TRANSCRIPT
Welcome to Vincent's pre-IPO briefing on OpenAI. Thank you everybody for joining us today. Thank you to our panelists here, Jan-Erik, who we'll get to in a second. My name's Eric Cantor. I'm the CEO of Vincent. We help individual investors navigate private markets and we're hoping to accomplish that mission to some degree today. I'm here with Jan-Erik Asplund co-founder of Sacra, the best private markets research company that we've found.
We're gonna talk about OpenAI. This is a business that's at the forefront of the industry, the forefront of the news, one that many of the almost 100,000 investors in Vincent's network today are interested in looking at. Some who already hold it and are thinking about whether it's maybe a time for liquidity soon, others who are considering a buy and others who are just along for the ride. Today's session, we're gonna keep it real concise to the point.
First, we'll walk through an overview of the OpenAI business as it stands today, little bit of background. Second, we're going to discuss its future prospects. Where could the business go? What are the implications for multiples, valuations, IPO? And third, we're going to open it up to questions. You, online, provided a number of questions before the call. I'm sure some more things are going to occur to you as you hear Jan's presentation.
So please, as things come to mind, just enter them in the Zoom app below. We will try to get to them time permitting. Before we start and kick this off officially, just one more reminder, nothing you're going to see here today is financial advice, right? You should be making decisions, consulting with your advisor based on your own situation, buy, sell, or otherwise. Just want to make sure that's clear. Great. So now let's dig in. Jan-Erik, welcome.
Tell us about OpenAI. Sweet, thanks Eric. Yeah, awesome to be here. And thank you so much for the very kind introduction. this is obviously one of the most exciting companies that's private right now and one that we've covered probably more than any other company over the last two years. So really excited to chat. Just start kind of a quick overview. If you don't know, OpenAI is the leading commercial AI company
The big breakout moment for them was really the launch of ChatGPT, which was two years ago this past Saturday, hit hundred million users in two months, making it the fastest growing consumer product ever in history. And today has 250 million weekly users. They hit 4 billion revenue run rate in September, projecting to hit 5 billion for the end of the year, which would be over 200 % year over year growth from roughly 1.6 billion at the end of last year.
diving deeper into kind of the revenue numbers that I just mentioned, it's a 75-25 split, roughly. 75 % of OpenAI's revenue comes from ChatGPT and 25 % comes from API. So just to break those down a little bit more, ChatGPT, you've probably used it to chatbot. There's a consumer use case and a B2B use case. So you might use it to write you a rap or a poem, but businesses can...
also use it enterprise plans where they can put proprietary documents into it and chat with them, do research, that kind of thing. So that's really the big cash cow. And then the API is kind of how businesses build custom apps using OpenAI's large language models. So for example, Klarna is one that's been in the news a lot as one that they do a lot of their customer service with bots that they put together with OpenAI APIs. And there's quite a few of these other names.
like that that are using OpenAI APIs in their products today, including Salesforce. So that is something that they're obviously pursuing as well. They have a big sales team sort of trying to sell these deals into big enterprises like Salesforce. On the other side, losses, this has been a big kind of topic around OpenAI. Huge losses, 3 billion net losses in first half of 2024.
The upside of first half of this year was that they only had cash burn about 340 million, which was less than they expected. And so they ended the first half with a billion dollars on their balance sheet. A lot of this comes back down to the economics of their deal with Microsoft too.
which is really a core part of OpenAI. So they did this deal with Microsoft, which started almost at the inception of OpenAI, but has gotten more more big over time. And the upshot is basically that OpenAI by working with Microsoft gets drastically subsidized compute. The recent lawsuit that you can read reveals that they were getting basically $50 million worth of Azure compute for every $10 million they paid.
And this was at the time seen as a really a life or death move for OpenAI if they wanted to be able to compete with Google specifically.
So far they've raised about 18 billion in total, much of that 14 billion or so from Microsoft valued at 157 billion. Some of the top companies and funds are on their cap table. One interesting point about OpenAI we'll talk about more is that they were a nonprofit at inception. They shifted to a capped profit structure in 2019 where investors get a hundred X return on their investment.
