Heeding the Feedback Loop: A follow-up to my Anthropic Valuation

0:00 Hi, welcome back. In this s session, I'm going to revisit a valuation I did just about a week ago on Enthropic ahead of its IPO and I'm going to draw on something that I developed while I was writing a book on how to convert stories numbers. The fifth step after you've told a story and value a company, the final step in in all of my valuations is what I call keeping the feedback loop open. What does that mean? Listening to people who disagree with you. learning from them but not always accepting what they suggest you should do. Now my anthropic valuation did draw some strong reactions both positive and negative. Some of you agreed with me, some of you did not. And as I mentioned at the time that I did that valuation, I'm not interested in selling my valuation to the rest of the world. I'm not an equity research analyst. I'm not an investment banker. I value companies for an audience of one.

0:55 And if you find it useful, I said take it with you. In fact, during the enthropic valuation, I said, "I'm going to open up the spreadsheet. Why don't you go in and change the numbers that you don't agree with me on and see what value you get?" Much of the push back that I got on the valuation was thoughtful and productive and I tried to incorporate into my thinking. Some of it is emotional. You know, emotional what sense? I don't like the company. I don't like AI. Therefore, I'm not going to invest in the company. Perfectly okay.

1:24 But some of it I think reflects a basic disconnect with markets and pricing and basic business sense. And I want to talk about that part of the feedback because it's not just about this company but almost every company that you value that you're going to run into this issue. Now as I go through this process, I'm going to start by playing devil's advocate with my own valuation. I find this a very useful exercise. In other words, value a company. Then put on a different hat and push back at your own valuation. Of course, to do this, you need two things. One is you need to understand how the inputs you have play out in value and how they connect to your story. That's part of the reason I keep my valuations parsimmonious. You know, when you have 400 line items, God knows what each input is doing. With my 15 to 20 line items, I know exactly where each input goes. And the second is it requires a willingness to be honest with yourself about which of your story lines are most fragile, more likely to be broken and weakest and which ones are strongest. So let's start the process by giving you a review of the valuation I put out last week. Now of course there's a complete summary of what I did last week. So nothing new here. So I'll review the story and my story. Anthropic is going to be an incredibly successful company in an AI space that's going to become much larger than it is today. In fact, in my story, Anthropic ends up with a trillion dollars in revenues, a mix of more premium price products than mass market, giving a margin of 32.5% which are just for regulatory drag of 5%. So basically 27.5% margins and a reinvestment that reflects compute power becoming easier to build and access in the future. $3 of revenues for every dollar of capital. I did include a cost of capital of 9.74% which I think was one of the biggest stumbling blocks for some people and I'll come back and talk about why I don't think it's the biggest issue in this valuation. You might agree or disagree with me but the bottom line was I came up with a value between 1.4 45 and $1.5 trillion for the company. That's an insanely large number, right? But still lower than the expected pricing range of 1.5 to 2 trillion. So as before we embark on looking at the weakest links, you know, a few details that I think I should emphasize. First is the value per share that you saw in that spreadsheet is a is a placeholder. Why? Because the share count has not quite been set yet. So I included the share count that you saw in the prospectus, but that share count will change before the bank has set an offering price. It's not going to change the aggregate value of equity, but the value per share is going to change. And it did include the restricted shares, and I will continue to do that when the share count changes. I use 9.74% as my cost of capital. And I might not have had a chance to go into the details, but I did include a cost of debt based on a doubleB plus rating, a below investment grade rating, which the Financial Times in a news story on Friday, 3 days ago said might likely be the rating going through, which is a below investment grade rating. I tried my best on compute commitments. is sprayed all over the prospectors and by my count about 450 billion in in compute commitments that are contractually sent and another couple of hundred billion on top of that which are commitments on which there's more flex the present value of the contractual commitments is about 373 billion I do include that as debt when I did my valuation now some of you did wonder where the dilution risk is I mean the delusion risk in the case of enthropic is going to show up in multiple places the restricted shares that they've already issued. I've already counted those as shares. You say, "What about the options they've given employees? I value the options and reduced the value of equity. Now, there are convertible bonds and again I've made the assumption that the convertible bonds will get the conversion is already captured in the value of the convertible bonds or they're going to get converted at the time of the IPO. Now some of you also I think prudently thought about the fact that in the future anthropic might need to raise capital and that'll cause more share count in a discounted cash flow valuation. You issue new new shares only if you have negative free cash flows and if you have negative free cash flows it's already incorporating value through the present value. So that's just know kind of tying up loose ends.

