Anthropic: The Fastest Growing Company in History

Artificial intelligence is changing business at an astonishing pace—but what does that actually mean for investors?

In this episode of Beyond the Noise, Max Clark and Josh Renfro examine the extraordinary rise of Anthropic, one of the fastest-growing software companies in history, and explore why its growth may represent a defining moment in the broader AI investment cycle.

Beyond the headline numbers, they discuss what investors should really be watching: enterprise adoption, recurring revenue, valuation, profitability, competitive positioning, capital intensity, and the enormous infrastructure required to power the next generation of AI.

They also explore the risks—including soaring compute costs, open-source competition, data security, and whether today's AI leaders can maintain their advantage over the coming decade.

Rather than asking "Which AI company should I buy?" this conversation asks the more important question: What does this moment tell us about where technology, business, and capital markets are headed next?

Transcript

Welcome to Beyond the Noise

Speaker 0:00

We're excited to be kicking off our new podcast Beyond the Noise. My name is Max Clark. I am one of the senior wealth advisors and partners at the Fox Alliance, also a member of our investment and finance committee. I'm joined here by Josh Renfro, our chief investment officer and one of the managing partners at the firm as well. And we're excited about the topics and things that we're going to be covering on this podcast. There's a couple of things that you're going to be seeing from us as we move forward here. The first is we want to cover topics of the day that in the sea of headlines out there are what we believe the most important things for our listeners to be focused on. And then the second is going to be a series on various investment topics that we believe are most valuable for our investors to be focused on and to be educated about as they go into their investing lives. Now, our first episode today, Josh, is going to be focused on AI and everything that's going on on that front. And specifically, we're going to be um diving into the details of one specific company, which has been in the news a lot recently, and that's Anthropic. That's right. Anthropic has seen some explosive growth. But before we get into that, I want to set the stage a little bit by talking about where they came from and you know what's kind of led to their presence in today's environment. So if we think about the founding of the company, it's led by the CEO Dario Amade. And he used to be the VP of research at their big rival, OpenAI. And back in 2020, he made the decision, along with his sister and I don't know, 10 or 12 other of the key players there at OpenAI to leave and to start anthropic. And from what we've been able to tell, there certainly was a bit of drama.

Speaker 2 1:43

Shark, Sam Alton

Why Anthropic Matters

Speaker 2 1:45

and Dario would have differences and uh and not like each other necessarily.

Speaker 1:49

Hard to believe. Hard to believe, honestly. Uh but it sounded like kind of the two main drivers there were one, it sounded like they kind of had a personal beef between one another. Uh Sam probably felt like he couldn't trust Dario, and Dario felt the same way about Sam. But then also, Dario had some concerns about security and safety, and obviously we've seen that become very apparent in the way that he is running uh Anthropic. But that started back in 2021. May of 2021 is when the company Anthropic was founded. And there is a pretty broad landscape of players in the AI world today. Maybe you could talk to us a little bit about that and how anthropic differentiates itself in that broad sea of players.

Speaker 2 2:32

Yeah,

Mapping the AI Landscape

Speaker 2 2:33

so if we look today at the sea of AI players that are out there, specifically within large language models, which is really what we're talking about here, you have Open AI, which of course is where most people first got exposure to AI in the first place. They, you know, ChatGPT. Yes, ChatGPT has become the verb, just like Google was the verb for the search engine era. And so um ChatGPT focuses on uh the retail customer, the average American. You know, if a person's gonna have a free version, there's a pretty good chance it's ChatGPT. Uh within that C, you also have Anthropic. Anthropic's version of ChatGPT is clawed. Um, you probably have seen some of their commercials, maybe you've even seen some of their Super Bowl commercials uh that they ultimately ended up doing, making fun of OpenAI and the ads that they were going to start doing. Um, you also have XAI, which is Elon Musk's um attempt at in the large language model race. They have been struggling. In fact, in Elon's own words, um they basically have had to start over from from scratch.

Speaker 3:32

You have to Well it's interesting with SpaceX going public, XAI is part of that. They I saw they just renamed their company to SpaceX AI.

