THE BIGGEST AND MOST DANGEROUS BUBBLE (Guest: Julien Garran)

Summary

• Market Outlook: Julian Garran discusses the current state of the market, highlighting what he believes is the largest and most dangerous bubble ever, driven by misallocated capital in sectors like AI, housing, and crypto.
• AI Technology: Garran argues that AI technology, particularly large language models, is overhyped and unlikely to deliver the transformative impact that tech bulls expect, due to its inherent limitations and high costs.
• Investment Strategy: He suggests that the misallocation of capital into non-productive areas is a significant risk, and emphasizes the importance of investing in areas with real economic returns, such as resources and emerging markets like India.
• Economic Risks: The conversation touches on the potential economic malaise that could result from the unwinding of the current bubble, which could challenge the globalist economic agenda established in the late 20th century.
• Commodity Insights: Garran shares insights from his past experiences in commodities, using historical examples to illustrate how market dynamics can lead to unexpected demand surges, as seen with iron ore in the past.
• Global Investment Themes: The discussion highlights the potential for significant growth in India's middle class, which could drive a substantial increase in resource demand, similar to China's growth trajectory in the early 2000s.
• Investment Caution: The podcast concludes with a cautionary note on the current exuberance in AI and tech stocks, suggesting that investors should be wary of overvalued assets and consider the broader economic implications of current market trends.

Transcript

Hit it. It's Thursday, October 15, 2025, episode 276. I'm Patrick Szna. And I'm Kevin Mure. This week, we have the great pleasure to welcome to the show Julian Garin from Macro Strategy. He explains why he thinks this is the biggest and most dangerous bubble we've ever seen. He walks through why he thinks not only is the pricing too optimistic, but why the AI technology will likely disappoint tech bulls. And he also hints at a trade that he likes a whole lot better. Then in this topy turvy market, the need for Patrick's crayon analysis has never been greater. The need stick around for talking charts. >> The need for those crayons is necessary. But listen, >> we're going to uh we're going to stick around here because we need to talk about this beer. Okay, >> Danny, what beer are uh we all drinking? We coordinated this. Uh, so what is it that we're treating? >> Just first of all, it's actually October 16th, not 15th. So just to correct you there, Patrick, just so you know. >> Well, you know what? That's my fault because I wrote the script for that and Patrick is like Ron Burgundy and he just reads whatever is there. >> Exactly. And secondly, >> literally Ron Burgundy. >> And secondly, this is Kev's pick. So this is mascot lightweight beer local. And I love love love the fact that we started this all together with the strongest beer picked by Batrick and now we're ending with a lighter >> pics by Kevin. So it just says it says everything, right? >> It just says everything. It's actually really nice. >> It's because he makes me drink this midday. So I don't want like a n and a half% beer. A nice three or three and a half or whatever this is. >> It's a lightweight like me. >> Lightweight. >> It's a lightweight. You know, one of my good pals used to say this when we would go out drinking. He goes, "Kev, you're a lightweight, but you never pretend that you're any different." So, that's good. >> All right, Kev, give us a disclaimer, buddy. >> Nothing in this podcast should be viewed as investment advice. Listeners should consult an investment professional before making any decisions regarding topics mentioned in the show. Side effects of too much huddle may include the precious metal mania disorder. That's the PMMD. [Music] >> Then the gold rush hysteria syndrome, the GRHS. I think a lot of folks have come down with that recently. >> And then finally, the silver squeeze psychosis, the SSP. >> Okay, so there's a lot of precious metals there, but uh none of them are the FOMO one where I sold too soon. And uh and >> well, I can tell you that a lot of people are experiencing that. >> Oh my god. All right. Uh let's let's get to the guest. >> It's our great pleasure to welcome to the show Julian Garin. Uh he is the partner in macro strategy. And Julian, thank you very much for making time for us today. >> Great to be here. Thank you. >> Um so, wow. when I read your piece, I was like, "This is got to be one of the most um I I don't know how to say it. Um the the piece that spoke to me the most about the AI bubble that we're currently in, and I'm just going to read what you said at the beginning here." He says, "This note seeks to set out the reasons why the AI bubble has grown so large and then to draw out the forces that will end the bubble and what the aftermath will look like. Make no mistake, I think this is the biggest and most dangerous bubble the world has ever seen. The misallocation of capital in the US, which also includes housing, VC, and crypto, is already 17 times the dotcom bubble and four times the 2008 real estate bubble. And as it unwinds, it will not just threaten significant economic malaise. It will threaten to overturn the entire globalist agenda that developed with the advent of Thatcher and Reagan in 1979 and 1982, accelerated with the fall of the Berlin Wall that fell in 1988 and sped up again with China's succession into the WTO at 2002. That is some piece and I it certainly got my attention and we're going to talk about that. But before we do, let's learn a little bit about yourself. Um why don't you tell us a little bit about your background? Did you you know how you got here, your journey and uh just kind of tell us let us know who you are. >> Yeah, sure thing. Well, I guess when when Rishi Sunnak was the chancellor of the exjecker here in the UK, the the finance minister, um my kind of running joke was that I had a better CV than him because we both went to Winchester and Oxbridge. um we both worked in top three investment banks and at hedge funds, but did he win the Winchester Economics Prize when he was 16? Uh no, he didn't. And so I think um history has proven that they they picked the wrong man for the job. But um but in in kind of um in finance in fi in finance um I started out as a commodities guy. I worked with commodities research unit. I was a then commodity strategist at ABN, Amro, UBS and and ran commodity funds on the buy side. Um, and I guess you for me I was always competing against kind of other firms which had much bigger teams and so on. So I was never going to beat them by having a more detailed nickel supply demand balance or or what have you. Um, so I was always trying to kind of turn it into a macro view or try try and kind of find some kind of angle. And I think that kind of the probably the best bit of work I did on the cell side was um was in the aftermath of the great financial crisis. Um, we'd identified that there'd been this kind of massive detocking taking place and most people were talking about say maybe a 5% uptick in in kind of copper or iron or demand or a 10% uptick. Um, but I use the example of the analogy of a wine seller. So if you imagine that in year one, you know, you've got 100 bottles of wine in your cellar, you drink a hundred, you buy a 100, and you're left with 100. So that's all fine. In year two, you're in a downturn. Um, so you decide to drink kind of 90. Um, but you only decide to replace say 50 of them. So, uh, you're down you're down to about a stock of about 60. Now the next year um you say okay well uh things are improving. I'm going to drink 95 but this time I'm going to buy 95. I'm going to replace everything. And suddenly your even though your year-on-year consumption is only up you know 6%. Um your purchases are up 60%. And and even more and so um that that kind of gave us the idea that something might happen that could be really quite exciting and unexpected. And then I worked with a guy who was um covering the the Brazilian iron ore maker Valet and he was saying well can we actually put these numbers you know into into a real world see what actually might happen to iron or demand out of Europe and so we did the back of the envelope numbers and I reckon that the demand for iron ore out of Europe should be up about 90%. Um so we called Valet um kind of that evening and said look you know this is what we're doing this is what it's telling us. are you seeing anything on the ground? And they said, well, funny you should mention that. We just did our quarterly kind of iron or demand kind of calls with all the European steel makers. And and the demand's up 100% and we're going to have to stop selling to China. And at that point, we knew it was real and and so we went we went out with that the next morning to to the clients and the market went mental and iron prices were up 40% in five weeks. Uh and the stocks went ballistic. So, so that was that was kind of fun. It was it was sort of moving it beyond a traditional kind of adding up all the and you know dotting the eyes and crossing the tees and trying to think a bit a bit more deeply about how the system works. So um just walk us through in terms of the timing like did you so you graduate from university what is that the the is it early 90s and or is it even earlier than that and like did you get a job straight on like on the sell side and uh was it always commodities? No, no. I had I had a bit of a So, so I started when I was at college, I was um very humanidocious. And so I I decided when I when I left university, what I wanted to do was work with homeless people. So So I did that for I did that for about five years in in Oxford and then in London. >> And then and then after that time, I was sort of realizing that I wasn't moving things forward that quickly. And and I kind of always liked markets. And so I started thinking about kind of what market, you know, I could re-engage my analytical brain uh and kind of and get into that and and I found probably the best kind of, you know, proving ground for that kind of thing, which was uh which was a commodities research unit. Um which had been going since the 60s. Um and that that was the perfect place to kind of learn learn the groundwork for how to properly analyze a market. Um and it was there you know I was kind of was getting interested in in these kind of um how how we could get how we were getting kind of supply demand balances wrong etc. and worked out uh an inventory kind of model that that was which effectively accelerated um the end demand effects and that was the beginning of of adding value in that way. Um so yeah so so I did that for about five years and then I joined ABN Amro as said head of commodity strategy but I I kind of moved on after sort of um five years of that to um working more on the macro side. So I became um the the chief kind of European equity strategist at ABN and then I also kind of worked as a a sort of macro kind of analyst on the sales desk before going onto the buy side um where I joined legal in general. It's a big insurer here in the UK um to both to run their global asset allocation and then to start their macro fund. The plan was at that point um to try and try and make kind of more than 10% returns completely uncorrelated with the S&P with with the idea that the average hedge fund or if you add all the hedge funds up together, they basically make you a return of the S&P times 0.5. So, so the argument is that's not that useful. You might as well just >> Let me just get this straight. You went from you went from like helping the homeless to working as a sellside commodities analyst basically. >> Yeah. And the clients have always been needy whichever job I've been doing. So you know >> well that was very honorable of you and I you you definitely have a unique uh background that way. In terms of the um you must have been there for the kind of boots on the ground for the bricks moment. What what was that like experiencing that kind of uh let's just say enthusiasm slashbubble like of of the China entering and and buying all these commodities. Can you tell us a little bit of you know maybe share a story or two about that? >> Yeah no sure thing and and well before it happened and and it felt it felt when it started happening it felt like redemption because before it happened we were um we were sitting on a mining desk. It was the European mining desk at ABN Ammer. We were next to the European kind of internet desk and these guys were kind of, you know, shouting their heads off about bubbles. You know, to the point you couldn't hear the microphone of so much feedback and we were sort of going, "This is nonsense. None of this is real what they do." And and um yet they were get, you know, the big bonuses were going in one direction and they certainly weren't coming in our direction. So, so, so when when we started seeing and and you during the downturn in two 2000 and into 2001, we started realizing that the next upturn could get very very interesting indeed. Um, and and the work that that I did on that was um used kind of scurve analysis. So, so it's this idea that um the intensity of use. So, when at at that point my report was called the $5,000 question, right? which was when someone in an emerging market gets to about $5,000 income PPP. Um they can start to afford to live in an apartment with uh kind of electricity and running water. And then the desire to step up and buy the modcons, the boiler, the kitchen set, the bathroom set, the the toilet and and and the moped and the car etc. is huge. You the the kind of incentive to better yourself to work hard is is massive. and you go through this incredible intensity of use kind of switch gears change and and and so we saw that coming um and that kind of explained this explosion in resource demand that we saw and and what's quite interesting in in the current world is that India um is in almost exactly the same position today as China was in in 20201 um in that it's got this burgeoning middle class population. Now the report is called the10 to $15,000 question because of inflation but the the idea is exactly the same. Um and and I'm using basically the same analysis that I used back in 2001 to argue that you know India could see just an extraordinary we could see an extraordinary increase in the intensity of use of resources as India's middle class moves into and then along this 10 to $15,000 income bracket and and just to put it in context um it took from about 1790 sorry 1890 to 1970 to get around uh 180 million Americans through the kind of whole ModCon kind of entering the middle class buying the the the you know getting the toilet and the hot water and stuff going through that whole middle class kind of process um that we're going to see about 240 million people in India go through that process over five years >> so so this is this is this is a potential massive global game changer. >> It's why do you think the market is not talking about that story? You think that that would be kind of front page story and it's it's not by any means. >> It it it isn't. And I I think part of the thing is and similarly they weren't talking about the story in China back in 2001. Um I I remember kind of like the front page of the Economist at the time saying that China, you know, couldn't couldn't expand because the banks were bust and corrupt. Um and and they they'd missed which may have been true, but it wasn't it didn't actually stop it didn't actually stop it from happening. Um and and so I I think it's a it's a sort of similar thing. The focus is elsewhere, but actually I think this is probably going to be one one of the most important and it's going to be the most important development in quantitive terms in terms of raising people's living standards. It's going to be the most important event in history in terms of doing that. if we see this happen over the next 10 years. >> Okay. So, let's get to the topic. Dujour um the bubble that we're in on the AI and I've had the good fortune to you know hear you speak about this. So, I know you're well verssed and you have a whole kind of presentation. So, instead