From Eyeballs to ARR: Buzzy Pricing Metrics and the AI Trade!

0:00 Hi, welcome back. This session is going to be about AI only indirectly and I want to talk a little bit about an IPO that's imminent and I'm talking of course about the anthropic IPO. As we talk about what the IPO is going to be priced at and this session I'm going to talk about the contrast between value and price. I'm also going to talk about pricing metrics. Pricing metrics that relate the price to something you can scale across companies. So let's start.

0:29 Last week, I was in San Francisco giving a talk to NYU Stern alumni about what else but AI. And on that same day, it was pure coincidence, October 15th of 2026, Enthropic had a preIPO meeting with prospective investors. Now, I'm not important enough or have enough money to make that list, but during the course of the meeting, Enthropic announced that its annualized revenue run rate. What is that? We're going to come back and talk more about this. It's called ARR. It's a statistic that both open AAI and Anthropic have adopted with Gusto and what it does is it takes a period the last week the last month and annualizes that number as opposed to what much of valuation much of you know much of financial analysis been built around what you had as revenues in the last 12 months. We'll talk about the rationale for using annualized revenue run rates the ARR and why pricing is being related to that number. In fact, during the course of the meeting, Anthropic also threw out an expectation that the annualized revenue run rate by the end of 2026 would be close to 100 billion.

1:36 Already, you can see this is a story not just of substantial revenues, but growing revenues. Now, in case you're wondering what's an ARR, why are we relating pricing to ARR? This is something you're going to see across time with young companies. When you try to relate price to something, you have to find with young companies something that might not be on the financial statements. I remember during the com boom that the market cap for.com companies was related to the number of website visitors. Along the way, Mary Maker came along and said it's not just how many visitors you have, but how much time you spend on how much time they spend on the website eyeballs or market pricing per eyeball. You're saying that's outlandish. In this session, I want to talk about first why young companies are more likely to be priced than valued. And I'll draw the contrast and why when you price young companies, you're likely to run into metrics you've often never seen before. So, let's set the stage. Let's look at companies across the life cycle and what sets young companies apart. This is something you're probably very familiar with if you watch some of my previous sessions, but I firmly believe that businesses follow a life cycle. Startups are like babies. Young growth companies are essentially starting to take off. Then you have high growth. Think of that as the peak of your life. Then you start to hit middle age, mature growth, and even more middle age perhaps spilling into old age with mature stable. And then you go into decline. Let's focus in on young companies and see what they share in common. Before we do that though, one of the questions you might ask is, "How do I know where my company falls in the life cycle?" You could take two easy routes. Both come with limitations, but they're very simplistic. The first is look at the age of the company, arguing that a company that's been around only 2 years is young, that a company that's been around 150 years is old. The problem with that is chronological age might not actually capture life cycle that well at many companies. You can have a young company that's in steady state. It had a niche business. It got into that business climbed in and especially with technology where companies mature very quickly. A a three or a four year old company can already start to look like a mature growth company. In contrast, if you're an infrastructure company, you could be 25 years old and just getting off the ground. So age is you know first step but it's not a very good way of classifying companies. The second approach which is just as lazy and just as easy to do is to look at what industry grouping your companies in and extrapolate from there. To be more specific the last 50 years tech has become the shorthand for growth companies. In other words if you're a tech company you must be a growth company. You can see the reason for that. For much of the last 50 years, growth in the market has come from tech companies. But the downside of classifying classifying a company based on which industry grouping it's in is you can have within the tech sector a mature or a declining company. I've described tech companies as aging in dog years. A 35 or 40year tech company is essentially analogous to a 200-y year manufacturing company. It's an old company. So both age and industry grouping are shortcuts that people use but they're limited. So how do you decide where a company's in the life cycle ultimately this is about operating metrics.

5:07 If I looked at your operating metrics I should be rever able to reverse engineer from those metrics where you are in the life cycle. Let's start with revenues. Young companies tend to have high revenue growth usually of small scale small revenues growing fast though sometimes you can have companies with larger revenues that are still growing quickly. So you have high revenue growth. The second is with young companies you will tend to be losing money. Not all of them do but most of them do because your business model is still in flux. It's still being formed.