Speaker 2 3:40

Yep, exactly. So you've got um XAI, not right now. You when also within this context, we have what are called frontier models, which are models that are on the leading edge. They're the best of the best. There are a number of different categories that they compete with right now. Uh both Claude and OpenAI fit into that on the to the frontier categories. Um XAI would not, or if it does, it's only touching one or two of those parameters. Um you have Meta's OpenAI or an OpenAI counterpart, which is called Llama, also a very much a laggard, not on the frontier today. Um, and then you have a couple of other other smaller players.

Speaker 4:15

Gemini's the other big one.

Speaker 2 4:16

Google's the other big one with Gemini. And then you have a couple of other smaller players that are on the fringes, but not really on the frontier with where things are at today. So OpenAI has made its money and its revenue and had its growth predominantly it's been mixed, but they've had a much larger chunk of their revenue come from consumer, from the average American spending, you know, 20 or 40 bucks a month on a chat GPT subscription. Claude is not nearly as well known. Yes, they have some um some presence with the average American, but that is not the line share of what they've done. What Claude has chosen, and Enthropic has, I'm sorry, Enthropic has chosen to do with Claude instead, is to focus on bringing their um services to the business customer. Uh and by far the most popular product that they've brought has been called Claude Code. Some of our listeners may have heard of Claude Code, but it's basically a tool predominantly used, as the name suggests, for doing actual coding and the creating of projects. Um so that's where the lion share of, uh, or at least where they've become really well known. They have some other products in security and other things like that, but Cloud Code is what they're definitely the most known for, and specifically their business enterprise products that they bring to the market.

Speaker 5:26

I mean, that's

The Enterprise AI Advantage

Speaker 5:27

where we've heard terms like vibe coding, right? Where you basically put into Cloud Code, hey, I want to create this app or this uh you know software or whatever it may be, and you vibe code it.

Speaker 2 5:38

Yeah, you're not you're not the stereotype of you know typing 120 words a minute and a whole bunch of other symbols and actually doing the code. You're putting into a a chat window what you want and you're letting the AI agent ultimately do it for you.

Speaker 5:49

You're using common language and it spits out the code, and it's been impressive what many people can um suddenly everybody's an expert at coding.

Speaker 2 5:58

Yes, yes, it's amazing. Yeah, I can do that, I can code that. Yeah, yes, yeah.

Speaker 6:02

Well, that is what's led to Anthropic's massive growth. Is I think uh to their credit, they saw an area where AI could just really make a significant impact and have a direct application, and they have led the field by far in that coding arena.

Speaker 2 6:21

They have. I mean, I was actually talking with a um, we're considering potentially using some AI just to improve our own processes, and we were talking to one particular um AI uh or coder, and his comment was, I am approximately 10x more efficient than I was before. It's not perfect. Yeah, it doesn't write everything, and it's you don't you still have to go through it to make sure everything is done correctly, but it is in the hands of a good and talented person, it can be radically transformative in terms of how much productivity you can do.

Speaker 6:52

Or like having a really good draft. You have a really good draft that you can then proofread and make sure that it's exactly done appropriately. Yeah. Well, let's talk about some of that explosive growth that Anthropic has seen. I mean, it has been truly out of this world in terms of what their growth has been really since they started. And more specifically, in just the last couple of years, two or three years, is when we've seen the major explosive growth. Let's get into some of the details of that. Talk to us about um, you know, kind of what we've seen from an annual recurring revenue, the actual revenue that is kind of a subscription basis that is happening every

The Fastest Growth in Tech History?

Speaker 7:26

year.