of me kind of peppering you with questions, I'm going to actually let you go. So, we're going to pull up the slides and why don't you just kind of lay out the thesis for you for us. I'll jump in and you know when when needed but otherwise you go ahead and tell your story because it's a it's a terrific story. >> Sure thing. Okay. So um I've I've been following AI kind of closely from from well well before kind of chatbt was was kind of launched kind of back in uh kind of November 22. Um I've been looking at how AI has been developing and and how useful it may or may not be. Um and um over that time I've come up with a kind of golden rule uh and especially since the advent of the large language models with chat GBT and Anthropic etc. And my golden rule is um if you use a large language model um to develop an app or a service um it'll never be commercial. You'll never be able to make a return off doing that. Um and there's a a range of reasons why that's true. Um the first set of reasons was that it was kind of built to fail in that the way that they've built it. It can't reach it can become a very good simulacum of language uh and to a lesser degree of pictures and a much lesser degree of for videos but it can't ever actually apply cognitive intelligence. Um and the places where people do find it useful which is kind of uh coders for instance accessing large amounts of data um or kind of in some cases people using the video functions to to shortcut a more expensive kind of image creation process. Um in those cases the the fact that large language models have hit a scaling wall um means that the cost of providing all the compute to service those needs is greater than what people are prepared to pay. So it's getting kind of squeezed from both sides and I think you know we're starting to see uh the fallout from this uh if you watch closely in that when Chachi BT5 was launched it was kind of widely perceived as being a flop because it wasn't significantly better than previous versions. Uh yet they'd spent on my estimates up to 80 times 80 to 100 times more compute on building it. So that that starts to raise questions and what we've also seen is that the developers of of large language models um have started talking about a pivot to uh you know naturally if something's not working. Uh if you're if you're a kind of California startup you pivot and and the the pivot was first of all to to kind of reinforcement learning that was in August. Um but Toby or the uh um the kind of computer engineer pointed out that to get one bit of knowledge if you like from reinforcement learning takes a thousand to a million times more compute than even a large language model which is using too much compute anyway. Uh so that was seen as too expensive and they're now moving to talking about using world models. So this is just two months later they've changed again talking about using world models which are like synthetic physics kind of uh correlated kind of models of how you move around the world. But again and I'll talk about this a bit later these these things are very limited. You can't learn how the world works from a model that doesn't include everything about how the world works. And so what I what I might just sort of highlight um is the elements that that make this happen. And within that I think the big issue is that this is a misallocation of capital. Um this is a huge amount of investment taking place uh that's not making a return on the cost of capital and that's hugely important for the economy for society as a whole because the capital stock is the foundation for basically the entire economy everyone's you know hopes and and and prospects all depend on a capital stock that makes a return. And so if you divert the size of funds away from the real economy into something that doesn't make a return, then you're basically taking away the prospects for the economy and for the world. And and that's clearly happening. When we look at bank lending this year, uh there's been no bank lending growth to the real economy. It's all been into the financialized economy to non-bank financial or non-depository taking financial institutions a chunk of which has been diverted into chips. Um and the way that we measure this misalgation is that we use a Wixle spritz and that Wixle was an economist kind of from a bit over a hundred years ago. He's Swedish. Um and he argued that um the key to capitalism, key to allocating capital um was getting a neutral rate of interest um which was you know the cost of debt for the average corporate should be a couple of percentage points above the growth of nominal GDP which is kind of the return on the economy as a whole. And if you set rates at that level um then business people who know their own business uh if they think they can make a return over the cost of debt they'll borrow and they'll expand. uh if they don't think they'll make a return, they'll run for cash and they'll shrink. And that way capital is allocated as well as it possibly can be uh by the people who are closest to uh their businesses and and their business's uh potential profits and losses. Um and that gives you not only the fastest growth you can get um but it also means the highest returns that you can get and it also means the biggest kind of surplus that you can use to invest in growing the capital stock to improve the economy for everyone going forward. Now the problem with this is that if you interfere with it and my view is that Powell interfered with it more than anyone else in history during 2001 um when he was running QE with zero rates um when uh the nominal economy was growing at 11%. Um and what that did was it created on the bottom right there um a wixle spread that had gone massively negative. uh that means it's hugely stimulatory to people and it incentivizes them to basically people in companies to gear up to buy and build assets. Um and the thing about this particular kind of post kind of COVID kind of stimulus and then gradual withdrawal of stimulus is a lot of the stimulus was kind of side pocketed and then bled into the economy a bit later. So about 1.3 trillion of the fiscal stimulus was saved and then spent down over 22, 23 and into 24. Two trillion of monetary stimulus went into reverse repo still there by the end of 22 came out over the next couple of years etc. So this thing kind of lasted for five years. Um now the danger with with this is that when you do this um the assets get priced to perfection but then later in the cycle when rates go up and growth slows and spreads widen uh the wixel spread goes above 2% which it's just doing right now. Okay. Um and at that point you start to see the assets impaired and the debt and equity behind them damaged. And that's the danger is that we're just getting into the zone now where the Wixle dynamics are going wrong. Um, and that's why we're kind of sounding the alarm because this is the first time since the pandemic that we're actually at a point where the Wixle spread is saying, look, this it's time to be careful about misallocated capital. Um and in terms of the size of the misallocation um it's now reached cumulatively about 65% just over 65% of US GDP. This is about 17 times bigger than the docom bubble and about four times bigger than the real estate and credit bubble that preceded the GFC. Now this isn't just tech. This isn't just AI. It's also kind of real estate. It's also crypto. It's nonfgeible tokens etc. But all of these things together uh create a very substantial risk to an economy that looks kind of plastered on the surface. Underneath there are all sorts of weaknesses there. Um now that gets me to the kind of issues around kind of um large language model AI and I'll I'll talk about I think I'll go straight into talking about um why they were built to fail um and then the scaling issues. And so in terms of why they were built to fail, basically what they do when they build a large language model is they um they give each word in in a language a vector. Um and a vector for you or I like if we were talking about where we are in the world, we'd need three numbers, our longitude, our latitude, and our height above sea level. And if I subtracted my vector from yours, we'd know exactly how far apart we were. So it's it's a useful kind of tool uh in a kind of narrow kind of uh application. Um what they've done with large language models is they've given each word say 60,000 words a vector. Chatb3 had uh 12,440 numbers or dimensions per vector. Um and then they train it. Um and what the training does it goes through a transformer. Uh so they basically train it on a large amount of data on the net goes through a transformer and the transformer works out the correlations between each of the numbers and what you end up with um is um a thing that can work out how words are statistically related to each other. So cat is quite closely related to dog because they're both pets. Both have four legs and a tail and so on. Um, both cat and dog are quite closely related to running because they both do that, but they're related obviously in a different way because that's what they do. Um, but dogs tend to run in the park and cats more likely to run in the garden and so on. Um, and what that then allows is for the large language model to come up with a good prediction of what the next word is likely to be in a coherent sentence. And it's allowed them to create a very effective kind of similacrim of language. But you can tell by the way that they've done it that that isn't actually language. That isn't a person with an intention of kind of transferring a meaning and trying to get something done to someone else. This is just a statistical process. Um and that's the sort of hidden heart of why this may be a problem. um and one kind of thing that sort of comes out of this. Now this is the one strand on which they're working. The other strand that they work along is um rope learning. So what people found was that um these large language models, you know, could create coherent sentences. Um but sometimes they would give good answers and sometimes they wouldn't give good answers. And so they've put in overlays and the overlays are kind of rope learning overlays. So there are companies that will you know go and find people in a distressed nation um uh pay them a few cents an hour to come up with model answers to likely questions. So they overlay those. Now the problem with that is um if you asked a kid um kind of you know what's the what's the structure of of an atom and that kid has access to a reference library. Does it matter um if the kid has a deep understanding of atomic physics and is answered based on knowledge or has just copied it out of the encyclopedia? Um and the answer is it does matter because you might be able to ask one more question like kind of why is ra uranium radioactive and it might be able to copy that out of the encyclopedia. But if you ask a a less common question like why is firmium radioactive or Americive which isn't in the encyclopedia the the encyclopedia copying kid is going to be all at sea. Um whereas the kid who's got a deep understanding of atomic physics is going to be able to answer your question. And because life is messy, um if you're outside the constraints of a very narrow application, um the chances are you're going to come across one of these types of unasked questions really quite quickly. And that's where these things start falling down. And so I gave the example. So once you know that's how they kind of work, um it's quite easy to think of questions that they're going to struggle with. So for instance, I asked um the um my image LLM um to do a picture of a photorealistic picture of a chessboard one move before white wins. Now this picture is a catastrophe. Number one, you could never you could never get a chessboard looking like that during a normal game. Number two, there's only seven squares on the front and back edges. This this is useless. Yet these LLMs have been trained on more chess books than any human could read in a lifetime and more chess games than any person could play out in a lifetime. And and why has it got it wrong? Well, it's got it wrong because that's not a question people ask. People might ask, can you play a game of chess? They don't you no one really asks, do put a chessboard up. I've not seen anyone ask that question before. And so it's it's lost because it's not got anything to correlate with. And secondly, even though it's been trained on all these books, it's not been trained to understand the meaning of the book to teach you how to tra play chess. It's been trained to understand the relationship between all the words in the book. And so this dual process of understanding the correlation of words and rope learning doesn't work for something dynamic like that. Um and the closer you further away you get from the abstract of language and the closer you get to the physical world reality and in a sense pictures are one step towards reality imm videos are another step towards and then you get the actual physical world to operate in. the closer you get, the worse it gets. And and for a long time when you asked um kind of uh the the video creating kind of LLMs to to do a video of gymnasts, they would have a real problem. I.e. the limbs would be all over the place. Extra limbs would grow during the video like in the image in the bottom right. Um because it doesn't fundamentally understand human physomy and that there are constraints on movement for human bodies. And so the first set of problems is that you have a limit to what this could be applied to. And and you know what can you use a regurgitation machine um that sometimes gets it wrong for? Well, you can you can use it for um you can use it for for instance homework. Um because you know homework is just regurgitation. Uh my daughter brought some homework home when she was 15 and and I recognized that it's exactly the same homework I had when I was 15. It's like Latin homework. It can change. Same textbook, same questions. Um and you look at that and go, it's got no value. So, so yes, they can regurgitate it. Um but it doesn't have you homework has no commercial value. And you'd argue all the value of homework is in the doing of it. Uh whether it be the learning, the discipline, learning the content, etc. And if you if you sidestep that then you destroy any intrinsic value in doing the homework. Um and then you say well how about coding? Um because that's sort of widely touted as the best use. But the problem is you you get kind of you know if you ask it to do something you can get it to you know recreate a video game like a shoot them up or breakouts or what have you. Um, but it it's kind of buggy and you know it's not it wouldn't fit in with your broader software stack. It's just copied the code that it's been taught on and so you couldn't commercialize it. Um, because the game's already out there. It's probably used copyright code, etc. Um, and so again, you run short of the requirements for a commercial, you know, a killer app, a very popular profitable app. Um now the the next issue is the issue of scaling. So, so um because of the the way these vectors work, if you want to go the next level and make your large language model more accurate and put more numbers in your vector, train it on bigger data and use more compute, the problem is that the benefit of having a bit bigger sort of um vector like 12,441 numbers in it is that your benefits from that are at best linear and not even that if you're running into the limit statist statistical limits of um of kind of of language. Um yet your costs you have to basically recalculate every single correlation within not just the oneword vector but in every other word vector and between every number of every word vector. And so your costs just go right through the ceiling while your benefits go flat and flat and flat. Now that didn't matter when all you were using was your PC at the lower numbers. because you know your hard disk and the compute kind of that you've got available to you in a sense was was just there. It didn't have any cost. But once you get into data center size requirements, the costs are just astronomic. And that's basically where we've got to. And you know, while that's a theoretical argument