? This is a revolution in the discovery of oil, this changes everything. I think there's a good chance that we're right. But I also am haunted by the AltaVistas and AskJeeves of the world. And I think being too early in too many things can be tantamount to being wrong.

Elad Gil

Yeah, I think directionally it's kind of clear that this is a big wave. And so I think everybody who's starting a company or investing in companies is thinking the thing they're doing is the right one. But the reality is, if you miss Google, then you missed most of it, right?

And so it's back to like, what are the characteristics that suggest that something is the right company? So you mentioned durability is an important thing. So say that you come in and you're about to lead a series B or C in a key AI company. What are the characteristics you look for? What do you view as those metrics of durability? This is actually Google and not Lycos, right?

Brad Gerstner

I mean, these things are some just enduring attributes. First, business is one of the hardest competitive sports in the world. So if you're not dealing with maniacs who are willing to sacrifice everything in life to win, then you probably, no matter how good your product is, no matter how good your idea is, no matter how good your head start is, you probably don't win. So at that stage, I have to know, what is the wedge that makes this group of humans specially entitled to win in this thing?

Then, of course, secondly, for Altimeter in particular, one of the things that we spend a lot of time thinking about is I think it's probably equally I don't know if you have parents who started a restaurant, but I always look at somebody who starts a restaurant. I'm like, that's one of the hardest things in the world to start. Emotionally draining work 24x7, kill yourself. Barrier to entry is low. It's probably as emotionally and intellectually challenging to start a restaurant as it is to start Google, right?

Is what we're betting on worth the risk and the grind and the probabilities associated with success? So I say if we assume everything you're saying is true, how big is the prize? Are we talking about a trillion-dollar prize, all of search? Or are we talking about a small prize? Right? And then obviously that factors how we think about the price of entry. Price of entry matters. It's got to be a fair distribution of risk and reward.

And then I would say another hallmark of Altimeter: we talk a lot about our culture of essentialism, the art of doing less better. And I encourage you all to read the book by Greg McKeown. It's a great book, but to me, there are different strategies and find your lane in venture.

I know people who place a lot of bets, right? And then they just see, they're like, okay, I'll pour water on the ones that are sprouting. We probably spend more time as anthropologists, researchers. We wait, we study, and when we see the real indicators emerge, then we're pretty aggressive in terms of really making that an important bet within the portfolio. So we have less ideas, but we have more wood behind each swing of the bat, I think, within our portfolios.

So those are some of the attributes, I think, that we think about. Are they killers? Do they have a great idea that if all the things are true, it's going to be a really big outcome? And then finally, are these the people that we want to be in business with for the next ten years? Right, integrity. Do they think there should be balance between a fair and equitable distribution of risks? Are they going to have our backs? Do we want to have their backs? Because this is a long journey, these are hard journeys, and life is short, so be in business with people that you actually want to spend time with.

Elad Gil

I think related to that, if you see the shift that's coming through AI societally, there's going to be potentially a lot of changes in terms of how we think about work and how we think about equity and sort of the nature of the economy and other things are going to shift over time.

And I think one thing that you've talked about a lot is really how technology and people working in technology should take a responsible lens on innovation, how they should engage with society, how they should try and push things forward. There's so much pushing technology and optimism forward.

How do we also make sure that society benefits holistically in that?

And one thing that you talked about recently that I thought was really interesting and inspiring was InvestAmerica, which I think is an exciting program in that direction. Can you talk a little about that? I think it kind of ties into some of these things.

Brad Gerstner

You know, first we're all super privileged, right? We're in this room, and we get to play this sport in one of those most magical times and magical places in the history of the world, not just the history of capitalism, like the history of the world. Picture this. We have an hourglass. Each grain of sand represents 10 million people. The bottom of the hourglass is nearly full. 110,000,000,000 people have lived on this planet and are now in the bottom of the hourglass. And for all of them or for 99.8% of them, they never saw a single invention during their entire life. Their life cycle was shorter than the invention cycle. And now we have a few grains of sand going through the hourglass every year, and we have only a few in the top, just 7 billion in the top. And we get to live at that moment in time where we're inundated daily with new innovations.

