보수적인 하워드 막스가 이번엔 AI에 의외의 낙관을 꺼냈습니다... 모른다면서도 돈을 넣은 단 한 곳은 따로 있었습니다 [구루 11월 - 하워드 막스] — Transcript
Full transcript
- 0:00To start with the conclusion, despite
- 0:01Howard Marks being a very conservative
- 0:03and cautious person, he shared some
- 0:05surprisingly positive remarks. The world
- 0:07moves along a spectrum from good to bad
- 0:09, but the market has a habit of greatly
- 0:11exaggerating this. Looking at past
- 0:13examples, when people become immune to
- 0:15worry, market sentiment becomes
- 0:16overheated. When it comes to fear, it’s
- 0:18not when the market fluctuates like it
- 0:20does these days, but when the market
- 0:21sees wild swings and people aren’t
- 0:23afraid—isn't that when the real crisis
- 0:25starts?
- 0:30>> Hello. This is Lee Ho-seok. Today, I'm
- 0:32joined by Eric for the first time in a
- 0:34while. Hello.
- 0:35>> Yes, hello.
- 0:36>> It's been a while, so what interesting
- 0:37topic should we start with?
- 0:39>> Well, isn't it a time of chaos?
- 0:42I've gathered insights on what
- 0:44perspectives the masters hold during
- 0:46these chaotic times.
- 0:48>> That's right.
- 0:49>> First, our first main guest is a legend
- 0:51worth listening to during times like
- 0:53these. Well, not just "worth" listening
- 0:55to. He’s a legend you absolutely must
- 0:57listen to.
- 0:57>> Right, right.
- 0:58>> It’s Howard Marks. Howard Marks is
- 1:00famous for his memos, but he’s also
- 1:02known by another nickname—the market’s
- 1:03thermometer, right? He has an incredible
- 1:05knack for judging whether the market is
- 1:08overheated, cooled off, or at a
- 1:09reasonable level.
- 1:11So, we’re going to find out how Marks
- 1:13assesses this current chaotic market
- 1:15situation.
- 1:17To start with the conclusion, despite
- 1:19Howard Marks being a very conservative
- 1:21and cautious person, he shared some
- 1:23surprisingly positive remarks.
- 1:25>> I also read his memo and thought, he's
- 1:27usually not the type to say things like
- 1:29this, but there were many parts that
- 1:31read quite positively regarding AI.
- 1:34>> That’s correct. Let's start with the
- 1:36conclusion first. He stated that some of
- 1:38the overheated sentiment in the current
- 1:40market has faded. But, this statement
- 1:42could lead to some misunderstandings.
- 1:45Is the cooling a result of hitting a
- 1:46peak and the bubble bursting on the way
- 1:48down, or was it a process heading toward
- 1:51a bubble where things just settled a bit
- 1:52?
- 1:55I've been tracking this, and Marks has
- 1:57never called a bubble.
- 1:59He's said it's overheated.
- 2:01He has said we might be in the early
- 2:02stages of a bubble, but since he’s never
- 2:04called a bubble, it’s not that we’ve hit
- 2:07a peak and are crashing down.
- 2:08>> It's not that.
- 2:09>> That's right. He said that the
- 2:10overheating has faded while we are in a
- 2:12phase heading toward a bubble. He also
- 2:14mentioned that the market seems unaware
- 2:16of what smart investors already know.
- 2:19Marks emphasizes that the world moves
- 2:21along a spectrum from good to bad, but
- 2:23the market has a habit of greatly
- 2:25exaggerating this.
- 2:27It has a habit of swinging between
- 2:29heaven and hell, and that’s exactly the
- 2:30atmosphere we’re seeing right now. The
- 2:32situation isn't actually that bad, but
- 2:34people are exaggerating it into hell,
- 2:36and smart investors are saying this is a
- 2:38market where you can spot that
- 2:40discrepancy.
- 2:41>> So, are you a smart investor? How about
- 2:44you?
- 2:45The key is understanding what smart
- 2:47investors know that the general public
- 2:49doesn't.
- 2:50>> Right, and they spoke quite positively
- 2:52about the stock market while saying that
- 2:53. Then there's the credit market, which
- 2:55is another important sector, isn't it?
- 2:57Since everyone has flocked to the credit
- 2:59market to invest in AI, and Marks '
- 3:01specialty is actually credit.