5:40 So you often lose money. You often need to reinvest large amounts to get the potential growth you see in the future which combined with the fact that you're not making money will give you very negative cash flows. So you have high revenue growth, operating losses, lots of reinvestment, negative cash flows. Those are features not bugs. And because those negative free cash flows have to be funded with external capital, there's a risk you will not make it. Failure risk is high. As you mature as a company, your revenue growth will tend to come down. Your operating margins will improve if you survive as a business. Your reinvestment will scale down as your growth plans come down and your free cash flow to the firm will turn positive. And your failure risk should mitigate as well until you get into decline as a company. And there failure risk can rear its ugly head if you borrowed a lot of money and you're now shrinking as a business.

6:33 Put simply, you tell me your operating characteristics, I'll tell you where in the life cycle you are. So, if you tell me your revenue growth is 6% a year and you have nice stable margins of 30%. You're not reinvesting very much and you're paying out large dividends, you are a mature company no matter what you think of yourselves. In contrast, you have high revenue growth, you're losing money, you have lots of reinvestment and you've negative cash flows, you're dependent on external capital. You are a young growth company and that's where I'm going to put you. So, with that lead and let's think about enthropic and open AI and where we would put them. You might know the answer already, but humor me. How old are they as companies?

7:13 Anthropic is about 5 years old, founded in 2021. OpenAI is 11 years old. There are almost no revenues in 2023. Anthropic at 100 million, OpenAI at a billion dollars. And their revenues have grown exponentially. You can see the growth from 2023 to 2024, 2024 to 2025. And at least early in this game, Open AI is running way ahead of Enthropic. In fact, if you look at revenues in 2025, Open AI was clearly three times the size of Enthropic. Then something clearly shifted in 2026. And Enthropic did not just catch up, it went ahead of open AI. We'll talk about how the AI business evolved and what allowed open Enthropic to get ahead. But if you look at the revenues in the trailing 12 months, Enthropic is now at about 35 billion in trailing 12-month revenues.

8:05 This is the last four quarters. And Open AI is closer to 20 billion. They're saying, "Well, that doesn't look like the numbers I've been seeing for these two companies." That's because both of these companies for much of the last year have been reporting what are called annualized revenue run rates or ARRS. And here's what you do. You take the revenues in the last month and you multiply by 12. Why would you do that? If your revenues are growing very quickly, your argument is trailing 12-month numbers will not reflect where I am as a company. So, I'm going to to annualize what I make as revenues. That annualized revenue run rate for Enthropic was about 67 billion in sep, you know, leading into October of 2026.

8:50 It's still preliminary. it might ship and open AAI is at 50. One more thing, the way these companies compute their revenues is actually different. We're going to come back and talk about why this makes comparisons messier. Enthropic reports what's what are called gross revenues because some of these revenues are companies where you have costs with these companies that you have to net out and Enthropic reports the total revenues and nets out the costs. Open AAI has switched reporting net revenue. So that you know the 50 and the 67 might not be directly comparable but at least on a generic level anthropic is at least as big and perhaps bigger than open AI. The revenue growth rate has been astronomical for Enthropic especially if you look at the last year it's less so for open AI but still impressive 120%.

9:38 So if you look in terms of the characteristics of young companies clearly both these companies have high revenue growth. We can debate how high the growth rate is. We can't debate that it is high. In conjunction with these revenues, these companies have been reporting losses. In 2020 2024, both companies lost money. 3 billion in anthropic, 5 billion for open AI. Those losses expanded in 2020 25. 8 billion in anthropic. 39 billion in open AI. didn't break it out, but some of that might have been one time extraordinary. In in the first half of 2026, anthropic report an operating loss of 8.7 billion was actually closer to 42 billion, but a lot of that was one-time non-cash. So, I'm taking the most charitable view of profits at these companies that I can. They're money losing companies.

10:31 Anthropic during its meeting on October 15th of 2026 made claimed that it made money in the third quarter of 2026 on an adjusted basis. I don't think even anthropic believes it made money. This adjusted is a often a fictional number. So I think we can we we can I I think clearly start with a presumption that these are high revenue growth companies with big operating losses. So they're young companies. You're saying so what when you think about putting a number on a company there are two ways you can put a number on a company you can value the company which requires understanding the business building up through revenue growth and cash flows and and risk to a value for the company that's what you do in a discounted cash flow valuation or an intrinsic valuation but you can also price the company by doing what by looking at what others are paying for similar companies. You think shouldn't the two be the same? Maybe in an efficient market, but the two can give you divergent numbers. And I suggest to people then when they sit down to put a number on a company, they make a decision. Should I value the company or should I price it? So let's take a closer look at what you would need to value a company. Now when you value a company, there are three crutches, and I'm using that word deliberately that we use to kind of pull ourselves up to value the company. The first is we look at the past financial statements for the company.

