Transcription
Venture capital as we know it could die. One of the major tech companies could take a mortal wound. The amount of debt they've taken on, the amount of assets they've taken on could actually be lethal. It will kill trust in the tech industry, which I actually think is a good thing. The tech industry needs a punishment. The people at the top need to realize that the path we are on is broken. Where I think most economic analyses fall down. It's this thing where they always assume that things will go good, that all this money can't be wrong. What if everyone is wrong? Because if everybody is wrong, I think we're on the path to one of the first real tech Great Depressions.
Hi, I'm Isaac and joining me on the tech report as always on a Friday is Ed Zitran. Welcome back.
>> What's up?
>> So, the five largest US hyperscalers are now spending more on capex than their combined operating cash flow according to Dodge Bank analysis. In other words, some of the largest and most profitable companies on the planet have become net cash borrowers to fund AI spending and they're only expected to spend more in the coming years. Meanwhile, developing and selling AI has never been profitable and this fact is kind of becoming a lot harder to ignore. Do you think there is an appetite out there to fund the AI buildout through growing and and increasing debt?
>> So, there is, but it's waning. The Amazon bond sale this week. Usually bond sales, pretty much across the board, kind of like IPOs, they get way oversubscribed. Like people make way more orders than they actually will think because there's only so much interest. With the Amazon bond sale, the $25 billion, which went out this week, they only got 1.6 times the demand, which is not great. The thing about the hyperscale capex situation that's really insane though is the fact that they are very likely spending more for the same because the cost of memory is going up aggressively. I saw a report, I forget from exactly where, that they expect the high bandwidth memory, so the memory on the GPUs, to go up by 90% year-over-year by 2027. That is an astonishing amount. The cost of RAM across the board is increasing. The cost of storage, these are all the things you need. So I think a single GB300, so an MVL 73, the 72 GPU racks, a single one of them has 17 terabytes of high bandwidth memory, which is, I mean, uh, it was hundreds over $300,000 worth, I think. And that's only increasing because of the RAM cartel of Marvel. Sorry, Mar, poor Marvel, Micron, SK Hynix, and Samsung. There is a, there's a tech company called Marvel. I'm not thinking about comics, but these companies, the RAM companies have limited space in their fabs that they're now dedicating a lot to high bandwidth memory, but because they are the only ones that really make it, they can set the price to whatever they want. So, yeah, not only are they becoming net cash borrowers, but they are committing themselves to a strategy where the cost of what they're doing is linearly increas- well, sorry, linearly maybe isn't appropriate, increasing based on the whims of three companies. And this is incredibly dangerous because, like you said, we have had no proof that this is making any profit. In fact, even if you remove the capex, why have they not broken out their AI revenues? Why do we still talk in terms of run rate? And they do that because the, well, the numbers are probably pretty bad. Deutsche Bank's Jim Reed also pointed out that as these companies are becoming net borrowers, the, there is a potential for inflation to rise, which would make this massive debt completely just absolutely make no sense.
How close are we to seeing a sort of a macroeconomic factor like that tipping the scale on the house of cuts?
>> We're already seeing it with the RAM. So the RAM increase in RAM prices, the only way that RAM prices are going to come down is if the memory companies lower them, which is not going to happen, or they increase capacity, which will take two or three years. So we're already seeing that inflation happening. The reason everything is being made, if you want to blame someone for all your stuff getting more expensive, it's Microsoft, Google, Amazon, Meta. It's the companies buying the giant data centers worth of GPUs. I calculated this week, I'm doing a big guide to RAM, that for a gigawatt data center, there's nearly $2 billion of RAM, of just high bandwidth memory, just just for across the GPUs. So, we're already seeing the inflation is happening there. We're also seeing it across the cost of talent. We're seeing it across the cost of materials. The cost of everything, really, is already going up. Inflation spikes, everything gets even more expensive, more expensive. The problem here is that hyperscalers have caused tech inflation through their kind of rapacious need for more GPUs. Had they not sunk all this money into this doom technology, this would not have happened. They have created what it took to make the memory cartel even more powerful. So, yeah, we're in a point where we're already seeing the signs of inflation within tech. And as inflation worsens, which it will, especially as we stay in Hormuz and whatever the hell it is that America is doing out there, things are only going to get worse there. And they're already guaranteed to get more expensive. This isn't a maybe this will happen. Uh, SK Hynix, Samsung, Micron have already said prices will increase. And the reason they know that they can do that is, well, they, it's just a dial they turn. They had 84.9% gross margins, Micron, on their last earnings. Like they can just turn the dial all they want. They could make it cheaper. They don't. Insane. It's insane. It's bad and everyone involved should feel ashamed of themselves.