5:59 So with that in place, let me take a look at my own entropic valuation and it's at the low end as I said of the offering price estimates and you know and and many of you asked am I not concerned? Am I not uncertain? And my response is of course I'm uncertain of course I'm concerned but what's concern going to do for me? Of course I'd rather be right than wrong but I'm going to be wrong 100% of the time. So, I want to list out the assumptions that I made in order of most important to least important. I list at least the top five.

6:31 The ones that worry me the most or I'm most concerned about where I could be most wrong to the ones I don't care as much about. Let's start at the top end of the of of the assumption ladder. The biggest assumption I made in anthropic valuation was assuming that by the end of the next decade, next 10 years, their revenues would build up to a trillion. Let's pause right there. Trillion would make them the largest company in the world in terms of revenues right now.

7:02 Might not be 10 years from now, but that's really big revenues. To make that, I made a series of assumptions and I decided to break down what those assumptions had to be and what could go wrong. So I started with the target total market for AI products and services. That's going to become between 5 and 10 trillion by the end of the 10th year. That requires widespread use across businesses, more employee replacement than tools and customers being willing to pay high prices to have access to these tools. What can go wrong? The product and service market might end up being more tool than employee replacement and only in a subset of businesses.

7:39 Then I assume that a significant portion of that market will be for premium products, frontier models where people will pay high prices. And there again, what can go wrong? The premium part of the AI market may end up being a much smaller percentage. Why? Because open-source models might be able to do a lot of what these premium models do and companies aren't willing to pay. I did also building on top of that, I did assume that Enthropic would be a winner, have a dominant market share in the premium AI market. partly because it is in the lead right now and perhaps that lead's going to allow it to consolidate market share. What could go wrong? The premium market could end up becoming splintered, perhaps even disrupted in which case entropic might be have a much smaller market share. One final part, I did buy into some the anthropic story that they could build on add-on markets.

8:32 They use the example of the drug market where in addition to providing AI tools to companies in that segment they are able to enter the market themselves at least for a portion of the market anthropic drugs. Well, what can go wrong? Maybe those markets those those businesses will require skills that anthropic doesn't have. I mean to give you an example Google has been able to build self-driving cars. But does that mean Google wants to be a ride sharing company? I don't think so. It's a very different set of logistical skills needed that Google doesn't have and the same thing might get in the way of anthropic entering other businesses. So my target revenues are built on layers of assumptions and if any of them or all of them fail, the revenues are going to be below 1 trillion. How much below will depend on how many fail. So that's to me the biggest assumption. The second biggest assumption to me is that governance here is not going to come back and bite us. Now for some of you, you might think that's not a number.

9:34 It's not a big deal. To me, it is a really big deal. And let me explain why. The story I'm telling for anthropic requires there's a plausible pathway to get there, right? But there are lots of things that can go wrong along the way. some temporarily, some permanently, which means management at comp at at Enthropic has to be willing to change and adapt and sometimes go back and reverse policies that they themselves adopted. That's really tough for human beings to do. Now, I know the push backers, they're really smart people.

10:10 That might exa be the exact problem. They're really smart people, but really smart people hate admitting mistakes. You need somebody acting as a check, somebody pushing management to change even though they don't want to. And my problem with the anthropic structure where you put all the power in the hands of seven people, the founders, is if there's a mistake and they all have the same world view and they come from the same background, they think the same way about the world and operate in an echo chamber, change is going to be really difficult. So to me that's the second biggest issue with the entropic valuation. It's not showing up in the numbers now. It'll show show up in how much below the value the price has to drop before I'm willing to buy. The third link the third big assumption is the profitability assumption. My target operating margin before I make the regulatory adjustment is 32.5%. But let's think about the assumptions you need to make to get there. The first is that with compute costs moving from the variable cost to the fixed cost column that the gross margins will remain high 80% or higher. What could change that pricing pressures could put the push the price of AI products down and the cost that go in might increase. So the gross margins might actually be less than 80%.

11:30 As revenues grow, we are making assumptions that economies of scale are going to kick in especially on the compute costs. So we need that economies of scale and if compute costs and employee costs grow at rates faster than we expect the economies of scale are going to get weaker. We also assume that anthropic can keep delivering these margins long term. Now one of the words that kept coming up in critiques was modes. What's anthropic mode? Now you know many of you pointed out that there's no visible moat. I I I tend to disagree. In technology, the modes aren't like modes that you find in consumer products which could be brand name or in commodities which could be cost advantages. Technology often first movers who can establish stickiness with their products where customers once they switch into your product find it difficult to switch out have significant modes. I mean, think of Microsoft beating Apple even though Apple had the better upfront operating system. I know I I know I'm biased on this one, but Windows came later and Windows eventually won because it became the default. The way Office got entrenched in businesses because once you're into Excel and Windows and PowerPoint, it gets sticky to switch up. My guess, you know, in my story, I'm assuming that Anthropic is going to develop stickiness by doing what? By having premium products that embed themselves in their customers data and usage and it becomes very difficult for customers to switch out because of course this could be a pipe dream. There might be no modes. So any of the three assumptions break down.