that you're going to hit a a scaling wall, and this is a chart from Toby or um the reality is that chat GBT3 cost 50 million, was released in in kind of November 2022. Chat GBT4 cost 500 million, was released in March of 2023. Uh they were planning to spend five billion on five and release it in Q3 of 2023. But when we got to that point, um, insiders and there's been reports from Reuters and the information talking to OpenAI insiders were saying that this thing isn't actually sufficiently better than the previous version to justify the 10 times higher cost. So they kept on running it on various training training runs. I estimate they used 80 to 100 times more compute uh on the ultimate five. They initially released it as 4.5 um than they did on four. And yet it's not much better than what's out there. And so you then have this problem that if you if people are using it rather than just for regurgitation, but they're using it to get hold of cheap compute, which is what some coders do. There's a subreddit kind of uh um kind of top 10 list of the people who've managed to extract the most amount of compute out of a 200 buck monthly subscription to Anthropics kind of um anthropics kind of coding app. Uh and and the winner or the winning kind of number at the moment is someone's got 10,200 bucks or so of compute out for 200 bucks. And so Anthrop is then like kind of hemorrhaging money from these guys and and this is so completely different from software as we know it where you know Microsoft the cost of Microsoft's first first Excel might have been a hundred,000 bucks but their second sale you know didn't cost them a buck and then and it goes down to you know less than a cent. Um this is the opposite of that. when you start scaling it, it actually costs an increasing amount. And so you can't get you can't get returns on that other wing of potential uses either. And that applies then to, you know, people making videos and images. And videos are just rife with with errors. And so you end up with the problem that people will just keep on redoing them and redoing them and redoing them until eventually they get something that doesn't look immediately like it's errorridden. But by the time they've done that, they've used massive amounts of compute. Which is why, for instance, Sora only allows you to do a 10-second video because if they allowed you to do anything longer, you'd get into all sorts of trouble. Likewise, you know, there's all sorts of questions that you can ask that that kind of get it working too hard. So, so you if you try and I tried one which didn't work, which is I said, you know, count from one to infinity minus one and it just said no, we're not doing that because that if they if they hadn't put a block on that then then open AI would have gone bust. this when the question but if you ask a slightly different question like um you know draw me the the architectural schematics for a new kind of 10 million person city just outside Dubai it's not necessarily going to know that it has to stop and so they have to put all these constraints on to try and stop wasting huge amounts of compute um and those fundamentally are the limits um and I might kind of pause for breath Sure, no problem. I'm going to jump in and because uh you know you you were doing great. I didn't want to bug you and uh interrupt, but one of the things that I think is amazing about your research and your your view is that a lot of folks will talk about the price of the stocks and the assets for AI as being overpriced. And there's very few people that are actually looking at the technology and saying, "No, not only is the assets overpriced compared to where they, you know, what the market is putting in there, but you're also saying it's not as good as everyone thinks and it's not as transformative. It's not as is as is as is as as revolutionary." And one of the things that I just guess when when people are listening to this, they're going to wonder, well, what does he see that the rest of the folks don't see? Like, why are you having this outside the box view about this? Whereas, meanwhile, everyone else is telling us that this is the second coming, the most important thing since like, you know, the wheel and yada yada yada. where where where do you think the disconnect becomes and how did we get here? um to to be honest I think I think this disconnect exists across a whole range of areas where especially in even even down to what I was saying earlier about the capital stock um and and misallocation of capital that there's very little understanding you know throughout most of school you're not taught that stuff um and and so people don't understand the foundation for why society and civilization actually exists and we're We're not very different. We're, you know, no more intelligent um than our ancestors who lived in caves in Europe. Um and um yet we live these very different lives uh both in terms of you know the uh the comfort that we live in but also the opportunity that we face and yet there's very little understanding in the public domain about what that is and why the capital stock and why a strong capital stock is is behind it. So, so I'm not surprised and and this stuff is, you know, it's partly because in the modern world we we have an attention deficit issue. Um, I was I was giving an interview with a a newspaper, national newspaper in the UK, uh, the Telegraph one time and I was chatting to one of the journalists there. Um, and they said that and I was asking, you know, what her what her kind of daily routine was. she was she was a junior and her routine was writing eight articles that she'd mainly kind of paraphrased from the from the wires and you're sort of going that's not journalism that's just that's just repeating stuff and and so if if you don't have the time um and if and in a sandbag culture it's quite hard to to get to the point where you're going so you've you've you use this thing right uh you know to to ask it's like an Aristotleian kind of set of questions to help someone get to the truth. And and so one one example was I've got a good friend um who started using it. So he he thought AIS had a series of uses and one of the things he was doing he didn't know how to use Python. So if you want to interrogate a large um data set um you kind of need to know how to use the Python coding language to do it. And what these large language models have allowed you to do is is interrogate the data set without having to learn Python. So he was doing that. Okay. >> Um and um and one of the issues that I have is that one of the reasons large language models are successful is because it looks about right. Most of the time the answer looks about right. And so it's very easy to go, oh well that is right then and I'll move on. Whereas if you actually go in and test it um rigorously, which takes a lot of time, it means you have to do all the work yourself in parallel, then you find out that it's not right all the time. And so he did this by accident because he he uploaded um he was uploading different data sets and asking the the um to to do correlations on them so he could look at how electricity prices would move with macro signals, for instance. and he did and by accident he uploaded the same data set a second time. uh but but it gave different answers and so he suddenly realized okay so all I don't know if all the previous answers were right because they look reasonable but then so does this and so so so you get stuck and and I think probably the best the the the best study that I've seen um was out of Stanford um kind of led by professor Frank Zoo um and it was called the the agent company and so what what he did was um he recreated a software company. Software compan is probably going to be the most accessible for a large language model to try and do the jobs and and he broke down all the tasks you need to run a software company and they had something like 486 or something. And then he got lot he got a range of large language model agents to do the tasks. uh and he found that the success rate ranged between one and a half percent which is obviously completion rate which was which was Grock which at the time Elamus was saying was Grock 2 which was the greatest LLM ever um up to about 34%. Now 34% would be useful if it was consistent um because you could then automate a third of all the tasks of the company and that would be valuable. The problem is it's not consistent. it's a different 34% every time you run the system because it's a probabilistic system and it runs into these issues where it kind of hits a question it's not been asked before. Um, which just happens in business. Um, and so that that for me is is it's because it looks all right that a lot of people accept it and because they don't have the time to test it, they accept it. >> The other thing if you take that time >> Yeah. Sorry. The other thing I thought was interesting when I heard you speak before was you mentioned that part of the problem is it it asserts these things. It gives you these answers and it doesn't put confidence intervals on it. So there could be something that has a 95% confidence interval and it'll look the same as something that it really just thinks has a 30% confidence interval. It's always forced to come up with an answer. >> Yeah. and and that's and and that's an issue of um the how how it was put into the public domain. So this was something that you know that the large language model is effectively aware um of whether it's been trained on an answer. Okay, so it it's got to kind of do two things when it tries to come up with an answer. Number one, it's got to try and find the meaning in the question. Uh, so it's got to kind of pause the question down and work out how closely connected that is to the right answers it's got or to the other training answers that it's got. And that's not always completely easy. Um and then it and and so once but once it's done that um if it's a pretty close correlation which in a sense it will know because it goes through a neural net and you get in a sense a bit like a neurons firing if you get a lot of neurons firing it tells you that's the likely thing that's going to be right but then it'll have others that are marginal but what the developers found was that if you have a system that comes out and and frequently says I'm not really Sure. Uh people have less confidence in it and it's just a it's a psychological thing that that people find unconfident people or other people who are unconfident less convincing than people who are confident. Now what we obviously in markets we we have to realize that just because someone's confident doesn't mean they're right and it could be actually more dangerous if they're confident than than if they're less confident. But um but that aside and and so they've so they've they've allowed that to run because otherwise less people would use it and they wouldn't get funding. And so in a sense it's disingenuous and and it's one of it's one of its major. Now, the other thing I've noticed is I I saw that um study or that paper that said that coders when you actually look at it, they weren't any faster using AI and that it was actually something like a it took them 120% of the time because, you know, they had to go fix the errors and it was to your point, it wasn't as good as they had expected. And and I just want to bring it back to this question like we're we're talking about all these things that are wrong with it yet we have Zuckerberg, Elon, all these folks just rushing out and spending huge amounts of money. So they obviously have a different view than than than you. Where does that kind of come from and and why are they doing that? Okay. So the in in the first instance that this you know when when you ask uh when you ask an LLM to do some coding for you it immediately looks impressive because you know within a minute of you asking the question or a couple of minutes it's written 200 lines of code and normally the code kind of does something. Um and so especially for a non-coder or a sort of amateur coder that's you know that's mana from heaven. It's been a lot quicker. The difficulty is then that it's buggy and that it's kind of potentially very open to hacking or disruption or just not consistent enough and what have you. So, it's it's useful if you're hobby coding like if you're kind of if you want to create a small app for your fossil collection or whatever you might have. Um but but it's but if you actually want to put this thing in the public domain, it's just nowhere close. and and if you want to connect it in with a broader kind of stack of of um of programs and models, it it's just not good enough. And so the it turns out to be a false economy. The speed at the front end is made up by all sorts of work you have to do if you want to commercialize it at the back end. Now there are some exceptions to this and and um and so this is where the question of how broadly useful it is comes in because you could argue quite reasonably there are some jobs out there that are kind of jobs and so and so parts of HR for instance if you're if you're the person who's in charge of sifting through the CVS um for for for someone yeah for a company How do they know whether you've done that job properly? Um they just get they just get the sort of shorter number smaller number of potential candidates. How do they know whether you've left a diamond in in the sort of uh in the dust bin? The answer is uh they don't know. And so no one can actually kind of work out whether these people are doing their job properly or not. Um there's an argument there's a there's a book out kind of last year um arguing that management consultants are mainly operating off hype and they're actually are actually a dead weight on society and aren't generating aren't generating useful work. And then there's also an argument um that monopolies um h are a problem because the the standard textbook understanding of a monopoly is that the monopoly can raise their price. the price in elastic. They raise the price and not that many people stop buying and so they get these monopoly profits. But it also works in another way which is they can cut their costs and lower their quality and not many people leave because they don't have a choice and so they can maximize their profits that way as well. And that's what ultimately I think is happening. I think that some jobs where people aren't testing or don't know whether the work is good or not uh like like HR or or consultancies or possibly marketing kind of programs and what have you. I think people are substituting with Um and so yeah, okay, but it's not it's not it's not great value. And then another area is if you if you're in a sufficiently strong market position that you can allow your quality to go down, then you can employ an AI to do that. And and you know what we're seeing at the moment is and Amazon yesterday said they were firing people. Uh Microsoft have been firing people and so on. And what seems to be happening is that they're losing people in the states. Um they're then using an AI to run a first pass on their work. Um, and then they're getting someone offshore to to basically clean up the slop and clean up the mess. Now, the end result of that is that the product isn't as good, but they don't really mind because their cost is less and they make more profit. And and so I think yes, it has those narrow applications, but when you look at kind of broader businesses outside the sort of um big seven, um you can see that adoption after a couple of years of trying these things, adoption is actually rolling off here on that chart on the right there. um it's actually declining while the likes of Meta and Microsoft you know may be in using it but not necessarily for for kind of um genuinely useful things just purely to cut costs even at the expense of quality. Right. Now, you mentioned Microsoft and one of the things that I've heard you talk about is the fact that Nadella is is recently stepped away from their expansion or their the pace of their expansion into AI. And I was wondering if you could just expand on that and why this might be kind of the canary, you know, singing in the coal mine. >> Yeah. And a sure thing. So the there's been and it hasn't just been the large language