And if you look at almost all the advancements of humans, literacy, longevity, quality of life, right? Average GDP on the entire planet, like, this is the best time in the history of the world to occupy the planet, and we're the ones who get to experiment, to invent the shit that moves it forward even more. And AI is unlocking that at an accelerating rate. All that is true and makes me a huge optimist, and that's why I'm passionate about what we do.

I think what we all do matters, but I'm also eyes wide open to the fact that what we do does not impact everybody equally, right? In fact, some of the things we do adversely affect certain people's human progress, by definition. So work with me on this, right? The Industrial Revolution. If you were in the business of making horse-drawn carriages, or you were a blacksmith repairing the horseshoes for horses, you were out of business. Now scale that by a million times. Think about the people who are going to be displaced by AI, right? Like what the Industrial Revolution did to craftspeople, we're going to do to white-collar employment with AI call center engineers getting 30% to 50% more productive. That's going to change the rate at which now there will be other new and great jobs.

I'm with Andreessen, and I think I'm in the optimist camp on this. But those dislocations happen fast, and then society rebuilds and backfills. So part of my thing, particularly for this country, because we're in the privileged position to do this, I think that everybody should share in the upside of this grand bargain that we're all working on together. And so it's a very simple idea. We have 3.7 million children born in this country every year. The federal government should set up a seed investment account for every one of those children. Seed it with $1,000 in the S&P 500.

Because today 70% of families and kids will never benefit from savings or investments. Warren Buffett calls compounding the 8th wonder of the world. You want to know why wealth gaps exist? Because the fact of the matter is, if you're a winner today, Mark Zuckerberg, he sells to 3 billion people. Rockefeller, the prior age of robber barons, had 20 million customers at most. So just think about the scale at which we're now operating. Think about the people who work in call centers or work in places that are going to be displaced by AI.

So we bring everybody under the tent, right? I think this is going to be a golden age for this country, a golden age for our markets. This is less than 1/100 of 1% of our national budget to give everybody skin in the game in which we're all playing. And if we can afford to spend another 60 billion on Ukraine, or 3 billion for 20 countries around the world every year in foreign aid, think about this. The 60 billion going to Ukraine. And that's a totally separate debate, whether you think that's worthy. But that would pay for 50 million children in this country, almost every child for the next 15 years to have a starting account, a seed capital account, right.

And if you grow that and compound that over time, if you start with $1000 and you add $750 every year. So I've talked to Dara Khosrowshahi at Uber. He's like, well, we'd probably match that for all the kids of our employees. Right? I was talking with Michael Dell. Yeah, that's an interesting thing. We would consider that. So now we have a match and parents can start saving for their kids. Then by the time you're ten years old, you have $15,000 in that account. But by the time you're 30 years old, you have $150,000 in that account. By the time you're 50, you have a million dollars in that account.

We have a crisis in this country that less than half of people under the age of 40 believe in capitalism. Okay? Pew and Gallup are showing polls that say just in the last three years, 10% more people lost their confidence in capitalism. We have to solve that crisis. Ray Dalio talks about the rise and fall of nations in this moment, but it's not static. There's a dynamic system that we can actually do something about.

And so we have a broad coalition coming together, a bipartisan coalition coming together to support the legislation. We've got the funding. Follow us. I brought a little take home I'll leave you with but follow it at InvestAmerica24 on Twitter and look for the ads that will start hitting this political season, and legislation to introduce next year.

And yes, I think to answer your question, we are privileged to do what we do. We have a voice. We have reach. I think we need to be part of these solutions, whether it's AI regulation, whether it's issues of national security, whether it's issues if we look at the regional banking crisis that we had earlier in this year. I think we do. And we should have voice, and we should use that to bring everybody along for the benefits that we're seeing.

Elad Gil

That's great. That's a very inspirational place to stop. So thanks so much for joining today.

Brad Gerstner

Thank you. Happy to do some Q&A. Thank you.