>> So, if we were talking about a group of companies outside of the AI hype, how alarmed do you think people would be to hear that the five largest companies on the planet pretty much had become net cash borrowers in only what, a couple years?
>> It's genuinely horrifying. The whole sales pitch of the Magnificent 7, in particular, Meta, Google, Microsoft, and Amazon, was that they were cash heavy, asset light. You invested in these companies because software can theoretically sprawl outwards and outwards and be perpetually worth more, and you can just keep raising the prices as they have. Not saying it's good, but it's a good on paper business. These companies have, in the space of really two years, now gone from being asset light, cash heavy to cash light, asset heavy with no real exit point. The amount of, what let's say they spent $765 billion this year, trillion next year, they're going to need four, five trillion dollars worth of profit very quickly within the next 10 years. And I mean specifically from AI profit improvements, improvements to previous products is not going to cut it anymore. We need something brand spanking new and it needs to print money instead of burning it. But that's kind of what happened here. It's really strange. There's, I don't see any historical precedent. It's nothing like the dot-com bubble. Nothing like that. During the dot-com bubble, none of these companies really overstocked like this. I guess you could say the memory crisis has similarities to 2021 with the supply chain crisis which caused chips to get backed up, I guess. But even then, the scale here is completely unprecedented. I mean, just even the dot-com bubble was spread out across tons of companies. This is relatively centralized. And the danger here is that because it's relatively centralized, it only takes one of these burks to to blink for this to all change. Because the crazy thing is, is that the reason all the RAM is so expensive is because of high bandwidth memory. And Nvidia, just Nvidia, accounts for about 65% of all high bandwidth memory, which means that if Nvidia stops selling GPUs, all of it crashes. All of it. It's just a, it's a load-bearing company. And yeah, it's really quite concerning. Concerning in a way that I'm not sure people even want to think about too much because Microsoft, Google, Meta, or Amazon, they will be fine at the end of this. They will. But their whims are currently changing the course of capital within, I mean, the entire tech industry and partially the world. And said whims are just a waste of money at this point. You can, people can talk about the key-jingling AI features they like in that particular week because they were told to like them all they want. I don't really care. Nothing that is, nothing that's been made so far or will be made is remotely worth any of this. And when they eventually pull the string, because they will eventually have to pull back on capex, the hyperscalers are going to pull the entire tech industry into a depression. And what's horrible is continuing only makes it worse. The longer this goes on, the more the bubble inflates, the more costs will rise. And it will take a long, long time for them to come down. And they'll do so only after the memory companies take billions and billions of dollars of losses.
>> A statistic that kind of really drove this home for me was that US stock valuations are now 41 times higher than their total earnings in the last decade.
>> And then you, yeah. And then you have the Bank of America warning that speculation is hitting extreme levels, is their word, which historically has led to what they also call a snapback, which is a bubble popping or repricing. Do you, do you think there would be as much buy-in to AI right now if it wasn't creating the illusion that economics sort of around the world and particularly in America, I suppose, were doing better than they are?
