YouTube2Text

'Big Short' Investor Explains How the AI Bubble Will Burst — Transcript

by New Money · 8,475 words · 1,461 segments · language en · Watch on YouTube

Full transcript

  1. 0:00Let's just imagine that Open AI fails.
  2. 0:03Could happen.
  3. 0:04>> The host of the Real I man Playbook
  4. 0:05podcast.
  5. 0:06>> know about you, but I know that I don't
  6. 0:08have a hundred billion dollars to spend
  7. 0:10on building that a sentence.
  8. 0:11>> [music]
  9. 0:11>> You may know our next guest from The Big
  10. 0:13Short.
  11. 0:14>> Your character
  12. 0:15>> You had to work that in there, didn't
  13. 0:16you?
  14. 0:17>> I did have to
  15. 0:18>> Do you like being described [music] that
  16. 0:19way?
  17. 0:19>> it's going to be on my tombstone.
  18. 0:20>> The whole United States of America would
  19. 0:22go into a recession overnight.
  20. 0:24>> Oh, yikes. Okay.
  21. 0:26>> Well, I I have to say you're out of your
  22. 0:27mind.
  23. 0:28>> Yeah, yeah, you're insane.
  24. 0:29>> To get my programming to impersonate a
  25. 0:31deity.
  26. 0:31>> This industry, despite all the hundreds
  27. 0:33of billions of dollars [music] that's
  28. 0:34been spent on it, hasn't
  29. 0:37>> Joining us right now is Steve Eisman.
  30. 0:38>> Let's bring in Steve Eisman.
  31. 0:41>> Steve, welcome back. You are known for
  32. 0:45spotting a bubble before anyone else
  33. 0:47does. Michael Lewis wrote a whole book
  34. 0:50on it. So, my first question to you is
  35. 0:53where on earth are we
  36. 0:55in AI right now?
  37. 0:57>> There was a great movie with Sean
  38. 0:59Connery, where he played this I can't
  39. 1:01remember the name of Finding Forrester.
  40. 1:03And I remember the young character asked
  41. 1:05him a very complicated question, and his
  42. 1:07response is as he's eating soup, he
  43. 1:10goes,
  44. 1:11It's not exactly a soup question. It's a
  45. 1:14complicated question.
  46. 1:15>> [laughter]
  47. 1:15>> I never forgot that line. I thought it
  48. 1:16was one of the best lines in in the
  49. 1:18history of movies. It's not a soup
  50. 1:19question.
  51. 1:21>> Not exactly a soup question, is it?
  52. 1:23>> This is how I look at it. The
  53. 1:24concentration
  54. 1:26risks here are all inspiring. You know,
  55. 1:29so if you take a step back and you
  56. 1:31someone said to me, "Why don't you
  57. 1:33analyze this software company? Forget
  58. 1:35about what it does. It's a software
  59. 1:36company. An established software
  60. 1:38company." And if it turned out that the
  61. 1:40company had thousands of customers,
  62. 1:44that would be great. If it turned out
  63. 1:45the company only had two customers,
  64. 1:47you'd say, "I don't want to invest in
  65. 1:49that because if something bad happens to
  66. 1:51one of those customers, this company is
  67. 1:54is dead."
  68. 1:55There's something of that going on in
  69. 1:57the whole AI story. So, let's start just
  70. 2:00with Nvidia. So, Nvidia, God bless them,
  71. 2:03and I own the stock, okay? When they
  72. 2:05reported a few weeks ago, I think the
  73. 2:07revenue growth was like it was like
  74. 2:09110%. It's a lot. So, let's just let's
  75. 2:12just take a step back just for a second
  76. 2:14and say to ourselves, wait a minute. The
  77. 2:16largest company on planet Earth just had
  78. 2:19Forget about earnings growth, which was
  79. 2:21great, too.
  80. 2:22Just revenue. Revenue growth of over
  81. 2:24100%? Like, that's insane. So, that
  82. 2:28would say the AI story is great. Until
  83. 2:30you read the 10-Q, which came out that
  84. 2:33night, and I'm going to impress your
  85. 2:35viewers by saying if they look at it and
  86. 2:36they they go to note seven.
  87. 2:38>> Oh, okay.
  88. 2:39>> Okay.
  89. 2:39>> Yeah.
  90. 2:40>> Note seven says that 70% of Nvidia's
  91. 2:44accounts receivable as of the end of
  92. 2:47July came from five customers.
  93. 2:49>> Mhm.
  94. 2:50>> Warning flag. Not the end of the world.
  95. 2:53Warning flag. Now, let's go to the
  96. 2:55hyperscalers. You're talking about
  97. 2:57Google, Meta, Amazon, Microsoft. And
  98. 3:01let's throw Oracle in for good measure.
  99. 3:03Massive companies, spending massive
  100. 3:05amounts of money by buying Nvidia's
  101. 3:08chips and everything else under the sun.
  102. 3:1070%
  103. 3:12of their
  104. 3:13AI revenue,
  105. 3:15which equals something like 25 to 30% of
  106. 3:20their entire cloud revenue,
  107. 3:23comes solely from Anthropic and OpenAI.
  108. 3:26>> Right.
  109. 3:26>> Let me just let me just say it again so
  110. 3:28your viewers get it. If you look at Let
  111. 3:30leave out Oracle for a second.
  112. 3:32Microsoft, Amazon, [snorts]
  113. 3:34Google, 70% of their AI revenue, which
  114. 3:38is equivalent to 25 to 30% of their
  115. 3:41total cloud revenue, is just from
  116. 3:44Anthropic and OpenAI. If you go to
  117. 3:46Oracle, Oracle puts out a data point
  118. 3:50called RPO,
  119. 3:53which is
  120. 3:54basically a form of backlog.
  121. 3:56>> Mhm.
  122. 3:56>> When they reported earnings last year
  123. 4:00in October for their August quarter,
  124. 4:04their RPO went from like 150 billion to
  125. 4:07like 400 billion in 3 months. It was a
  126. 4:10massive jump. a massive jump. I mean,
  127. 4:12people went insane.
  128. 4:13>> Yes.
  129. 4:13>> And the stock
  130. 4:14>> went crazy, didn't it? Yeah.
  131. 4:16>> went from 230
  132. 4:18to 330 like in 2 days.
  133. 4:20>> Yep.
  134. 4:21>> And then some of the sell-side analysts
  135. 4:23were very good who did some digging came
  136. 4:25out with reports that said
  137. 4:2750% of that RPO is just from Open AI.
  138. 4:32>> Jeez.
  139. 4:32>> Now Now, today go Oracle's is over 600
  140. 4:36billion and it's still like 50% is from
  141. 4:39Open AI. Basically, 50% of future
  142. 4:42revenue of Oracle is from a company that
  143. 4:45loses money like crazy.
  144. 4:47>> Well, that's what I was going to ask
  145. 4:48you.
  146. 4:49Maybe you can explain how this works,
  147. 4:51but from my perspective, I don't
  148. 4:52understand where this money is coming
  149. 4:55from.
  150. 4:55>> We're coming to that.
  151. 4:56>> Okay.
  152. 4:56>> Okay, let me Let me just finish the Let
  153. 4:58me just finish the chain.
  154. 4:59>> Yeah.
  155. 4:59>> So, now we come all the way down to to
  156. 5:02Anthropic and Open AI. And my view is
  157. 5:05the entire chain from Nvidia to the
  158. 5:07hyperscalers all the way down, the whole
  159. 5:10chain rests on the future health and
  160. 5:12growth of Anthropic and Open AI because
  161. 5:15they're creating the commitments to the
  162. 5:16hyperscalers. If they don't grow and
  163. 5:18have the money to pay for those
  164. 5:20commitments,
  165. 5:21well, then the whole chain slows.
  166. 5:23>> Yeah.
  167. 5:24>> So, that's the risk. I would say between
  168. 5:27the two, probably Open AI is the weaker
  169. 5:29entity, but it's not clear
  170. 5:31cuz really we really don't have enough
  171. 5:32numbers. The only thing we do know, cuz
  172. 5:34The Wall Street Journal reported this,
  173. 5:35so I'm assuming it's true, Open AI had I
  174. 5:39something like 6 and 1/2
  175. 5:42billion
  176. 5:44in revenue
  177. 5:46in the second quarter of this year.
  178. 5:48>> Okay.
  179. 5:49>> Anthropic was at 11 plus.
  180. 5:52>> But this is just revenue.
  181. 5:53>> Just revenue.
  182. 5:54>> Okay.