13:17 Unit economics are not as good. economies of scale don't kick in as much as expected and profits are not sustainable, the target margin is not going to be 32.5%. It's going to be a number much lower than 32.5%. Now, compute power is the big investment that that the anthropic is going to have to make. And here again, I've made some assumptions to get to my base case, which is $3 of incremental revenues for every dollar of incremental capital.

13:47 Now I you know the way I get there is I first assume that through the computer agreements Enthropic is already entered into it's locked in enough compute power capacity to meet requirements in the near future. It doesn't have to do this in the next year or two. Why does it matter? Because right now compute power is in a crunch as revenues grow. I'm going to assume that Enthropic will not just be able to add compute power but at costs lower than it is paying today as chips become cheaper and easier to access. Of course, that assumption could break down if chip costs, you know, chip cost might decrease, but AI agent power costs and other costs might go up to offset.

14:26 And anthropics ambitions leading into new businesses like drugs, we have no idea what the reinvestment there is going to be. Maybe it's going to be higher, maybe it's going to be lower. So those three assumptions about base compute capacity, growth reinvestments in other reinvestment are giving me the three $3 of revenues for dollar and capital. But if that's not the case, anthropic will have to reinvest more to deliver the same revenues. Which brings me to the risk factor. If you remember when my did did my anthropic valuation, there were three ways risk played in. One is through a cost to capital of 9.74%.

15:05 I know I know I'll come back and talk more about that. The second is a failure risk which is hey Enthropic is still a young company things can go wrong regulatory risk loom and attach a failure risk of 10%. I'm going to argue later that Enthropic is no core weave that this is not the kind of company where there's a 40 or 50% failure risk. You might disagree but my failure is so I I am okay with the cost to capital from the failure risk. The risk I feel least comfortable about is my regulatory risk assumption, which I assumed would knock 5% off the margins, but who really knows? I mean, political change pressures might build on those regulations, make them much more ownorous than we think they're going to be today. But that's a risk that I worry about the most. So, with all of that, you're saying, if you're so uncertain about anthropic, why bother valuing it?

15:59 Why would you ever, you know, value company when there's so much uncertainty? I've thought about it a lot because I have invested in companies where I face a lot of uncertainty and I've you know considered what it is that allows some people to invest in others to hold back and when you look at investments with a lot of uncertainty there are three pathways you could take one is you can say I'll avoid them that's what a lot of conservative investing suggested right don't buy companies where you don't know what the company does traditional old-time value investing the second is to use a margin of safety another value investing device. What's the margin of safety? You get a value 1.5 trillion for entropic.

16:37 Knock off 40%. Don't buy it unless it's 900 billion lower. I can understand the intuition and logic for using a margin of safety. But I'm going to present you with a tradeoff on margin of safety that might keep it from becoming too big a number. And third, I'm going to talk about using the law of large numbers to kind of allow you to take risk you otherwise wouldn't have. saying what are you talking about? Let's hold off on that for a moment. But let's talk about the margin of safety. As I said, the margin of safety comes about when you value a company and you price it. The price has to be a certain percentage under the value. 20% 30% before you before you'll buy it. Now, when I first present you that proposition, you're saying I'm going to make it 80% lower.

17:23 More is better, right? Not necessarily. Because here's what happens when you increase the margin of safety. use 30 instead of 20, 40 instead of 30, you reduce the likelihood of what I call type one errors. Type one errors are when you buy the wrong stock or fail to sell the right stock. You're avoiding that mistake, but you're increasing the likelihood of type two errors. When you put a margin of safety of 40%, there are going to be investments you should have made you're not going to make because you set the margin of safety too high.

17:53 See, how do I know what's right for me? First, it depends on your risk aversion, but it also depends on what your portfolio looks like right now. If your portfolio right now is 50% in cash or 70% in cash, your margin of safety has been set too high because that basically means you're not investing in anything because you're too you're too risk averse. And are you earning a return on your portfolio that is more than you'd have made investing in index funds? So those things should drive your margin of safety. Higher margins of safety are not necessarily better than lower margins of safety. Now do I have a margin of safety for anthropic? Probably. But more for the the governance risk than for uncertainty. And here's why I can live with uncertainty.