Speaker 2 7:26

Yeah, and so that's I guess the first thing. Their business model is generally recurring revenue. Um, you know, whether it's the Claude subscription for the individual or for the business, they have this recurring revenue, which is what the ARR abbreviation stands for. Yeah. And so their growth, and this is the really the heart of the story, and what is is bizarre, and this is why this is a headline worthy of being thought of a lot, is the growth we've seen on a recurring revenue basis has been astronomical. In 2024, um, Anthropic's revenue was sitting approximately around $400 million. Yeah, they'd grown what good? I mean, once again, it started in their first fund right fundraising was in May of 21. So by 24, they'd grown to 400 million. That's impressive. That's impressive growth. But that's nothing compared to what happens next. From 24 to 25, they 10x their growth. They blow investors' minds. That's crazy. Nobody was underwriting a 10x kind of growth rate. That's not the kind of growth you normally see in those situations. So they go from 400 to 4 billion. So most people are like, okay, now we're going to start to see a more tepid, normal, still very impressive, I'm sure, 15, 20% growth. That's what most people are underwriting. Well, we get from 2025 to 2026 and we see another 10x growth. 4 billion to today, we're targeting a run rate now of 47 billion dollars. It's crazy. Astronomical. Yeah. Absolutely astronomical growth. Like literally, once again, in the in the venture capital world, the the investors that they spend all their times around what they we call in the world these unicorns, these um over billion dollar companies with massive growth rates. This is something that has literally never been seen with any major company.

Speaker 9:07

To put that in a little context, that means that their annual recurring revenue is growing almost $100 million per day. Per day. Per day. Per day. That's crazy. Um, if you think about it in the context of some of these other big companies that we think of, you think, oh man, Google for sure grew that fast, or NVIDIA or whatever it is, right? It's nothing like that. No. If you think about NVIDIA, which is, you know, the apple of everyone's eye right now, uh, it took them 25 years to reach $50 billion run rate. 25 years. 25 years.

Speaker 2 9:39

Versus like two years.

Speaker 9:40

A year and a half.

Speaker 2 9:40

Yeah.

Speaker 9:41

For uh Anthropic. You compare that to Google, Google was nine, Amazon was 13. I mean, this is unheard of in terms of the speed with which they've And this is something that most people don't even realize.

Speaker 2 9:54

It's like, you know, for the average American as they're going around in their day-to-day lives, you know, this is one of those news headlines that if they saw it all, it just blipped by.

Why Investors Should Pay Attention

Speaker 2 10:03

Yeah. But it truly, from an investment standpoint and a really a company standpoint, is a history-setting moment that deserves a moment of pause just to think about what is the impact of this. Yeah. And I do think it's important just to think about what is the, what does this really signify? You know, we've talked about before how we do believe that medium and longer term AI will be radically transformative. It's one thing to say that. But the question is, when do you actually start to see the proof that that's likely to happen? And I think the the level of growth that you're seeing with anthropic is really kind of the first green shoots that point to that broader trend. Because, you know, someone has to, if for it to really be transformative, you have to see everybody really leaning into it. And I think that's what you're seeing. You're seeing a large number of companies that are really fearful of missing out, or they just see dramatic opportunities to use the tool to begin to get efficiencies, whether it's operating efficiencies or reducing costs or whatever the case might be. I think this is the proof that you're seeing companies really begin to invest in these things.

Speaker 11:02

Seeing real widespread adoption of it. Yeah. And I think that's the key thing there is companies, right? This is not just uh everyday person saying, oh great, I'm gonna commit $20 or $200 a month or whatever it is. This is companies committing billions of dollars in some cases for their uh employees to have access to this type of.

Speaker 2 11:23

And I think part of the reason for that comes back to this idea that once again, with the what Claude Code, specifically here with Anthropic, has unlocked, is they have made software developers much more efficient. Yeah. Um and I'll give you, I'll just give you one practical example. Even with our own company, you know, we have internal financial planners, um, and we've created our own proprietary process where we do financial planning for clients. Well, that's a very labor-intensive process. Um, it's literally all done by hand, individual person analyzing a situation, coming up with a series of recommendations. Um, we could take a decent chunk of what we do and put it into a computer program. There still would be a good amount of manual changes and corrections along the way, but it could speed up our financial planners in a meaningful way. But guess what? The price tag on doing something like that probably exceeded a million bucks or at least a half a million bucks. Wasn't something that was worth investing in. Well, we're currently in conversations right now about considering hiring someone to actually do it because this is now a cost-effective thing that could make our own financial planners a lot more efficient and being able to crank out a lot more financial plans for customers.