And there's reporting now that they're transitioning into a traditional, very investor-friendly, for-profit, normal corporate structure. But we'll talk about that more.
Cool. And then on the secondary market, it's interesting prices really dropped in a big way at the end of 2023. This is when Sam Altman was fired by the board. It's hard to imagine almost, but there was a genuine panic around the company. There was a big planned tender offer, which was canceled. Microsoft was reportedly questioning the whole partnership and the stock was as a result on this sale.
implied valuation of about 92 billion. So since then, the share price on the secondary market has climbed back up to 280. So that's roughly a 30X multiple on 4 billion in ARR as of September. So basically 160 billion just above the last primary valuation and about double, almost double since all that drama went down.
Amazing founding journey to get to where we got to here for this company. It's an extraordinary story. mean, putting this all together, mean, keeping in mind that this company has doubled since the board governor, the CEO. But if you take all these things that you just talked about, like, know, board firing founder, CEO, capped profits, nonprofit.
executives leaving, coming, all this stuff. If you put that in a normal startup, like any company we've seen for the last 10 years, investors would think that I can't touch this. This is, mean, and also a strategic deal with it, a significant strategic player in the space, Microsoft that gives up a lot of future options for your path. All of this would be a no-no, but in this case, seems to almost like.
propel forward the narrative about this company. I do you think as investors looking at this, all of this, I mean, I want to say the word hair, because in a normal business, you'd say there's so much hair in this deal. This is so complicated. Is this somehow creating like a story that makes people want to invest more? Is this a risk? I if you were going to buy this company at this price and you thought it was a good price, you need to discount it based on all of this sort of drama? Or how do we think about that as investors?
Yeah, it's really interesting. And there's so many other kind of, it's probably one of the most dramatic companies out there in tech. You know, another aspect is all the high profile departures, right? You've had a founding team of roughly a dozen people and half of them now work at Anthropic. And, you know, a fourth of them have started their own companies to try to build artificial general intelligence before OpenAI does.
It all adds up to this crazy kaleidoscope of factors. I think one takeaway for me is that they really speed ran the process of building a huge tech company. The speed at which they went from... The company was founded in 2016, but when the launch of Chet GPT is when they began as a legitimate product tech company. And they've gone accelerated so fast and the company has changed so much.
that in a way that they've compressed a decade of drama into two years. So I think that is part of it. And I think there is reason to look at that and question especially around the corporate structure, which I think is why they're trying now so fast to accelerate to a for-profit structure. That's definitely a key one. But I think a lot of it is to the kind of unplannedness of it all.
like check GPT was never meant to be a Google search product. It was a tech demo, which they released on a whim and took off. And so I think there was a lot of, they were not expecting this to happen. So the mission of the company continues to be AGI, artificial general intelligence. Do you think that's, is that just a rallying cry, like a big hairy vision, or is that really what
all these pieces that they're putting in place are building towards. And does that again matter for investors? if they get there not?
Yeah, it's a good question. think one aspect of it is that there is a religious motivating factor behind the company and behind the now 2000 people they have there. you you can't really ask for a better, big, ambitious mission to drive, you know, 2000 people towards a goal. So I think it's very powerful from a sort of company culture perspective to have folks focused on a
massive vision like that. I don't think they necessarily need to usher in the singularity and develop a self-improving AI that's better than humans at everything in every domain for investors to get a return. Because I mean, we're already seeing the adoption of AI in both B2B and B2C contexts, far outpacing the early adoption of the internet or cloud or mobile.
and generating real revenue and cost savings. So I'm not going to speculate necessarily on whether AGI is near. It could be. And that's certainly something they talk a lot about and what a lot of the super bull case around open AIs based on, but I don't think it's a prerequisite to building a successful company. Cool. We're going to get to that bull case in a second. I just have one more question before we start looking forward.
which is really about, you mentioned the costs, right? The company is doing 4 billion, but they're spending like triple that. The Microsoft deal was really because everything's so expensive, I need some strategic deal to just be able to build. I mean, does this company get profitable? Is there a level of scale or is it the tech is gonna advance like solar has and at some point in the future, this stuff is free or cheap?