18:44 If I had a put, if I had a million dollars to invest and that's all I had and you ask me, would I buy anthropic with the entire million? Of course not. Absolutely no. Even if it's undervalued, why? There's too much uncertainty. The more concentrated your portfolio, the more difficult it is going to be for you to pull the trigger on stocks like anthropic where there's a lot of uncertainty. Some of you might know I don't believe in concentrated portfolios. I spread my risks. I know that makes me chicken. I'm okay with that. But what spreading your risk allows me to do or you know allows you to do is essentially buy a stock like anthropic. It's one of 20 one of 25 stocks. And here's how the law of large numbers comes into play. You're wrong on each and every one of those 25 stocks.

19:36 Some more, some less. But if on average you're bringing something to valuation, that's all that matters across your portfolio, you're going to be okay. It's not having a concentrated portfolio that allows me to invest in companies with a great deal of uncertainty. So now let's turn to your push back feedback. As I said, I structured my spreadsheet to allow you to disagree with me. In fact, I created a Google shared spreadsheet where you could go go in use my spreadsheet, put in your inputs instead of mine and come up with values. There about this morning 70 and it's still a live spreadsheet. So, if you want to go in and add your name and numbers to it, please do so and you can use my spreadsheet as a tool. Here's what the distribution of values look like. About maybe 65% 70% of you felt my value was too high.

20:34 perfectly okay right because that's what the whole point of this exercise you know and there were I think only two out of the 70 valuations came up with values greater than two really so basically you think I'm skewed upward that maybe I am biased upward that could very well be the case but I then also started looking through the comments and I looked at why you disagreed with me okay so if you break down the valuations values too high outnumbers value is too low and if if you're within 10 15% of my value I'm going to say you agreed with my valuation so I looked at when the value is too high and focused on that component what you thought the reason was the most common was your gut you basically said there's no way an anthropy could be worth that much or any company can be worth that much I don't even know how to deal to address that because I'm I don't know what's in your gut but the second biggest reason for disagreement with me was my cost of capital. You felt my cost of capital of 9.74% was too low. And I find that a little disappointing because and and not surprising both disappointed but not surprised. Disappointing because cost of capital is never the biggest number in your it should never be the biggest number in your valuation. The fact that you focused in on that disappoints me because they're far bigger inputs that drive value a lot more. But I understand. I understand because for many analysts and appraisers, the cost of capital is the center of the DCF universe. It's the D in the DCF. And people obsess over the number. They finesse it and they fight, you know, straw men. It's a CAPM. it's beta you know and most of all I think for many analysts it becomes the receptacle for their hopes their fears their fantasies everything goes into the cost of capital so again this is not so much about anthropic but in general when you do discounted cash flow valuation I want to start by first giving you an intuitive feel for what the cost of capital is what is it trying to measure without using models and without using equations I call the cost of capital the Swiss army knife of finance. Let me explain. I teach three very different classes in finance. A corporate finance class which is about businesses and first first financial principles associated with managing businesses. How to invest, how much to finance, how much to borrow and how much equity to use, how much to pay in dividends. Corporate finance is a player there. It's a player because it becomes a hurdle rate that we use to decide whether to take a project or an investment. It's an optimizing tool to decide what the right mix of debt and equity is. It's it becomes your metric for determining whether you should be returning more cash to your shareholders or reinvesting back in the business. It's everywhere in corporate finance. Then I get to my valuation class and there we're looking at the same business from the outside in. The same cost of capital now becomes the discount rate we used to value the company. Cost of capital is everywhere and depending on where you use it, people often have very different views of what it should be. So let's look at the components that go into cost of capital before we delve into what a reasonable number might be for a company. The cost of capital is a weighted average of what? the amount of equity you have times the cost of equity and the amount of debt you have times the cost of debt.

24:16 Cost of equity has a risk-free rate base both the cost of equity and the cost of debt. But to get to a cost of equity at a risk premium is a function of both a market number and how much risk you're safer your company is relative to the market. And for a cost you add a default spread based on how much worry you have about default. And of course it's a tax issue to the extent that interest is taxdeductible, cash flows to equity are not you have one minus tax rate there.

24:45 That's a cost of capital calculation. Now the question is how do we estimate this number for a company? Now as you know as you take books and you read through them you know I think that sometimes in the process of developing models and equations we've we've removed the intuition and common sense from the estimation. Here's what a cost of equity is supposed to measure. Forget about all the models you learn. It's supposed to measure what an investor can make on an investment of equivalent risk elsewhere in the market. Let me repeat that again.