Speaker 12:26

And that's the real value of it, is we're seeing the cost come down. And as a result, you're seeing more and more people be able to do that.

Speaker 2 12:33

We would never considered using it before, and now we're potentially a customer that ultimately would indirectly be using cloud code through a software developer.

Speaker 12:41

Right, right. So we've seen Anthropics explosive growth from a revenue standpoint. I think that's also meaningfully impacted their valuation. Let's talk a little bit about that because that has been crazy to think about.

Can a $1 Trillion Valuation Be Justified?

Speaker 2 12:55

It has. So for perspective, 15 months ago, so I think it's May. March March of 25, yeah. They come out, they raise a large amount of money at a $65 billion valuation. Okay. So once again, this is a company that was started in May of 21. Right. So that's like early investors, whoever invested in that $124 million round back in May of 21 is sitting very, very happy at that point. But so May of 25, 65 billion, their most recent round was over 900 billion. I think it was like 965 billion dollars, almost a trillion dollars. And it's because of that exponent, that exponential growth, that 10x growth from 24 to 25, then from 25 ultimately um into 26. So we've seen this massive valuation move. So a like a 12x return in the span of like 15 months. Or yeah, 15 months, 14 months. Yeah. So huge, huge growth uh that we've seen on that side. And now we're seeing Anthropic has um ultimately filed its S1 with the SEC, um, which is their indication to go public. Yes, it's their desire to go public. Here's our information, is there a problem with this? And so they're looking to go public, and that's rumored to be probably late summer, early fall um at this point.

Speaker 14:14

This is momentous times, yes, for sure. So the other thing that I think is important to note is as they're in this process of being about to file filing to go public, is they're projecting for this next quarter that they're gonna hit their first profitable quarter. And they're they're going from negative profitability to potentially over $500 million of profitability in a quarter, which is in a big.

Speaker 2 14:50

Because to grow that much, you have to put everything back into expanding the business. And so there's typically nothing left at the end because you're just putting everything back into growing the company. And that's what makes this moment so astounding. Once again, if we look back at companies like Amazon, NVIDIA, and Google, these companies were four years, five years, seven years in before they actually had their first quarterly profit. And we're looking at really within four years of officially their first fundraising round, and then really even call it within two to three years of actual serious product development, actually having their first quarterly profit while having a growth rate that surpasses anything we've ever seen.

Speaker 15:34

Yeah, I think here's the amazing thing about that. They're profitable and their revenue is so significant. There is only one other software company out there that has higher revenue than them, and it's Microsoft. And we've seen they've been there forever. They've been there forever. Yeah. It's it's an amazing thing to witness. Hey everyone, if you're enjoying the insights that we're bringing you on this channel, do us a huge favor and hit the subscribe button on Apple, Spotify, wherever it is that you listen to your podcast. It takes two seconds, it's completely free, and helps us continue to bring you high quality content. Now, with all of that optimism and excitement, the

AI's Biggest Bottleneck

Speaker 16:11

real question at this point is where do we go from here? Are they likely to continue this significant growth? So I think it's important for us to talk about some of the threats that are at the you know facing them as they go into the future.

Speaker 2 16:26

And one more thing, really quickly before we hit there, because it's obviously really important. But one other thing that is also just staggering about this business. So there's this concept with software businesses or recurring revenue business, and it's called net dollar retention.

Speaker 1 16:38

Oh, yeah.

Speaker 2 16:38

Which is when you think about the existing customers, are they staying with the same amount, are they spending the same amount of money with you? Are they reducing it because they don't like it as much, or are they growing? Right now, Anthropic's net new dollar retention is 500%.

Speaker 1 16:55

Wow.

Speaker 2 16:56

Five, like I I can't even begin to fathom that. 500%. That means literally if a company was spending a million bucks a year ago, they're spending five million bucks today.

Speaker 17:05

And we'd expect them next year to be spending.