  183. 5:55>> Open AI lost something had costs of
  184. 5:58something like 12 billion. And so the
  185. 6:00way the math worked was in 3 months Open
  186. 6:02AI Open AI's revenue went up a billion.
  187. 6:05>> Yes.
  188. 6:05>> And its costs went up 3 billion, which
  189. 6:08which we like to say is upside down.
  190. 6:11You want the reverse, not the former.
  191. 6:13They want the other way to go.
  192. 6:15>> Great business model.
  193. 6:16>> And and their revenue grew 18%
  194. 6:20in 3 months, whereas Anthropic's revenue
  195. 6:23grew over 100% in 3 months. So they're
  196. 6:25the weaker company at this point. That
  197. 6:27could that could change. My my only
  198. 6:28point is that this whole this whole
  199. 6:31industry makes me nervous because let's
  200. 6:33let's just imagine that Open AI
  201. 6:36fails. Could happen.
  202. 6:38The whole like the the whole United
  203. 6:40States of America would go into
  204. 6:42recession overnight if this would And
  205. 6:44now eventually
  206. 6:45there'll be a lot more diversification
  207. 6:47and there'll be a lot more companies,
  208. 6:49but that's going to take time. So, you
  209. 6:52know, within the next year or so those
  210. 6:54companies have got to stay healthy.
  211. 6:55That's where I think we are.
  212. 6:56>> So there's a concentration risk.
  213. 6:58>> There's a massive concentration risk.
  214. 7:00>> And that brings me to the to the
  215. 7:01question I I was asking before. My
  216. 7:03understanding is that well, we just
  217. 7:06spoke about these Open AI and Anthropic
  218. 7:09are not making money, they're losing
  219. 7:10money.
  220. 7:11>> Oh,
  221. 7:11excuse me.
  222. 7:13To say that they're losing money would
  223. 7:15be kind.
  224. 7:16They bleed money.
  225. 7:18>> Yeah.
  226. 7:18>> What's the nicest way
  227. 7:19>> They lose a lot of money.
  228. 7:20>> We're pre Yeah, we're pre-profit.
  229. 7:23>> [laughter]
  230. 7:24>> They they make money if you exclude all
  231. 7:26costs. That's how they like to think
  232. 7:28about it.
  233. 7:28>> that's we should start reporting that
  234. 7:30metric.
  235. 7:30>> Yes, we should. [laughter]
  236. 7:32>> Um so where where does the money come
  237. 7:34from?
  238. 7:35>> Well, that's actually a very interesting
  239. 7:36question. I had thought
  240. 7:39that most of the company most of the
  241. 7:40money was coming from venture capital,
  242. 7:43and that actually happens to be not
  243. 7:44true. Most of the company is coming from
  244. 7:47Amazon, Google, Microsoft, and Nvidia
  245. 7:52investing in these companies, and
  246. 7:53SoftBank.
  247. 7:54>> Okay. So, taking equity stakes in the
  248. 7:57>> Taking equity stakes when they raise
  249. 7:58capital, and they those guys have ponied
  250. 8:00up money. Whether they want to continue
  251. 8:02to pony up money, I don't know.
  252. 8:04>> That sounds like a circle.
  253. 8:06It sounds like money or commitments are
  254. 8:08going one way, and then commitments are
  255. 8:09coming back
  256. 8:10>> Yeah, it does have that tendency, does
  257. 8:11it?
  258. 8:12>> This kind of
  259. 8:13reminds me of the It's actually a scene
  260. 8:15from The Big Short with your character
  261. 8:18>> You had to work that in there, didn't
  262. 8:19you?
  263. 8:19>> I I did have to work. I remember you
  264. 8:22saying, "Oh, it's kind of like CDO A,
  265. 8:24and then they put parts of that into CDO
  266. 8:25B, and then those two get put into CDO
  267. 8:28C."
  268. 8:28>> There's some similarities. Obviously,
  269. 8:31but in their defense, there is a
  270. 8:32circularity to the financing, but as
  271. 8:35long as Anthropic and OpenAI keep
  272. 8:37growing very, very rapidly, and people
  273. 8:38keep giving them money,
  274. 8:41the chain will hold. It's if one of
  275. 8:43those two companies really messes up,
  276. 8:46and people people pull money, or don't
  277. 8:49want to invest in anymore. That that's
  278. 8:50when the chain doesn't hold. That's
  279. 8:51where the concentration risk problem
  280. 8:53comes in. If the industry was much If If
  281. 8:55this If If If I had said
  282. 8:58instead of 70%
  283. 9:01of hyperscaler
  284. 9:04AI revenue comes from Anthropic and
  285. 9:05OpenAI, if that number had been 10%,
  286. 9:10we'd be having an entirely different
  287. 9:11conversation because there uh clearly
  288. 9:13there are a lot more customers out there
  289. 9:15of size.
  290. 9:16>> Mhm. So, that brings me to another
  291. 9:19argument that I've heard you make on
  292. 9:21your podcast and with your guests is
  293. 9:23that if some
  294. 9:24>> Let me just plug that podcast for a
  295. 9:25second. It's called The Real Investment
  296. 9:27Playbook.
  297. 9:28>> yeah, absolutely. Very It's very, very
  298. 9:30good, and it's gaining a lot of
  299. 9:32traction. I I think it comes down to the
  300. 9:33authenticity of it.
  301. 9:35>> Well, appreciate that.
  302. 9:36>> I really enjoy it. I think it's very
  303. 9:37very high quality.
  304. 9:38>> My wife and I work on it
  305. 9:40every day.
  306. 9:41>> Both of you guys? I didn't realize
  307. 9:42>> my wife is my partner.
  308. 9:44>> All right. Okay.
  309. 9:45>> And
  310. 9:46so she does, God bless her, all the
  311. 9:49editing.
  312. 9:49>> Oh, really?
  313. 9:50>> to give you an insight into things. So I
  314. 9:53am a very linear thinker.
  315. 9:56Which which the way I would define it is
  316. 9:581 2 3 4 5 conclusion. Got it.
  317. 10:03And too often I write that way.
  318. 10:06So I'll write I'll do we have this thing
  319. 10:07called the weekly wrap which we put out
  320. 10:09Friday where I sum up the week.
  321. 10:11>> Mhm.
  322. 10:11>> And too often when I write it I'll I'll
  323. 10:14I'll I'll do 1 2 3 4 5 6 and and Valerie
  324. 10:19my wife who's who's the editor will
  325. 10:21always say you buried the lead again.
  326. 10:24And she'll flip it.
  327. 10:26>> okay.
  328. 10:26>> And so because she says you know most
  329. 10:28people never get to the bottom. You know
  330. 10:31people get your conclusions at the
  331. 10:32bottom and you takes it takes 10 minutes
  332. 10:34to get to the bottom. So she she she
  333. 10:36edits it and she runs the business.
  334. 10:39>> Right. Okay. That's cool. I didn't know
  335. 10:41that. That's that's a little behind the
  336. 10:42scenes. I like that. Okay, but back to
  337. 10:44back to AI train of thought before I
  338. 10:46forget my train of thought again. So
  339. 10:48going on from what you were saying if
  340. 10:50something happens
  341. 10:52to OpenAI or Anthropic
  342. 10:55that's where the problems could be at
  343. 10:57this moment in time
  344. 10:59when we got the concentration risk.
  345. 11:00>> Yeah.
  346. 11:00>> How do you see
  347. 11:02things like
  348. 11:04Deep Seek's new model or Kimmy what is
  349. 11:07it called now? Kimmy 3
  350. 11:08>> best name in the biz.
  351. 11:10>> Kimmy K3. I like that. It just rolls off
  352. 11:12the tongue.
  353. 11:12>> So this is where the the industry has I
  354. 11:15think real weak business weakness. Not
  355. 11:18not con-
  356. 11:19Leave aside concentration risk. That's
  357. 11:21its own risk.
  358. 11:23Here's the business risk. There was
  359. 11:26something going on for a while with
  360. 11:28which is called token maxing.
  361. 11:31Which is where for lack of a better
  362. 11:33term, I work for a company
  363. 11:36>> [clears throat]
  364. 11:36>> and I'm the I'm the I'm a software
  365. 11:38engineer
  366. 11:39and I've been told by management, you're
  367. 11:41to use AI 24/7.
  368. 11:44Whether you need it or not.
  369. 11:45>> Just do it.
  370. 11:46>> Just do it.
  371. 11:47>> Right.