25:24 The cost of equity in an investment should reflect what investors can earn. It's an opportunity cost on an investment of equivalent risk. Key word is equivalent risk. It is subjective. It can vary across investors based on risk aversion. So you're saying I can make up whatever I want, right? Not so fast because there are reality checks. You got to live in the world you're in, not the world you wished you were in. So when I ask you what the cost of equity is for a typical US company, I'm going to pause on this slide before I give you an answer. I want you to think about what number comes into your head because now I'm going to come back to you with the reality checks. The first reality check is history. What does history show us? History shows us how much you as an investor could have made investing in equities.

26:13 Here I have, you know, historical returns on stocks and both all stocks, small stocks, treasury bills, T-bonds for the last 100 years pretty much through 1928 through 2025, last 50 years and last 25 years. Over the last 25 years, the average returns you'd have made investing in equities collectively in the US is 8.74%. Over the last 50 years, that average return is a little better, 11.92%.

26:46 Over the last 100 years, it's 10.13%. You think, so what? You walk around telling me that the cost of equity for an average firm is 15%. Take a look at history. You never would have been able to deliver that 15%. Now, one of the issues with average returns is they vary depending on what the T- bond rate is. So, one of the other calculations I've made here, and you can see that in the little table to the right, is the risk premium you'd have earned over equities over the last 100 years. Between 1928 and 2025, US equities on a geometric average basis, compounded basis, would have earned about 5.54% more than the debond rate. Say, what do I do with that?

27:30 What's a debond rate right now? About 4.9%. 4.9% plus 5.54% gives me 10.44%. If history were entirely guiding you, you should be getting cost of equity for average firms what 9 10 maybe 10 12%. Not 12%, 15% or 18%. And here I would add a cautionary note. This the US was one of the most successful equity markets of the 20th century. I'm giving you the returns you'd have earned on stocks in the most successful market. You're saying so what this is selection bias. You could argue that you're overestimating what you learn over the next 10, the next 20, the next 50 years.

28:13 So I'm going to give you a second test you could run. Every month at the start of the month, I compute an implied equity risk premium for the asset provider. You say, "What the heck is that?" It's just a very simple way of backing out given this given the level of the S&P 500 what you can expect to make as returns on stock. So the S&P 500 on September 30th, 2026 was selling at no 7,651.54. You estimate expected cash flows. I know I know those estimates could be wrong. Hang in there though.

28:46 You estimate the cash flows you'd receive in dividends and buybacks and you solve for the discount rate that you'd earn. At the start of October 2026, that number was 8.99%. In fact, I just updated the November number, it's 8.76% or I'm sorry, 8.83%. 8.99%. You're saying, "What am I going to do with that?" At the start of October 2026, US equities were priced to deliver a return of 8.99%.

29:18 Hey, there's your cost of equity. You got to live in the world you're in, not the world you wish wished you were in. You might tell me you want to make 15%, but you're in a market with stocks in the aggregate, a price to deliver about 9%. Now here again, I can subtract out the T- bond rate, come up with an equity risk premium at the start of October, that number would have given me about 3.7% over the T- bond rate.

29:45 Now you say but there have been times in the past where this expected return would have been much higher right yeah let's check here I've computed that implied expected return what you saw in October going back to 1960 have there been times where the cost of equity for US companies have been 10% or higher yes but they were all before 1994 and if you tell me your cost of equity is 15% the only times in history where US equity ities have delivered an expected return greater than 15%. We're in the late 70s and early 80s. And you know why they deliver that return? Cuz t-bond rates were 10% 11% 12%.

30:26 So when you give me a cost of equity 15 to 20% for US stocks maybe are stuck in time in the 1970s or early 80s, but US equities have not delivered an expected return greater than 10% in 32 years. So that's the second stop. you're still not convinced, right? So, I'll give you a third way where you can decide whether your cost of equity makes sense. You could take a very simple, in this case, a dividend discount model with a little bit of algebra and trust me on this one. And recognizing that what companies can pay out as a function of their growth rate and return equity, you can write the forward PE for the market as a function of the growth rate in earnings, the return in equity in companies and the cost of equity. Let's throw some numbers that I think are pretty reasonable. Now, let's say the nominal growth in the US economy is 5%.

31:22 Okay? Say the average return equity across across firms is about 20%. That's a generous estimate. If your cost of equity is 15%, your PE ratio should be roughly 7.27 for the entire market. When was the last time US equities traded at a singledigit PE? And given that the PE today is closer than 20, if you truly believe the cost of equity is 15%. You should have none of your money invested in equities or you have to revisit your cost of equity assessment. There's no other choice here. So already you can see that if you're looking at an average company in the US, you're looking at no 8% 9% cost of equity and a cost of capital that's lower because debt is no on average at least cheaper than equity.