Speaker 2 17:08

Yeah, if that rate continued, we would expect that that to jump to 25 million bucks. So that's another thing that's just really mind-blowing here is when you think about this from a net dollar retention standpoint, a return on compute, they're having massive, um, massive numbers here as well.

Speaker 17:23

Yeah. So that ties in very closely to the threats, right? Because we're seeing all of these really positive things. And I think one of the biggest threats that we see for them currently is how much money they're having to spend to make this happen. You know, when we were talking about AI, what is the input that they have to uh spend? It is that they have to have computing power. Well, one thing that's interesting about Anthropic is they don't own any of their own compute.

Speaker 2 17:50

Well, and just to back up, so what is compute power? Just to make that in tangible terms, you know, at when we, you know, when each of us use a laptop or a personal computer, um, there's a certain amount of compute power that you have literally within the system. Well, to do AI and to do it well, this is why there's all this obsession about data centers. Right. These these data centers connect, you ultimately connect with that network, and you're using the power of that data center, the electricity, the water, everything that goes into that to ultimately then do these massive computations with uh GPUs or other or TPUs, these different um computing tools in order to generate those outputs.

Speaker 18:27

Yeah, that's what we refer to as compute. That's really the data center. That's right. That's the brain behind uh the AI that's going on. That's right. Um, so what's interesting is Anthropic doesn't own any of that. They don't. They basically rent it from companies like Google, Amazon, and one real big news story recently was that they started paying SpaceX one and a quarter billion dollars a month. One and a quarter billion dollars a month to rent space from his Colossus uh data center, basically. And so that's you know, we'll see how all that can continues to play out. But in order for them to continue training models and to continue getting better, they have to have more compute. And if you've got um uh customers that are spending five times more on compute next year versus this year, you gotta have more compute power. And so that means their cost is going to continue to grow. The the real question for them is gonna be is the cost of that compute per token that's generated or per you know uh amount of data that's analyzed through AI, is that going to come down in the future? Because we know for sure that their demand's gonna go up. The question is, is the cost of that compute gonna come down? And that's the real question. I think a big threat of theirs is they have to spend so much on what we refer to as CapEx, right? Capital expenditures on getting access to new computing power that if we see any wavering in this growth rate, it has potential to wipe out their profitability that they've just recently gotten access to. So that's something that we're certainly gonna have to keep an eye on from a threat standpoint.

Speaker 2 20:11

So, Max, I think one other thing that's really important just to take a step back on is when you look at most other major innovative moments of the last 25 years, you think about the personal computer revolution, you then think about the internet, you think about the mobile device, you then think about uh going out of the cloud. Specifically, the last three all have one thing in common, which was that these business models were very low cost.

Speaker 1 20:36

Yeah.

Speaker 2 20:37

Okay. You to for the mobile device, for cloud, you had massive profit margins in these businesses, meaning for asset light. Yes, they're asset-light businesses. You didn't have to have a lot of capital expenditure in order to scale the businesses. And as a result, they were massively profitable. And the ability to scale them quickly was much. It was plug and play, basically, right? It was. That is not the case with AI. AI is not a capital light business.

Speaker 21:04

It's interesting because you kind of think of it as though it is, right? It's the cloud, it's you know, not a very capital-intensive business, but you're right, it absolutely is.

Speaker 2 21:12

It is. These data centers cost a tremendous amount of money to build the electricity that you have to pay for. It's like what like think about it this way: when Google built the search engine, okay, for every additional user that used the search engine, what did it do? It reduced the overall cost. Right. You have these fixed set of costs, and there are there are some variable costs, some things that will increase with the number of users, but by and large, you build it once, and then it's just the maintenance and upkeep. And so there's no, as you add additional users, you're also adding meaningful additional costs there. Not true with AI. For every additional user that performs a search query, because of how expensive compute is, um, it does radically increase the cost, which comes back to your point. This is not a business where you Know you build it once and it's done. Right. This is a business where the variable costs do increase along the way.

Speaker 22:05

It's like having so much space on your computer. You fill it up, you can't just add more. You've got to get a little bit more. You gotta have an extra hard drive or another computer or something else in order to make that work. Yeah. I mean, that's that's a huge uh reality of this AI world.