11:54 last 5 years, last 10 years. What are we hoping to learn there? How quickly revenues are growing, what the unit economics are by looking at gross margins, how profitable the companies by looking at operating margins, perhaps even learning about how the leverage is affecting the company through net margins. We're looking at the financial statements, looking for clues about the company's business model, its profitability, and its value. Now, of course, we're a mature company with a long history. There's a great deal you can learn by looking at Exxon Mobile's history. But can you learn very much about a young company which has been around 2 years or 3 years? Probably not.

12:32 So already you can see that crutch becomes much more difficult to pull yourself up on if you're looking at a young company. The second thing you need when you value a company is knowing what the total market size is and what your market share is and what the competition looks like. easier to do again if you're an established company with an established business model working in a market where you can see the market size already. But if you're young company that's building into an evolving and still unspecified market, much more difficult to do, right? So the second crutch becomes more difficult to get as well. And finally, when you value an individual company, a steel company, it's good to know what margins look like across steel companies. When you're looking at a retail company, what do turnover ratios look like across retail companies? When you look at risk, you can look at cost of capital across companies. It's nice to have those industry averages. Again, that's easier to get for some companies than others.

13:30 And if that crutch disappears, and with many young companies, it will. You have no crutches to use. So, what does that imply? In most young companies, people don't even try to value the company and they give the excuse of I don't have history. I don't have know what the market is going to be. Basically, they're saying there's too much uncertainty here and I don't have my usual crutches to make my estimates. Fair enough. And in fact, if you look at valing open AI and enthropic, you can see the uncertainties you're going to face. First, when you look at their current financials, there's not much history there. two years, maybe three years, and you're going to be looking at companies with growing revenues, but the rest of the income statement is a complete mess. You see, what about the future? Well, we're debating still what the total addressable market for AI will be. I know bankers have thrown out trillions of dollars, 20 trillion, 25 trillion, 30, but it's still very much in play. And even if you buy into that total market and give it a number, we still don't know what the competitive advantages will be for the companies that win in this market. What will allow them to earn more than their cost of capital? Much more uncertainty about the value of growth. If you ask me how risky are these companies, what can I show you? Again, two years of financials, we don't know whether these companies are cyclical or non-yclical. How risky they are. There's no price history. There's no financial history.

14:57 And finally when you think about nirvana in valuation which is when will I become a mature company both anthropic and open AI have significant compute cost commitments. These are contractual commitments they have foru for computing power. Now you're saying so what lots of companies have commitments for the future but in a money losing company in an evolving business there's a risk of failure. That might mean you never get to nirvana. Already you can see that when you sit down to value these companies, all of these challenges are going to come up and next week when Anthropics prospectus comes out and as you look at the prospectors, you're going to be faced with all of these challenges with one more thrown into the mix.

15:42 I've heard that Anthropic's plan is to have two classes of shares, high voting right and low voting right shares, which means at the end you also have to deal with corporate governance and different voting rights and how you would value one one class of shares relative to the other. No wonder most people don't even try to value young companies. In the case of entropic and open AI, I will concede up front that most people will probably not even try. You say, what are they going to do instead? You're going to be tempted to rely on people you think are better informed than you are.

16:15 In this case, venture capitalists. Anthropic and Open AI have had multiple rounds of venture capital. Anthropic is at 19. Open AI was at 10. Enthropic has already raised 135 billion from venture capitalist. Open AAI 185 billion. Lots of big names in those VC rounds. In its most recent pricing, which was in May of 2026, Enthropic was priced at 965 billion. You saying, "How did the VCs come up with that?" We'll talk about it in a moment, but 965 billion.