- 3:03To start with the conclusion on credit:
- 3:05it’s in good shape.
- 3:07He stated there is no concern of a
- 3:09credit crunch or the credit market
- 3:11collapsing, as some people worry about.
- 3:14He pointed out that just like the stock
- 3:16market, the credit market also has
- 3:18cycles, and he explained this using this
- 3:20graph.
- 3:23Looking at the current cycle, he
- 3:25mentioned we were near a peak until
- 3:27recently.
- 3:30That’s because there are signals that
- 3:31appear at the top.
- 3:35Since the market is so good, everyone
- 3:38wants to lend money, which attracts more
- 3:40people, and that creates a momentum
- 3:42effect that generates returns, leading
- 3:44to a point where people are practically
- 3:46begging to lend money, which lowers
- 3:48lending standards—that’s a classic peak
- 3:50signal, and that’s the atmosphere we had
- 3:53.
- 3:55Because of AI, the mood was "AI will
- 3:57definitely succeed, just lend the money
- 3:59," but haven't there been some negative
- 4:01events recently, like the CoreWeave
- 4:03situation? As a result, doubts have
- 4:05started to emerge in the market.
- 4:06>> Hmm.
- 4:08>> The tone has shifted from "AI cannot
- 4:10fail" to "maybe there are some risks,"
- 4:12and he says this cooling off has
- 4:14actually moved us down from the peak
- 4:16into a much healthier situation.
- 4:19>> Doubt is a good sign. I think this is an
- 4:21important point you should write down.
- 4:23The market is doubtful? Well, then
- 4:25that’s a good sign. It repeats this
- 4:27cycle of going up and down, and since AI
- 4:30has such high volatility, when it drops
- 4:32, it feels like the world is ending. But
- 4:34that process of doubt—if AI wasn't the
- 4:36real deal, you'd have to keep doubting
- 4:38it, but if it is, that doubt is actually
- 4:40a positive indicator.
- 4:41>> He also talked about the AI industry. He
- 4:43made some very positive comments
- 4:44regarding the industry as well. The AI
- 4:46cycle is still in its very early stages
- 4:48. They say it’s like being in the third
- 4:50inning of a baseball game.
- 4:53The reason is that the infrastructure
- 4:56built so far—like data centers and
- 4:58semiconductor plants—was designed with
- 5:00future demand in mind, and that
- 5:02anticipated demand is now emerging, with
- 5:04strong demand still ahead and many
- 5:06large-scale construction projects
- 5:08remaining, so they view the AI cycle as
- 5:11still being in its early stages.
- 5:15>> Yes.
- 5:16>> And in terms of corporate adoption, if
- 5:18you look at computing demand, it’s
- 5:21currently concentrated in a few frontier
- 5:23labs, and the cost of AI adoption
- 5:25relative to S&P 500 revenue is only at
- 5:28about the 0.1%level.
- 5:31That’s why we’re just at the beginning
- 5:33of the adoption phase, and since Max
- 5:35manages an investment firm, he must be
- 5:37using AI quite a lot, right?
- 5:39He’s found that companies need to
- 5:41completely reorganize their work
- 5:42processes and organizational structures
- 5:44to adopt AI more easily. Also, they need
- 5:46to feel the cost-effectiveness for
- 5:48full-scale adoption to happen; looking
- 5:50at things like Palantir’s earnings
- 5:52reports and seminars from domestic tech
- 5:53firms, while people used to focus on how
- 5:55much AI was used or how many tokens were
- 5:57consumed, now they’re asking, "What did
- 5:59you actually build?" They’ve started
- 6:01checking for cost-effectiveness, asking
- 6:03what progress has been made.
- 6:04>> That’s right.
- 6:05>> Even Amazon used to base performance
- 6:07evaluations on token usage, but they’ve
- 6:09scrapped that and changed their metrics
- 6:11to ask, "So, what did you actually
- 6:13achieve?"
- 6:15>> I think the point that inference costs
- 6:17are at 0.1%is the key.