  372. 11:47>> What happened was last year
  373. 11:51OpenAI and Anthropic were dramatically
  374. 11:53undercharging
  375. 11:54for for their services. Then they raised
  376. 11:56prices because they were so
  377. 11:58undercharging for the cost of tokens, it
  378. 12:00was killing them.
  379. 12:01>> Right.
  380. 12:02>> So they increased the prices that so
  381. 12:03that the the customer was bearing more
  382. 12:05of the cost of the token.
  383. 12:07>> So you get people in, once you got
  384. 12:08people in, raise the price.
  385. 12:09>> Raise price. Uber, I think, went blew
  386. 12:12through its entire AI budget in like
  387. 12:14three or four months.
  388. 12:15>> Oh, yikes. Okay.
  389. 12:17>> Okay. And then this was some some other
  390. 12:19company that I read about spent $500
  391. 12:21million before they even knew they had
  392. 12:23spent $500 million.
  393. 12:25So what's happened is
  394. 12:27people have gotten a lot more
  395. 12:28cost-conscious.
  396. 12:30Token maxing has gone away and people
  397. 12:32are using these open weight models a lot
  398. 12:36more. You don't need a Cadillac for
  399. 12:38everything. You know, so people will use
  400. 12:42Anthropic and OpenAI's models only for
  401. 12:44the
  402. 12:45super important tasks. Everything else
  403. 12:48they'll use Chemik3 or whatever. So what
  404. 12:51I like to say about this is that this
  405. 12:53industry, despite all the hundreds of
  406. 12:55billions of dollars has been spent on
  407. 12:56it, has no moats.
  408. 12:58>> Mhm.
  409. 12:59>> There are no moats. This is You know,
  410. 13:00Google with its search until very, very
  411. 13:03recently
  412. 13:04that was a moat. I mean everybody used
  413. 13:07Google. Like I mean
  414. 13:08>> 90%
  415. 13:09>> 90% of planet Earth used Google and no
  416. 13:11one even think about it.
  417. 13:12>> Mhm.
  418. 13:12>> Cuz cuz it was just better and nobody
  419. 13:15could could approach it.
  420. 13:17You know, here one day it's Gemini, one
  421. 13:20day it's ChatGPT, another day it's
  422. 13:22Claude. They they they just rotate.
  423. 13:25>> Mhm.
  424. 13:25>> And so well, we'll come to it I have
  425. 13:29about my conspiracy theory about the end
  426. 13:30of the world.
  427. 13:31>> Conspiracy theory, all right.
  428. 13:33>> So, this this this this dovetails into
  429. 13:34my conspiracy.
  430. 13:35>> Okay, okay.
  431. 13:36>> So, as every as all your viewers know,
  432. 13:38the world's going to end.
  433. 13:40>> At some point it has to.
  434. 13:40>> At some point it has to. Maybe it could
  435. 13:42be 5 billion years from now or it could
  436. 13:44be you know in a couple of weeks.
  437. 13:46>> Yeah, CNBC and Wall Street Journal I
  438. 13:47think it's next week.
  439. 13:48>> think this entire
  440. 13:51AI is going to end the world is a
  441. 13:54complete subterfuge.
  442. 13:55>> Okay.
  443. 13:56>> And what I think is really going on is
  444. 13:59that these companies are nervous.
  445. 14:02They're nervous that token maxing has
  446. 14:03ended. They're nervous that there are no
  447. 14:06moats. They're nervous that these they
  448. 14:08these Chinese open weight models are
  449. 14:10taking massive market share. So, they're
  450. 14:12manufacturing
  451. 14:14a hysteria. Which what they're hoping
  452. 14:16for is for the government to come in and
  453. 14:18regulate the industry.
  454. 14:20And they think
  455. 14:22that by minute they could manipulate
  456. 14:24that regulation
  457. 14:26to create a duopoly.
  458. 14:28>> So, by
  459. 14:28>> That will create that the regulation
  460. 14:29will create the moats. The regulation
  461. 14:32will say no Chinese AI models.
  462. 14:33>> Ah. It's too big a risk.
  463. 14:35>> Ah, I see.
  464. 14:35>> And then all all of a sudden there's a
  465. 14:37moat
  466. 14:38>> Yeah.
  467. 14:38>> that didn't exist before.
  468. 14:40>> A legal barrier.
  469. 14:41>> That's what I think is is is is actually
  470. 14:43happening here.
  471. 14:43>> So, you think that's the reason behind I
  472. 14:45cuz I've noticed Elon he's always
  473. 14:47talking about it but recently the
  474. 14:50Anthropic CEO Dario
  475. 14:53>> And they're also talking about slowing
  476. 14:54down.
  477. 14:55>> Yep, slowing down.
  478. 14:56>> That cannot be that cannot be taken
  479. 14:57seriously because if you really really
  480. 15:01really if you were if I was Dario Amodei
  481. 15:03and I really really really thought
  482. 15:06that
  483. 15:07my product
  484. 15:09is dangerous
  485. 15:11and I really need to slow down
  486. 15:14I'd postpone my IPO.
  487. 15:16>> Mhm.
  488. 15:16>> You'd have to postpone your IPO.
  489. 15:18>> You could just do take the steps
  490. 15:19>> Take a step back and and you know I'll
  491. 15:22fix fix what we need to fix, and we'll
  492. 15:24come back. Are they pushing by their
  493. 15:25IPO? No.
  494. 15:27You know, Elon Musk had a very funny
  495. 15:29quote the other day where I think he did
  496. 15:31it on X where he said something like um
  497. 15:33I'm going to paraphrase it. I don't have
  498. 15:35the quote exactly, but he said, "This is
  499. 15:36some messed up 4D chess where you're
  500. 15:40saying that the that your product's
  501. 15:42going to end. Oh, and by the way, how
  502. 15:44much can I allocate to you for the IPO?"
  503. 15:47>> [laughter]
  504. 15:48>> Yeah, you know, you know, seriously, I I
  505. 15:50I take people seriously when they put
  506. 15:52their their money at risk.
  507. 15:54>> Right.
  508. 15:54>> You know, this this this statement about
  509. 15:56the slowdown is is just whole part of
  510. 15:58this hysteria that they're trying to
  511. 15:59manufacture.
  512. 16:00>> Right. So, they're aware that there may
  513. 16:02not be moats. And it kind of for for
  514. 16:04those that don't know, explain what you
  515. 16:07what you mean. We're talking moats
  516. 16:08specifically in
  517. 16:10not hyperscale, in the LLM providers.
  518. 16:12Because they're
  519. 16:14they're I guess they're still
  520. 16:15>> The hyperscalers have moats.
  521. 16:16>> Okay. Can you explain
  522. 16:17>> So, so what's a moat?
  523. 16:19>> Yeah.
  524. 16:20>> Your grocery store doesn't have a moat.
  525. 16:22Because somebody can open up a grocery
  526. 16:24store across the street tomorrow.
  527. 16:26>> Mhm.
  528. 16:27>> But there are some businesses that have
  529. 16:31real moats
  530. 16:32real moats around them.
  531. 16:34>> Like a competitive advantage.
  532. 16:35>> It's but it's a competitive advantage
  533. 16:38that is eternal. Or or at least very
  534. 16:41long-lasting. Like Nvidia makes GPUs.
  535. 16:44>> Mhm.
  536. 16:45>> Well, nobody else really makes GPUs.
  537. 16:48That's a moat. Certain software
  538. 16:50companies have you know, when if if
  539. 16:53you're Salesforce or ServiceNow, which
  540. 16:56are two massive software companies, your
  541. 16:58product is embedded in the companies
  542. 17:01that you service. Like like
  543. 17:02>> Yeah.
  544. 17:03>> Those companies that use your your
  545. 17:04product,
  546. 17:05>> [snorts]
  547. 17:05>> they've used it for so long, they can't
  548. 17:08function without your product.
  549. 17:10>> Hard to switch away.
  550. 17:11>> and get to try and switch out of that is
  551. 17:13brutal.
  552. 17:14>> Yeah.
  553. 17:15>> That's a moat. There's no
  554. 17:17hyperscaler. What's the moat? Well, I
  555. 17:19don't know about you, but I know that I
  556. 17:21don't have a hundred billion dollars to
  557. 17:23spend on building data centers.
  558. 17:25>> Okay.
  559. 17:25>> You just don't have it.
  560. 17:26>> So, there is some moat in terms of size.
  561. 17:29>> So, just in terms of size and money, I
  562. 17:31mean, all the only certain companies
  563. 17:33that can actually build data centers.
  564. 17:35>> Mhm.