Or is there a chance that, you the business could be amazing, but this company will never figure out how make money and therefore like Wall Street will never accept it.
Yeah, it's certainly possible. The company itself projects profits about four years out from here, four or five years out. So they have a model built around the idea that they will generate profits one day. A lot of the costs so far has been the cost of doing the pre-training on these massive models, which requires hundreds of thousands of GPUs, which are very supply constrained because there's one company that
is basically the bottleneck for this entire industry. So a lot of the work that you see happening at Google and also OpenAI is on trying to unlock that bottleneck and have custom shifts. Google already has custom shifts. OpenAI has been poaching engineers to work on their own hardware. You see a lot of effort from the sort of hyperscalers, Microsoft, Amazon, Google, to work on this. So there is a...
vision on the hardware side for this to get cheaper. There's also the product side is very unclear whether more and more GPUs, i.e. mounting losses, is the way to get to a better model. And it's looking like that might not be the case. I think the cost structure is really hard to project out when you're dealing with so much uncertainty.
Got it. So you're getting us into the structure of the industry that's evolving and where the company could go. So why don't you tell us what lies ahead for open AI potentially. Yeah, let's do it. So we saw this with computing in general. We saw this with the internet and mobile, that the big infrastructure players capture a lot of the initial value that's created. like Cisco, for example. And then the big value in the long run.
goes away from your Cisco's and towards your Apple's, Google's, Facebook's, Microsoft's that are really aggregating users and aggregating value around their products and their platforms. We see that today with Nvidia, also companies like CoreWeave who are sort of the initial beneficiaries in a big way of AI, also with the sort of hyperscaler clouds that make the LLMs accessible to third parties.
In the long run, we see that value similarly going out to the platforms and apps that aggregate people around their use of AI. I think companies that are purely in this model layer are, you know, it's going to be hard to capture value there. You could think of it almost as like the CDNs of the early internet. You know, the first ones were novel. You could get content delivered anywhere, you know, the world quickly. But then we sort of got to parody. There was a lot of competitors. It got commoditized.
And so that's really why a lot of the value of OpenAI, maybe the majority of the value is in the fact that they have chat GPT, which 250 million people every week talk to. And for Anthropic also, the value is that they have Claude, and they also have a state of the art API product. So yeah.
Yeah, to say more about Anthropic. Anthropic has emerged as a big competitor to OpenAI, but if you look at trend data, what people are searching for, chat GPT is so far above every other AI product that it almost makes them insignificant. People don't talk about plotting something, you're talking about chat GPTing something. Google is definitely a big other competitor as they have.
both a line of near state-of-the-art models with Gemini, and they're also sort of working really hard on the hardware side. And then you have Meta, which has their open source models, which are also getting closer and closer to parity.
Yeah, so this chart really puts the challenge in perspective in terms of the scale of the companies that are going up against these NVIDIA's, Meta's, Microsoft's and Google's. But one thing I wanted to point out was that the left y-axis, OpenAI is growing three times faster than Facebook was at IPO and about 50 % faster than Google at IPO. And so if you take that 30x multiple, which sounds high and compare it to Facebook at 28x, it doesn't actually look
all that crazy, you Facebook has gone from a hundred billion at IPO to 1.5 trillion today.
Cool, so what's next for OpenAI, right? ChatGPT, like I said, was basically this product demo. It's not clear that ChatGPT will sort of be their Google search, right? Their way to drive monopoly cashflow, you know, at scale. So they're really hunting right now for that app too. One possibility is that it's ChatGPT but better because right now it's just a chat bot, but every AI lab is working on agents, especially
looking to 2025 and beyond, a chat interface maybe, but that can do things for you, book hotel rooms for you, et cetera. So that is one possibility, but they're also looking at AI hardware. They hired the iPhone designer to build them a device. They're doing search across both consumer and enterprise. They're building a web browser. So they're looking all over for kind of what could be the big.
cash cow of the business in the future.