- 6:19I mean, look at it this way. If S&P 500
- 6:21companies’ revenue doesn't drop, it will
- 6:24continue to grow. And they will spend a
- 6:26portion of that revenue on costs. But
- 6:29the current spending on AI is only 0.1%
- 6:32. How much do you think they could end
- 6:33up spending? When you think about it,
- 6:35people are worried that a significant
- 6:37portion of labor costs, raw material
- 6:39costs, and other expenses could be
- 6:41replaced by AI. Whenever I talk about
- 6:43this, I tell people to pick one side or
- 6:45the other. Isn’t it contradictory to
- 6:48worry about losing jobs to AI while
- 6:49simultaneously thinking this expenditure
- 6:52won’t go up? I mean, just pick one.
- 6:54Either don’t worry about it, or consider
- 6:57that if the current AI expenditure is
- 6:590.1%of revenue, hitting 1%would be a 10
- 7:02-fold increase. And 2%would be a 20-
- 7:04fold increase. So, looking at the speed
- 7:06of current progress, you might think, "
- 7:09What should we do?" but I feel like you
- 7:11should either stop worrying or just
- 7:13commit to it; you have to pick one. Yes
- 7:15.
- 7:15>> He also talks about what comes after
- 7:17adoption. As mentioned earlier, having
- 7:19used it himself, he speaks in terms of
- 7:21the commercialization of AI models.
- 7:23Going forward, performance within the
- 7:25same job groups will inevitably become
- 7:27similar, so the differentiator will be
- 7:29how humans use AI to make judgments;
- 7:31that gap in quality is where corporate
- 7:33competitiveness will be decided. That’s
- 7:35why people talk so much about how the
- 7:37market has entered the AI adoption phase
- 7:39. He emphasized that in this adoption
- 7:41phase, companies that utilize AI
- 7:43effectively will be the ones that truly
- 7:45stand out.
- 7:45>> That’s good.
- 7:46>> Yes. Regarding the possibility of a
- 7:48bubble, while he spoke positively, he
- 7:50suggests that when discussing whether a
- 7:52bubble is coming or the current status,
- 7:54we must break AI down into three
- 7:55categories. First, technology, capital
- 7:57expenditure, and stocks. As for the
- 7:59technology, it is not a bubble. It is an
- 8:01undeniably transformative technological
- 8:03innovation, and as for capital
- 8:04investment, it is unclear if that is a
- 8:06bubble. Because although people worry
- 8:08about the large amounts being spent,
- 8:09>> Hmm.
- 8:10>> if the earnings generated are just as
- 8:12large, then it isn't a bubble. Therefore
- 8:14, the correct answer is that we don't
- 8:15know yet, and regarding whether stocks
- 8:17are a bubble, as I mentioned before, the
- 8:19overheated state seems to have calmed
- 8:20down significantly. Meaning it’s not a
- 8:22bubble.
- 8:22>> Yes.
- 8:23>> He diagnosed that the current bubble
- 8:25progression is relatively healthy. Yes.
- 8:27At the same time, if he talks so
- 8:29positively about AI, how is he investing
- 8:31? Regarding AI investment, he said that
- 8:33while the situation is healthy, he is
- 8:35staying out because he doesn't know.
- 8:36That’s what he said. Of course, I will
- 8:38mention this later. He does say that. He
- 8:40claims that AI is the most uncertain
- 8:41technology he has seen in his 60-year
- 8:43career. It’s difficult to predict how
- 8:45far the technology will advance, how the
- 8:47industrial landscape will change to
- 8:49determine winners and losers, and
- 8:50exactly how the business models will
- 8:52work to generate money from AI. For
- 8:54instance, Anthropic is receiving a lot
- 8:56of attention right now, but if you ask
- 8:58what new business they will be in five
- 9:00years before their IPO, no one can
- 9:02answer. Hmm. It’s that transformative of
- 9:04a technology, and since Marks is a value
- 9:06investor, wouldn’t you agree? He needs
- 9:08to be able to forecast what future cash
- 9:10flows will look like. You need to build
- 9:12scenarios, but AI is difficult to value
- 9:15because of the many uncertainties
- 9:16mentioned earlier. He mentioned that
- 9:19data centers and infrastructure are at
- 9:21least a bit better. Does that mean you
- 9:22shouldn't invest? That's not it. Because
- 9:24this is such a revolutionary technology
- 9:26and, as we saw earlier, we are still in
- 9:28the early stages of the cycle, so it's
- 9:30not about not investing, but the
- 9:31important thing is to invest while
- 9:33acknowledging there is much you don't
- 9:35know. So, don't just think, "Oh, this
- 9:36seems like it will do well." Don't go
- 9:38all-in with leverage just because you
- 9:40think it’s bound to go up; don't do
- 9:41things like that.