  565. 17:36>> They're just that expensive. LLMs, the
  566. 17:39creation of the models is expensive, but
  567. 17:41there's so much competition.
  568. 17:44And
  569. 17:45there's no loyalty. Like, you know, if
  570. 17:47you're a if you're a software developer
  571. 17:50and you're using Claude, if tomorrow
  572. 17:52another LLM shows up that's better than
  573. 17:54Claude, you switch.
  574. 17:56>> Yep.
  575. 17:56>> You're you're switch you there's no
  576. 17:57there's you're not stuck. That's the
  577. 18:00problem with the LLM model.
  578. 18:01>> Got you. So, there's no moats in LLMs.
  579. 18:04>> Yes.
  580. 18:04>> Hyper-scalers, they have I guess they
  581. 18:06have diversified business models. So,
  582. 18:08there's
  583. 18:08>> And they have moats, but they're
  584. 18:10dependent upon the LLMs.
  585. 18:11>> True.
  586. 18:12>> cloud businesses. That's their weakness
  587. 18:14right now.
  588. 18:15>> So, okay, so there's kind of a argument
  589. 18:17for hyper-scalers for and against having
  590. 18:19a moat. You kind of in some ways there's
  591. 18:21the scale and the cost and the barrier
  592. 18:24to entry moat of of the investment, and
  593. 18:26then there's also the dependency on the
  594. 18:28people that are buying the
  595. 18:29>> They have no moats.
  596. 18:30>> Yes. Okay.
  597. 18:31>> That's the problem.
  598. 18:32>> Interesting. Okay.
  599. 18:34>> And the other and the other problem with
  600. 18:35the hyper-scalers, this may be
  601. 18:37temporary,
  602. 18:39but then again, maybe it's not, is, you
  603. 18:42know, Microsoft, Amazon, Google, three,
  604. 18:46four years ago and way prior to that,
  605. 18:49these companies were incredibly
  606. 18:51profitable, but even more importantly,
  607. 18:54they just threw off cash like crazy. I
  608. 18:57mean, so much cash they didn't even know
  609. 18:59what to do with it. So, they just bought
  610. 19:00back stock because they literally didn't
  611. 19:02have any enough investments to pour back
  612. 19:04into their own businesses. Today,
  613. 19:07because of the incredible amount of
  614. 19:09money that there's that they're spending
  615. 19:11on these data centers,
  616. 19:14their cash flow is gone.
  617. 19:15>> I noticed that.
  618. 19:15>> some cases negative.
  619. 19:16>> Negative now. Yeah. Yeah, very little
  620. 19:19>> I mean, Google raised equity capital 85
  621. 19:23billion. I mean, if you would said to me
  622. 19:24a year a year two years ago that that
  623. 19:26hey Steve, I'm going to make a
  624. 19:28prediction. Google
  625. 19:30which hasn't raised capital since it
  626. 19:32went public is going to raise 85 billion
  627. 19:36not in debt
  628. 19:37in equity capital. I would have said
  629. 19:39you're out of your mind.
  630. 19:40>> Yeah.
  631. 19:40>> Yeah, you're you're insane. Like like
  632. 19:42what are you talking about? They
  633. 19:43they create 85 billion in cash in like
  634. 19:47overnight. Like why what would what
  635. 19:49would possess them to raise equity
  636. 19:50capital? Well, world changed.
  637. 19:52>> It's a very dramatic shift.
  638. 19:55Uh I I feel, you know, as, you know, an
  639. 19:58investor that's held Google for probably
  640. 20:01eight years.
  641. 20:02>> Right.
  642. 20:03>> The company that I hold now is very
  643. 20:04different to the company that and that's
  644. 20:08an interesting point. We've seen the
  645. 20:10market get a little bit jittery with the
  646. 20:12amount of spending that's happening. It
  647. 20:15could be a payoff. Maybe there's not you
  648. 20:17>> By the way, let me just interrupt you
  649. 20:18for a second. Let's go back to Oracle.
  650. 20:20>> Okay.
  651. 20:21>> Cuz after that I didn't finish. After
  652. 20:24Oracle,
  653. 20:25every people said that 50% of Oracle's
  654. 20:29RPO the the backlog is from Open AI. The
  655. 20:32stock which had gone from 230 to 330
  656. 20:36over the next two or three months
  657. 20:38went to 200. And today it's 150. And
  658. 20:41what's fascinating fascinating
  659. 20:44is Oracle just reported and the numbers
  660. 20:47are pretty good.
  661. 20:48And the stock was up 4 or 5% after hours
  662. 20:53and was up 7% at the open
  663. 20:55and closed down on day.
  664. 20:57>> Right.
  665. 20:57>> And I was
  666. 20:58and there was no news. So I don't have
  667. 21:01like I don't have like a news
  668. 21:02explanation. Like there's nothing
  669. 21:04happened, but clearly people are very
  670. 21:07nervous about Oracle because Oracle got
  671. 21:09downgraded in its debt rating. It's like
  672. 21:11triple B minus by S&P, which is like, I
  673. 21:14think, maybe just one level above junk.
  674. 21:16>> Mhm.
  675. 21:17>> So, people are nervous about Oracle and
  676. 21:19how much debt they have.
  677. 21:20>> Well, it seems investors are nervous
  678. 21:22about
  679. 21:23all of these hyperscalers now that are
  680. 21:25investing literally hundreds of billions
  681. 21:27like it
  682. 21:28>> Hundreds.
  683. 21:28>> Hundreds of billions. It it's just it's
  684. 21:31insane.
  685. 21:31>> Yeah. It's The numbers in you know, the
  686. 21:33numbers are just so big. I think the
  687. 21:36number that I heard this year is that if
  688. 21:37you just look at the hyperscalers, they
  689. 21:40will spend $700 billion
  690. 21:42on AI CapEx this year.
  691. 21:44>> Mhm.
  692. 21:45>> It's like that's such a huge number.
  693. 21:47It's hard to even get your mind around
  694. 21:49it. It's so enormous.
  695. 21:50>> I'm interested in your perspective on
  696. 21:53what Michael Burry has been saying where
  697. 21:55he's concerned that the data centers are
  698. 21:58taking too long to come online. They're
  699. 22:00buying so many chips. He his opinion is
  700. 22:03the chips become obsolete way faster
  701. 22:05than the depreciation schedules allow.
  702. 22:07>> I I I understand his argument. So, so
  703. 22:09let me re
  704. 22:11give it his it's it's due first.
  705. 22:13>> Yeah.
  706. 22:13>> What he pointed out last year, I think
  707. 22:15in November, was that the hyperscalers
  708. 22:18had changed the depreciation schedule
  709. 22:22of the chips from three to four years to
  710. 22:25like five to six years.
  711. 22:27And if you did like a cal- I can't
  712. 22:29remember exactly what the calculation,
  713. 22:31but but it's an it's an enormous
  714. 22:32increase in profitability just from the
  715. 22:35change in that accounting because by by
  716. 22:37changing
  717. 22:37>> Right.
  718. 22:38>> It's like
  719. 22:39>> Click.
  720. 22:39>> By by um by changing your depreciation
  721. 22:42schedule from three to three to four
  722. 22:44years to five to six years, by
  723. 22:45definition, your depreciation expense,
  724. 22:48which you report, is going to be lower,
  725. 22:50all other things being equal. He also
  726. 22:51said that, you know, there's so many new
  727. 22:54chips coming that they become obsolete.
  728. 22:55Where I think
  729. 22:57he's wrong for the moment is that there
  730. 23:01is such demand for chips right now that
  731. 23:05there's still huge demand for the older
  732. 23:07chips whose price has gone up with all
  733. 23:09the other chips.
  734. 23:10>> Right.
  735. 23:10>> So,
  736. 23:11I I I think with all due respect to
  737. 23:15Michael, I think his argument is too
  738. 23:18academic.
  739. 23:19>> Okay.
  740. 23:19>> Like
  741. 23:21put this way. If AI succeeds because
  742. 23:23Anthropic and OpenAI, you know, grow
  743. 23:25like crazy and the hyperscalers do well,
  744. 23:28etc., etc., it's not going to matter
  745. 23:31if the depreciation schedule changed
  746. 23:33from 3 to 4 years to 5 to 6 years.
  747. 23:35>> Right.