model companies that that have caught the public attention. What we've also seen is the cloud companies have been doing very well and they've booking been booking large amount you know very large revenue growth and Microsoft obviously is right in the heart of that. Um now what Microsoft has been doing is basically been um selling uh open AAI um kind of compute time in return for equity and then booking that compute time as revenue um and then because OpenAI has been going through um kind of funding rounds at higher prices um Microsoft can book that as marktomarket rise in the value of their investments and that just looks fantastic. Fantic. However, um probably no one else outside of a Open AI um knows more about Open AI than Nadella, given the large equity that Microsoft holds in the in the business. Um, and he was surely completely aware that uh, Chhatcha BD5 was meant to be coming out in Q3 of of kind of 2023 and that they had to keep spending more and more on it to try and get it to be sufficiently better to be worth releasing. And yet that wasn't working. And then he's also aware that um you know if you did happen to make a better LLM um that was a step above everything else that was out there um that there are systems that you can employ which is what kind of Deepseek did the Chinese um company which is you can use an LLM to create the synthetic data to train another LLM that's almost as good but which is much uh more efficient. And so you kind of lose your moat. Even if you were able to do it, you'd lose your moat and they weren't able to do it. And so my view is, well, he keeps promoting this because it's good for Microsoft. If Open AI keeps raising more funds at ever higher levels, it's good for their investment. But I think he's deeply aware that um that they've hit a scaling wall. And that's why he announced a Microsoft pivot and also said at the same time uh kind of four or five months ago that OpenAI was pivoting to products rather than training because clearly they're not seeing the benefit from training anymore. And then the problem then is that the products aren't much good because right you use you know the these agents that that um that open AAI launched you know sort of amusingly sort of one day after deepse the deepseek announcement earlier this year um its agent called operator um kind of I I watched the sort of launch of that uh and they were talking about what it could do. It's me, it's meant to be able to, you know, order your groceries, buy your tickets to a ball game, get you a pizza, and all that kind of stuff. Uh, and they got about sort of 20 minutes into this presentation and then showed it against benchmarks and it was completing these tasks one-third of the time. And you're sort of going, "That's terrible." Like if you if you order a pizza and you know and it and and then it doesn't it either just doesn't order it which is just a waste of your time or else it sort of orders it and it's delivered to your other house or whatever your home not your office or what whatever it does and then you're in kind of you know you're in kind of uh in this hell of trying to of trying to get your money back. And so that's that's clearly what's been happening. And ultimately what it reflects is that the ecosystem the AI ecosystem uh is reflective of this lack of ability to make returns in that only really Nvidia making the chips and maybe to a much lesser extent some of the training companies doing these model answers are making any money at all and everyone else is losing money hand over fist. Okay. >> And that's why Yeah. That's why that's where we get the next issue of price. >> I was going to talk about this because I think that we should bring it to some numbers because you do have not just a skepticism about the um the technology, but you also have a skepticism about the numbers and there's a lot of great stuff that you you've highlighted there. One of the things that you said is if you're a large data center, you can't make money. You're pretty well guaranteed a loss. Could you walk us through the math for these folks and what the problem is there? >> Yeah. No, sure. Sure thing. So, um, basically it it's possible to work out. So, so kind of if you think of the strata of of different AI companies, you've got Nvidia making the chips, you've got these data center, public data centers like coreweave and lambda uh along with data centers within the majors, but you don't get much data out of those on what they're doing. Uh and then on top of that you have the train the the developers like OpenAI and Anthropic and then on top of that you get some software companies that are providing kind of front-end things using like a like a kind of um coding app using a large language model. So if now if a an ecosystem is is healthy like say the residential property market in London is healthy you'd reckon you should be able to buy an apartment or a house and rent it out uh and make a small profit. Okay, so that would be indicative of a of a healthy ecosystem. And likewise, if the AI ecosystem is healthy, these companies, Corewave, Lambda, etc. Hyperbolic should be able to buy the chips at rent them out um and make a bit of a profit. Um now we can work out whether they are or not. Um because we know that the cost of a Blackwell is around 50,000 bucks. Um, we know it costs an extra 25,000 to put the thing into a data center because you not only have to build the data center, you have to build an incredibly uh advanced cooling system because each chip uses about the same power as four houses. So that's and you've got you've got kind of anything from 8 to 32 to 64 of these things in a stack. So that's some proper cooling you need. Um, and then and then you've got to provide the electricity and you've got a CPU to manage the workload and you know and some and some people working in the data center to make sure everything's going okay. Um, so that's your extra 25 grand. Now when you build a business like that, you'd be going okay well number one I've got to cover the and then and then the next issue is how long do they last? Now, I think this is the sort of Achilles heel of the of the whole ecosystem, which is that these chips uh you know, they're not like fiber optic cable that lasted for 30 40 years. These things degrade at about 1% a month. Um and um there are new versions and and kind of Nvidia has promised new, faster, more advanced versions on an annual basis. And so the value of these chips is actually should depreciate much faster than the six years or so that the majors and some of these kind of um some of these data centers are using. I and I think a fair number would be about two and a half years. Um now you can work out because you've got to pay the cost of debt. um you're expecting a return maybe 10% a year because you've got the potential for voids and for burnouts you can work out and that's what I've done in this chart on the left there how much and and also you add in the cost of electricity how much should they be renting it out to make a a reasonable 10% annual return and the answer is six bucks 31 an hour um now I waited all year to try and find a publicly available blackwell because there were lots of delays because of problems with the cooling. But finally in July um Lambda started publishing um kind of rental prices for Blackwells. Well, it's three bucks 79 an hour. Um so so yeah, you're not making a return. You're making 25% loss. And and that that's not that's not a business. and and the kind of the only way that becomes a business is if there is some extraordinary breakthrough with large language model training that it becomes you know an AGI that it becomes actually conscious and so the demand for training goes through the roof again. Now back in back in Q2 of 2023 uh the demand was going through the roof because everyone wanted to build an image AI. Um, so at that point an H100 was renting for about eight bucks and you could make your money back in about 15 months and then everything else would be gravy. So that was, you know, that at that point if those prices were going to hold, that looked like a decent thing to do. But today with that absent, and I think it stays absent, you're guaranteed a loss. And so at that point you're going this is this is just heading for a brick wall and I think the costs are going to you know it's going to end up costing these companies a lot of money because they're suddenly going to have to start showing a realistic depreciation schedule at which point their losses are going to go through the roof. And that that was why in in 2001 when the um when the economy had a relatively mild recession, a sort of 1% recession, um corporates saw profits go down kind of 35 to 40% roughly and that was because of all these writeoffs that took place around kind of.com infrastructure investment. I think the same thing is the set is set up um when this begins to roll over as well >> and and that was the vendor financing and like could you talk a little bit about that that period with the Cisco and the Nortells and the Lucence of the world and and how that might look like what we might see going forward here. >> Yeah. And a sure thing. So um the interesting thing about 2000 so so kind of the tech best happened in two stages. Um the first stage was that after pumping liquidity into the system um because of concerns around Y2K, Greenspan then started taking it out in February of of kind of 2000. And so you saw the NASDAQ peak in March. But um the process of investing in the internet at the corporate level continued um and during that period um Cisco which was the leading company providing internet infrastructure all the routters and switches um it was kind of ramping up its vendor financing and so its vendor financing its receivables went up about 138% over the course of about 15 months. Um but then by the time we got to November, um Cisco started to realize that their customers who they were offering decent credit terms to were running short of cash. Um and so they started getting nervous and then they pulled the plug and then the whole thing went and even though Cisco was a great company and very well managed, its share still went down over 90%. Now, when you look at um what's going on at Nvidia, um it kind of puts that in the shade in that Nvidia's receivables are up 626% over 30 months, which is an extraordinary amount. Now, obviously, the natural push back to that would be, yeah, but their revenues are going up fast. And that is true. They are. But um the um back at the beginning of that period um their receivables were around 55% of quarterly revenue. And given a receivable normally has a three-year kind of time frame, that's a reasonable comparison. But now they're 85%. Um and what Nvidia has been doing is selling chips uh and renting them back. And the question is why are they doing that? Um and and what are they doing with you know with that compute? Um and the answer is that they were setting up a process basically this global model stuff um so that people could train kind of robots and driverless cars on these models um with a view that that would help them they'd be able to do it quicker than real life training. Um but the danger here is that those are quite small ecosystems. you know, you don't have that many companies trying to train driverless cars. Uh you don't, you know, the the robot ecosystem, like the autonomous robot ecosystem is small. Um and there's a big problem with these models, these real world models, is that, you know, even if you're a pixel out or pixel different, they can learn very different things. So they're actually very fragile. Um and a further problem is that um they don't understand how objects work which is kind of quite fundamental. So so the mo for most of western civilizations they call it from the beginning from kind of year zero through to about 80 years ago um it was driven by a western philosophy that defined objects as being kind of discreet. uh you know they had physical properties like if it was dry there where it wasn't wet and if it was hard it wasn't soft and it wouldn't disappear if it went behind something and so on and and people were very happy with that and then H high digger came out about 80 90 years ago and said well just a minute actually the way you should define an object is how it's used um how people use it how it could be turned into a tool of some kind and that to understand an object in that way requires a much deeper understanding of how the world works and how humanity interacts with the physical world and it's much much more difficult to achieve. Those things aren't in the world models. They're they're just they're using the old kind of philosophy. They're not using a new one that understands how things can be used. And so the problem with all of this is that yeah that's not that's not a big market is and and there's a real question about whether Nvidia is able to able to kind of use that compute or whether it's just in a process of kind of roundtpping to try and keep the game going and that's the concern is that is that that process of of kind of roundtpping which I I did an example for coreweave when after the core IPO in March uh of all the various investments that have taken place either via magnetar capital uh so corweave for instance bought into a magnetar capital fund that geared up raised more money and then bought uh bought compute from corweave magnetar also organized all the um they by the way invented the co so they got some history um they also um kind of organized the bond like extraordinary kind of yields um that Koree has used to to to speed up. Nvidia kind of basically kind of um invested a billion uh on the eve of the IPO to help get the thing done. In return, Corewave is but bought 7.5 billion of GPUs from Nvidia who's then promised 1.3 billion of takeoff for compute uh over the next four years. And you look at all that and go that this is this is starting to raise red flags. And then the the final kind of red flag is is what's happening to insider selling. Um and you know Bloomberg has a useful kind of tech GPTR where you can look at insider buying and selling of a stock. Uh those are in well it's all one way for Nvidia all year. You can get a similar picture out of Coreweave once they were able to sell um kind of after a certain gap after the after the IPO. you kind of go, well, if this was real, if this was actually going to be a a killer app and and incredibly useful and transformative and so on, you'd you'd want to stay in there. You wouldn't want to be selling handover fist. And and for me that's a final set of red flags that sort of compounds the issue not just of the price and whether these guys can make any money and so far they can't but then also about the questions the deeper questions about whether this is ever going to be more than narrowly useful. >> Okay. And um we're getting late so I'll just ask you a couple quick ones here. What would it what would be the signs that you're wrong? like what would you worry about if you saw XYZ happen that would make you reconsider your thesis? >> Um yeah, and there's there's two or three ways I could be wrong. Um and and the most fundamental way that I could be wrong is if uh the new pivots that the companies are making towards um trying to find different overlays that that kind of uh work as an intelligence. Uh if one of those actually works then then I'm wrong. I can't see it. I don't see how they do that. I think that the work required to create a system that would actually be able to use world models effectively is a long long way away. Um and it's as much an issue of understanding as it is of brute force and investment. Um the second way is that someone manages to work out um some kind of useful way of using it that I haven't thought of before. Um, and and probably a bit more than just a sort of fad game. You you'd need to be able to come up with an app that's wildly popular and profitable and doesn't run into scaling issues. So, if one of those appears, uh, then then I'm going to be wrong because then you're going to start needing to use the compute again to a much greater degree. Uh, we haven't seen it. Um, I think one of the big problems with the AI ecosystem is everyone's marking their own homework. You don't, you know, you