16:47 Open AAI had 10 rounds. Most recent round was in March of 2026. It was priced at 852 billion. Open AI is a pending round where it's being priced, at least the rumor is, at around 1.4 trillion. All of these were taking the VC capital raise and we're extrapolating from them which is a little dangerous but clearly there's some pricing out there and you're saying well given that there's so much uncertainty why shouldn't I go along with how venture capitalists have priced these companies after all they have access to far more information than I do and they're much smarter than I am. Well let me take you inside venture capital valuation and kind of strip it to its basics. Venture capital valuation doesn't deserve to have the word valuation in it. It's forward pricing.

17:36 Let me explain what's called venture capital valuation. Here's what you do. You project out a point in the future. Completely arbitrary. 3 or 5 years from now. What a metric for the company will be. So let's take an example. Let's suppose it's revenues 5 years from now. You expect Anthropic to have $300 billion in revenues 5 years from now. You apply a multiple to those revenues. See, how do I come up with that? By looking at publicly traded companies perhaps later in the in the life cycle.

18:06 And you say, look, you can look at the MAG 7. They roughly trade at five, seven times revenues. 5* 300 billion is 1,500 billion or 1.5 trillion. That's my exit year pricing. And if this were truly a VC round early in the process, you will then discount that pricing back to today at what looks like a discount rate, but it's really nothing to do with expected returns or risk and return. It's a madeup number. It's a bargaining number. And you'll often see VCs use numbers like 30, 50, 70% to discount this forward pricing back. So stripped to its basics, VC valuation is a forward pricing discounted back to today at a number that makes that number lower because as a VC you have an incentive to make that number lower. That's in a normal VC round. With enthropic, so there's not much information there, right? You're not learning much about the company from a VC valuation or pricing. With Enthropic and Open AI, you're going to learn even less. And here's why. They're not projecting out five, seven years from now because, you know, they're basically they can see an IPO is imminent. They're projecting out to what that IPO date will be and what they expect the pricing to be. Of course, they will have to discount that pricing for the months or perhaps the year they have to wait. And what do they have to discount for? That something bad could come out about the company. That something could delay or maybe cancel the IPO. So let's take you back to May of 2026 when you saw 965 billion pricing for Enthropics. How did they come up with it? At that time the expected pricing on anthropics IPO is 1.15 trillion. Today it's closer to one and a half or two trillion but then it was 1.15 trillion.

19:58 The estimated pricing VCs were using was about $150 billion less. You saying why $150 billion? Why not? You're making up stuff at this point. Might as well make up 150 billion. We attach too much value to what VCs think about a company because at its core VC pricing, you're guessing a market pricing and you're reverse engineering from it here. So, as you look at the the reality of anthropic going public, we can almost guarantee it's going to be price not valued. And at this point, it's perhaps useful to stop and think about what goes into the pricing process. This generic three steps that drive it. The first step is you choose a multiple. You say, "What does that mean?" You can't compare market price per share or market caps across companies because the market price per share can be changed by just altering the number of shares having a stock split. So what do you do? You divide the market price by something something that companies share in common. So when you look at a multiple the numerator is always a market number market cap. It can be market which is market value of equity. It can be market value of equity plus debt. Market value of the firm or what's more commonly used is market value of equity plus debt minus cash enterprise value. They're all measures of what the market thinks about your company. That's a numerator. In the denominator, you choose a scalar.

21:26 What is a scalar? You can relate the market value to the revenues that a company has. You think why revenues? Well, desperation might drive you up the income statement. Every company you're looking at or most companies are losing money, you climb the income, you lose use revenues. You could use earnings. Those earnings can either be to equity investors, net income, earnings per share or to the entire company operating income. You can use cash flows. For equity investors, that's net income plus depreciation or maybe free cash flow equity. But for the entire firm, it can be operating income plus depreciation, EIDA, or free cash flow of the firm. Or you can use book value because after all, that's what accountants tell you your company's worth. Either at the equity level, book value of equity or book value of the entire business or invested capital. Numerator is a market value, denominator is a scalar. Now you're saying what if my company's pre-revenue then you try to relate the pricing to something that might deliver revenues for you in the future users subscribers eyeballs.