  748. 23:35>> At the same time, if if OpenAI fails and
  749. 23:39the and the whole chain goes in reverse,
  750. 23:41we'll have a massive correction which
  751. 23:43has nothing to do with the depreciation
  752. 23:45schedule. I don't I mean, I think what
  753. 23:47he's
  754. 23:48deep down what he's just trying to point
  755. 23:50out is maybe there's something wrong
  756. 23:51here, but I don't think what the the
  757. 23:53thing that he's pointing to as being
  758. 23:55wrong is what's going to is is important
  759. 23:57enough.
  760. 23:58>> Right, there's bigger factors at play
  761. 23:59than depreciation.
  762. 24:00>> Much bigger factors.
  763. 24:02>> Okay, interesting. So, we've spoken
  764. 24:03about
  765. 24:05no moats, we've spoken about China
  766. 24:07coming in potentially being competition.
  767. 24:10Another headwind that I've been trying
  768. 24:12to wrap my head around more is is the
  769. 24:14power element as well. Because this is
  770. 24:18another one of those big things that
  771. 24:19we're talking about earnings and chips
  772. 24:21and this and that. But when I started to
  773. 24:24look at power, I I think it was Elon
  774. 24:27Musk's interview with The Economist that
  775. 24:28opened my eyes up to it. He said, "China
  776. 24:31has a chip problem. The US has a power
  777. 24:33problem."
  778. 24:33>> He's right.
  779. 24:34>> However,
  780. 24:35in his view, China can solve its chip
  781. 24:38problem. It might take some time, but a
  782. 24:40harder one to solve is the power
  783. 24:41problem. Because power's physical
  784. 24:43infrastructure, it takes a long time.
  785. 24:45>> Correct.
  786. 24:46>> I don't know if I have the expertise or
  787. 24:48the understanding to know how big of a
  788. 24:50restraint or a bottleneck power in the
  789. 24:52United States is actually going to be.
  790. 24:55>> Get in line. Nobody Nobody knows.
  791. 24:57>> Right. I keep looking, you know, I I
  792. 24:59there are people who say that that that
  793. 25:01things are slow. There are other people
  794. 25:02who say things are fine. I can't I can't
  795. 25:06nail it down yet.
  796. 25:07>> Right. Okay.
  797. 25:08>> I mean I do know that the companies that
  798. 25:10are involved with power
  799. 25:12are doing great.
  800. 25:13>> Mhm.
  801. 25:13>> Like GE Vernova for example. I mean that
  802. 25:15stock's gone nuts, you know. I'll pat
  803. 25:17myself a little bit on the back. I
  804. 25:19bought that stock really early.
  805. 25:20>> Oh, really?
  806. 25:20>> But but I I I bought it cuz a sell-side
  807. 25:22analyst I'm very friendly with told me I
  808. 25:23should buy it. I just took a flyer on
  809. 25:25>> Yeah.
  810. 25:25>> But um
  811. 25:26>> For those that don't know GE Vernova,
  812. 25:28tell me what they do.
  813. 25:28>> me tell you it's a very interesting. So
  814. 25:31GE used to be composed basically of
  815. 25:34three massive divisions, health care,
  816. 25:37>> Uh-huh.
  817. 25:38>> aerospace, where they basically make the
  818. 25:41jet engines for planes and then they
  819. 25:42service them,
  820. 25:44and then call it energy. If you ever saw
  821. 25:47a a jet engine
  822. 25:48>> Mhm.
  823. 25:48>> and looked at it at a gas turbine, which
  824. 25:51is what goes into a utility that creates
  825. 25:53electricity, they look exactly the same.
  826. 25:55It's just that the gas turbine's much
  827. 25:57bigger.
  828. 25:57>> Yes.
  829. 25:59>> But it's basically the same technology.
  830. 26:01So the energy division of of GE
  831. 26:04makes gas turbines. They have all this
  832. 26:06electrical equipment that they sell, and
  833. 26:08then they have a wind division, which
  834. 26:10does terribly.
  835. 26:11>> Mhm.
  836. 26:12>> Now, this will show you how like fast
  837. 26:14the world can change. The the energy
  838. 26:17division was created when I think around
  839. 26:202015 or so, GE bought a company in
  840. 26:24Europe called Alstom. Now Alstom did was
  841. 26:27an energy company that also created gas
  842. 26:29turbines, and GE had a business that
  843. 26:31created gas turbines, so they merged
  844. 26:33them together.
  845. 26:34Just in time for the entire gas turbine
  846. 26:37business to fall apart.
  847. 26:38>> Right.
  848. 26:40>> And this is why Immelt lost his Jeff
  849. 26:42Immelt who was the CEO of GE finally
  850. 26:45lost his job, cuz that was like enough
  851. 26:47already. So eventually all three
  852. 26:49divisions got spun out. So there's GE
  853. 26:51Healthcare, and I think its symbol is
  854. 26:53GEH.
  855. 26:54>> Okay.
  856. 26:55>> And there's
  857. 26:56the energy division, which is called GE
  858. 26:58Vernova, which is GEV.
  859. 27:00And then there's GE, which is the
  860. 27:03aerospace division.
  861. 27:04>> Got you.
  862. 27:05>> A year before [snorts] GE Vernova got
  863. 27:07spun out, so that would have been like
  864. 27:112023
  865. 27:12maybe or 2022.
  866. 27:15If you were to read sell-side reports
  867. 27:17upon about the industry,
  868. 27:20the energy business was so bad
  869. 27:23that they ascribed negative value
  870. 27:25to G- to to Vernova.
  871. 27:27>> Negative value to the energy business.
  872. 27:28>> That it was worth negative three I think
  873. 27:30one guy wrote it was worth negative
  874. 27:31three billion.
  875. 27:32>> Oh my gosh.
  876. 27:32>> In terms of a sum of the parts analysis.
  877. 27:34>> Right.
  878. 27:35>> Now, what's happened is even prior
  879. 27:39to the whole data center thing,
  880. 27:41electrical
  881. 27:43production in the United States finally
  882. 27:45started to increase for the first time
  883. 27:47like in 15 years. Now, add on top of
  884. 27:50that,
  885. 27:51the
  886. 27:52data centers, and you're talking about
  887. 27:55US
  888. 27:56electricity growing 3 to 4% per year.
  889. 28:00Now,
  890. 28:01that may not sound like such a huge
  891. 28:03number,
  892. 28:04but 3 to 4%
  893. 28:06off of the base of the United States
  894. 28:09is the equivalent of like two large
  895. 28:11cities.
  896. 28:11>> Okay. It's a lot.
  897. 28:12>> It's huge.
  898. 28:13>> Yeah.
  899. 28:13>> You know, a company like GE Vernova has
  900. 28:15backlogged like two two that 2035. That
  901. 28:18that's how crazy things are.
  902. 28:20>> Cuz this is how these data centers are
  903. 28:23being powered. It's with these gas
  904. 28:25turbines, right?
  905. 28:26>> Mostly.
  906. 28:27>> Mostly.
  907. 28:27>> And there's some alternatives. People
  908. 28:28are talking about nuclear and there's
  909. 28:30and there's a company called Bloom
  910. 28:31Energy which makes its own little
  911. 28:34turbine
  912. 28:35uh that that you could hook up to a to a
  913. 28:36data center, but most of it's going to
  914. 28:38be through gas turbines.
  915. 28:40>> Mhm. Well, that's what um
  916. 28:43Elon had to do with the um Colossus data
  917. 28:45center that he built in Memphis. The
  918. 28:48grid was too slow. It was not ready. So,
  919. 28:50he he ended up getting 35 of the
  920. 28:53portable gas turbines.
  921. 28:54>> Right. They hooked it up to his that
  922. 28:56hooked it up.
  923. 28:56>> Hooked it up. He created his own mini
  924. 28:57power center like on-site right next to
  925. 29:00Okay.
  926. 29:01>> So, so most most of GE Vernova is these
  927. 29:05>> Well, it's the gas turbines. It's all
  928. 29:07the electrical equipment. Think about
  929. 29:09it. I mean, there's
  930. 29:10you know, you're not just when you're
  931. 29:11building a new utility plant, it ain't
  932. 29:13just turbine. There's all this other
  933. 29:15electrical equipment that's got to get
  934. 29:16hooked up. They make that, too. And the
  935. 29:18wind business will always, I think, be a
  936. 29:20crappy business. And and that that So,
  937. 29:23what?
  938. 29:23>> Right. Okay. Very interesting. Very
  939. 29:25interesting. But, you would say the
  940. 29:27thing to look out for in GE Vernova's
  941. 29:29case is the gas turbines. Is that the
  942. 29:31core of that business?
  943. 29:32>> Yeah, I mean, that's the core. And then
  944. 29:33you just want to look at the orders.