Audience Speaker

So many questions. That was amazing. Thank you. I guess one, to sort of where you ended with America's place in the world. I think it sort of has potential to be a really special period for capitalism. But we also have this problem you mentioned about people not having faith in capitalism and people not having faith in America. And we have this kind of ballooning debt problem, two wars on two fronts happening. You can see how these things might overstretch America economically, and most importantly, the faith in the US.

And so how do you think that plays out? Where do you think AI plays into it?

In theory, winning the AI deflection makes America sort of gain its place in the world. I just would love to hear your thoughts on that.

Brad Gerstner

I would say first, and you probably heard us debate this a little bit on the All in Pod, where I'm affectionately the fifth bestie, which means I'm not good enough to be in the top four, but if somebody goes down, they pull me off the bench. What I would say is this. I think having traveled the world and looked at entrepreneurial ecosystems in China, in the Middle East, in Europe, etc. I think one of the things that's deeply underappreciated about this country is the genetic imprint. This country was founded by entrepreneurs. It was founded by risk-takers. The move west was risk-takers.

Every step along the way was basically an entrepreneurial journey. We set up our entire system. Even if you think about the system of bankruptcy as an incentive system to be an entrepreneur that even if you try and fail, we value you so much that you don't have to ruin the rest of your life. Thomas Jefferson declared bankruptcy four or five times, right? Like, that was a deal we made in this economic system.

So first, I think that exists. I think the pulse is as strong as ever. With that said, I don't think there's a natural entitlement forever. I think you need to recognize this specialness and you need to protect it. I look at it and I say, you've probably heard us talk on the pod. When you think about the dollar's reserve currency, when you think about interest rates, these are all relative questions. Relative to who else? The yuan. Right? I was just in the Gulf countries where they've been pegged since the early eighties to the US dollar. And I asked the heads of the Fed in Bahrain, in Saudi. I hear a lot about the yuan. Would you consider repegging? They're like, are you kidding us? Right? Like, the yuan is manipulated. The US has got the most diversified strong economy in the world. The future of technology is happening in the United States.

At the end of the day, why do you invest in a company? Because you're like, fundamentally they have the best developers, they're building the best products. They may have a bad quarter, but it's an engine of innovation that will continue to grow. And I think that is the perception I have of this country. And again, there are things we need to do better. We can't afford $2 trillion annual deficits. We have a $34 trillion national debt. Right? Like, we need to be serious about addressing these things, but we can. We will, and we should.

When I think about things like InvestAmerica, you also can't be against everything we do, spend a lot of money. But we have a choice. Do we need to spend $5 billion on the next weapon system? Or how about seeding every child in America to bring them into this system of capitalism? And so I think we need a fresh look, right, top to bottom, at where our spending is. And it's going to mean, listen, prior generations borrowed from you guys and borrowed from your kids. That debt is going to have to be paid, right?

And so the good news is we are among the few countries on the planet that have a free cash flow coming out of this innovation super cycle. We're the largest producer. We're now a net exporter of oil, like we have a lot of natural gifts, right? And the intellectual and ingenuity is the greatest one. So I'm optimistic about that. But we can't turn a blind eye. We can't take it for granted. I think the state of California has done too much about that. They need to get it back on the right track. And I think that listen, the last ten years, I said, time to get fit to Mark Zuckerberg. I said it to everybody in Silicon Valley. I can write the exact same letter to the United States federal government. It's time to get fit, rein in the excess of the free money period. Right? And this is just basic math. We can do this.

Elad Gil

I think if you look at our world and data too, there are two things that stand out. One is what Brad said earlier, where if you look at almost every aspect of humanity in terms of global poverty, child education, education of girls, death by disease, famine, etc, we're basically living through the best period of time that humanity has ever seen.

And then if you ask people in different countries how optimistic they are about the future, the poorer and worse off the country is, the more optimistic the people are. And when you get to the west western Europe, the US, etc. They're the most pessimistic about the future.

And you could argue that there are two reasons for that. One is viewpoints around participation and capitalism or other things. But I think a more important one is really good times tend to distract people from what really matters, and things get too easy and you take your hands off the steering wheel. And I would argue the last decade we took our hands off the steering wheel in all sorts of ways, the fiscal responsibility side, but also how we think about the roles of universities, how we think about the world philosophically. And so I think there's going to potentially be a hard reset because you start squandering all the positives over time.