none of the stuff whenever they claim successes and things, none, none of it's been something in the public domain that you can just go and test for yourself because all of these things end up running in to some kind of problem. Um, and then I guess the the most likely way that I'm wrong is that this just carries on longer than than I would expect and it's sort of starting to do that already and and and that's difficult because because that's the that's the bit and you know I I can point to there being less VC funding available uh and I can point to there being you know having been being a very narrow group of investors uh and and It looks like, you know, even sort of country investors are struggling. So, for instance, you might expect the Saudi Arabian sovereign wealth fund to buy in, but Saudi Arabia has had to cut social benefits because they're not making enough of return on their oil sales. So, they're not going to have an infinite amount of money to spend. Now, Syoshi, who put in a very large commitment kind of late last year, early this year, has been struggling to fund it. he's had to borrow against his his shares in order to do the first trunk. He's got another two/3s to go. It's not obvious how he does it. And now we're sort of in in this sort of environment where we're having all these kind of cross deals where Open AI has taken over from Microsoft offering sorry um Nvidia's taken over from Microsoft offering Open AI compute for equity and and and all of those things. And he's going, "This feels very last gasp." But that doesn't mean it doesn't go on for a bit longer. It just everything I've seen is telling me that there are so many red flags. I don't want to invest in it. But you have to be if you're a much shorter term trader. If you're a family office, you can go fine, I'm going to walk away. I'm going to invest in resources in India because that's the thing for the next 10 years. Um, but if you're in a three-month kind of investment pod, um, you you've got to be much more careful about about trading it. And and I would have all sorts of stops and and technical kind of trading systems in place to make sure I didn't lose too much money while being skeptical of this thing. >> Got it. Well, listen, Julian, I could do this all day, but unfortunately, we're running out of time. So, we're going to finish up here with our desert island trader edition. Um, as people know, this is something if you were put on a desert island, um, much like the BBC show, we're going to make you pick three different albums or bands and then the a a traitor through history that you would most likely want to share that time with. So, let's start with your three albums or bands. What do you got for us? >> Okay. Well, um, so so I'm I'm I'm into music. When I was 17, I traveled around the States with a friend kind of searching for the blues. We we spent the whole of the kind of um Austin Blues Festival at S at at Anton's Blues Bar. Um and I saw Hubert Sumlin, who's the guitar player for Howling Wolf. >> Okay. >> Um kind of uh play live there. Um so so I'd probably have to pick uh one of the Howling Wolf albums because of that. Okay. Um, I guess and now now my other two choices probably going to be a bit sort of, you know, defensive in that I love I love Kind of Blue, the the the Miles Davis album. I think it reminds me of of kind of reading books on Sunday afternoons and I would def I would definitely have that there. And um >> and then Blonde on Blonde by Bob Dylan. I think >> great albums. >> Electric. Yeah, just extraordinary. Um Yeah. And so so those are the three albums. And then and then as a trader it would have to be Jesse Livermore just because >> um reminiscence of a stock operator was was kind of the most influential book on in in my kind of trading lifetime. And and I think the the bit about it I like the most is the chapter about this guy. He's he's sort of moved from the bucket shops which I guess the equivalent of of kind of spread betting trading. >> Yeah. >> And into a sort of a more established office. And there's this old guy called Mr. Partridge who has his kind of chin sort of stuck on his chest. Um and and Livermore's trying to work out why he's not making money like he used to make in the the shops and and you know because he's always missing the trade somehow and then he's watching this guy and and this guy's um you know people coming up to you he'll have a big position in whatever it is uh US steel someone say I've just heard this about the combine and they're going to be selling some and and Partridge would say yes thank you thank you very much it's very good information I'm sure you're right sure you're right and then not change his position stay long and and and the reason he'd do that is because he was looking at the longer term trend and and it was Livermore learned from that that it's the long trend that makes you the big money and and that's why I'd have him on my island. >> Oh, that's awesome. That's a terrific uh answer. All right, so let's learn a little bit more about Macroscope. Tell us about your shop, who it um appeals to, like uh how they can learn more information. Give us the whole spiel. Yeah, I sure think. So, um, so the three, um, kind of writing partners, um, kind of James Ferguson, Andy Lease, and myself. Uh, we also have, um, kind of a sales team, Michael Wilson, and Victoria Wilson, and so on. Um, but the three kind of um, analysts, if you like, writing partners, um, we've all had kind of 20 years plus of experience kind of on the buy side and the sell side. Um, and we were all ultimately too grumpy to work in big businesses. Um, and and even even too grumpy to work uh in the same office. We all work in separate offices. um because because we just have this natural tendency to to be independent and and not to want to be constrained and and so I I reckon if I'd stayed at my last job, I've been at macro strategy for about a decade, I'd probably been sacked about 15 times by now because I'd have I'd have just I'd have just ended up with a different view than the China economist and they wouldn't have wanted me to publish on that and so on. And so basically what we do with with that experience we we kind of do bottomup work. We use public domain work to try and identify how we think the world's working. And when that's different from how the market thinks it's working, then it starts getting interesting. Um, and especially when we, even though we allow each other to be independent, you know, I'm independent of Andy and James, when we happen to end up on the same page, but having started from different premises, then the message is even more powerful. Um, and so that I think is something that just has to be lacking um, in investment banks because you have to tow the line in anything that's not your direct kind of view. Um and um so what we do is we we kind of write a thematic daily. Andy writes that. Uh we each publish a couple of pieces uh per month like independent kind of thought pieces per month. Um and myself and my colleague James Ferguson do kind of update videos on the back of that. Um and then we um yeah and and and that's the kind of primary product that that we put out. Obviously we're free for meetings etc. as well. Um, and I think it's kind of the fact that we're independent, uh, sometimes contrarian, but not always. Um, it's when when you have that position that the potential big trades, the sort of 10-year trades, um, they can come to the four. And I think I think we're just about to enter one. I think we're in an investment regime shift where we're going to move away from kind of the disinflationary boom of of tech and the US dollar tending to win and we're going to move to much more of a reflationary and inflationary environment over the next 10 years where the big winners are going to be India, Vietnam, resources um uh um and even bluecollar workers within the states. Um, and I think that kind of those kinds of conclusions aren't really possible if you're working in your silo in an investment bank. >> All right. So, where can they find more about it? And um, is it um geared towards family offices, institutions, or is there a retail component? >> Um, there's not a real retail component. Um but what what we and we basically have about 220 clients and and there's a mixture and most is most is I'd say 80% is hedge fund institutional um there's about 10% family offices 5% ultra high netw worth individuals and 5% like central banks and company corporates um and yeah if if you uh look us up on uh macrostrategy.co.uk UK uh or send either myself an email of Julian julen macrostrategy.co.uk or our head salesperson Michael Wilson at michael@macrostrategy.co.uk. Um then you can get in touch and we'll set you up with a free trial and a look at our stuff to see if it suits you. >> Right. So go to macrostrategy.co.uk to learn more about the product. Julian, thank you so much for being on the show. I really appreciate It's a pleasure having you on. We'll have to get you back on to talk about India and the the next big trade. >> Yeah, definitely. >> Thank you very much. >> All right. Cheers. Bye then. >> All right, Patrick, talking charts. And just for a reminder, we're doing this Thursday afternoon. >> Yes. >> So, it's uh it won't be the typical Friday afternoon when we do this. We're a whole day early. So, if uh I don't know if Friday ends up being another ripper and we end up looking like idiots or if we get the the Black Friday or whatever, just be aware. >> Oh, listen. It's Thursday when we're recording this and we can only talk about the charts from what we see. So uh you know what um first of all uh you know when we were doing this session when I was uh a few weeks ago we were talking about how this market was at ridiculously elevated levels but it was in a bull trend and it was uh you know completely resilient. Nothing could take this market down. No government shut down no nothing. But it uh Trump did it and Trump uh started well obviously China started it and Trump reacted with the rare earth's uh announcement and the market goes out and has a peak to trough uh drop of 4%. Now that wasn't what was on the end of the day but it was the the beginning drop on the downside. Since then we've bounced. There's a couple dynamics going on. First of all, it's OPEX week. Uh, and we're still stuck in the 6,700 strike zone. There's a lot of, uh, gamma pinning that can still go on, but a lot of that starts to roll off tomorrow. Um, and so the fact that the market bounced a bit here, uh, into this opex is very much in line with that kind of a a scenario where the the options flow could have driven this. Uh but we're now literally going to go into next week without that being a factor. And so one of the things that we want to talk about is did something change LA uh uh when when this drop happened here on Friday like uh it was something different. Well, at minimum what I feel like we've done, Kev, is we've transitioned into uh a distribution cycle. And the way I frame it at at Big Picture Trading is we look at markets in accumulation cycles, bull cycles, distribution cycles, and then bare cycles. So there's actually four phases, not just bullish and bearish, but there's uh there's transitory periods where they're transitioning from bull and bare. And uh an imperfect example of that is topping formations and technical analysis where if we look here back in November through February, the market stopped making higher highs. traded si sideways. All the rallies kept sorry all the the sell-offs immediately rallied but it was always met with distribution and the market stopped making any progress higher. And what I feel like uh we have at minimum uh seen here is the beginning of this kind of period where the stock market this rise of five months in a row just relentless rises. It's transitioning now to uh a period where we're going to see market topping formations uh develop that will inevitably lead to a deeper correction uh on that. What's what's your thoughts? Well, I can't disagree with any of that except for the fact that you blamed it on Trump. And I have a little bit of push back on that because although I agree he was the catalyst, I would argue that the extended overbought nature of the markets where the CTAs were all like 100th percentile longs. The fall control had gotten to a point where it was implicit in the pricing was a continued low low V environment. um the massive amounts of enthusiasm out there in terms of folks just biting at the you know the bit to to own stocks. This uh belief that it was impregnable and there was no way that we were going to get a decline. It was just waiting for something and >> but but it wasn't the government shutdown like there was there were headlines that could have been the trigger. I think though that the government shutdown I think the market has realistically understood that that doesn't really mean anything. Like it's the people asked me about the government shutdown a couple weeks ago and I said it doesn't matter and I and I don't believe it does. Maybe at some point it'll matter but it used to be a bigger deal. They used to worry about like them not um being able to to go over the debt ceiling limit and for them to do a technical default and there was all sorts of other things. I I think the markets just completely understands that this is just, you know, Washington politics and they ignore it and they just kind of dismiss it. But I I I agree that although the Trump news people I'm I'm sure people listening will say, "No, no, the Trump news is a big deal." And I and I and I get it. It is a big deal. But there was going to be something coming along that the market was going to use as an excuse and it happened to be that in another, you know, two months ago that Trump news might have just got shrugged off because it wouldn't have been so overbought and it wouldn't have been a situation where the vault control folks would have ended up being selling where the CTAs weren't 100% long where every >> But Kev, I personally think that we haven't seen a CTA trigger yet. That sell-off there was obviously a reactionary one that uh was uh where a lot of kind of um traders with incredibly tight stops that were looking for a pivot were leaned into. But usually CTAs need more than a 3 or 4% market drop. And even then that 3 to 4% drop would have to actually sustain. I mean, we basically are only one 2% off of the highs at this moment. um in order for us to see true uh kind of um downside volatility where the CTAs are triggered at minimum we should be breaking down uh greater than even 5% where we start doing real technical damage on the charts where all of their uh quant models and other things that they're using will actually create the cell cycle that that sells. So, I actually don't think we've even seen the quants make any meaning or certain quants the CTAs make any meaningful pivot yet. >> Well, maybe you're right on the CTAs, but like the V control, if you pull up the S&P 500 V control index and you just pick like a random I'll just I got it in front of me. Let's just pick the 10 V the 10 target V um exposure. It was N7 on Friday. It's down to 72. >> Yeah. like it it degrossed um kind of 25% >> right >> in in in 2 days or 3 days. So, it's it's happening like it there was some selling and and and not only that, there was a lot of folks that were long. And then the interesting thing about it, I was talking to a friend about this and we were kind of noticing the fact that over the past couple weeks, we've been getting a lot of flack in terms of people calling me the, you know, grumpy old man yelling at clouds is what I've called myself, like a a go YAC. And I was looking at the chart and and yes, you could make fun of us and say that this is the case and you know you're missing out all the you know the the rally into the end of the year but in the space of like a day everyone that had bought in those previous two weeks was offside. >> Yeah. >> It doesn't take very much. And that's the scary thing about this environment is that we could turn around tomorrow and we could be down another, you know, like let's just say another 3% or something like like what we experienced last time. And then all of a sudden all the buying everything that you've you know all