  945. 29:35>> Yes.
  946. 29:35>> Which I think in the last quarter they
  947. 29:36were up like 85%. Something insane.
  948. 29:38>> Well, I I
  949. 29:39>> I mean, they they I mean, the there's
  950. 29:40those gas turbines like in the room that
  951. 29:42we're in, it's like it's like five of
  952. 29:45these rooms combined is how big these
  953. 29:47things are. They're they're huge. They I
  954. 29:49mean, it takes years to build them.
  955. 29:51>> Yes. Well, that's what I was going to
  956. 29:52say. I think I read something that their
  957. 29:54backlog is stretching out to past 2030
  958. 29:58or something like that.
  959. 29:59>> Yes, it is.
  960. 29:59>> Which is just
  961. 30:00>> Well, cuz people want to they want to
  962. 30:02line it up as much as they can.
  963. 30:03>> Mhm. But, is that even more of a
  964. 30:07uh an argument for this power problem if
  965. 30:09>> [laughter]
  966. 30:10>> if people making orders now, we want
  967. 30:12these turbines now. And hang on, well,
  968. 30:13we've got 2030. You want to
  969. 30:16hang on to
  970. 30:16>> I I just don't know. I really don't
  971. 30:18know. I don't have enough information.
  972. 30:19>> Yeah. I I can't remember who it was that
  973. 30:21you interviewed, the man that knew uh it
  974. 30:24was about power, but I found that was a
  975. 30:26really good interview. I might leave it
  976. 30:27linked um on screen right now, but I I
  977. 30:29thought that was really, really good
  978. 30:30explanation. Okay. So, we've covered a
  979. 30:32lot of headwinds
  980. 30:35when it comes to AI. I think you you
  981. 30:38went on the record saying that if you
  982. 30:39try and predict what's going to happen,
  983. 30:40you're a fool. So, don't don't try and
  984. 30:42predict it. I think there are a lot of
  985. 30:43people out there that feel compelled to
  986. 30:46look at these AI plays, to to look into
  987. 30:49the realm of AI. If you're
  988. 30:52If someone comes to you and says, "Oh,
  989. 30:53look Steve, I really got to get in on AI
  990. 30:55somehow." What are What are some of the
  991. 30:58maybe safer ways to play AI? And what's
  992. 31:02What would you say are the high-risk
  993. 31:04ways to play AI, talking about just
  994. 31:06investing in in stocks?
  995. 31:07>> I would play
  996. 31:09I wouldn't invest in an LLM cuz I just
  997. 31:11think as I said, there's no I would not
  998. 31:13>> not.
  999. 31:13>> cuz there are no motes.
  1000. 31:15>> Yeah, that makes sense.
  1001. 31:16>> I might be a little wary of the
  1002. 31:18hyperscalers at this point just because
  1003. 31:20their businesses have They've lost all
  1004. 31:22their cash flow. But, I would be looking
  1005. 31:24at the companies that are getting that
  1006. 31:26cash flow.
  1007. 31:27>> Right.
  1008. 31:27>> So, you know, that would be Nvidia.
  1009. 31:30You know, you maybe you would you'd want
  1010. 31:31to own Micron, GE Vernova,
  1011. 31:34um
  1012. 31:35Arista Network, Cisco. And then if you
  1013. 31:38want to get into the industrial side,
  1014. 31:40you could talk about like an Eaton,
  1015. 31:42which puts this electrification company.
  1016. 31:45That's what I would do.
  1017. 31:45>> more it's more picks and shovels
  1018. 31:47>> Picks and shovels.
  1019. 31:48>> Right, as opposed to the flashy software
  1020. 31:50side.
  1021. 31:50>> Right.
  1022. 31:51>> Okay.
  1023. 31:51>> The software you know, the whole
  1024. 31:52software industry I mean, I'm sure
  1025. 31:54you've heard the word SaaS-pocalypse.
  1026. 31:56>> I have, yeah.
  1027. 31:56>> Um
  1028. 31:57>> [laughter]
  1029. 31:57>> And I I I
  1030. 31:59I I All I know is there will be software
  1031. 32:02companies that will have problems.
  1032. 32:04>> Mhm.
  1033. 32:04>> Because you know, take this new agentic
  1034. 32:07AI um use, I think it's called what Meta
  1035. 32:10put out. You know, if I want to book
  1036. 32:13a flight, I say to my muse,
  1037. 32:16Ooh, that's good. I say to my muse,
  1038. 32:19"I want to book a flight to Miami
  1039. 32:22on such and such a date.
  1040. 32:25Book me in the best hotel in Brickell."
  1041. 32:27>> Mhm.
  1042. 32:27>> Okay.
  1043. 32:29And it goes and does it. Well, how does
  1044. 32:30it do it? It goes on all the travel
  1045. 32:32sites
  1046. 32:35and finds the best price and books it.
  1047. 32:38>> Mhm.
  1048. 32:39>> Well, that kind of makes the travel
  1049. 32:41sites worth less.
  1050. 32:43Cuz you're not going to you're not going
  1051. 32:44to bookings or Travelocity or what
  1052. 32:46whatever directly anymore.
  1053. 32:50You're not their customer anymore.
  1054. 32:52>> Your AI is
  1055. 32:53>> Your AI AI my AI agent's my customer.
  1056. 32:56And he and he does the work.
  1057. 32:57>> sidebar ads don't work on the AI.
  1058. 32:59>> like that, I think stuff in the payment
  1059. 33:01world could get dicey. But on the other
  1060. 33:03hand, you know, software that's deeply
  1061. 33:05embedded in enterprises is probably
  1062. 33:07okay.
  1063. 33:07>> Well, that's what I was going to ask
  1064. 33:08you. It sounds like the most important
  1065. 33:09thing to look at is the switching mode.
  1066. 33:11And how how resilient is the switching
  1067. 33:13mode?
  1068. 33:13>> is it?
  1069. 33:13>> Okay.
  1070. 33:14>> I mean, you know, for bookings, what's
  1071. 33:15the switching mode? You know, I go to I
  1072. 33:17go on the bookings website and I book a
  1073. 33:19trip. So now I don't go on the bookings
  1074. 33:21website. I my agentic AI finds us the
  1075. 33:25best deal.
  1076. 33:26So that kind of makes the the travel
  1077. 33:29online
  1078. 33:30companies worth less, I think. It's a
  1079. 33:33little early, but I think that's a
  1080. 33:35possibility.
  1081. 33:36>> Yeah, okay. So I guess
  1082. 33:39another argument that I've heard, I'm
  1083. 33:41interested to hear your overarching
  1084. 33:43thoughts on this around AI, is people
  1085. 33:45are very fast to liken it to 1999.
  1086. 33:49>> Mhm.
  1087. 33:50>> A tech bubble 2.0.
  1088. 33:52You know, that's that's what the media
  1089. 33:53will say.
  1090. 33:53>> Right.
  1091. 33:54>> I'm very interested in what your
  1092. 33:57thoughts are on this.
  1093. 33:59I have my own opinion, but I'm
  1094. 34:01interested to hear what you think. We're
  1095. 34:03in the same setup, new technology,
  1096. 34:05speculation in financial markets.
  1097. 34:07>> Yeah.
  1098. 34:08>> Similar setup. Is it different this
  1099. 34:10time? Is there anything fundamentally
  1100. 34:11different?
  1101. 34:12>> I I don't know if it's different or not.
  1102. 34:14I think it's too early. I mean, if
  1103. 34:16OpenAI or Anthropic fail, you'll have a
  1104. 34:19real correction and then the next
  1105. 34:21generation of people will come up and
  1106. 34:22pick up the pieces. I don't know if
  1107. 34:24that's going to happen or not. So I
  1108. 34:25don't I just don't know.
  1109. 34:26>> Fair enough. Going back to the investing
  1110. 34:28argument, I'm actually interested cuz I
  1111. 34:30didn't ask you last time and I had some
  1112. 34:33subscribers that are interested in
  1113. 34:34understanding how you actually go about
  1114. 34:37your investing. Not stocks, not what
  1115. 34:38stocks you're picking or anything like
  1116. 34:39that, but when it comes to the real
  1117. 34:42Eisman playbook, what is the real Eisman
  1118. 34:44playbook? Like how do you analyze
  1119. 34:46companies? Is it Do you stick within a
  1120. 34:48circle of competence? Do you go down
  1121. 34:49rabbit holes? Do you look at certain
  1122. 34:51financial metrics that you really love
  1123. 34:53to see or not?