22:28 So the first step is picking a multiple and as a substep you might have to make a decision about know what the timing for that that number is. You can use revenues in the last fiscal year, revenues in the next year, revenues in the year 2031. You're saying which is right. It there's no right or wrong. You just have to pick a common scalar. Each will give you a different answer and you'll have different things to control for. So the first step is picking a multiple and picking a scalar for your companies. Second step is finding a pure group because the essence of pricing is finding other companies just like yours and seeing what investors are pricing that company. That's going to be messy, right? You might look at other companies within your sector. You might look at other companies within your geography, other companies of similar size. You can already see that there's a subjective component to this as well, but you'd end up with a pure group. So, you picked a pricing multiple. You pick a pure group.

23:26 And then you compare the pricing multiple across the pure group. And you might discover that your companies priced at 15 times earnings and other companies in your peer group are priced to 20. and you get ready to buy the company and say, "Wait, it could be because my company has lower growth, the higher risk." You control for differences. You pick a multiple or standardize the price. You find a pure group and you tell your story. That's the pricing process. Let's see how this process would play out for young companies in general and for Enthropic and its IPO in specific.

24:03 If you look at pricing metrics across the life cycle, the numerator is a market value. So if you're publicly traded, it's a market price. If you're privately held, it might be the VC pricing. But in the denominator, you need a positive number. You need a number that's a plus, not a minus. And already you can see with young companies, the metrics are going to shrink, right? Because once you get below revenues, almost everything in that income statement is a negative. The EBIT is negative. The EIT is negative. The net income is negative. So what are you going to do? You're going to see a lot of revenue multiples.

24:37 And you're pricing it based on potential market capital access. No. So you're going to see a lot of revenue multiples. Young growth company. With startups, you're going to look at pre-revenue multiples, drivers of revenue, subscribers, users, eyeballs, total addressable market. If you're as you move across the life cycle, your choices will proliferate. So as you go to high growth, you'll still see revenue multiples, but you're going to start to see some earnings multiples. With mature growth, you're more likely to see PEG ratios or forward PES. With mature companies, you have you can pick a dozen different multiples and they should all work. And finally, in decline, you'll see book value multiples, price to book kind of come to the surface. The pricing metrics will change and your peer groups are going to change. You can't compare young companies to mature companies because then the young companies are always going to look expensive to you. So you're going to find other young companies. And if you're privately funded, you might be looking at other young private companies with VC funding.

25:39 If you're publicly traded, you might find other young companies that are public. But again, the peer group is going to shift as you move through the life cycle. So as you move through the life cycle, you're going to see pricing dominate early on. The metrics you're going to see are going to be very different. than you than the metrics you see for mature companies. Let's look at the potential pricing metrics anthropic and open AR. I'm going to play their game of annualized the ARR the annualized run rate and revenues. September of 2026 that number was about 65 67 billion for Enthropic. For OpenAI it was 50 billion.

26:18 I could look at what multiple of that number I'd be paying if anthropic is priced as it's rumored at 1.5 trillion be 23 times ARR. Is that a good number? Is that a bad number? Well, you know, in this case, one of the problems is the only real comparable is open AI which hasn't gone public yet. And if you look at its base value of 50 50 billion ARR at 1.5 trillion, open AI would be more expensive than entropic. But here I want to go back to something I mentioned earlier on. Anthropic and OpenAI's ARR might not be directly comparable because one uses gross and the other uses net.

26:57 And this is something we'll have to resolve very quickly before we start using these multiples to compare across companies. In fact, if you're concerned about using the ARR from the most recent month and you want to go forward, well, you can use ARR at the end of 2026 estimated to be 110 billion at Enthropic and 70 billion at open AR. So now your multiples at 1.5 times 1.5 trillion looks at looks more reasonable if you can call 13.6 times 2026 revenues reasonable or 18.18 if you use two trillion as your market cap. Or you could keep going, right? You could go to expected revenues in 2030, 300 billion by some counts. Well, 1.5 trillion is now only five times revenues or even a 2 trillion, 6.67 times trillion. I'll make a prediction. You're going to see a a lot of anthropic bulls use forward numbers. Really, really forward numbers.

27:54 You know why? Because it makes your metrics look more reasonable. It's it's as simple as that. So that's, you know, everything below your revenues is going to be negative. It's going to be tough to find earnings multiples at work here. But there's a second choice you could make. One of the issues that both anthropic and open AI are running into is compute power is becoming scarce. So one way you can market yourself as an AI company. Look, I have a lot of compute power lined up.