Cool. And then just to sort of out the thesis, right? The upside case for OpenAI is that they've captured lightning in a bottle with ChatGPT. It's the Kleenex of AI and they have a huge lead. Especially when people start getting the new iPhone and using ChatGPT on mobile, know, the trajectory to 1 billion weekly users is not hard to imagine when you think about being on every iOS device. On the flip side, so...
OpenAI's brand recognition is really good. Google and Meta still have better brand recognition. companies like Facebook, we're not the first mover. So sometimes you can be a second, you can be second or third to the market, be better and sort of learn from the early mistakes that people make and get better. yeah, there's also the risk that there is that sort of AI winter risk, right?
Progress in LLMs has been insanely fast over the last two years, but we are seeing reporting now that there's kind of this plateau on the scaling methods that they've been using to date with LLMs. Part of that is because they are trained on public data, which is running out in the sense that they've already trained basically on all of it. So we are seeing new approaches there, which TBD,
what the results will be. they could have launched GPT-5 while I was talking. So we'll have to check on that. And then lastly, there's this unique corporate structure. So you don't actually own equity in OpenAI. You own profit shares with 100x capped returns, although that cap increases every year. But still, OpenAI hasn't generated profit yet, as you pointed out.
And while there is this plan now to transition, it's not clear that that will necessarily work. There's a lawsuit filed by the head of the Department of Government Efficiency, Elon Musk, to stop that. And the terms of how your shares would convert also is not totally clear.
All right, so we'll end off with some scenarios. Bullcase for open AIs, basically that they're on a Google-like trajectory. They're currently growing a little faster than Google at IPO. And this is the case that they continue to grow at that same rate. And basically hinges on them finding some way to develop monopoly profits. It could be what they're looking at right now. And they're hiring people around this idea of putting ads in chat GPT.
in some way, shape or form. It'll have to be different from what Google does, obviously. Then there's also this hyper-bull case, like I'll it, because in a way, this is a very binary investment. There is a scenario, and this depends a little bit on your level of belief in AGI. But if OpenAI was the first lab to develop AGI,
In other words, in artificial intelligence that surpasses human ability in a generalized range of domains, that could be the biggest business of all time. It could be generating its own businesses, which then each one generates monopolies in industries that we don't even know about. that is kind of the high upside, perhaps more unlikely bit. Although I will say for both the bull and hyperbull case,
there is this possibility that the company succeeds and most investors don't succeed if this capped upside holds. But then the base case we put together was basically that OpenAI remains this number one, number two, foundation model company. Chatt GPT maintains kind of this position of supremacy as an AI product. OpenAI hits 15 billion in ARR.
over the next five years. But it's not quite the speed of advancement that they're talking about and that investors are hoping for. And then the bear case. The bear case is basically that we are hitting the limits on what LLMs and what AI in this paradigm can do. And in this world, we see these other open source models and other company models catching up
possibly surpassing OpenAI. And what you get basically then is that the hyperscalers that have all the compute and the resources end up capturing most of the value. Or the alternative view on the bear case is basically that it just keeps costing more and more to train these models and revenue doesn't catch up to losses. So those are kind of the basic scenarios.
tough company to do scenario analysis for, but I think the binary nature of the bet is definitely key. And yeah, that's about it. yeah, ready for some questions. Great. Thank you for all that information, trying to synthesize and process it. We've got a number of investors on the call, so ideally they can drop some more questions in, but we've collected a few. Let me try to run through as many as we can. You talked about Elon Musk.
you know, who's now we have emails coming out about Elon's vision for AI and how the founding of the company went down. And now, of course, there's a lawsuit. also got Trump now moving into the White House, which could affect policy. mean, all these personalities around and in competitive roles and governing roles is. Do you think that has implications for this business or is it just the ship just keeps moving forward regardless of the the noise around it?