  1124. 34:54>> Well, a couple of things. I'm very story
  1125. 34:56oriented.
  1126. 34:57>> Story, okay.
  1127. 34:58>> not a quote-unquote value player, I
  1128. 35:00think.
  1129. 35:01>> Okay.
  1130. 35:01>> You know, if stocks are cheap, they're
  1131. 35:03probably cheap for a reason.
  1132. 35:04>> Okay.
  1133. 35:05>> You know.
  1134. 35:05>> But I can't see you being someone that
  1135. 35:06will grossly overpay for something
  1136. 35:08either.
  1137. 35:09>> Uh
  1138. 35:09>> I see you as a Personally, I see you as
  1139. 35:11very rational.
  1140. 35:12>> you know, I I mean I would have loved to
  1141. 35:13have owned Palantir, but I won't buy it
  1142. 35:15now because it's so expensive.
  1143. 35:17>> Yeah.
  1144. 35:17>> Um
  1145. 35:18>> So, story, but rational.
  1146. 35:19>> story story, but rational.
  1147. 35:21>> Yeah. Okay. Okay. [laughter]
  1148. 35:22Is there anything uh Are there any kind
  1149. 35:25of uh metrics on your checklist or
  1150. 35:27anything that you love to look at that
  1151. 35:29might be a red flag? I'm just interested
  1152. 35:31to see like what you really look for. Is
  1153. 35:33a moat like a must-have for you or
  1154. 35:36>> Not necessarily. I like a moat.
  1155. 35:38>> Yeah.
  1156. 35:38>> That's why I own Moody's, for example.
  1157. 35:40>> Yeah.
  1158. 35:41>> Um that's why I own Visa.
  1159. 35:43>> Mhm.
  1160. 35:44>> But um it's not a complete requirement.
  1161. 35:47So, no.
  1162. 35:48>> Is there anything that you particularly
  1163. 35:50hate to see in a company? What's what's
  1164. 35:52What are some things that will instantly
  1165. 35:54turn you off?
  1166. 35:55>> Management selling stock.
  1167. 35:57>> Okay.
  1168. 35:58>> I generally don't like cyclical
  1169. 36:00companies.
  1170. 36:01>> Okay.
  1171. 36:01>> Cuz
  1172. 36:03then you're just predicting the economy.
  1173. 36:05And I mean there are exceptions, but I I
  1174. 36:07I I I like companies that have a real
  1175. 36:10story with real growth
  1176. 36:12tailwinds.
  1177. 36:13>> Right.
  1178. 36:13>> That's what I like.
  1179. 36:14>> Okay, interesting. Hey, do you mind if I
  1180. 36:16finish off by asking you some questions
  1181. 36:18from the audience? Sure. Is that all
  1182. 36:19right? All right, I had a quick screen,
  1183. 36:22but uh I might have forgotten some.
  1184. 36:24>> [laughter]
  1185. 36:25>> All right, let's Oh, this is a really
  1186. 36:26interesting one. did want to get your
  1187. 36:27opinion on this. US debt. This is such a
  1188. 36:31an interesting topic and it's very very
  1189. 36:34very highly covered. So, it's at $40
  1190. 36:37now. The average interest rate on it has
  1191. 36:39gone from 1.77% 2020 to 3.45% today. The
  1192. 36:44interest expense has risen from 523
  1193. 36:46billion a year to now 1.22 trillion. Is
  1194. 36:49that something investors need to be
  1195. 36:51genuinely worried about? Is there a real
  1196. 36:54risk of a debt spiral in the future?
  1197. 36:57>> I mean, all other things being equal, I
  1198. 36:58wish the deficit was smaller.
  1199. 36:59>> Yeah.
  1200. 37:00>> Um
  1201. 37:01I I have my doubts about a debt spiral.
  1202. 37:04Number one, we are the reserve currency
  1203. 37:06of the world.
  1204. 37:08But maybe even more importantly, US
  1205. 37:10Treasuries
  1206. 37:12are the financial system of planet
  1207. 37:14Earth. So, just for example,
  1208. 37:16banks all over the world do something
  1209. 37:19called repos, where they lend to each
  1210. 37:20other overnight. They do it through
  1211. 37:23overnight Treasuries.
  1212. 37:25So, as long as the US Treasury is the
  1213. 37:29backbone of the financial system of the
  1214. 37:31world, I tend not to worry about the
  1215. 37:34deficit too much, although I'd like it
  1216. 37:36to be smaller. If there was an
  1217. 37:38alternative, we'd be in trouble.
  1218. 37:40>> Okay.
  1219. 37:41>> But there is no alternative at this
  1220. 37:42point.
  1221. 37:43>> Let me ask you this. The Fed just raised
  1222. 37:46rates for the first time in 3 years. Are
  1223. 37:48rates
  1224. 37:50uh likely to be a showstopper for the
  1225. 37:52market and in particular the AI
  1226. 37:53narrative?
  1227. 37:54>> The rate to look at is the 10-year. It's
  1228. 37:57the long-term rates, not the short-term
  1229. 37:58rates that the Fed does, because that's
  1230. 38:00where people borrow.
  1231. 38:01>> Okay.
  1232. 38:02>> You know, the Fed is just
  1233. 38:03the Fed funds rate is the rate at which
  1234. 38:05the Fed lends to banks overnight.
  1235. 38:07>> Okay.
  1236. 38:08>> You don't have access to that. Neither
  1237. 38:10do [laughter] I.
  1238. 38:12Um
  1239. 38:12>> That's the side.
  1240. 38:13>> Yeah. So,
  1241. 38:15you know, just today, for example, the
  1242. 38:17markets rallied because despite the Fed
  1243. 38:19raising rates, the 10-year yield went
  1244. 38:21down.
  1245. 38:23I'm getting the feeling that 5% is sort
  1246. 38:25of the Rubicon for the market. And um
  1247. 38:28as long as we're below that, we'll
  1248. 38:30probably be okay. But if something were
  1249. 38:32to happen and and and we blow through
  1250. 38:34that, I think we get a correction.
  1251. 38:36>> Okay.
  1252. 38:36>> That's just my guess. I actually thought
  1253. 38:38originally the number was 4 and 1/2% and
  1254. 38:40I was wrong.
  1255. 38:41>> Okay.
  1256. 38:41>> But five feels more
  1257. 38:43>> Do you know where it is now? It's around
  1258. 38:44there, isn't it?
  1259. 38:46>> 4.98, okay.
  1260. 38:47>> But it was over 5% yesterday.
  1261. 38:48>> Yes, okay.
  1262. 38:49>> And it's come back down.
  1263. 38:50>> Okay.
  1264. 38:50>> People are watching every tick.
  1265. 38:52>> Inter- Yes, interesting. I'm interested
  1266. 38:55to hear Steve's thoughts on Here we go.
  1267. 38:57Rising yields of long-term Treasuries.
  1268. 38:59Is Scott Bessent's buyback plan really
  1269. 39:01designed to increase liquidity in older
  1270. 39:04long-term bonds or is the government
  1271. 39:05trying to manipulate long interest rates
  1272. 39:08to ease their interest problem?
  1273. 39:09>> The latter.
  1274. 39:11He's trying to buy long long-term
  1275. 39:13Treasuries to bring rates down to ease
  1276. 39:17the cost of of money for the United
  1277. 39:19States of America. He's failed miserably
  1278. 39:22at this point.
  1279. 39:23You know, rates are higher than when he
  1280. 39:25made his announcement. And I have a
  1281. 39:27suspicion that he's going to come with
  1282. 39:28something else because he cuz well, he
  1283. 39:30he he announced 6 billion. 6 billion is
  1284. 39:32nothing. It's a $40 trillion deficit.
  1285. 39:35>> That's That was my thoughts.
  1286. 39:36>> So, I I I don't think he's an idiot.
  1287. 39:39So, I think he's going to come with
  1288. 39:40something else. What else that is, I
  1289. 39:42don't know.
  1290. 39:42>> Okay, so a different plan of attack to
  1291. 39:44do the same thing.
  1292. 39:45>> Yes.
  1293. 39:45>> Uh okay. Okay. I've always been
  1294. 39:47skeptical about precious metals, but
  1295. 39:49Steve's recent discussion with Porter
  1296. 39:50Collins and Vincent Daniel made me
  1297. 39:52second-guess that. I would like to know
  1298. 39:54if he has any conflicting feelings about
  1299. 39:56precious metals or is he still firmly
  1300. 39:58opposed to the asset class?
  1301. 40:00>> I'm not opposed, but I don't I've never
  1302. 40:01owned it.