I think that if the hyperscalers had not invested so much in GPUs, that none of this would have taken off at all. I just, deadly serious, I just don't think it would have. Because the only reason the AI bubble is even relevant at all is, well, it's two reasons. One, media manufactured consent. The media hook, line, and sinker took this. They slurped down the slop. They were ready to manufacture consent across the world. They were ready to say, "This is the most important thing," whether or not it was was irrelevant to them. But had there not been a massive capex bubble, had there not been all this money being sunk in, it would have been kind of hard to actually point to any number going up. And the only way to make a bubble inflate is to have a number go up. The problem is, is that at some point, as I've been saying forever now, at some point, this needs to result in something. It has to be more tangible than them saying, "Revenue has grown for some reason. We're not really sure what it could be. Perhaps it is AI." Or of course, it's AI. Can we say how much? No, no, no. We couldn't possibly. The fact that it became a speculative stock bubble is what is going to make this so much worse. But really, it is this memory crisis now that I keep coming back to because there is really no fixing this now. Now that the hyperscalers have gone so deep into this, now that they've created these capacity crunches, now that they've made these situations where basically effectively every electronic is in low supply because they're building so many data centers, there is no pulling out of this. And this would not have gone on so long had they either given up earlier, which they should have for the sake of their companies, everyone else, or had they just, I don't know, found an alternate solution, but they didn't want to because they're, they're out of other ideas and they know the market is full of people that don't really understand what they're talking about. So, they're just doing what they always do, which is spend more money, hire people, and fire people, which is why thousands of people lost their jobs at Xbox this week. Sachin Nadella is a scumbag. He is a, and a scumbag. Just to be very clear about him. He has no idea what he has at Microsoft and he'll tear that entire place down so he can buy Sam goddamn Altman another three GPUs. Miserable bastard.
>> Hard to, hard to follow that one, I must admit. But bring it, bringing it back to what we're talking about. In some further analysis from Deutschbank's Jim Reed, he says that well, he thinks that LLMs have a great potential for increased productivity, but that possibility is years away from now and only would only come after technology is embedded and effective use case has been found. But the thing that seemed missing to me from the analysis was that frontier model developers need to be able to reach that point. But from what I'm seeing, it doesn't seem like anyone has the patience to wait years to see a return on investment.
>> I mean, also, what is the, like, can you imagine if you went to work and you went to your boss and say, "Mate, you're going to have to give me three years until I make you any money."
>> Uh, I can't really do my job yet, but with enough experience, I know there's no proof, but with enough experience, I think I could be even better than you at your job." You get fired. In this case, it's even in this analysis, they are like, well, of course, we see the potential of LLM. Why? Why? To your point, Isaac, it's like, to get them to the, what they are describing is this theoretical thing of LLMs being useful and affordable. So, this is not just a software breakthrough. It's a hardware breakthrough. Multiple generations of hardware breakthrough that will only get funded as long as the AI bubble exists. Kind of what you were saying. The software breakthroughs will only happen if money comes through to the AI labs to keep doing it. But also, all of this assumes it's possible. All of this assumes that this is something that will actually happen. They are just, we are a trillion plus dollars into this and we don't really, we have proof that it can do something. It can generate code, I guess. But as far as like a business that makes sense, as far as a product that can be consistently used and reliable, we're still nowhere near. We don't have, we have increasing amounts of proof that if you throw a lot of compute at this, it can do things a bit better, but a bit better in very specific ground. Grok 4.5, new model came out, Elon Musk, and apparently it was competitive with Fable. Other than the fact that in small print it said, "Yeah, however, the training data accidentally included the test it was doing. We have no idea how this might affect it." It's always that. So, the point I'm making is, is that all of these assumptions, they're always like, "Well, we need companies to change around what LLMs do. We need LLMs to become better in some indeterminate way." No. The way that we need LLMs to become better is that they have become a different business with better economics, with more, with full reliability. The things that LLMs cannot do and have been proven to never really being able to do. And I think that all of these analyses make that kind of fundamental mistake. They always come from the place of, well, it will work out when it works out, how it will work out, just naturally assuming that. And it's where I think most economic analyses fall down. It's this thing where they always assume that things will go good, that all this money can't be wrong, despite the fact the most useful economic analysis I think right now is, what if everyone is wrong? Because if everybody is wrong, I think we're on the path to one of the the first real tech Great Depression. Because the dot-com bubble was bad, but there were still Microsoft, Amazon, Google. They, they were in their earliest days, so they were just kind of plugging along. They hadn't got to the point where they were so deeply relevant to the stock market. Google, Microsoft, Meta, and Amazon are now right at the top, and they're going to take an absolute kicking after this. But on top of that, there is nothing left for them. They don't have a next big thing. And when people realize that, that's killing venture capital, who is already going to suffer because most AI startups are going to zero. It's going to kill any debt going into tech because can you imagine after the data center collapse, can you, do you think anyone's going to be like, "Oh, let me fund the next big tech thing?" It will kill trust in the tech industry, which I actually think is a good thing. The tech industry needs a punishment. It needs to suffer. It needs to realize it, the people at the top need to realize that the path we are on is broken and it's wrong and it's not leading to building the future, it's leading to a re-entrenchment of the status quo and empowering the largest and richest companies in the world. Or at least they used to be until they took on so much goddamn debt.