Yeah, that's a good point. One thing that's really hard is denoising it because there's so much noise around the administration. Elon also has a competing AI company, XAI, which I think might be, I think I might've seen there at like hundred million revenue run rate. they're further behind, but they're obviously well positioned with Elon's family of companies to approach being a
being a real competitor. You know, there's been reporting that, you know, OpenAI, and they've talked about this, that they're in deep discussions with the Trump White House, you know, of pre-election and post-election, because the, you know, the Trump line on AI seems to be that we need to win and China can't win, right? And so as part of that, OpenAI is a crucial partner.
because they are the leaders. I see it as just like OpenAI has basically been leading the space from the beginning. They kicked it off and they will continue to be important, especially in a sort of security, national support context because of that and because of the importance of maintaining kind of domestic, maintaining our supremacy in AI. So I don't think that,
Elon and Doge will necessarily be sort of sufficient to destroy OpenAI.
Yeah, I mean, it does seem like this sector of all the sectors in tech has this likelihood of becoming tightly linked to national security, which probably restricts the ability of new people to enter. And it probably, again, reinforces the big players, which I think, like you said, leapfrogged into that, even though they're not part of the FAANG. mean, they might be.
We'll get to that one in a second. Let's just talk more about the competitive. If we can just grab that competitive slide again, showing the growth rates of Anthropic. I mean, long story short, this is a highly competitive market with tough competitors. OpenAI is one of them. Is there room, maybe the other slide with those bubbles for the growth rate that you showed, is there room
for, you know, how many these companies can win, I guess? Is there a TAM where all these companies do a trillion dollars and everyone's happy or is somebody have to be on the short end of this stick? And I guess the defensibility of each business model is part of that question. Can you compete, get ahead and win here? And how many of these, at least the upstarts will still be here in five years?
Yeah, I think just building a better or just building a foundation model, you know, what OpenAI and Anthropec started with is it seems unlikely to be sufficient because we've seen meta release an open source model, is, you know, trained on the same vague number of parameters and it has near, you know, it's almost as good as your GPT-4 or CLOD 3 models. So
I think just building a model is unlikely to be sufficient, I think you are seeing kind of the, you're seeing the battle lines of cloud replicate in AI. So obviously Microsoft saw that they were falling behind Google correctly and acquired this huge stake in open AI so that they would be able to catch up. You're seeing Amazon doing the same kind of deal now with Anthropic, which they've identified as.
most likely their way to stay competitive on AI. And the same way that you're getting Apple's putting chat between their phones, there's discussions now and plans to put Anthropix models into your, what's it called? Amazon, know, Echo or Alexa, the thing you put on your kitchen counter and talk to. So I think a lot of it will hinge on whether or not these companies can build
or get embedded into products that have huge reach and huge usage, and also maybe a hyperscaler with a lot of compute to support them. But standalone model seems challenging to sustain.
We have time for two more questions. Let's jump to one that just came in from the investors watching. Do you have a sense from Vinay, do you have a sense of how well the ChatgPee subscription for the enterprise users are doing? Are there specific types of enterprise customers that choose the subscription model over the Azure OpenAI solution? I think that this is really about sales. Who's buying the product, who's using it? And I think what Vinay is getting is that there was quite a novelty factor initially, right? know.
myself purchased it when I first came out to see what is it all about. And that, sounds like 75 % of the revenue is still, meaning a $3 billion run rate of people that are just paying $20 to do small things, maybe consolidate some memos, write a poem for their partner's birthday, this or that. And then there's this enterprise segment. And I had heard that they just ramped up to like 300 salespeople, meaning there are some
big intense sales activities going. I'd also heard that a lot of enterprise customers are still in the novelty phase. They're playing with it, but they're either doing little plugins, maybe in my product, you can do some usage of chat GPT, but nothing that's like super mission critical for the most part. Any comments on that? mean, just because it does seem like the growth of the API business depends on nailing some use cases there. So any thoughts on that or the point that he raised about?
Azure versus Subscribing Direct.
Yeah, definitely. So on Azure, you know, there's definitely a dynamic where companies that need a set of security assurances, it's very similar with Anthropic and using them through AWS or directly. There's both a security reasoning to use Azure if you already have existing, you know, agreements with Microsoft and potentially, you know, doing an agreement with a company like OpenAI would be challenging.