Speaker 2 22:21

Great point. I think let's even back out a little further. So if we even think about like a non-paying customer, even a paying customer with Anthropic or even a ChatGPT, we're experiencing this across the board. Um, these companies have a scarce resource, which once again is compute. It's the data centers and all the infrastructure that ultimately creates the output from your search query that you make. And we're already seeing this with paid models. Hey, you've exceeded your use for your subscription for today, if you'd like to put your credit card in and you can pay for more. Why? They're not just doing this to make more money, yes, they are, but they're also doing it because they legitimately do not have enough compute.

Speaker 22:58

There's a constraint there.

Speaker 2 22:59

There's a constraint there to handle

The Open Source Challenge

Speaker 2 23:01

the existing users. That's the first thing to be aware of is like legitimately, this is a finite limited resource. And to put this in perspective, if you look at every area that impacts AI, the picks and shovels, the chips, the memory, um, energy, the harnesses, the energy, all of these things are strapped. And like we're already one or two years out where everybody who's producing it, whether it's Taiwan Semiconductor and the chips, or it's Micron with the memory, or you name the carrier, they are completely sold out for the next like two years at this point. Um, there's just a limited amount of new data centers that are going to come on board, which is ultimately the blood of the entire AI network. On top of that, you have these large communities that are pushing back against new AI centers, AI data centers being built. So that's one other constraint that before we even get to the cost of compute, it contributes to the cost of compute because when you only have so much supply coming in and the demand is so high, that's what's causing the price of compute and therefore the price that OpenAI and Anthropic and others are having to charge to be a lot higher. So I would agree that the cost of compute is definitely one piece. But the first thing you said I think is also really important, which is they don't own any of it. So think about this. Anthropic right now is relying on Colossus with XAI for a decent chunk of their compute. Notice the length of that contract.

Speaker 1 24:26

Yeah.

Speaker 2 24:26

It's a quarter by quarter contract. This is not a long contract. Every 90 days. Every 90 days, Elon could say, I need all of it myself.

Speaker 24:33

What a strange dynamic, too, right? That a competitor is leasing this compute to you. To uh a leading competitor.

Speaker 2 24:41

And they're making a lot of money off of it. There was actually rumors recently that um Meta, now or formerly Facebook, um, is potentially leaning into getting into the compute business and just because they've struggled to create a uh frontier large language model that's hey, let's use this data center stuff that we've built instead and sell the compute to those who need it. Um hasn't actually happened yet. But the point is this Anthropoc could lose Colossus like that, right? Um, which is a meaningful chunk of the compute power they have. So that is going to be a meaningful challenge for them is can we actually get the compute that we need? Um now they are partnered with Amazon and Google, but that is a meaningful, I think, um, restraint or threat that potentially could be uh a challenge for them.

Speaker 25:25

Concentrate on future growth. For sure. Yeah. So let's jump into uh another one of the big headwinds or threats that we see for AI, or or really for anthropics specifically. But I think this applies to all of what we would describe as the frontier models, those that are at the cutting edge of the AI development. Um and that would be open source models for people that don't you know swim in this space as frequently as maybe you and I do. Um I think it's important to distinguish between what a frontier model is and what an open source model is.

Speaker 2 25:55

Or a closed model and an open source model. Closed model, closed model and open source, right? Yeah, so closed source models are the ones we've mentioned here. You pay for them. Yeah, you pay for them. You know, that they they they're owned by somebody. You don't have the freedom to go in and adjust the code or make changes and those sorts of things. So claw uh anthropics models, open AI is auto able, open AI's models are that way. Gemini, um, Gemini, etc. Um, you have other models, um, probably a name that no one's heard of, but like Kimmy would be a good example of an open-ended model, um, an open model rather. Open source, yeah. Yeah, an open source model. And so these are models where literally it's like you you can download these models, you can put them on a personal computer, and you can change them. You can literally say, hey, they're they're not connected to a data center somewhere. You have to power the usage yourself.