the that's this this buying and this accumulation as you point out has been distribution. It it's been people's the the strong hands selling it to the weaker hands. And I I'm I'm very skeptical that we're going to get this rally that everyone's so convinced of into year end. Like tell me somebody who does like now finally some folks are getting a little more worried about the rally into year end. But let's just take a week ago before Trump said anything. >> Did you know anybody that didn't think we were rallying into your end? >> Yeah. But you you know what's what's interesting about that is is that uh when you put it in the uh frame of uh the pain trade. So the pain trade being how can the market make the most amount of people wrong as possible and use that as your starting base case. The reality is when we were back in the summer in July and August, the one thing and even I was guilty of it where a lot of people the seasonality of the market means September and October are are uh the highly vulnerable periods for the market to turn. I could imagine there's a there was a very large amount of hedging being done into the traditional seasonal period of September, October. Here we are October expiration. So, anyone that was back in August buying an October hedge is literally about to roll off at the time when you need the hedge the most. >> Good point. >> And um >> great point, Patrick. And >> and probably not going to buy it again because they're so convinced we're going rallying into year end >> or or simply that the risk is that risk window is passed. So it wouldn't it be ironic is that uh literally we go past the September October hedging window where everyone's seasonal from a seasonality was hedged up to and suddenly they don't rehedge into November and then they get their feet cut out from underneath them. It would be interesting if uh if all of those uh hedges all rolled off like that, right? >> Yeah, great point. I never thought about that. And so that's why I think it's going to be really interesting to see what happens next week once we're past this monthly opex uh on that uh on that option front. The um the >> by the way before we continue I I have a very smart subscriber client that I have the privilege of talking to. I remember going into September and I was, you know, I've been worrisome slashbearish uh worried slashbearish way too early. Um cuz I've been at least saying that for the past month. And he said to me, it just feels like everyone's worried about the seasonality. And I remember him specifically saying that to me and I'm thinking to myself, ah, you know what, maybe he's right. And I think you are correct that there was too many folks talking about that. Goes to show you, Patrick, if everyone's worried about something that's the time that's the least timely you should be worried about it. >> Absolutely. It's uh you always have to look at where what is the consensus view and recognize that the tail risk is always on the consensus pivoting. Yeah. >> Right. And uh but what's interesting about this price action and the way the market dropped here, let's look at just a chart here. this way this bounced it reminds me a lot of what happened back over here uh in over Christmas this we had that December 18th um market drop day where uh the market basically uh went from peak to to trough about almost a 4% drop about 250 S&P points uh you know Trump was already in office at the time but the market was rallying up until that moment and then what happened in the the few days that followed was basically a rally that got the market almost back up to its highs. And what was interesting here is that the market proceeded to have numerous attempts to rally back up to their highs, creating a very classic topping formation and a distribution cycle along the high. Now, I'm not necessarily trying to forecast that that has to repeat itself, but it is an important lesson that just because we're uh bearish and we feel that the outcome will be to the downside, sometimes the market takes a little bit longer than most people expect to for the downside to really uh crack. And uh and I'm curious whether or not we're going to be able to sustain the first wave down in this market. Like I would it wouldn't it be an interesting scenario if this market came in with like one of these uh less than 10% drops like a 78% drop let's say 6,300 or something like that. The market then proceeds going into late November December for a double top retest and everyone just gives up on this hedging Everyone just like says you know what this hedging doesn't work. I'm spending all this money. the market never really followed through and I never saw my payday and then and then really after putting in some double top some topping thing then we have a Christmas massacre in December or something >> like uh >> you're you're saying Santa's bringing a bad bag of coal >> I I'm not forecasting this I'm just going through the pain trade scenarios of like what could what could basically happen to make the most amount of people get it wrong because a lot of times a market tops. Um, everyone like it's very hard to catch a market top because the anx at the top the anxiety and confusion gets to its maximum. Fake out moves, volatility going the wrong way, uh, quick drop, big buy on dip opportunity, you faded this and that. What happens is you just get so overwhelmed by what's going on that you can't step back and see what's actually about to happen. And most people are never prepared for when the actual turn happens. >> And and Patrick, to that point, it volatility often increases. And you and you alluded to that. You know, you get a 5% down move and then the the you get a 5% rip right back up and everyone thinks, oh, look, it was a bad it was a buy on dip and then that sags and then everyone's still in buy and dip mode and they don't recognize that that the uh nature of the market has turned. I completely agree and and you've really highlighted how the the increase in volatility makes it really tough to you know filter out through like you know to to eliminate the noise and to see what's happening. >> Absolutely. But what we do have is a VIX that's now hovering legitimately above 20. So now uh the V premium of buying insurance is is definitely picked up. So, so now, so now, uh, the downside hedging is getting more expensive. Uh, and, uh, and but this is to your point though, uh, we're going to have heightened volatility. The swings are going to get bigger, and the question becomes is like, will all these vault targeting funds have to start degrossing? Will the CTAs start getting signals? When will the the dealers start having their gamma flip levels uh, really get tested? there's going to be a a number of stress points that will uh will inevitably emerge that uh that we have to be aware of. One thing though, you know, while I think that next week can be quite volatile, I I just think that anyone who's planning their big short on this market really has to step back and reflect that, you know, if you think it's just going to happen in a 30-day window, uh you may be uh the market may be setting up to frustrate the out of you. uh and uh you you have to have the right time horizon to see through what will be inevitably a turn that is very likely going to be a big downside move. Yeah. But uh but it's just going to shock everyone how long it takes to get there. >> I remember our buddy Morris Sachs uh used to talk about that. He says in bare markets, bulls and bears, neither one of them make money. And the reason he would talk about that is because you would get these vicious rallies. people would get, you know, bearish and they get short and then the thing would rally up on them and stuff like and it's just it's a difficult game like no doubt about it. >> So couple of things I want to highlight about uh we've discussed last time about the markets. First of all, the breadth of the market continues to deteriorate. During that drop on Friday, we got to 35% of stocks uh above their 50-day moving average. I mean 65%. Basically two out of three stocks in the S&P were actually below their longer term trends. Uh they were downtrending basically. Now it's back to 47%. But here we are a market a stone throw away from its all-time high and you have more stocks in a decline than are rising. The fact that we have seen this type of deterioration in breath happening throughout the last couple months um is very typical along market tops which is is that there's a thinner and thinner leadership of stocks that are driving the momentum and the broader base of stocks stop participating and taking flows and um and this continues to be evident and this is like that that kind of market rotting is there like while we did have a little December event in breath You can observe that the breath kept deteriorating over and over again until we got the liberation day drop which got us down to like only 5% of stocks above their uh 50-day moving averages. Point though is is that the breath kept getting worse and worse for months in that deterioration and this is clearly evident here. Uh what what is interesting as well is is that like for instance today we got the news on the regional banks the Xeon and uh things like this but the regional banks were pretty much in a nice clean bull trend higher highs higher lows uh and then they came for a direct retest of their previous highs. you know, I this pretty much qualifies like a double top to me. And uh and we now have a legitimate reversal breakdown where it's no longer even holding a Fibonacci zone. And so what's interesting though is a lot of the major banks um actually had earnings beats, right? And uh let's I'm just going to use Bank of America as an example, but like we've had a couple of these banks act the big reg uh not the regionals, but the the large uh money center banks had these nice beats, but the regional banks are under stress and uh and I'm trying to determine whether or not these the way these financials are behaving is a tell in any way. What's your thoughts there? Anything? >> I think that they're behaving like crap. like pull up the KRE. You look at it, it's breaking below. >> Um, it's moving average. It's looking like it broke support. It looks terrible. And a lot of it has to do with this first brands thing. And a lot of smart folks are are messaging me saying where there's smoke there's fire and this is just the start of a lot of credit issues. I know as well um in terms of plumbing, we're seeing some signs of stress in the repo market for the SRF getting taken down. There's all sorts of worrisome signs out there, Patrick. Like it's kind of it's it's >> it's kind of amazing at how complacent everyone is about about the risks here because to me it feels like >> when you're making money, it's easy to be complacent. This is the nature of of how uh investing works. It's like when you're making the money, you generally are not worried. It's when you're losing the money, then you ask like what the going on, right? >> Yeah. I guess >> it seems to me that when an assets become more expensive, you should be more worried about it going down, not less. And when asset becomes, you know, cheaper, you should be more, you know, cons more willing to buy it. Like you know what's the line from Warren Buffett? It's like be greedy when others are fearful and be fearful when others are greedy. >> Sure seems like to me that this is the time to be fearful, not the other way around. So just a a quick thing and we're I don't want to pivot to bonds just yet, but one of the interesting things that we've talked about uh in the past was the correlation of junk bonds to um uh to equities in the sense that you know there's been a lot of 5% market corrections where junk bonds haven't reacted but almost uh all of the 10 plus% market corrections that the like deeper more pronounced corrections almost certainly We have C uh junk bond credit also declining. It was interesting is what kind of a a solid bull run that junk had for the last four months, right? Like it it was just an epic run and that Friday was the first time we've seen uh a junk bond start to react. Now I this is not a bearish chart like KRE yet, but this is the thing to watch in the next couple weeks. like are we going to see that junk bonds are going to start to get heavy here and start to roll uh because I still think that if this is going to be destined to be a greater than 10% correction that the junk bonds will give us a tell here uh on that do you agree with this? Oh, Patrick, couldn't agree more. And uh in my little recap that I send out, I I highlighted the fact that recently Howard Marx, a guy who knows a thing or two about investing, um pulled the trigger on his put provision to Brookfield on his firm, meaning that they sold the rest of their firm to Brookfield. And the thing about it that was kind of really interesting was it was only a year or two ago that Brookfield thought that the guys from Oak Tree would never do that, that they would always own their piece. And here they are, you know, sticking uh Brookfield with their with their credit um kind of management firm. And it just conjured up the same thoughts about or the reminding me about Samzel selling equity office properties at the all-time high in REITs. And I went and pulled up the charts. I was actually shocked. Do you know that he the equity office properties came off the board the day that the the Dow Jones REITs index topped and it was straight down from there. And so when I look at this, it just reminds me of it like smart grizzly, you know, um, wise old guys like Howard Marks are selling. Meanwhile, everyone else is getting all excited about it and saying what a great investment opportunity is and he's selling. And one of the things that you should always remember is watch what they do, not what they say. And when I see things like Howard Mark selling, it makes me think that this bond this credit bond market might be in trouble. >> All right, you heard it from Kevin. I love it. I never knew that. And I I'm glad you brought it up. I think that a lot of our listeners will appreciate that little insight. All right. So, listen, we got to move on because uh that we it's clear what we're thinking of the on the equity markets, but we got to talk about some of these crazy bubbles going on um in uh in the markets. First of all, actually, before we get to the precious metals, I wanted to quickly talk about uh the other thing that I think is the most important thing to talk about is the dollar. Uh but you're shocked, right? No. But >> for those who don't know, Patrick Eustus put that always in the top three things. Um, but he doesn't talk about it as much anymore. You you've you've kind of >> gone off of that. >> Sure. But I actually think if the US dollar is going to rip here, it matters. But what I was actually particularly wanting to bring up was the fact that I'm shocked actually where we are on implied on uh the currency options like the euro fu uh on the euro futures going out five months. We were down to like 6.6 implied. Um, and you know during higher V periods like 7 to 9% are incredibly common. And so like we're trading on uh some what I think is a a relatively cheap V uh on um on the currencies and the one thing I'm thinking I don't know you can check on your Bloomberg whether you >> I'm pulling it up right now. I think you're right. Are you pointing you out the money? >> Yeah. But are you >> six month I but six I did a six-month Euro dollar at the money and it's 6.8% and it's at the low end of the range. >> Yeah. Low end of the range. Exactly. And so one of the things I'm debating here is whether or not this V is right priced. Like when you put all these event risks and all of this different things that can happen in the markets um that the the chances of outsiz currency moves uh increases and yet they're pretty much giving you the convexity here pretty cheap. Uh and I'm curious you know like whether or like can the euro in five months move more than two cents on the downside. I mean it it even a 50% retracement here I'll put up the euro. Even a a 50% retracement in the euro is down to 112. Uh you know that's a that's a 500 pip move on the downside. To me I just feel like uh that these options you don't even have