>> I mean, Elon Musk recently joined other other companies in capping its AI usage, uh, limiting employees to, I think it was a $200 a week budget. Do you think that this is kind of a sign that the incorporation of AI is slowing down?
>> Yeah.
>> And then if so, who's going to pick up the tab for all of the the data centers being built? Assuming anyone has enough money to hand over that much cash.
>> Well, that's, you've kind of noted the most important story at the moment, which is Anthropic and OpenAI over $1.1 trillion of spending commitments across compute, just compute. Anthropic just signed a 20-year agreement for 400 megawatts of space from Cipher Mining or Terra Wolf, one of those two non-comp. But all of this, all of these commitments have been made under the assumption that growth is infinite, that they will always need more compute. But it's kind of a, kind of a paradoxical situation cuz they both need to make things cheaper and they need justification to spend so much money on compute. And it really comes down to, are they just, do they just believe that they're going to have a ton of customers so they can lower the costs, but also their customers are saying they'll spend less. So what's going on here? Like, what's the plan? And the answer is, I think everyone needs to accept that the reason there is no answer here is no one had one. I think from the very beginning, their only plan was to spend more money. When that didn't work, well, they spent more money. And when that didn't work, they kept spending money. And they kept getting, they kept getting talent, they kept paying for talent, they kept trying new industries, they try, they try, they try under the assumption that something would change. The reason that Anthropic releases a new product every two days, other than the fact they're just vibe coding that crap. They just, they're slopping it out. Uh, anyone actually use Claude Design? No, I didn't think so. Um, they're just throwing these stuff out. It's because they're waiting for something to hit. They're waiting for something to hit. They want something to hit so bad. Oh, GPT 5.6 Ultra Soul Luna, whatever. Who cares at this point? Oh, I've read a, I've read five different posts on Twitter that are 900 words long that say, "Oh, GPT 5.6 six Ultra Soul, whatever. It changed my life. This is the moment." I've read one of those posts. Every single model launched since GPT4, I have read a post by someone telling me that this has changed everything and this is the beginning of AGI for years. I'm tired of it. I'm tired of it. I think everyone else is tired of it. It's just kayfabe. It's wrestling talk at this point. They just, they're just doing a bit. But none of this makes any sense. None of it. I saw a report earlier that actually had way more like a 100 gigawatt worth of GPU sales from Nvidia through 2027, which is three times what I had anyway. Who is going to pay for them? Aquaman? Like, what, who is, who is paying for all of this? Who is the customer? And when you ask analysts and boosters and AI boosters in particular, they always just say, "Someone will pay for it. Some, the demand's there. The demand's off the charts. It's crazy. The demand's off the charts." The demand better be off the charts because I, I mean, 100 gigawatts would be a trillion dollars of annual compute spend. The entire software industry, every dollar in it, including Microsoft, Google, Amazon, Meta, is I think $750, $800 billion annually. So you're saying that just the software industry is going to double in the next four years? How? Who's paying for it? Well, AI is so useful. Shut up. Give me a real answer. Because I feel like the critiques of my work, getting a little personal here, come from a place of like, "Oh, well, what didn't you think about this? Didn't you think about this?" You ask an AI booster to think about the simplest thing and like, no, it'll work out. I'll just spend the money. Money come from it'll be fine. Don't worry about it. At some point, you got to worry about it. You got to worry about the fact that we don't really have a business model here and that we don't have the actual demand. And this is really serious. Like, everyone right now, you could double today's compute demands, which I think are somewhere between 50 and 100 billion dollars of annual spend, including OpenAI and Anthropic. You could double that and that will not even be halfway to fulfilling the or substantiating the demand for the data centers being built in the next two years. So, yeah, good luck with that, everyone. No one has a plan. No one needs a plan. Just keep hoping, I guess.