32:16 So let's talk about the cost of debt. Again, it's shrouded in mystery, but the cost of debt is the rate at which you can borrow money long term today. I know a lot of the discussion of cost of debt at Enthropic has come from the fact that many of these AI data centers have used insane amounts of debt and you're worried about default risk. I understand but Enthropic is not coreweave it's not a company where default risk should lead your story but I understand it's become part of the discussion and you're saying well why aren't you factoring in default risk? I am through the failure risk. But the question is what is it going to lead to in terms of a cost of debt? Some of the comments suggested that it'll cost entropic 10% or more to borrow money.

33:05 I seriously doubt it. I mean I you know they there might be other companies in the space where the cost of debt is 10% or higher. But the I think one way to think about the cost of debt is what kind of rating do you think Anthropic is going to have after its IPO. So here's why ratings matter. As your rating decreases, your cost of debt goes up. There's a default spread. So if you think a is going to be double C-rated, of course, all of this. But remember, when you have a double C rating, that's really concerns about default that are pushing up your cost of debt. As I mentioned, the s the the rating that Anthropic is likely to have after the IPO is closer to double BB plus where default spreads are not going to be monstrously large. They're going to be large. The cost the cost to debum ended up using for anthropic is about 6 and a half% pre-tax. But I think that's a pretty reasonable judgment.

34:08 So when you see costs of debt out there of 10 15 20% a lot of that is due to failure risk and I separate failure risk. In other words even if you ask me to value core we I'm not going to use a 15% cost of debt the cost cost of capital I'm going to use a much more reasonable going concerned cost of debt maybe seven or 8% but attach a really large failure risk. If you take my cost to capital of 9.74%, my failure risk of 10%, the effect I'm having on value is pretty close to what I'd have got if I just used 11% cost to capital. So if you think the 9.74% cost of capital is low, perhaps you're not factoring in the failure risk issue, which I think is the issue that you should be fighting on if you're worried about, hey, there's too much debt and they might not make it.

34:57 Which brings me to the cost of capital. Start of every year I create a histogram of cost of capital for US companies. This is September 2026 version. What does that tell me? It tells me what a typical cost of capital is in every sector. Right? These are just US companies. I've also computed the first decile, the first quartile, the third quartile, the ninth decile. Let's take energy. The first decile is 7.16%. My ninth my median is 7.6%.

35:32 And my ninth decile is 10.96%. 80% of energy companies have cost a capital between 7 and 11%. That's it. 7 and 11%. That's a pretty tight range, right? Let's take technology companies. The first decile for cost of capital for technology companies is 9.88%. The first quartile is 9.92%. The median is 10.75%. Third quartile is 12.3%.

36:03 Ninth decile, the 90th percentile is 14.5%. Terms of cost of capital, technology companies, 80% technology companies have cost of capital between 9 and 12.3%. Now, do you see why I think finessing the cost of capital is a misdirection of resources? there's not much room to run. Now, the way you get from the numbers we had for the average company to these cost of capital is you have to have measures of relative risk. We'll come back and talk about it. But the point I'm making is once you become public, the range on the cost of capital becomes pretty tight. And once it becomes pretty tight, finessing it is not going to do much in terms of adding to your value or making your value more precise.

36:50 Now, as you look at entropic, is it possible that they're riskier than I estimated? Maybe we should put them at the 90th percentile. Give it a cost of capital of 12%. But you can. I'm okay with that with two caveats. One is go back and fix that failure rate. Setting the cost of capital at 12% and setting the failure rate at 10% will really mean you're using a cost of capital of 13 1.5 or 14%. The second is if you're doing a valuation, remember your cost of capital will change over time. Why? Because your story is changing over time. With Enthropic, what's happening over time, at least in my story, is the company's going from less than a hundred billion in revenues to a trillion dollars in revenues. From losing money to making insane amounts of money. Be consistent.

37:36 Your cost of capital, even if you think it's risky today, shouldn't stay at that same level through. If you factor in both issues, the failure risk capturing that default risk you're worried about and the change in cost of capital over time, you can see why, at least for me, the cost of capital is not the hill I want to die on. It's not a number that is really driving my valuation. Now, you'll say, "What if Enthropic doesn't scale up its revenues? Won't they go won't they have a higher cost of capital?" Hey, trust me, if Enthropic doesn't scale up revenues and doesn't make money, it'll die on its own. You don't need a high cost of capital to destroy its value. Its value is going to very quickly turn negative.

38:18 So let's summarize the cost of capital issues because I think they're widely held. There are many analysts and investors out there who say what's wrong with using a higher rate as a hurdle rate. Higher is better. It's more conservative. And with individual investors and privately owned businesses, I probably am not going to push back. So if you as an individual investor say my hurdle rate is 25%. I'm going to point out the consequences of that and move on. It's your money and you're a privately owned business. I'm going to do the same thing. I'll talk about the tradeoffs for you, but you know, it's your money. You can do what you want. But if you're an institutional investor or a publicly traded company, I think there's a lot of gaming of this and a misunderstanding of cost of capital and hurdle rates can have real consequences.