28:23 So at least in that with that logic, companies that have lined up more compute power should be priced higher than companies that have less compute power. Early on, OpenAI had an advantage on this dimension with Enthropic and Enthropic has managed to catch up over the last or was trying to catch up. So you go back to 2026, OpenAI was spending 15 billion on compute cost. Anthropic was spending only 6 billion. that compute cost in 2028 is expected to be 111 billion in open AI, 28 billion in Enthropic.

29:00 But there is also a plan spend that the companies are being pretty open about. In fact, since I did this table, Enthropic has upped that number from 518 billion to 600 billion. This is their plan spending to accumulate more compute power and they have contractual commitments of more than 400 billion already. Now normally this is a negative right that you have all these commitments but if you're enthropic or open AI you could market this as a strength and say look look how much compute power we lined up and I can see the possibility of people saying well Enthropic looks cheap because it's trading at you know 3.62 times compute cost commitment relative to some other company that's trading at six times. I know it sounds like a reach and it is a reach, but with young companies, you're going to see a lot of reaches. So, I wouldn't be surprised if you start to see some investors at least talking about enterprise value to compute cost or compute power at these companies that one company looks cheaper because you're getting more compute power in return for the market cap.

30:09 And in fact if you look at compute capacity which is in gigawatt you basically can relate your market cap to how much compute capacity they have. And there again you can see as the you know you want a lower number rather than a higher number. And finally this is not something you see much talk about. We'll talk about why that might be. Early on these companies were measured based on how many people were using their AI agents and open AI was winning that game.

30:39 In fact, in September of 2026, OpenAI had 1.2 billion active users and Enthropic had 250 million active users. You could divide the market cap by active users and Enthropic looks like I mean OpenAI looks like a bargain, right? You're saying, why aren't people talking about that? Because until 2024, this might have been a good proxy for expected revenues. But in 2025, Enthropic broke the glass on this. Basically, by shifting away from individuals to enterprises and away from subscription models to usage models. So you if you look at that $6,000 per user in September 2020, it's much higher than the 1,250 per user for OpenAI.

31:24 Anthropic will justify it by saying we make a lot more per user because that user is an enterprise spending tens of thousands of dollars with us and therefore we're a better bargain already. you can see before you you know you you see the IPO that you know you're going to see you know people making the argument anthropic is cheap relative to open AI you know based on the annualized run rates and open AI bulls are going to say no open AI is cheaper based number of users maybe you know compute capacity right now so you're going to see people pick their poison and push that to the forefront I'm not saying one is right or wrong for the moment, but you're going to see metrics that you've never seen before.

32:10 Now, as you look at these these divergences, one thing to remember is bias does drive pricing just as it drives valuation. One of the reasons I get push back when I do intrinsic valuation is you're biased in your forecast of growth rates and margin. I agree completely, but that bias shows up in pricing as well and it often shows up in the metrics you use to make your case. That's the pricing metric. Let's move on to a peer group comparison.

32:38 Let's go from easiest to most difficult scenarios for peer group. If you're pricing a company where there are dozens or even hundreds of companies in that space that are publicly traded, your life got a lot easier. Why? Because every one of these companies has an updated price and a standardized metric. They're all covered often by the same accounting standards. That's the easiest scenario. If you're pricing a company where there are lots of other companies, but they're privately held, things get messier. Why?

33:08 Because you don't have an updated price. You have a price from the most V recent VC round, but that might be 3 months ago, 6 months ago, 9 months ago. VC pricing is always going to be messier even when you have lots of companies because the pricing is stale. And finally, let's think about the most difficult scenario in pricing. You're pricing a company that's privately held and there are only one or two or three other companies that are also privately held. Now you're in trouble, right? And in many ways you can see this is the challenge you you will face with open AI and anthropic. In fact, when anthropic goes public, what exactly are you going to compare it to? You can say open AI but open AI is still private. You'll have to compare it to a VC pricing and is that truly reliable? You say, "What about SpaceX?" Well, SpaceX is publicly traded, but it has two other businesses.