- 9:42>> When we hear the word "uncertain," we
- 9:44often see it as a negative. People only
- 9:46think in the negative sense, but we also
- 9:48have something called "upside risk." So
- 9:49, saying "I don't know what will happen
- 9:51with AI" can mean AI might not be able
- 9:53to do anything, but conversely, it can
- 9:55also mean I don't know what AI might be
- 9:57capable of. It’s that kind of talk as
- 9:58well. Watching the interview, what I
- 10:01felt about Howard Marks' remarks is that
- 10:03he is truly and thoroughly humble about
- 10:05AI technology.
- 10:08I don't know how far this will go.
- 10:09That’s why he says he won’t even short
- 10:11it. Because shorting it is dangerous.
- 10:14Who knows what will happen? That was one
- 10:16of the most impressive things from his
- 10:18interview, I think.
- 10:20When talking about Anthropic, he said, "
- 10:22I have no idea what Anthropic will be
- 10:24selling in three years."
- 10:26Because three years ago, their revenue
- 10:28was zero.
- 10:30But look at where they are now, and I
- 10:32can't even imagine what they’ll be
- 10:33selling next. So, obviously, you
- 10:35shouldn't short it. Yes.
- 10:37>> If so, regarding the fields he is
- 10:38investing in, as mentioned earlier, he
- 10:40is investing in data centers.
- 10:42>> I see.
- 10:42>> He stated that he is investing in
- 10:44related projects. When you look into it
- 10:45, he approaches this very conservatively
- 10:47and cautiously as well. He doesn't just
- 10:49invest in any data center; since there
- 10:51are a ton of people saying they'll build
- 10:53them, he said he only invested in
- 10:55projects where he believed a limited
- 10:56number of data centers would actually be
- 10:58built, considering geographic traits,
- 11:00local systems, and regulations.
- 11:02Also, considering the Marks I have known
- 11:05so far, he made a very bold statement,
- 11:07as he rarely uses the word "must."
- 11:10He attached the condition "must" to
- 11:13power. Power will "must" be a bottleneck
- 11:15. He said that AI infrastructure will
- 11:17increase by three to ten times in the
- 11:19future, so where will we get three to
- 11:21ten times the power? So, another memo
- 11:23was uploaded this morning.
- 11:26I saw that he even scored the
- 11:27predictability of these AI technology
- 11:29fields.
- 11:32Electricity is the most predictable
- 11:34sector.
- 11:35Score.
- 11:36>> The fact that power will be in short
- 11:38supply is highly predictable.
- 11:40However, I personally find it difficult
- 11:42to use the word "certainly" here.
- 11:45The reason is that AI infrastructure
- 11:47will likely require ten times more power
- 11:50.
- 11:52While current technology requires this
- 11:54much power, new, emerging technologies
- 11:57could lead to memory or other
- 11:58innovations that don't need as much.
- 12:02>> Power efficiency is key.
- 12:04>> Since we don't know how much improvement
- 12:06will come from those areas, "certainly"
- 12:08is a tough word, but under the
- 12:10assumption that there are no such
- 12:12technological surprises, then yes, it's
- 12:14certain.
- 12:15Yes.
- 12:16>> As I mentioned earlier, Marks is a
- 12:18credit investor, isn't he? He must be
- 12:20doing a tremendous amount of deal
- 12:22sourcing.
- 12:23I suspect he made this judgment after
- 12:25seeing things on the ground.
- 12:26>> The answer lies in the field.
- 12:28>> That’s correct. Regarding the bubble,
- 12:30will a bubble come in the future? Didn't
- 12:32the Fed say something about that
- 12:33recently? They claimed that raising
- 12:35interest rates had the effect of cooling
- 12:37down market overheating, but Marks
- 12:39doesn't agree with that.
- 12:40>> Aha.
- 12:40>> A bubble cannot be prevented by policy
- 12:43alone. Because it is a matter of human
- 12:44nature, specifically irrational
- 12:46exuberance, it cannot be avoided or
- 12:48stopped; haven't we seen recent
- 12:49phenomena like sharp drops? Sharp
- 12:51fluctuations. There must be these kinds
- 12:53of corrections to let the heat out and
- 12:55allow things to last longer.