39:08 I think the biggest culprits are what I'm going to call professional investors. They're really not much professional about what they do, but they manage other people's money. mutual funds, hedge funds, private equity, VC and especially VCs and private equity promise returns to the moon. If you get a chance, check out what VCs demand as a target rate of return on their angel financing or the the V the companies invested 25% 30% 40% 50% at least that's what they claim their target rate of return is.

39:46 and PE private equity investors when they try to attract invest you know people to invest with their with them present you with internal rates of return in their investments of 20 25 30%. Both VCs and PES investors are lying. How do we know they're lying? Because we can see their actual returns. This is a table that you know that and it's updated through September of 2026 of what the returns have been on private equity and venture capital, those professional investors who claim to deliver the moon relative to the S&P 500 and the NASDAQ.

40:25 You say, why am I paying these guys 2 and 20 or whatever you're paying them? For the most part, they actually perform close to or under what the market delivers. So there's many a slip between the cup and the lip. You're going around telling people your target rate of return is 25%, you're delivering 12%. Where's the leakage? It comes from the fact that you're pumping into the discount rate things like failure risk that ultimately drag your return down.

40:54 They might cut these guys some slack because they're just trying to sell stuff, right? The problem though is when you set a target rate that's too high as a VC or a private equity investor, you will overreach cuz to try to deliver 25%, you got to do things that break the rules of investing. You're going to go for that 100 bagger cuz you think that's the only way you can deliver what you promised. That's the pathway to investment danger. So I think there's a consequence here for professional investors. He's saying what about publicly traded companies? There are quite a few CFOs who comment on my entropic valuation and said 9.74%.

41:37 We use 14% as our cost to capital company. In fact, I think it's pretty routine in many businesses to set hurdle rates 2 to 3% higher than justified by market numbers. And if you push them, they say, "What's wrong with that? I'm just trying to be careful." Here's the problem, though. If you set the hurdle rate too high, you will turn away more investments. And for companies that have hundreds and hundreds of investments, it's not a big deal, right?

42:06 Early in the life cycle, you can get away with it. But as you age as a company and your investment choices become rarer or more or more scarce, this is going to come back to bite. You will underinvest. There's a potentially second consequence that's even more dangerous. Let's say your true cost of capital is 9%, but you set it at 15%. Nobody's going to be able to find projects that pass that 15%, right? And managers want to take projects, divisions want to take projects, and you're stopping them with this really high hurdle rate. So, you know what they're going to start to do? They're going to start gaming the numbers, pushing the growth rates up, margins up, cash flows higher. Why? Because they got to get past the 15%.

42:52 Nobody takes negative net present value projects on paper because people play with the cash flows and make the cash flows high enough to justify it. As CFO, you might say, well, I know people bias their cash flows upwards, so I'm going to increase my my hurdle rate even more from 15 to 18%. At which point, managers push up their cash flows even more. And by the time you're done, you're taking madeup cash flows and discounting them at a madeup discount rate. I don't even know what that net present value will give you.

43:21 And finally, you know, when you look at those costs of capital I showed you for business and say that looks really low, 8%. Because the cost of capital across the market is about 8 to 9%. Right? You think that's before you decide it's too low, take a look, and this is something I do every year. So this is not just this year. It's been true for for 15 years that I've been doing this. I look at the return on capital and cost to capital for every publicly traded company in the world.

43:50 And I look at the percentage of companies that deliver a return equity greater than the cost of equity. It's kind of a minimal standard for delivering value or return capital greater than the cost of capital. Take a look. Less than 30% of companies globally are able to beat my cost of capital, which you might have thought were too low. Can you imagine what would happen if you raise the cost of equity to 15%, the cost of capital to 12 a.5%?

44:17 You probably will get 5% of companies breaking through. Companies have a tough enough time already trying to beat what I'm estimating as a cost of capital before you put arbitrarily higher numbers as a hurdle rate. Final lend, let's talk about individual investors and privately owned businesses. Now, as an individual investor, you can say, look, my hurdle rate is 25%. Who knows where that number came from? It might have been a number you just made up, a number you heard Warren Buffett mention as a reasonable hurdle rate 30 years ago. Whatever it is, you say, "What's wrong with it? I can demand whatever." You're right. You can demand whatever you want as a rate of return as long as you understand the consequences.