Speaker 26:47

Um I think that's an important distinction is that these models are trained. They're basically uh one of the most uh time-intensive and money-intensive aspects is training these models, getting them to the point where they have you know looked at all of this data and have a good understanding and can respond accurately. You can then take that trained model and download it effectively to your own server and run it yourself internally. And then you can get updates as things change.

Speaker 2 27:14

Maybe you maybe if it's an if it's an open model, yes. Yes, yes, and so you have the ability to give updates, but at that point, you are now responsible for the compute.

Speaker 1 27:22

Right.

Speaker 2 27:22

You are powering that yourself either through your own personal device if the if the model is small enough, or you're seeing a lot of companies now do what is called creating these um on-premises compute models, basically where you have your own compute, you house it internally, no one else is getting access to your data, and you can use these open source models. So that is a real threat to anthropic because one, if you're open source, you're not having to pay a subscription. You and a lot of these, a lot of the open source models, many of them have come from China, although there's some in the US, um, are they not at the frontier, they're not as good as Claude or OpenAR as some of the other frontier models, but they're 80% as good or 85% as good. They're maybe six months behind the the latest models.

Speaker 28:07

And so the question is if you can get those for 15, 20 times cheaper, I mean if you got to pay for the hardware or whatever it is, but yeah, you're gonna have to pay for the compute cost yourself.

Speaker 2 28:17

But if you don't have to pay a company for that, then for tasks that don't require the leading edge, the frontier, the absolute best, you already are and probably will continue to see groups figure out, okay, what tasks do I actually need to use the frontier models for, like a Claude or an Enthropic, or can I use an open model instead in order to actually get this done a lot more cheaply? And so that's definitely another major threat to anthropic is the right is right now, even today, if we look at um open source use versus the closed source, closed source might be 20 or 30 percent of all queries. So it's actually not that big. But so far, if we look at the profits, the actual money that's been made, 80 plus percent of that is flowed through to the closed models. But um, if some of the open models could start to become more profitable and even pull more of that usage away to the open source models, that could obviously be a threat to anthropic as well.

Speaker 29:15

So it's interesting we talk about open AI. What you're telling me is open AI is actually closed.

Speaker 2 29:21

Yeah, that's a whole nother way. I don't think we should even go down the aortable, but yes, no, they're not. They're a closed source model, they weren't originally supposed to be. That was honestly part of the beef that Elon had with them, and one of the reasons why he sued was because it was supposed to be an open source model, so there couldn't be a major monopoly that controlled AI, and it's turned into a monopoly that controls AI at this point.

Speaker 29:42

Well, let's talk a little bit another nuance of the open source world that I think is really starting to come to light more and more today is it's not just the cost savings that you can get from using an open source model, it's the security and safety of your own data. Sure. Right? Because one of the things

Who Owns Your Data?

Speaker 29:58

that I don't think a lot of people think about when you type your information into OpenAI or Anthropic, they get that data now and they can analyze it and see it. And so what we've been seeing from a lot uh probably over the last year or two is Anthropic will serve some big company that say is producing legal documents or uh you know lots of different things, whatever it may be. And then miraculously, a few months later, they come out with their own product that is a direct competitor to that.

Speaker 2 30:28

Well, I mean, Cloud Code is actually a perfect example of that. Yeah, you had absolutely cursor and others that were had they were building these coding agents. Um they were clients of Anthropic, and Anthropic realizes, man, this is a massive revenue center. They come out with their own product, they make it, they eventually get it, make it better, and they actually at the end of the day end up shutting Cursor off from using their their model at all. And suddenly, who wants someone who was once their customer is now their business.

Speaker 1 30:55

Right.

Speaker 2 30:56

And in to be real, if you think about this, you saw this with Amazon as well. Yeah, Amazon has this full marketplace they allow everyone to sell, and suddenly you see Amazon is, you know, the product you were selling, Amazon now is selling a product that's exactly the same, that's 20% cheaper. Right. And they're just identifying what are the businesses that have the most profit and revenue potential.

Speaker 31:15

It's not a new model.