And then there's a sort of a latency question as well. it's going to, know, companies can find that enterprise scalers better performance going through kind of the Amazon or, or Microsoft versus going directly through the, the, the open AI or Anthropic. So that's one element. I think on the chat GPT enterprise, think the strength that they have is that they were first and you know, it was novelty, but also that has a durability and
You know, we've seen all these PwC and other firms that are doing these massive consulting rollouts where OpenAI is basically doing this channel partner sales through firms that help companies implement ChatGPT. And if you're familiar with companies like Glean, Harvey, these are building highly vertical AI apps for the enterprise. Glean is like
is like a chat GPT like product that integrates into all of your SaaS tools. So you can search across all of them and take action. And that stuff is really powerful and looks competitive and actually is competitive with chat GPT enterprise, but it's also niche and chat GPT is the thing that everyone knows. And so I think it's strength is that, and that will sort of give it some staying power and give them time to make it better, essentially.
because they can't really, you can't rest on your lores there. It's amazing listening to what you're saying just to realize how early we are in the evolution of this. And yet we've got a business here run rating $4 billion. So it's like, that's just pilots of tests and novelty. Imagine where they can get to when the real use cases start to get firmed up in the enterprise. All right, we have one more question. So let's get the slide on the...
scenarios again, because you outlined a, let's call it like two bull cases, right? We have like a 3X and a 10X, right? We're talking about price today in the 150 billion range, maybe 157. So you got 3X on the bull case that gets us to 450 would be a good pre IPO investment though, maybe not as game changing as some are hoping for. And then you've got like the hyper bowl, which is like a 10X talking about a $4 trillion company.
Where does the IPO come into play here? Because of them, I mean, one of the things I'm thinking of, if I want to buy this and, I can, I can say, think this thing has a great chance of the hyper bowl. want to, I want to buy it. Cause I know I can 10 X and my only downside is one X. So I think it's an asymmetrical, I think you said binary play, but when can I expect liquidity? mean, what would be a reasonable time or set of considerations for a business like this to actually go public?
Yeah, it's a really good question, especially thinking back to, we've talked about Google a bunch. And the thing with Google is that they went public in 2004. if you bought it at IPO and you held it until now, so, you know, 20 years, that would have captured 95 % of the value or more that was ever created in internet search if you just bought it at IPO. So, you know, in a way that's a case against venture.
but also we're living in a different world. Companies that going public, that companies that are going public often have some, you know, there's some external reason why they have to. OpenAI is not in that position. They have unlimited access to capital effectively in the private markets. They haven't had any issues raising money so far. And...
It's a company that everyone wants to invest money in. I don't see any external pressure coming to make them go public anytime soon. But if you listen to OpenAI investors at funds, they do talk about, they're eager for it to go public. There's a narrative that it would be good for a company that's this kind of potentially systemically important to the country to...
allow access to regular investors, but it's really hard to sort of predict because of the way that private markets work and the fact that they don't need to go to the public markets to raise money. yeah. It's a problem. How much they raised actually? How much have they raised? Yeah, I think like around 18 billion. Right. So an IPO is going to be a massive IPO, maybe the biggest ever.
So, you know, we'll have to look forward to that. It sounds like you're not thinking that's 2025, but.
Seems unlikely, but in the next two or three years, maybe. Yeah, it's really hard to say. I think it would be. I mean, I just wonder if there would be, you know, this kind of pressure as a company that's so important to expose, you know, to go to go public and be transparent. I don't know if it would come from the government, but. I think, yeah, I'm really excited for the S1 to drop and.
I think it'll be definitely an event, but it doesn't feel like there's any really mounting pressure to go public. So it's hard to say, but maybe three years, maybe.
All right, well, this has been a great discussion. We've gotten a lot of information here. Gonna keep thinking about it and processing it, but for now we're gonna sign off. So I just wanna thank you, It's a great, great presentation. Thanks to everybody who joined us and stuck with us for the last half hour or so. We will follow up with a recording via email and we will hope to see you at our next pre-IPO briefing. Thank you and have a great day.