- 12:57He concluded the interview by saying
- 12:59this.
- 13:00If I may add a few of my own thoughts.
- 13:02>> It means those who need to be shaken out
- 13:04should be shaken out.
- 13:05>> We were talking about bubbles, right? I
- 13:07personally believe that a bubble is born
- 13:09only after experiencing a crisis. I
- 13:11actually really like looking at old
- 13:13articles. I search for them. The period
- 13:16being compared to now is the dot-com
- 13:17bubble, isn't it?
- 13:19This graph is the NASDAQ chart from 1996
- 13:23to 1998. The black box is the period
- 13:25when the LTCM crisis occurred. At that
- 13:28time, when the LTCM crisis broke out, it
- 13:30fell 30%from its peak.
- 13:32>> It dropped 30%.
- 13:33>> It went from 2,000 to 14,000, so it fell
- 13:3630%, but for reference, it's almost
- 13:3830,000 now, right? Yes. That is correct
- 13:41. Hasn't it fallen about 10%from its
- 13:43peak now?
- 13:43>> 27,000 or something like that.
- 13:45>> If it's 30%, based on a rough
- 13:46quantitative standard.
- 13:47>> It goes down to 22,000. 22,000, or
- 13:4921,000. I think that could actually
- 13:51happen. Yes.
- 13:52>> So, it fell 30%like that, and if you
- 13:53look at the articles from that time,
- 13:55what was the mood? The financial market
- 13:56is ruined.
- 13:57>> Aha.
- 13:58>> Oh, what are we going to do? Back then,
- 13:59the Fed had to urgently summon
- 14:01commercial banks to create a bailout
- 14:02package. So the consensus was, "Yeah,
- 14:04the internet is great, but the stock
- 14:06market is broken." It's over. "I don't
- 14:08care about anything else anymore."
- 14:10That was the prevailing sentiment. "The
- 14:12market is doomed."
- 14:13>> Mhm.
- 14:14>> But what happened after that? This is
- 14:16the NASDAQ index leading up to the 2000s
- 14:19, specifically the peak in 2000. In the
- 14:21year following the LTCM crisis, the
- 14:24NASDAQ 100 rose by 102%. The index
- 14:27itself. Looking at things like this, we
- 14:29usually discuss market cycles in terms
- 14:32of market sentiment.
- 14:34The stages of that sentiment reflect Sir
- 14:37John Templeton’s famous quote: "Bull
- 14:39markets are born on pessimism, grow on
- 14:42skepticism, mature on optimism, and die
- 14:44on euphoria."
- 14:47So, the market cycle is determined by
- 14:49how much worry disappears.
- 14:53Usually, we tend to think that market
- 14:55sentiment overheats in a somewhat linear
- 14:57fashion, just getting better and better
- 14:59.
- 15:00But looking at past cases, market
- 15:02sentiment overheats when we become
- 15:03immune to worry.
- 15:04>> Wow.
- 15:05>> Even back then, people thought, "We
- 15:06survived the LTCM crisis, so there's no
- 15:08need to worry about the market breaking
- 15:10." They said, "We’re entering a new
- 15:11world," and that's how it ended up like
- 15:13this.
- 15:13>> Wow.
- 15:14>> So, you shouldn't be afraid when the
- 15:16market is volatile like it is today; the
- 15:19real danger is when the market swings
- 15:21wildly and people are not afraid.
- 15:23>> Uh-huh.
- 15:23>> That's when I start thinking that it
- 15:25might be a real crisis.
- 15:26>> Hmm.
- 15:27>> That’s all for today’s content.
- 15:29>> Yes.
- 15:31Listening to Howard Marks’ insights
- 15:33really calms the mind and helps me
- 15:35resolve how I should approach my
- 15:36investments.
- 15:38So, I really enjoyed hearing from Howard
- 15:41Marks.
About this transcript
This page contains the full transcript of 보수적인 하워드 막스가 이번엔 AI에 의외의 낙관을 꺼냈습니다... 모른다면서도 돈을 넣은 단 한 곳은 따로 있었습니다 [구루 11월 - 하워드 막스] by 이효석아카데미, generated from the public captions YouTube serves with the video. The transcript has 2,956 words across 492 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.