34:03 The space launch business and the satellite internet business acting as noise. So, here's what I'm going to argue. I'm going to argue that pricing doesn't require that you stay with other companies just like yours. It often just means picking a pure group that investors in your company will be comparing your company to. So, I might be getting ahead of myself on this, but my guess is if Anthropic goes public, 1.5 trillion, 2 trillion, you pick a number that it's almost automatically going to get slotted in with the Mag 7 and perhaps it's SpaceX. You could talk about these are different companies. I agree entirely. Apple is a very different company than Enthropic. But let's look at what the numbers will play out like because you're going to see the arguments based on these numbers.

34:52 Collectively, if you look at the Max 7, Alphabet, Apple, Amazon, Meta, Microsoft, Tesla, and Nvidia, and they're widely varying numbers across the companies. Collectively, these companies trade at about 8 to N times revenues. Enthropic if it goes public at 1.5 trillion will be trading at 48 times 20 the trailing 12-month revenues and about 15 times revenues in in 2026. And you know you can use best case worst case but essentially even the best number you put on if if it's anthropic is going to look expensive at 1.5 trillion 2 trillion if you use current revenues.

35:37 Now wait there's a way in which you can make this play in your favor in entropic ball which is project out revenues to 2030. Remember you know anthropics projected growth will give them 300 billion in revenues in 2030. Most of the max 7 don't have that kind of growth. It is entirely possible and I didn't try because I'm not going to be in the business of forecasting 2030 revenues for all of the all of the max 7. You could perhaps make the case with enterprise value to forward sales in 2030 that anthropic is cheap relative to the max 7. You're saying this is messy.

36:12 You can come to whatever conclusion you want. Hey, welcome to pricing. You walk into the game with a bias. You're going to find a pricing metric that backs up your bias. Now, as you use these pricing metrics and much of the world is built on pricing and you ask me which metric is the best one, there are two simple tests I will run. One is you know if if you use this metric how many companies are you going to get in your sample if you lose half your sample it's not a good metric for you if you're valing young or pricing young biotech companies don't use price earnings ratio 3/arters of your companies are money losing this is a very simple test don't lose much of your sample the second is once you pick a metric you got to relate it or connect it to value why because those connections will tell you what you need to control for there is no best metric.

37:02 There's a metric that you will use and as long as you control for the things that relate that metric to value, you're going to be okay with that metric. You see, what are you talking about? Let's start simple. Let's think price earnings ratios. The scalar is base your earnings. What's the connection between earnings in the most recent 12 months to the value of a company? Well, the earnings on most recent 12 months become the base on which you grow to get future earnings.

37:30 So the first connective variable here is expected earnings growth. Higher earnings growth will give you higher future earnings. So holding all else constant. Higher growth companies should trade at higher PE ratios than lower growth companies. Okay, that was straightforward. But you don't get paid out future earnings. You get paid out future cash flows, which is whatever is left over after reinvestment. You're saying, how do I know how much a company has to reinvest? Companies with higher return in equity can deliver the same growth with less reinvestment than companies with low return equity. So second stop, companies with high return equity will have higher PE ratios.

38:08 Growth matters. Your return equity matters. And finally, it'll be nice to have predictable future cash flows, right? Risky cash flows are worth less. So higher risk, which translate into higher cost of equity, will decrease your price earnings ratio. How does this play out? You're trying to estimate the price earnings ratio for a company looking at a peer group. You got to control for differences in growth and risk and return equity to make that judgment. Now, there's zero chance you're going to be using PE ratios for enthropic, right?

38:38 Or open AI. It's going to be a revenue multiple. So, let's look at the connection between revenues. And I talk about base year revenues. You can use annualized, you know, ARR, you can use trailing 12-month revenues, your starting point. Let's see how revenues in the most recent period connect onto value. First, those revenues become the base on which you apply revenue growth to get to future revenues. Higher revenue growth holding all else constant is good. It'll push up your re enterprise value relative to revenues.

39:09 But revenues themselves don't deliver value. You have to look at profits. So your expected profit margins which come from unity economics and economies of scale will determine what your future earnings are. And as with PE ratios, you don't get to take earnings out of the company. You got to look at what you need need to reinvest to get your revenue growth. That reinvestment if it's done efficiently have a higher expected return on capital, a much higher sales to capital ratio will translate into much higher EV to sales ratios. And finally, risk is a given.