Brad Gerstner

You can have hard resets or you can grow into it. And I think if we do this responsibly, we can grow into it. Remind you it was only 1999 that we had a surplus in this country, right? And I was with Bill Clinton this summer, and I said, you know, how were you able to govern in those moments? And he's like, because I was the governor in the state of Arkansas, and we had people who lived up in the mountains. He's like, I had to go into rural places and talk to people who had guns on the counter, right? And I couldn't ignore those people. I couldn't be an elitist. Like, I had to bring everybody together.

And what this country needs, it's so evenly divided, we need people who can bring people together around. Pragmatic solutions. InvestAmerica. I've started now four companies. I've been in politics for a long time. It has one of the highest product market fits I've ever found. Republicans and Democrats all listen for a minute and they're like, why has this not happened? People on the Hill, why has this not happened? Let's do this, right? And so that's just a pragmatic solution.

I went there with my son this summer, my son Lincoln, and we met with Speaker McCarthy. And as we walked out of an hour meeting talking about how to make this happen, I turned to him, he's 15 years old, and I said, every single thing that's ever happened in this country started with a conversation like that. A concerned citizen, a concerned community member, they started the conversation just like every company in Silicon Valley started with one person pulling out a napkin. Lots of people may have had the idea, but one person started. Right. And those are how movements are started.

That's why I think it's so important that we not only build the engines that drive the country forward, but we bring that same passion and intellect to the much bigger issues that the country faces.

Elad Gil

I think I just got convinced to run Brad for Governor.

Brad Gerstner

No, I'm not doing that.

Audience Speaker

Thank you so much. This has been awesome. Longtime fan and student of Altimeter.

I have eight questions that I'm going to pick one from, and I'll preview the second. The second takes on OpenAI 30 billion and 90 billion. What I'm actually asking you is your latest 2023 framework for dealing with uncertainty. You've said if it's not a fast yes, it's a quick no. And then I’ve also read Essentials twice on your recommendation. Thank you. And I'm just curious what your 2023 framework for uncertainty is, given you're investing in AI where there are literal funding risks. Can we get to GPT5? Like the Oppenheimer situation. Then you also invested in Grab. You took Grab public and 25% of their GMV was, I think, in Indonesia. So I'd love to hear your 2023, now we're a year or two in AI, what's your latest framework for uncertainty as a top investor on the planet?

Brad Gerstner

Well, I think that it's a rare breed that can be both a venture capitalist and a public market investor. When you think about venture capital, in many ways, it's the art of the possible. And when you think about the public markets, it's the art of the probable. Right. But we often just look through the lens of why the All in Pod grew out of poker. And why am I heading there tonight? And do I look forward so much to our Thursday night poker games? I mean, everything in life is a distribution of future unknown probabilities. And that's fun, right? And the same framework applies in venture that apply in public markets, that applies in poker.

Right. And we're continuous learners. And how can I, as an analyst, gather more and more? How can I make this single processor have more data, and run faster cycle times to make better decisions? As I sit here at the end of 23, I'm just sobered and reminded that ultimately, all valuation frameworks end up in the exact same place. It's a reasonable multiple relative to the risk-free rate of return of future expected cash flows. That's it.

Revenue multiples are total bullshit unless they result in future free cash flow that you're applying a multiple to that compared to the risk-free rate of return, providing you compensation for the equity risk that you're taking. Right? And so I would say I know people who, after moments of great distress, get paralyzed by fear. So it really helps to be an optimist. Buffett's like I've been an optimist for 50 years, and it served me well. I agree. It's really important. If you get up every morning pissed off about the world and think that tomorrow is going to be a worse place, you're going to be a terrible venture capitalist, and you're not going to be a very good public market investor. But I will hire you to be on my short team, to look at good short ideas, because that is just constitutionally something that is different.