  1303. 40:02>> Yep.
  1304. 40:02>> I've never owned it. I It's not
  1305. 40:05something that has ever really enticed
  1306. 40:07me one way or the other.
  1307. 40:08>> Similar thinking to Warren Buffett. It
  1308. 40:10just sits there and looks shiny.
  1309. 40:11>> does nothing.
  1310. 40:12>> Does nothing. It's not productive.
  1311. 40:13>> It It has value cuz people say it has
  1312. 40:15value.
  1313. 40:16>> Okay.
  1314. 40:16>> That's all. So, it doesn't pay you an
  1315. 40:18interest rate.
  1316. 40:19>> Similar argument, I'm guessing, to
  1317. 40:20Bitcoin and anything else that sits
  1318. 40:22there and
  1319. 40:22>> Oh, definitely.
  1320. 40:23>> baseball cards, blah, blah, blah,
  1321. 40:24whatever sits there.
  1322. 40:25>> is worse.
  1323. 40:26>> Bitcoin is worse? Yeah.
  1324. 40:28>> Bitcoin is worse because it trades
  1325. 40:30inversely to its own thesis.
  1326. 40:32>> Yes, I have noticed that. That is quite
  1327. 40:34strange.
  1328. 40:34>> So, for those for your viewers, what I
  1329. 40:36What I mean by this to explain, what I
  1330. 40:38mean by that is people Like if you went
  1331. 40:40to a Bitcoiner and you say, "Dude,
  1332. 40:43why do you want Bitcoin?"
  1333. 40:44The answer you would invariably get is
  1334. 40:46that that fear currency, which is
  1335. 40:48government currencies, have been
  1336. 40:49debased, inflation is coming, and you
  1337. 40:52want to hedge against this, so buy
  1338. 40:54Bitcoin.
  1339. 40:55>> Mhm.
  1340. 40:55>> Okay. That
  1341. 40:57sounds reasonable. The problem is that
  1342. 40:59if that were the case on days where
  1343. 41:02people worry about inflation, rates are
  1344. 41:04going up, and the stock market goes is
  1345. 41:06down, Bitcoin should be up.
  1346. 41:08And on days where Nasdaq is up like
  1347. 41:10crazy and rates are down,
  1348. 41:12Bitcoin should be down.
  1349. 41:14But it would do the opposite.
  1350. 41:15>> Yes.
  1351. 41:16>> So, So, you know, you you say that
  1352. 41:18Bitcoin is going to go up when because
  1353. 41:20when the sky is blue, and then the sky
  1354. 41:22is blue and it goes down, you know, why
  1355. 41:25do I own it? You have no thesis.
  1356. 41:27>> It seems to be just a speculative asset.
  1357. 41:29>> like it's a way to speculate about
  1358. 41:31speculating.
  1359. 41:32>> Speculate about speculating.
  1360. 41:34>> Yeah.
  1361. 41:34>> I like that. All right, let me ask you
  1362. 41:36this. Dear Steve, you seem to be very
  1363. 41:37much centered on the US stock market.
  1364. 41:39Have you ever tried to broaden your
  1365. 41:40investing or trading horizon
  1366. 41:41geographically? Do you have any interest
  1367. 41:43in companies held outside the US?
  1368. 41:45>> Great question. When I used to run my
  1369. 41:47hedge funds, I used to invest overseas.
  1370. 41:49>> Mhm.
  1371. 41:49>> And I found that there was no night and
  1372. 41:51there was no day.
  1373. 41:52>> Okay.
  1374. 41:53>> So, for the last many, many years, all I
  1375. 41:55do is the US, and I'm perfectly happy.
  1376. 41:57There's plenty to do in the US. I don't
  1377. 41:59feel the need to invest overseas.
  1378. 42:01>> Fair enough. Plenty of opportunities
  1379. 42:02here.
  1380. 42:02>> Plenty.
  1381. 42:03>> I like it.
  1382. 42:03>> What's his view on when all these wars
  1383. 42:05would end? I guess particularly the Iran
  1384. 42:08war. Do you see inflation coming down
  1385. 42:10anytime soon?
  1386. 42:11>> I have no more insight into the war than
  1387. 42:13anyone else.
  1388. 42:15So, I can't answer the question.
  1389. 42:16>> Fair enough.
  1390. 42:18Well, I think that is just about all
  1391. 42:19we've got. Is there one more Is there
  1392. 42:21one or a set of numerical indicators
  1393. 42:23whether related to interest rates or
  1394. 42:24inflation or unemployment or otherwise
  1395. 42:26that Steve could see as being the
  1396. 42:28tipping point for the US to take its
  1397. 42:30medicine with cutting benefits or wash
  1398. 42:32spiking rates up or whatever he thinks
  1399. 42:34that medicine might be. I've seen I've
  1400. 42:36seen it said that oil will rise until
  1401. 42:38stocks fall meaning that the US won't
  1402. 42:40get out of Iran until stocks really take
  1403. 42:41a beating. But, this is a much broader
  1404. 42:43question about getting the debt back
  1405. 42:45well under control. I guess we kind of
  1406. 42:47touched on that with the with the
  1407. 42:4810-year Treasury.
  1408. 42:49>> I mean, all I would say is that's a
  1409. 42:51total political question and there's no
  1410. 42:53political appetite in Washington right
  1411. 42:55now to cut the deficit
  1412. 42:57a dollar.
  1413. 42:58>> Mhm.
  1414. 42:58>> By either side.
  1415. 43:00>> We'll see where that goes.
  1416. 43:01>> Yeah.
  1417. 43:01>> I think that's just about how we
  1418. 43:02finished our last [laughter] our last
  1419. 43:04interview. And I was like, on that
  1420. 43:06cheery note.
  1421. 43:06>> Yeah.
  1422. 43:07>> So, again on that cheery note, Steve,
  1423. 43:09thank you very much for for coming on.
  1424. 43:11For those that don't know, well, we
  1425. 43:12talked about it earlier, but the Real
  1426. 43:13Investment Playbook is uh is what you're
  1427. 43:15currently working on. That's your
  1428. 43:16podcast. Can you tell us a little bit
  1429. 43:17more? What can people expect from that?
  1430. 43:19It's on YouTube. I guess you can get it
  1431. 43:21on podcast platforms as well.
  1432. 43:22>> do two two free podcasts a week. So, one
  1433. 43:25is an interview.
  1434. 43:26So, and then
  1435. 43:27the other one is a market wrap where on
  1436. 43:30Friday I put out like a summary of the
  1437. 43:31whole week. And then if you're willing
  1438. 43:33to pay for the paywall on Substack, we
  1439. 43:35do an additional podcast, which is
  1440. 43:37sometimes an interview. This week was
  1441. 43:39part one of two-part master lecture of
  1442. 43:44how to analyze banks.
  1443. 43:45>> All right.
  1444. 43:46>> Yeah.
  1445. 43:46>> I'll check it out.
  1446. 43:47>> ever But, you have to subscribe.
  1447. 43:48>> Okay. I will subscribe. [laughter] No
  1448. 43:50free lunch.
  1449. 43:51>> No free lunch there.
  1450. 43:52>> lunch city here.
  1451. 43:54We'll stay
  1452. 43:54>> Yeah, but if you want to know how to
  1453. 43:55analyze banks,
  1454. 43:56>> go there.
  1455. 43:57>> that's where you should go.
  1456. 43:58>> Awesome. Steve, thank you very much for
  1457. 44:00>> Thank you very much.
  1458. 44:00>> Appreciate it.
  1459. 44:01>> Great. Bye.
  1460. 44:03>> Yes.
  1461. 44:05You're the man now, dog.

About this transcript

This page contains the full transcript of 'Big Short' Investor Explains How the AI Bubble Will Burst by New Money, generated from the public captions YouTube serves with the video. The transcript has 8,475 words across 1,461 segments, with the original timestamps preserved so you can click any line to jump to that moment in the embedded player.

What you can do with it

Use the transcript to take notes, quote the speaker, build a study guide, generate a summary with ChatGPT or Claude via the YouTube Summary tool, or export it as a timed subtitle file with YouTube to SRT. You can also re-open it in the transcriber to choose a caption track available on the video.

Free YouTube transcript tool

YouTube2Text is a free YouTube transcript generator — no signup, no daily limit. Paste any YouTube link and get the full transcript instantly, with timestamps, click-to-jump, caption language selection, AI prompts for ChatGPT, Claude, and Gemini, and exports to TXT, SRT, VTT, or Markdown.