to make a crazy call on the currencies and take huge delta 1 risk on this. these options are actually think are very reasonably priced for the conditions of uh that we're going into. You know, I think that if things really got spicy, there's no reason why the dollar can't have, you know, a 500 to a,000 pip rip uh just based on flows. Yeah, Patrick, I've been arguing this for a while and it's just so far it hasn't worked. But I I continue to believe that the the foreign exchange currency va market is one of the most underappreciated asset classes out there and that at some point the ne this this trade war will turn into a currency war and when that happens you're going to want to own these vs and I think you go out as long as you can you own the far-dated stuff so you don't get you know the theta monster burning you you know like scratching at your door all the day and you just own this thing cuz I'm with you. I I suspect that we're going to get some wild crazy moves and I don't know which ones in which way, but I just I suspect that there's going to be more of all in the dollar than we think. >> Yeah. Yeah. All right. So, let's talk about what I what I'm sure everyone wants to talk about, and that's what the freak is going on with gold. Uh like what a like it's parabolic. I know you've been calling for a decade this $1,000 update. Uh, well, you know, >> you haven't even known me for decades. So, don't. So, don't >> dude, since I remember you, like >> I didn't always say that. That was my thing. We' have a $1,000 up week. I don't think we're going to get it. I because I feel like this is it. And I don't feel >> No, I mean, look, if we got a $1,000 up two months. Yeah. >> So, okay. Like, >> I was a little hyperbolic. I was trying out for the Macro Voices show at the time. >> Yeah, exactly. But but we are we are now actually in the accelerated parabolic phase of uh of this advance. And the one thing that you know we've talked about parabolic advances numerous times on the show over the last five plus years. And um and th this one is just not going to be any different. It's not that gold isn't going to eventually be 10,000 or 15,000 over a long enough time period as a huge, you know, fiat currency debasement is happening over the long term. But when things accelerate in these parabolic fashions, it attracts substantial amounts of leverage uh in futures markets in option gamma and all these different things that actually add like fuel accelerating the the underpinning bullish things like central bank buying. And uh and inevitably uh on the other side of this they're often pullbacks are as violent as the rises that that preceded them. And uh and so this gold rise now that we're entering parabolic, who knows? We could see 4500, 4,800. We could see uh we could see 56,000. I'm not necessarily thinking that crazy, but but like with these parabolic rises, they just keep going and shock everyone how far they go and how fast they get there. But, you know, like I I want to go back and look at the rise of gold in that kind of phase leading up into 2011 where it was a pretty bullish kind of parabolic rise period on gold. And when gold made that rise from like 21 to 2600 in that final stage of that uh thing, the subsequent market drop in the next month had it give back the entire initial rally uh or the last phase of the rally. Point being is that the drop on the downside was arguably even more violent than the final rise. uh and these things can turn and I just don't think gold will be any different and it won't change any of the bullish reasons to be bullish in the long term. Like it's we're about to see some huge gold ball and uh I think that um when this blows off, we could easily have even uh 500 to even 700 on the downside just from uh flow dynamics. Yeah. Um, I don't really have anything to add. At this point, every bin's become a gold expert. Um, we're at this stage where it's become manic. Will it go keep going? Who knows? I don't know. I feel like for me, the value ad is talking about it before these things happen, not during them happening. So, you know, for the gold bulls, enjoy the ride. Let's hope it keeps going. And I'm not disagreeing that the it's getting more and more dangerous. And the only thing I would say is if you're looking at your position and you look at it on a V uh basis, you should probably be making sure it's not growing too much cuz not only is it going up in terms of your notional amount, it's also going up in terms of the V. So that would be the only thing I would tell people is, you know, great. I hope everyone's long and I hope everyone uh, you know, practices safe gold investing. you know, the way that the way that we're approaching it at Big Picture Trading is we're um uh we're converting. So, uh a lot of our LEAP positioning. So, we we were up like 300 400% on a bunch of these leaps that were uh going on the upside. And the problem is is they these things are all approaching delta 1. Yeah. You know, like they're literally synthetic equities. And the problem is is that you're therefore taking you lost all the convexity of the LEAP and and you're now basically taking on almost all that long exposure direct if there's downside volatility. And so what we've been doing is actually closing the LEAP positions and replacing them with bull call spreads going only even one month out. And so what happens is you can reconstruct the uh the payoff structure where you know you could still get 2:1 and 3:1 payoffs on um on bull call spreads continue to participate on the upside but having locked the vast majority of your gains in and um and that's sort of the way I'm looking at it. You know you don't want to short this bubble. uh but you you need to have some sort of uh of um way to reset the asymmetry of the trade to be able to keep participating on the upside. >> Right. And one of the things that um I do that uh in terms of trying to keep your bend with like the convexity that you talk about >> is like let's just imagine you were buying this back in November of 2024 when gold was 2600. Let's just imagine you bought a $3,000 call, right? and you were buying two-year $3,000 calls. And the reality is that now, as you rightfully point out, Patrick, that $3,000 call is almost a delta of one, meaning it's just trading one for one with the gold price. So, you're you're stuck with this in thing where you've lost all the bend. It's just become a straight one. So, one of the things that I do at times when I'm thinking about taking a long-term position is that back when I was buying the the $3,000 call, I might buy some 3000s, some 3200s, some 3500s, 3750s, 4,000s and that way if I'm re like and they get progressively cheaper and cheaper and cheaper and that way if I if it keeps going I don't lose the bend if you understand. >> Yeah. >> Because you still have some new ones. So that's just something to think about. >> Do you do you do you trim do you trim off the deep in the monies and just keep riding like you almost reset your delta dollars in terms of what your exposure is? >> So yeah, and sometimes that's what I'll do. And so I'll buy Yeah, that's one of the things I like to do. >> I thought someone I'm not sure what a Christmas tree is. I don't know. >> No, but that's the upside down. That's I think the Christmas tree is like the you know I'm not going to pretend like I uh I'm like you're you're basically staggering on the lower strikes by getting larger and larger on the farther and farther out of the money strikes. >> But um I guess anyways we got to come up we got to come up with a name for this. But that's that's what I like doing because I think that losing your bend is actually more of a problem and it's a good problem to have but it is a difficult problem. And that's in essence what's happened to you right now is you've lost your bend. And not only that, because it's gone and shot up here so fast, fall is more expensive now to pay up for. >> So that's why if you buy it there. Okay. So next crude oil quite bearish. >> Oh, I'm loving this bearishness. >> I'm loving it. >> Well, listen, you could you could love it. I know you're like the emperor in in Star Wars. You just like let your hate flow through me like whatever you do. But um for me uh what I only care about is when do we transition from distribution sorry bearishness to accumulation like where where is that kind of basing thing? Right now this is still uh unfortunately in outright uh bare mode which is like lower lows persistently in in sell mode. And um and I'm curious like whether or not this is simply going to retest previous lows. Like where where are we going to bang out the bottom? Like I get the long-term bull thesis. Uh I know so many of our guests come on and and share this idea. The key thing here is just about uh when does it transition into that next bull phase? Right now, this unfortunately uh doesn't have any major support lines anywhere here. Uh we could we could easily still shave off, you know, 10% on the downside on this. Now, will it be a epic buying opportunity? I'd like to think yes. Uh but uh you right now on the short term, it's there. It's hard to make any bull argument other than everyone's bearish. >> Yeah. Well, you know me, that's often enough. >> That's often enough. >> But but I'm with you. Listen, and I'm I'm I'm gonna I'm picking away. I'm starting to think about things and how to position for this. And what I'll probably do is start picking away at it. I won't buy like my full position. And then I'll when you tell me that things have gone to the positive from the tech side, >> technical analysis side, then I'll jump on and do it in a big way. >> I like it. I like it. the um uh I'm I'm just trying to cherrypick some of the commodities we're talking about. Uh will natural gas here hold the support line? This is obviously in line with the oil thing. Um look, there's always room here. I'm using the UNL which is the strip just so that we avoid having to play the term structure game. um you know will is there still possibility that a a head of an inverted head and shoulder happens where this thing still has one more round of weakness for a couple weeks. Uh it's plausible. I still think the bottom and that gas is going to come in here. Uh but the fact that we gave back the entire bull breakout in one fluid motion like this is a short-term setback. It means that the starting gun did not fire and the race did not begin and we're still in some sort of basing accumulation phase of this security. And so we're going to be watching closely uh what what this does here. But on this sets it back for the short term. Maybe it's going to have to take November or December to start turning this chart back up. The other thing, Patrick, that I think is interesting is that folks will say to me, "Well, Kev, you're bearish on the economy, so why are you bullish oil?" And um they're kind of correct if you think about it from like an economic point of view. If I think the economy is going to be weaker, why why would I think that oil is going to go up? And and there's obviously a lot more going on, but I do believe that there's a certain amount of correlation in games that are played with the people's books. And so if we got a situation where we did get a major correction in the stock market and the and the economy turned, I suspect that investors would sell all the stuff they're long and go out and buy all the stuff that they're underweight. And as strange as it sounds, I think economic weakness could actually see the nat gas and the oil market going up. >> But do you think nobody is that thesis? Is that thesis on um the commodity or on the equity? >> I think it's actually both. I don't think it's I like I think it works on both. And and there's lots of times if you go look at NA gas, it seems to be negatively correlated to the stock market. And I don't I don't understand why. I don't really have a ton of theories, but uh you'll see times when the stock market gets into real troubles that that cash goes bid for the day. Um, it could be one of these things that it's a kind of a carry trade. The hedge funds end up being short because it's a little bit like shorting VIX futures. It just pays over time. >> I'm not sure, but I definitely think that there's there's something there and you should keep your eyes on that. >> There you go. All right. But what one thing I wanted to touch on here, platinum and platium, both of them have literally finished measured moves and these parabolic rises. What's interesting is that whe whether it's silver, platinum, platium, or gold, they actually are all doing the exact same thing at the same time, which to me means that I think that they're all going to correlate. When we see uh a short-term blowoff top on this, then they're all going to move together like that. I don't think that the any of these are marching to the beat of their own drum anymore. >> Okay. >> The uh >> I can't wait for you to get to bonds. >> Well, okay. Okay, let's get to bonds. Actually, by the way, before we do, let's talk one more thing. Wheat. >> Oh my god. What? >> I guess that's cuz they can't sell anything, right? Like American farmers are basically getting shut out around the world and we're looking at wheat priced in like for delivery in the US. >> And so it makes sense that none of these are going anywhere. I'd be curious if you went and looked at a more global wheat like if there was is there a Brazil wheat soybeans. A great example of that is let's say looking at copper in the US which is obviously had the tariff pop and drop. Yeah. Versus you look at the um LME gradea copper and in London and I mean we're virtually at this beautiful bull chart on on global copper prices. And you know when you're talking the reflation trade if you're looking at the HG you're not really getting a good pulse on what what global copper is doing. You're getting messed up with all the com >> Yeah. >> Uh that was driven by uh by all the uh >> But it's made even worse because at least copper you can store. >> I think >> I think stores but does it store forever? >> Yeah. Yeah. >> Like gold. >> Yeah. Yeah. >> Okay. >> Yeah. Yeah. >> Some things you can't store forever like that they oxidize or whatever. I'm not sure. But anyways, you can store for a while. >> That's why you can use it for copper pipes and like this. Oh yeah, that's right. Okay. It's got a long long life. um the um uh but anyway, so let's talk about bonds, >> okay? >> And let's talk about >> I have something that I want to I I want to get off my chest eventually. >> The 10-year yield under 4%. >> Okay, it's under 4%. And more importantly, like okay, other than the one brief day where it closed uh uh um below 4% or actually never even closed below 4% back during the liberation day. It just traded below 4% and opened below 4% but it never actually closed there. And so the so if we close here, this is the lowest closing price on the yield in 2025. You have to go back to like September, October of last year to see lower yields. So bonds are clearly uh on the uptick. But what's interesting is that all of this chaos has certainly li uh um stirred up the sturd traders. Uh the uh because you you have the um uh the the 2026 December uh uh silver future right back to its highs. Yeah. You know, and that's uh what we're seeing here is is that literally it doesn't matter whether you're talking 2-year bond, 10-year, or the 30-year, you are seeing money flowing into bonds straight out. Okay. So, Patrick, as you know, for the past couple of months, I've been telling you that I am flat. You know me. >> Don't Don't say it. Don't say it. >> No, no, no. I am not I am not long bond so you don't have to worry yet. Okay. >> But I am just laughing myself silly because >> I remember back in January of 2022 or February, I wrote this piece and it was really, you know, bearish bonds and it was all about the reflation trade and all about what potentially could go wrong. And people told me I was an idiot. I got all this terrible flack and there was just like, "No, you don't understand. Rates always go down." yada yada yada. Okay. Um, fast forward to today and by the way, you know, 2021 and 2022 were the two worst years or 2022 ended up being the worst year in the bond market I think might have been ever in a century or