But at the end of the day, Warren Buffett, one of my heroes, guys like Paul Reeder, value investors, guys like Seth Klarman, I'm an optimist, but I'm also kind of a value investor. Like, I need to understand why the distribution of probabilities is fair. So in the public markets, last year you could have bought Facebook, which is one of the greatest businesses in the history of the world, trading at six times EBITDA, right, at nine times fully taxed earnings. Mark gave you the blueprint on December of 22. He said we're going to get fit. You knew how much they could save. You knew how much margins could expand. You knew how much he was spending in the Metaverse. So that is a puzzle I love putting together and I love the conversations I get to have with some of these incredible founders who are still running some of the biggest businesses on the planet.

But we've actually, relative to our peers, we've done very little in AI so far this year. And I have to tell you, it's very hard because I'm as enthusiastic about all these companies and founders at this moment in time as anybody. But at the same time, when I look at some of these opportunities, let's just take OpenAI, because I think it has the best team in the business. GPT4 is clearly the best if not one of two of the best models that exist out there today. But if you're asking me to invest at $85 billion to buy common stock, that's capped on the upside, right?

Now we have a deal. It's like a poker hand. And I can say, okay, I know what my upside max potential is. Right? I can then plot a distribution of likely outcomes. How many companies have ever in the history of Silicon Valley been worth durably more than $100 billion? What do I have to believe to be true?

Okay, so they have two businesses, an enterprise business and a consumer business. Well, in the enterprise business, they have to go head-to-head with Microsoft, right. They have to go head-to-head with Google, head-to-head with AWS. So we went out and we talked to 200 customers and just like, what are the customers telling us? Right? And the customers are telling us we can't send our data to OpenAI. Right? Like, a public company can't do that. So we're going to work with Azure OpenAI, which, by our accounts, benefits Microsoft, and doesn't benefit OpenAI? Or we're going to stay at Amazon or we're going to work with Snowflake.

We have to apply some discount rate to our model. About how big are they likely to be in enterprise, how hard is it to challenge the biggest incumbents? You got to go build a sales force like who's building it? On the consumer side, which is where I get a lot more excited. about OpenAI, is I say you are the verb. I would rename the company ChatGPT. I would say we're not going to spend a minute thinking about enterprise, we're going to recut our deal with Microsoft and just get some of the goodies that come out of enterprise. And I would go all in on building the world's best fiduciary agent to serve all of us in a way better than ten blue links, right? Because there's a trillion-dollar prize if you can displace Google at the top of the funnel.

But it is going to be a heroic battle. And Google's not asleep at the switch and Meta is not asleep at the switch. And you got new incumbents like Inflection, you got folks like TikTok. You know, Bing is building a really good agent. So there are going to be a lot of people in this game, right? And if you want to win that game, which is the biggest prize in all of technology, you better be singularly focused with the best team in the world and building the best model.

So, again, at that, I like all these guys, they're friends, I have deep respect and it's really hard to say no, but at the end of the day, I'm building a portfolio to manage investment returns, right? And we've compounded at some of the highest rates, I think, on the public markets and in the venture markets now for 15 years. And I didn't get there by just investing in things that I thought would be sexy to have in my portfolio or logos that I wanted to collect. I have to see math that ultimately works.

And I reserve the right to change my mind on OpenAI if we learn additional facts or if the price was less, if the conversation today was 10 billion or 20 billion, if I didn't have uncapped or I didn't have a capped upside, if I wasn't buying a common security. I mean, there are a lot of things that you could change about that equation that would absolutely send me running in the direction of wanting to invest in them.

And so you can go through everything in the AI stack and I would say I wish we had put more money into the ground. It's been difficult to find things that I thought had an equitable distribution of risk and reward in AI. But I do think coming out of the summer we're starting to see a bit of a reset in AI because ultimately these things are like if you're a software-oriented company in AI, you ultimately are a software company with software like economics that has to sell into the public markets. And we know what those multiples are. If you're a consumer-facing company, then we've invested in lots of consumer companies over the years, from Google to Booking to Zillow to Bytedance/TikTok to Facebook. We know the way that works, so we can just pencil it out.