45:04 If you demand a 20% cost of equity, I can tell you with assurance there isn't a single firm in the S&P 500 that would be undervalued. And even you look at the entire market, the companies that look undervalued with the 20% cost of equity will be basket case companies on the verge of going under. So if you follow through with the 20% cost of equity, you should have nothing invested in equities. Say what's wrong with that? Where are you going to go? If your unreasonable hurdle rates carry over into other asset class, you say, "I will not invest in real estate unless I make a 15%."

45:42 Cost of equity or in whatever other investment, you're not going to invest in any risky asset class. You're going to be sitting on cash or worse still, you're going to become target for scams. Why? Because with 20% cost of equity, you know who can beat them? Scam artists. YouTube celebrities, a friend of a friend with some investment scheme that claims to make 20 and you're going to fall for it because you've demanded too high a hurdle rate. So, I will leave the judgment to you. I mean, as I said, you can set whatever hurdle rate you want, but there will be consequences.

46:21 Let's talk about privatelyowned businesses that set hurdle rates too high. Cost of capital of 18, 20, 25%. When you do that, you're essentially setting a hurdle rate that you can't find investments that'll beat basically you'll you'll end up become investing too little and opening the door to more aggressive compar competition often from publicly traded companies. You will lose market share for all the wrong reasons. But there's a bigger there's to me at least a bigger danger which is when you get ready to sell your business. If you've misestimated your cost of capital, you've overestimated, you're going to undervalue your business and sell it for too low a price. And unfortunately, this process is facilitated by private company appraisers who, in my view, just make up costs of capital that are outlandishly high. They use what I what they call the buildup approach, the cost of capital.

47:17 And here's what they do. They start with the traditional cost of capital that 9% you saw for public companies and then they start adding premiums premiums which are thinly sourced like the small cap premium a premium that hasn't existed in close to 40 years since 1981 45 years now they add a small cap premium and they'll add a liquidity premium where they'll double count they push up your discount rate because of liquidity and then proceed to knock your value down for an liquidity discount.

47:48 And then worst of all, they will add another premium that they can't even come up with a good term for. They'll call it company specific risk premium. It's just madeup number. Don't believe any backing that's given for it because it's made up. What are they trying to do? They're trying to get to a cost to capital that's high enough that their gut feels okay with it. The end result is you're valuing private business with cost of capital of 20 22 25%. You're lowering the value. And if you're selling your business to a public company, you're giving a gift to that company because the minute that public company buys your business, all of these premiums disappear.

48:32 So whether you're an individual investor or a privately owned business, think about your hurdle rates. They matter. They change the way you behave. They change the way you run your business and most of all they change the way you think about what the value of your business is. So in conclusion, I revisited my anthropic valuation not because I want to defend myself. There's nothing to defend, right? I've come up with the value. I'm going to invest based on that. I do think the key assumption that I made is the revenue assumption which means all of the information I'm tracking is really about will the revenues come through and I you know the third quarter revenues came out I think today that's not the number that's going to make or break the endgame here and I would I think a healthy discussion of anthropic will be around those revenues which will also be a discussion of what's AI going to do is it a tool pool or a replacement all those questions that we raised in the previous session.

49:36 You're welcome to push back with any cost of equity you want. And I hope you don't take this session as a direct assault on your number. It is just to point out that when you come up with a number for the cost of equity, there are you you have to make sure that that number passes the history test. Historically, stocks have earned about 9 to 10%. The implied expected return step where where stocks haven't earned expected returns in double digits since 1994.

50:07 And let's be consistent with what you think is a reasonable PE ratio of the market. You can't tell me your cost of equity is 20%. And the same bread say you'll buy stocks the PE ratio is 15. The two don't go together. Now, of course, if you're using a 15, 18, 20% rate of return, it might make sense if you're valuing your company in Turkish lamb and quatta, that table you saw was in US dollars. If you're doing that table in rupees, it might be 2% higher across the board. In Swiss Franks, it might be 2% lower across the board inflation adjustment. So, if you're valuing a company in Turkish lamb and quatcha with inflation rate can be 10, 12, 15%, of course, you can use a 20% cost of equity. If you're valuing a company in Afghanistan or South Sudan, yeah, country risk is daunting or you could just be making up stuff.

50:58 Now, do I wish I could make 20% on my equity investments? Absolutely. But I also wish I could drink a milkshake with every meal and not have high glucose and high cholesterol or get older without any of the symptoms of aging. But just because I want something doesn't make it so. So, I hope whatever you think about the cost of equity capital, this session will lead you to rethink that. Think about what's reasonable, what's what's justifiable, because you just can't make up a number and stick with it because it'll have consequences for you that you might not like.

51:34 Thank you very much for listening and I hope you found this useful.