something. >> Yeah. So, it's it's it ended up being those folks were really wrong. I am laughing myself silly because I go and I look on Twitter and someone says you should buy bonds, okay? And the the kind of debasement degens jump all over them and tell them they're going to be so wrong and that there's like we're going to reflate and all this stuff and it just looks like the mirror opposite image of what we experienced in 2022. Back then, everyone was completely convinced that you had to own fixed income and that there was no way that rates were going to stay, you know, go up and all that stuff. And now it's the exact opposite. I saw Bob Elliott get there and he said something and people were literally like laughing at him. And anytime I see a group of traders getting laughed up on on Twitter for their views and and their outside consensus, I know we're getting close. >> Yeah. And I I don't know if I can bring myself to own bonds, Patrick. And maybe I won't do it just to help you out, >> but I I I want to I want to own bonds. >> I I actually want to own bonds. >> Well, listen, just compromise. Just at least throw you throw some TUA in your account or some leverage two years because then you know what? Because then you're playing Fed policy and not uh and not really the long >> There's a lot in there, too, though. So, the one of the other things that I was laughing about was um so I have this this credit trade spread on >> that hasn't worked. And so I'm basically short um the CBOE um HY and IG bond products >> and I'm long the the the fives against it in essence so that I isolate and just end up with just the only thing is credit. And a couple of times, Patrick, I felt like dirty about like having the the the fives on cuz like I'm technically long, although there's no real duration risk. So I was like So I was like, "Oh, maybe I should take it off." cuz like I'm like sitting there and and now I'm looking at the P&L and I'm like, "Oh my god, am I so thankful I didn't screw that up by going and taking that like lifting the that part of the hedge because there's nothing goes more wrong than when you lift a leg trying to fix a problem, a bad trade, right? Like you know, like >> but uh anyways, the long and short of it is Patrick, I'm with you. I I don't know how or why, but I think that the bearishness on bonds is just overwhelming and that we are going to be surprised that they do much better than folks expect. Well, I mean, I just feel that while, listen, while I respect the conversations that you have often with some of our, you know, really smart guests and you guys go through the hypothetical scenarios of when stocks, bonds, and the dollar can all go down at the same time and all these different theories. Those theories some point may come to fruition. And I'm not saying that you can't be right at some point, but there are general intermarket rules and relationships uh that you know take unique circumstances for them to diverge. And typically the flow into uh into bonds as a safe haven during a a rotation when equity when pe when big uh smart money is taking money off the table on equities. this natural flow happens and um and so you know anchoring that off the idea that bonds can go down at the same times as stocks some uh will almost kind of fooar you when with the most simple intermarket relationship that more often not works >> well and I I'll push back a little there because I actually think that when we do get the next one they're going to run ahead and they're going to go and price in way more Fed cuts and the long end is actually going to underperform >> and I do believe that. So ultimately I'm not a fan of going and buying 30s because I suspect the rates are going down. I'd much rather play for the fact that we have 100 basis points of tightening priced into the next year and that when things really do finally roll over, we're going to quickly go to 200 basis points or something like that, right? which is something you've talked about and you were correct to highlight as that eventually what happens is the market runs ahead and then the Fed has to catch up >> and that'll be what happens and I still do agree that the long end isn't the safety valve that it used to be. So I I I'm not going to I'm not going to go all in and say you got to buy 30s. Having said that, one of the things that you know Bruce Cner says what I look for is a market that is not confirming consensus. And to me, consensus right now is that the debasement trade is going to make it so that you are going to lose money in treasuries. And Patrick, you pointed it out. It's behaving well. It's not behaving like everyone that's making fun of Bob Elliott on Twitter is act like is is is and would guess. Yeah, fair enough. Um, you know, uh, the last thing I want to, uh, pivot to here is looking at global equity. And what I, uh, particularly want to note is at least as of this moment, this natural correlation of where kind of, uh, all global equity markets tend to correlate with the US market in some cases. It doesn't seem to be evident yet. like the NIK for instance had a quick drop uh on on Tuesday almost immediately recovers. It still feels like the Nikkay is being well accumulated bought on dip and still working. You know, you look at the Euro stock that was a clean 50% retrace right down to its 50-day moving average. uh got immediately bought on dip and more importantly uh is legitimately holding out of the trade range that's established over the last year. These suddenly we have a bunch of these global equity markets and they still look like these are pretty darn bullish charts that actually can continue higher. And I don't know how to reconcile it because I'm so nervous about what's about to happen in the S&P that I'm finding it hard to get myself to just look at the chart for what it is. The chart looks bullish. Like I, you know, and and I'm s sort of like talking myself out of of like taking positions here. >> Yeah, I listen, I hear you. Go do a spread trade then. >> And and Patrick, you're you're spot on correct. And I wrote that piece where I highlighted that the MSCI World XUS had three times the sharp of the S&P 500 over the past year. And the reality is that the rest of the world looks a lot better than the US does. And to me, what's happening is we're getting a global reflation. Um, all the world is is is doing fiscal stimulus because Trump has in essence necessitated that they do it. And I and I think that's a good thing. A lot of these countries were starving their their economies of fiscal stimulus. And so what we had is this global fiscal stimulus and all the whole world is reflating and US stocks are going up at the same time and everyone thinks that they're leading and they're not. They're not the leaders and that's >> well the AI ass bubbles is >> sure's not but not the US stock market like you have you have to isolate the the one kind of little pocket but but like listen I agree with you and I want to highlight like okay so like you can just use the EM as an example sorry uh which has got um China in the basket but like look at the way that's ripped this year like it's just >> did that just go right to to the to the tick almost of that moving average. >> Yeah. >> Wow. >> Right. And just tested and when but China still looks like look how fast China recovered off that after the dip. Look at the way pres so India uh just consolidated for a couple months and just doing a breakout candle like it's ready to begin its next leg higher. You take uh the Ebo Vita in Brazil. I mean it hasn't >> you did that really well. >> Uh it Thank you sir. uh but uh it came right to this moving averages. Is this going to be a buy on dip in in this space? You take uh ch look at the way Mexico comes right to the moving average right to uh just a basic 50% retrace uh breakout candles like this is things ready to go. >> It's almost like technical analysis works. >> Okay. Of course it does. Of course it does. This is why we talk charts, Kevin. But listen, but the point I'm making though is that these global indices are uh or these global different markets are actually way more bullish looking. And even though we're uh I'm so kind of like my anxiety level on that S&P is so high and I'm so bearish there that I'm I'm having a hard time carrying any heavy longs in these other markets. >> Got it. I understand. That's why they're so bullish looking. That's why trading is so tough, man. >> If it was easy, everyone would do it. >> Is there Is there something, Kevin, you ask this to guest all the time. Is there a question or a chart I should have looked at or asked? >> Oh my god, you actually listened to my interviews. >> I am barely. Barely. >> No, I want I want to go back to trading. This is the middle of the day, so let's call it a day. >> All right. Well, listen. You know, the fact that we recorded this on a Thursday guarantees that some show is going to happen tomorrow, right? >> We're going up or down. Up or down 200 S&P points. One of the two. >> Up or down? We just don't know which one. >> We don't know which one. It's going Listen, you got to keep your mind open for up two. Like it could possibly happen. >> All right. Listen, bare market, bull market. We're just happy you guys spend some time with us on this ride. So now, stick around for the after show. Okay. Daniel, by the way, uh Daniel, before we start, uh I see you're in your van. >> Is it down by the river? >> It's down by the ocean. >> No, he probably too. >> You know what? I know the song down. >> It's on a song. It's It's a skit from SNL. Chris Farley. He did the He did He lived in a van down by the river. >> Oh, okay. I I'll take that one. Maybe it was about me. So, you're down by the ocean, not down by the river. So, what's everyone's I'm going to go first. I love the beer. Nine and a half. This is lightweight. It's awesome. It's perfect. This is truly a sessionable beer. I am going to give it literally I'm going to do the amateur score of 5.0. There is nothing special about this beer. This beer is as ordinary down the center. It is totally drinkable and so ordinary. Like I I >> just like me. It's just so ordinary. >> This is the most ordinary beer that I've had. I'm giving God 5.0 right down the center because it >> it just deserves that. >> I was actually going to give it the exact same score, which is really so bad that I'm going to go home and drink wine because I need to get that taste out of my mouth. Oh, okay. Oh my god, that's worse than me. I was trying to be neutral. I was going to be neutral. You're got to give it an under five. >> It was awful. It was not nice at all. No. >> Well, localal. >> I mean like like co light was would be maybe where it something similar, but even cos light's better than that. That's >> No, come on. You You take that back. You take that back. Oh, >> I don't mind the 5-0 call, but calling this worse than Kors Light is too far. >> I used to That's That's a ser That was a serious blow. >> Yeah, that was >> Listen, I just need to do a shout out. Listen, I just need to do a quick shout out. Listen, so one of our Huddle uh listeners, uh Chris, uh he goes under the Twitter handle CCE and um he uh knew I was out in Lisbon. was visiting and so he went out of his way to reach out to me. He he uh is a beekeeper and uh and makes his own honey and he just wanted to drop off some honey for me and I thought that was so awesome that a listener >> so kind is he American Indian where is he from? >> Oh my god, I'm embarrassed to give you all the detail. I don't know all the details. >> Is he North American? >> I think he I think he came from You know what? I'll get you the details. I I should have been prepared. I don't know the answers. But anyway, he was visiting to Lisbon and he just went out his way and I was grateful. So, thank you, Chris. Uh, on that >> he's a beekeeper. Let me just look here. Looking up. That's his handle. CCE. >> Hold on. I I'll find it. Hold on a second. >> You're like the worst shout out guy around. >> Well, I just wanted to give say thank you. And I wasn't prepared to answer. >> If you're going to be Thank you, you got to help the guy get her some booths and like some ratings and like Okay, I'm looking I'm looking I'm looking it up. >> Have you tried the honey, by the way? >> Uh, I didn't want to open it until Oh, it's uh the Twitter handles once CC. >> Okay. >> And does it say where he's from? >> Uh, I'm looking. Hold on. >> I don't look at Patrick's messages just in case. >> Oh. Oh, there you go. I'll take that on the other side. on his shirt. >> He punts options a bit, keeps bees a bit, rides bike a bit, tests cricket van. Oh, no. He's not ex life monkey. No, no, no. He's he's British. >> Yeah. >> Or he's somewhere from Europe. >> Yeah, >> he's an old pit trader. No wonder we I love the guy. >> There you go. >> He's an ex pit trader. He's like he keeps bees. It's awesome. >> Awesome. Anyway, you know what? I'm going to try it out. Thank you, Chris, so much for uh for sending that out. Okay. Um, so now we got to talk South Africa. >> Yes. >> Okay. So, >> so did we I can't remember. Did we You said we're going to go to the sharks. Have we talked sharks already? >> Oh, yeah. We Yeah, we did. We talked to sharks. >> So, we talked sharks. >> But I went to the safari. >> Yeah. So, now we got to go to the next thing that the Playboy did. >> Listen, we we we we did the whole thing like we we went uh uh parasailing. We did uh the um uh drifting race car drifting uh uh like it's some kickass but we got to talk the safari. So, we went all the way to Krueger Park. >> We flew all the way down to the the main park and I my mind was blown. I did not know what to expect, but this was a freaking complete Jurassic Park experience. Uh, and the the one moment that actually was the absolute craziest was um, okay, cuz we got to see everything, lions, everything. We got to see them all in in this great event. But there the one moment was we were driving down this trail and there was a rhino right on the path. >> Yeah. >> And the drivers like all calm. And by the way, okay, I got to deviate for a second. We had a driver with the coolest freaking name ever. >> Okay. What is His name was Panic. >> Come on. >> Literally spelled exactly panic. That like what? Who's mother? What mother? >> No, but was that his actual name? >> That's his actual name. >> Like we we checked his ID. Like we we called him out on it. Like that's >> if he has kids, they're going to be panicking. >> They're going to be panicking. Yeah. >> Okay. So, so Panic, the driver, is driving us down this road uh down a trail and uh we turn and there's a rhino and there's this huge monstrous like one ton monster of a rhino sitting there. And at first everyone, he's calm, everything's fine, you know, we could all take pictures and like this. And then the rhino turns and looks and he goes, "Oh shit." And and he and he and panic panics. Yeah. And he hits reverse and and I'm like like I'm thinking the entire Jurassic Park scene when the the they're being chased by the uh T-Rex. Like I'm like we're about to be charged by a rhino cuz like when Panic is panicking something is going astray. And >> well big shout out to Panic that he kept you guys alive. You think he's a listener? >> No. >> No, he's not. But listen, I do want to say anyone uh that ever wants one of the coolest life experiences like going uh to the savannah and being out there in that nature and seeing the the big five. Like what's crazy is is that the lions are so um climatized to the vehicles that they were walking within a meter of the car and they did not give a that you were there. Like and and the coolest part was we saw these three lionesses actually hunting. Uh, and while we weren't there for when they made a kill, the uh our other car full of other members uh thing actually got to see um them actually uh rip apart the prey and all run off with chunks of and like that. Like like they got to see like the real safari ass Okay, so listen, >> become a big picture member so that you can go on safari and see the poor gazelle get ripped to be pieces and run off like >> absolutely. >> Okay. Well, on that note, we won't beat that. So, everyone uh safe uh trading. Thanks for tuning in and we'll see you in a couple weeks. >> Take care, everyone.