And so my suspicion is that some of the lather will go away over the course of the next year or two. But in the short run, we're probably overestimating it. And we're in a bit of a little bit of hypishness around prices, but over the medium to long term, we're probably underestimating it. The winners, trust me, the winners of this will generate a lot more than $100 billion of enterprise value from the prize that they're going to win. There are trillion dollars in prizes to be won by the winners in this and so earlier stage, if you're investing like this guy's lucky enough to in the earliest stage, those are a different type of bet, different size checks.

But as you scale up and the price scales up, you have to reorient. So an angel check, where I'm writing 50 or $500,000 into a company, the only thing I'm underwriting there are these killers. And it's a generally good idea. And so you could place a bunch of those bets on the board. But if you're writing a 100 million or a 300 million dollar check into a company, then you have to know that there's, I don't even call them venture companies anymore at that point, I call them quasi public because they get shopped around to a bunch of people, right? We all have big stacks, we can all write the checks. And so at that point in time, there's much less arbitrage in the market, right? The market becomes way more efficient. And now I have to compete. Just like in the public market, it only takes one person to price that last security. And so I don't know who that person is. I have to price it myself and say, are we set up asymmetrically to win?

And so hold these two truths that are intentioned just like they were in 98, 99. This will be bigger. It will have more impact on humanity than probably everything that came before. But it doesn't mean you have to own it all now. This will play out over decades, not days. And so that's the way we're thinking about it. It's been a lot of fun and I envy the stage at which you guys all are. I'd pay a lot of money to go back and play 20 years over again. But follow the InvestAmerica thing. Get in, let your voice be heard. And thanks for having me.

Elad Gil

Thanks so much.

MY BOOK
You can order the High Growth Handbook here. Or read it online for free.

OTHER POSTS

Firesides & Podcasts

• Apoorva Metha: Starting Instacart

• Satya Nadella: Building Microsoft

• Clem Delangue: Hugging Face, Open Source, AI

• Dylan Field: Figma, AI & Design, Education

• Reid Hoffman on AI, Big Tech, and Society

• Sam Altman, CEO OpenAI

• Emad Mostaque, Stability.AI

• NoPriors AI Podcast

Markets:

• 2023 Unicorn Market Cap & Rise of AI

• AI Regulation

• Early Days of AI

• AI Safety: Technology vs Species Threats

• Startup Decoupling and Reckoning

• Defensibility and Competition

• AI Platforms, Markets, and Open Source

• Changing times (or, why is every layoff 10-15%?)

• AI Startup Vs Incumbent Value

• AI Revolution - Transformers and Large Language Models 

• Startup Markets Summer 2022

• False Narrative Around Theranos

• Index Companies

• Defense Tech

• Collaborative Enterprise

• Industry Towns: Where you start a company matters

• Markets are 10X Bigger

• Hot Markets 2019

• End of Cycle?

• Machine Learning Startups

• 3 Types Of Platform Companies

• Defensibility and Lock-In: Uber and Lyft

• Uber And Disruption

• Who Cares If Its Been Tried Before?

• The Road To $5 Billion Is A Long One

• How To Win As Second Mover

• End Of Silicon Valley

• Social Products

• Hot Markets For 2015

Startup life

• Capital Efficient Businesses

• Startups Are An Act of Desperation

• Back To The Office

• Hiring Executives and Bad Advice

• When executives break

• Fear of Sales

• A brief guide to startup pivots

• 4 Ways Startups Fail

• Founder Investors and Scout Programs

• Better Meetings

• Magic Startup Moments

• Founder Investors & Scout Programs

• Jobs, Wozniak, Cook

Co-Founders

• How To Choose A Co-Founder

• Unequal Cofounders

• How To Fire A Co-Founder

• Founders Should Divide and Conquer

Raising Money

• Preemptive rounds

• Don't Ask For Too Much Money

• Building VC Relationships

• Founders Should Divide And Conquer

• Lead VC Vetos

• What Is A Good VC?

• How To Choose The Right VC For You

• Signs a VC Just Isn't That Into You

• Series A Crunch

• Questions VCs Will Ask You

• How To Raise A Successful VC Round

• Differences Between Funding Rounds: Series Seed, A, B, C...