In your newsletter this week, you kind of make the point that the AI industry is not too big to fail. It's not a financial system like the banks that kind of are needed for a lot of things. And I suppose I'm, I'm also wondering what a bailout would actually achieve in the first place other than just giving taxpayer money to Nvidia. I mean, it, it wouldn't change the fact that it costs more than it makes and OpenAI or or Anthropic or whoever would just be back asking for more when the bailout runs out.
>> So, I want to start with something. People keep saying OpenAI is too big to fail, blah, blah, blah. This is an intellectual crutch. It is a way of not thinking about the actual problems. OpenAI, Anthropic won't be bailed out. They could get some government funding through a stake or a sovereign wealth fund. Shouldn't happen. Maybe it will. People say, "Oh, Trump's just going to do what they want because it gets bailed out." Stop. First of all, stop giving them the idea, but also stop giving up so easily. Also, that's a lazy way of looking at it. So, Great Financial Crisis is not what most people thinks. There was the T.A.R.P. program, uh, Troubled Asset Relief Program that existed to buy dodgy mortgages and securities and bonds off of the banks, off of the, also bail out the auto industry. There was $40 billion that was meant to help foreclosures. It didn't do anything. But that wasn't really what the bailout was. The real bailout was something called the P.R.D.C. and the T.S.T.S.L.F., the Term Light Term Security Long. Basically, it was feeding banks money to take their bonds so that they could get overnight funding. Banks are funded through overnight funding through the repo facility. Hundred billion dollars a week. That was because if they didn't do this, the banks would run out of money. Like that is as simple as it is. The bailout was not for stocks. The bailout was not for really for companies other than the auto and the bank industry, which were critical to the American infrastructure. OpenAI and Anthropic are not critical to anything other than the single most annoying hype cycle in history. Nvidia, Meta, Google, Amazon, Microsoft, they're not dying of this. Nvidia's revenue could be cut by 90% comfortably, maybe more. They are, they were making single-digit billions of dollars in 2023. They could easily return to that. Nvidia has bet everything on this. I get why they did it. Nvidia isn't dying, though. Now, could Jensen Huang need a pardon one day? I don't know. I hope Nvidia isn't selling GPUs to China and they find proof of that because that would be illegal. I don't, I'm not accusing them. I'm just saying there's a lot of articles that suggest that. So, a bailout of the AI industry would likely be one of a few things. It would be a bailout of the data center loans. This would be probably the most historically unpopular government program ever. It would make 2008 look like a, a beloved socialist policy because can you think it, like the only industry I think hated more than the banks would be data centers and this would be bailing out data centers for the banks. Insane stuff. Truly won't happen. Also would be hundreds of billions of dollars. Just not going to happen. Now you could say, well, Trump will get, Trump will get bribed. Trump will get bribed. The US government will happily let companies die. They do it all the time. And also, fundamentally, the AI bubble is a speculative stock bubble. It is the crazed valuations based on a non-existent industry, an industry that pretends it's big by spending a lot of money, but it's actually maybe a hundred billion dollars of revenue a year, including compute costs. So, it's actually quite small. But say Trump, in this theoretical scenario, Trump stops OpenAI dying once or Anthropic dying once. That doesn't change the fact that the stock market is moved by vibes. It's moved by investors reacting to things. I don't think the investors will react well to a bailout. And indeed, when it gets to like Oracle, for example, if Oracle stock crashes and Larry Ellison's margin loans get called in, while they might bail out Oracle itself, they're not going to help Mad Larry. Can't, I don't think the US government is going to bail out tens of billions of dollars of personal loans. And if they do, wow, will that also be unpopular? But I really just don't think it's possible. People have this weird kind of defeatist attitude when it comes to Trump and bailouts