39:42 Higher risk in your cash flows will translate into higher cost of capital, lower value, revenue growth, margins, reinvestment risk. Now, I've already given you the ammunition for that sales pitch you're going to hear on Enthropic, right? Enthropic is trading at let's take 15.28 what you know the number that we saw for trailing 12 month revenues. 15.45 times revenues. But that's okay. It's a good company. You're going to start with revenue growth. You're going to ask about margins. You're going to ask about the efficiency with which revenue growth can be delivered, which is primarily the investments they have to make in compute cost as they go along and the risk. You think that's a pain in the neck. That's why I'm not valuing companies. Hey, guess what? When you price companies, you're making exactly the same assumptions I will have to make to value the company. You're just making them implicitly.

40:38 So as we look at Enthropic's potential IPO, you know, it is clear that for Enthropic to have the 1.5 to 2 trillion in enterprise value, it's revenue growth has to be not just high but but through the stratosphere because you got to make that that 70 billion or 65 billion in in in annualized revenue run rate. It looks like a big number but that number has to become hundreds of billions of dollars very quickly. So you need to hear a story about the market for AI and why they can grow. The second is there's got to be a part of the story about profitability. You know unit economics and economies of scale and that story has to then talk about the gross margins that anthropic has and also about the economies of scale which are primarily the compute costs growing slower than revenues. The third is a risk story about not just operating risk but the fact that you have these these compute commitments not content commitments but compute commitments that you've got to meet which creates financial risk and regulatory risk growth profitability risk have to be part of the pricing story and to make a case for entropic to be worth 1.5 or 2 trillion you need high revenue growth with operating margin surging to 30 35 40%.

42:00 and favorable unit economics and economies of scale driving those margins and a risk profile that quickly converges on average. So regulatory risk rec finds its feet. Is there a plausible story? Of course. So clearly there's a story to be told but there's a lot of work to make that story into into reality. So as Anthropic heads towards its IPO my suggestion is you keep your eyes on the price. I'm reminded of the as you listen to stories of anthropic, you're going to get from both the bulls and the bears, you're going to get distractions.

42:35 They'll take a strand of the story and make it the central part. So, if you're a bear and you hear a lot, you might hear a lot of talk about the depreciation of comput the size of the compute commitment and how it might put anthropic into the ground. You know, if you're an anthropic bull, you're going to hear hear a lot of talk about gross margins being 80% and how big the the total addressable market for AI is. I'm not saying any of these trends by themselves are bad, but I'm saying it's, you know, it reminds me of this fable. I think it's an Indian fable of these six blind men who are taken to to visit an elephant to to and each of them touches a different part of the elephant and extrapolates from there what the elephant actually looks like. Remember they're all blind so they have to extrapolate based on what they felt. The guy who felt the trunk describes it as you know very differently than the the guy who felt the the head or the the tusk or the tail. And in many ways, that's what you're going to see with the analyst giving you pieces of the hole.

43:41 So your job in a sense is to fill out the rest because if you're an anthropic, well, your story has to start with AI being a really big market and anthropic being one of the big winners. Because to get the pricing you to to get the $ 1.5 or 2 trillion pricing, you need high revenue growth, great margins. If you're a bear, your story will start with AI as a market not taking off either because of you know regulatory push back or u or AI products and services being unreliable and enthropic itself as a company being you know constrained by the fact that it has weak corporate governance and you don't trust Dario to run the company and the pricing that the bears will use is they say look at the multiple of revenues you're paying and there's no way you're going to make it back. Look at the size of the compute commitments.

44:34 There's no way you're going to cover that. And I think that now both sides are telling you a part of the story. And you have to then decide where you're going to go. It might look like coping out, but I don't know. I don't think I can put myself in either of these groups. And you're saying that's a cop out. You got to pick. No, I don't have to because I don't like pricing. I don't feel enough faith in pricing to be able to pull the trigger. I need to get my hands around the details of all of these things were skating around in pricing and I hope to do that when I do an intrinsic valuation of enthropy. You see what's holding you back? Well, I don't have the full prospect. I've got these leak numbers right now and I do hope that I will get that. You know, the the the rumors are it should be coming out any day now, tomorrow, early next week.

45:28 And I plan to do my intrinsic valuation and I plan to do it well before the bankers attach an offering price because as I pointed out, you know, bankers don't value companies, they price them just like VCs do. There's not much you can learn waiting for the offering price to come out. I hope you found the session useful and I thank you very much for listening.