• Financing Approaches Most Likely To Kill Your Company

• Party Rounds: How to Get A High Valuation For Your Seed Startup

• 20 Questions To Ask Yourself Before Raising Money

• The 7 Types Of Angel Investors

• Fundraising Will Take You 3 Months

• How To Sell Secondary Stock

Capital Efficient Businesses

2023-12-05 06:25:10

For the last 5 decades[1], every wave of technology-driven companies has had amongst them highly capital efficient businesses. The capital efficiency tends to reflect that (a) people really want your product and will pay you for it and (b) the founders are cost-conscious and frugal, and do not overhire.

Indeed, Paul Graham from YC has developed the metric of “default alive” to reflect capital efficiency as a core sort of startup metric.

Capital efficiency has existed in roughly every technology wave. Many of the largest, more important companies in the world started off highly capital efficient. Inded, capital efficiency tends to reflect an especially strong business model (and in some cases founder). Examples include:

• Microsoft was bootstrapped in the 1970s and did not raise any venture capital until a round right prior to their IPO, when Bill Gates wanted a VC for his board and that came with strings attached for the VC to invest. The investment went straight into the bank account and remained untouched.

• Dell was bootstrapped off cash flow in the 1980s until a similar pre-IPO round.

• Yahoo and eBay famously did not touch their early venture capital funding in the 1990s - they were run so lean and profitably that they did not need to raise large sums.

• Google raised a single round of traditional venture capital, before doing a pre-IPO round with Yahoo! and others.

• Instagram was just a dozen or so people before being acquired by Facebook, while Zapier only ever raised a $1.3M seed round before bootstrapping from then on. These companies were founded in 2010s on.

• Midjourney (founded in 2020s) is rumored to be entirely bootstrapped today.

Why are some businesses capital efficient?

In general there tend to be two drivers of capital efficiency.

1. Customers will pay (a lot) for the product. The “capital” side of capital efficiency is often a proxy for both product / market fit and an intense customer need. Customers are willing to pay up for a product that is important to them, and there is insufficient competition in the market to commoditize pricing or destroy the category (so the product is somehow differentiated). Pricing is often a proxy for value & differentiation of a product.

2. The company is run efficiently. During COVID roughly all tech startups lost their way on spending. Capital was flowing freely and teams often rapidly and dramatically over hired, boosted expenses on things non-crucial for the business, and spent wastefully. The most capital efficient businesses tend to be frugal and have a low cost approach to the world. Salaries are lower to help make equity more valuable. The founders and employees of these businesses treat the dollar spent by the business as their own money (which it is, as they are shareholders in the business). They realize that profitability gives them infinite runway and enormous freedom on decision making and future path optionality.

Frugality has felt like a lost art over the last few years - hopefully it is recovered.

When to bootstrap?

Too few silicon valley (or NY or other cluster) based technology companies bootstrap.

1. If you can grow organically and optimally without hiring a massive team and increasing expenses it is great to do so!

2. If your company is a cash versus equity business, you should bootstrap.

3. If your company is growing slowly and will never hit venture scale, you should bootstrap.

When to raise money?

Venture capital is typically used to either:

1. Build out or prototype something. This may be something inexpensive but for some reason can not be bootstrapped off of customers (e.g. a new SaaS product) or is something capital intensive that may have a giant market on the other side. The later includes things like building rockets for spaceX, or biotech drugs.

2. Scale something that is working. For example, you want to add sales or go-to-market functions to sell faster/better, or your consumer app is growing like crazy and you want to be able to add more compute to serve users. Uber needed to raise billions to both scale rapidly and beat out global competition.

3. You need the valuation for external uses. E.g. M&A or hiring (there are other ways to do this too).

In general, if you are not prototyping / proving something works, or scaling something that does work, you should not raise money.

One could argue that while too many SV/NY/cluster-based tech companies raise money, too few outside of major tech clusters do. In many cities and regions people bootstrap for too long, do not scale quickly enough, or do not think about time to winning in a big market. It is possible that non-cluster tech companies in the US end up scaling fast too infrequently.