Speaker 2 31:16

This is something that is happening. This is something that's happened over the last 20 years a lot with the internet and personal computing and everything else like that. And so that is a legitimate, another legitimate threat right now is anthropic and uh open AI both are viewed as the big, you know, the big thief that can come in and basically take the data that you are allowing them to get access to, use it to create their own product and competition to you. Um, and that's a big problem. You know, for there will be some companies where the information edge isn't as important, but um, you know, we listen to the all-in podcast a lot, and Friedberg mentioned that within the life sciences area, yeah, companies have been given the opportunity, whether it's with OpenAI or Anthropic, to participate in a um a group sharing program of all the data of different companies together. And many companies are saying, no, yeah, no, I don't want to give you access to my proprietary data because I'm I think in many cases they're afraid of this exact thing happening.

Speaker 32:10

I mean, life sciences is a very uh highly specialized. Well, highly specialized, but very you you differentiate yourself from you have very important IP. Yes. And it's not about hey, I do this a little bit better than somebody else. It's hey, we've got a breakthrough that nobody else should know about because it's you know specific to our company. The other thing that you've seen is you're starting to see more and more of this play out, is just recently Alex Carp from Palantir went on um CNBC and talked for 20 minutes about this very thing and saying, why are companies doing this? This is a a real risk. And so I think we're gonna see more and more of that trend potentially pick up, which isn't just I absolutely think you will.

Speaker 2 32:51

The reality is like you you can't even if a company like Anthropic or OpenEye in you know assures you that they are not using. I promise I'm not taking your information. There simply is far too much risk. Well, you have alternatives. You yeah, there are alternatives, you know, and so the interesting thing is in some ways we may see the reversal of a trend. For 10 years, there was this move to moving everybody's data and computing to the cloud.

Speaker 1 33:17

Yeah.

Speaker 2 33:18

So that people anywhere could access it. And, you know, we've seen a lot of people in the industry say that, hey, I think we're actually gonna see the other trend happen, which is instead we're gonna go back to having more of the racks and the compute power be located on premises for really two reasons. One, so that these larger companies can't access my data. Uh, and two, for security reasons. That's the other kind of big driving factor that you're seeing on that side.

Speaker 1 33:41

Yeah.

Speaker 2 33:42

So as we wrap, I think kind of just as we think about um this moment in history right now, the biggest takeaways are one, we are seeing within AI some truly earth-shattering growth rates.

Speaker 1 33:56

Yeah.

Speaker 2 33:56

Growth rates we have literally never seen before. Um, and so

What Comes Next for AI Investing?

Speaker 2 34:00

the question naturally is will those growth rates continue? Will we continue to see that same movement? I personally think even if anthropic doesn't directly benefit from the same parabolic growth it's been experiencing, we're gonna see. You're gonna see it. The question is maybe maybe they're not the beneficiary, maybe maybe more open models um ultimately end up getting to capture it instead. But I think that's the biggest um bellwether that you really see, or the the signal that you see within this story is the growth is happening. You're seeing companies explore it, you're seeing them lean. AI is being adopted, it is being adopted, yeah, and anthropics parabolic growth on the revenue side is really just is the is the final piece that shows that that's actually happening.

Speaker 34:42

And I think the exciting thing that we're all gonna get a chance to see is we're gonna get more visibility into all of that here in the next you know couple of quarters when they ultimately go public, because then all of their data or all of their books and everything is gonna be made public, and we're gonna get a chance to kind of see what's more happening under the hood.

Speaker 2 35:00

I do want to make one clarifying point. So none of this is investment advice. We're not suggesting you invest in the Anthropic. The reality is while they have had astronomical growth, the valuation is also quite high. Right. You know, they're doing $50, $47 billion of revenue. Their current valuation

Closing Thoughts

Speaker 2 35:16

is $965 billion. So it's 20 times revenue, not profits. So this is a this is an expensive multiple, which you would expect to pay for a company. Um, but I do want to be clear this is not a recommendation to buy them.

Speaker 35:29

Absolutely, absolutely. Well, it's gonna be interesting to see how all of this plays out. Uh, it sure is an exciting time to be a lot. That's for sure.