The dangers of capital efficiency

Occasionally, you also see an SV/NY/cluster-based company that is growing really well and has turned profitable, and then forgoes building against and winning in their category. Sometimes this is the right thing for founders to do, and sometimes it reflects a lack of know-how, ambition, or aggressiveness. Sometimes, it just shows the founders had a bad experience at a company that scaled for no good reason and ruined the company culture, ability to execute, and products. The wrong lessons may be learned from bad growth and bad execution. It is so rare to actually build something that people care about, that it feels like a shame to not go win when you can - but obviously it is up to each founder and team to chose their own path. Some companies that focus on or hit profitability early then forgo winning the market - their focus shifts too much on maintaining cash flow versus growing faster to take the market.

Generically, startups are rewarded for progress per unit time, versus progress per unit dollar (all else being equal within a given burn multiple range).

Hopefully more capital efficiency returns to technology now that ZIRP and COVID policies are (roughly) a thing of the past. Many of the most important companies in the world started in a capital efficient state.

NOTES

[1] An emerging meme today is that with AI, the costs of starting a company are lower as you can augment people with AI from day 1. Past variations of this thought has occurred for many macro waves including distributed talent (“talent in India is cheaper, so your company can be too!”), cloud services (“cheaper to not build your own data center!”) and other shifts.

Notably all these shifts have impacted the costs of doing business. However, the impact tends to come with scale and the need for more people intensive operations. I am very bullish on e.g. AI applied to private equity and buyout models.

MY BOOK
You can order the High Growth Handbook here. Or read it online for free.

OTHER POSTS

Firesides & Podcasts

• Apoorva Metha: Starting Instacart

• Satya Nadella: Building Microsoft

• Clem Delangue: Hugging Face, Open Source, AI

• Dylan Field: Figma, AI & Design, Education

• Reid Hoffman on AI, Big Tech, and Society

• Sam Altman, CEO OpenAI

• Emad Mostaque, Stability.AI

• NoPriors AI Podcast

Markets:

• 2023 Unicorn Market Cap & Rise of AI

• AI Regulation

• Early Days of AI

• AI Safety: Technology vs Species Threats

• Startup Decoupling and Reckoning

• Defensibility and Competition

• AI Platforms, Markets, and Open Source

• Changing times (or, why is every layoff 10-15%?)

• AI Startup Vs Incumbent Value

• AI Revolution - Transformers and Large Language Models 

• Startup Markets Summer 2022

• False Narrative Around Theranos

• Index Companies

• Defense Tech

• Collaborative Enterprise

• Industry Towns: Where you start a company matters

• Markets are 10X Bigger

• Hot Markets 2019

• End of Cycle?

• Machine Learning Startups

• 3 Types Of Platform Companies

• Defensibility and Lock-In: Uber and Lyft

• Uber And Disruption

• Who Cares If Its Been Tried Before?

• The Road To $5 Billion Is A Long One

• How To Win As Second Mover

• End Of Silicon Valley

• Social Products

• Hot Markets For 2015

Startup life

• Startups Are An Act of Desperation

• Back To The Office

• Hiring Executives and Bad Advice

• When executives break

• Fear of Sales

• A brief guide to startup pivots

• 4 Ways Startups Fail

• Founder Investors and Scout Programs

• Better Meetings

• Magic Startup Moments

• Founder Investors & Scout Programs

• Jobs, Wozniak, Cook

Co-Founders

• How To Choose A Co-Founder

• Unequal Cofounders

• How To Fire A Co-Founder

• Founders Should Divide and Conquer

Raising Money

• Preemptive rounds

• Don't Ask For Too Much Money

• Building VC Relationships

• Founders Should Divide And Conquer

• Lead VC Vetos

• What Is A Good VC?

• How To Choose The Right VC For You

• Signs a VC Just Isn't That Into You

• Series A Crunch

• Questions VCs Will Ask You

• How To Raise A Successful VC Round

• Differences Between Funding Rounds: Series Seed, A, B, C...

• Financing Approaches Most Likely To Kill Your Company

• Party Rounds: How to Get A High Valuation For Your Seed Startup

• 20 Questions To Ask Yourself Before Raising Money

• The 7 Types Of Angel Investors

• Fundraising Will Take You 3 Months

• How To Sell Secondary Stock