where they're like, well, 2008 was very traumatic and it was traumatic, way more traumatic for the regular people than anyone in the banking industry. But that was a fundamentally different thing to what the tech industry is going through now. That was a situation where the collapse of AIG would have collapsed commercial paper. Commercial paper loans were held by money market funds throughout the world. So if those collapsed, then the literal underlying fabric of how money was moved and people's holdings would have just detonated. And they already dropped because the markets died. There is no situation like that here. The closest we'll get maybe is private credit. And private credit itself, that's because they're funded a lot by pensions and insurance funds. Even then, I have real hesitance to say whether there'd be any bailouts there. I just think the people want to default to that. The reason that they do is because it's much easier to say, "Bad thing I know the shape of will happen. Bad people get money to keep doing bad things." Also, a bailout of OpenAI wouldn't solve the problem of how do they get more money in like six months.
>> Yeah. But nevertheless, people want to default to this because it's much easier and it's actually intellectually stimulating and say, "Oh, I know what's coming. I know the bad thing that's coming and I can be ahead of it and I'm smart." Not saying that they're not smart. I'm just saying that this is an intellectual crutch to not think about the other thing, which is a stock market crash realistically, and then also a period in tech without growth because at some point everything slows down and at some point Microsoft, Google, Meta, and Amazon have to accept and will see that their core revenue streams will slow. And it will happen. The only thing that grows forever is cancer. And so thinking about that, thinking about an actual collapse requires you to say, venture capital as we know it could die. One of the major tech companies could take a mortal wound. The amount of debt they've taken on, the amount of assets they've taken on could actually be lethal 10 years time. But still, people don't want to think about the complexity of that because that means a fundamental reimagining of the tech industry. And I don't think people are ready to think about that. It's easier to just say, "Bad thing that we know today will continue."
>> I mean, looking back at some stuff from history, you also in your newsletter make the point that the dot-com bubble and the AI bubble is not, not the same. You mentioned it earlier as well. I mean, not least because of the infrastructure, fiber has has a use even today. Some of the data centers that are being built or about to be built or planned to be built are going to be out of date before they've even, before they might have even started construction. But aside from, I suppose, a rising and burgeoning laser tag bubble, what, what do you think would, what will actually happen to the hundreds of billions of dollars of infrastructure that's being built to run LLMs? Is it just going to be just empty warehouses and scrap metal? Surely there is some use to find for this that somebody could, I don't know. It's just, what, what happens to it, I guess, is the question.
>> Depends how much of it is built. So on the hyperscaler level, they are going to do everything they can to use the GPUs. Even if it means charging basically nothing for them just to make sure that they don't have to take cash impairment. They would rather take an operating loss off of these things than take the billions of dollars of impairment because of the useful life for these things, because they need them to last six years so that they can just take them off the books. If they can get them six years, it'll be fine. The stuff that isn't built for them is the problem. Because if they get to a point where they pull back on capex and they have to accept that they're not installing what the millions of hundreds of thousands, truly don't know how many is sitting in warehouses, they will have to take an impairment on those. They'll have to sell them off. So the answer for what happens with GPUs, and I know I'm going to get an email from someone saying there are actual uses for GPUs, not at this scale. There's GPU-driven analytics, as in like very big database searches. There is 3D modeling. There is scientific imaging. Those are not, those are multi-million dollar industries for an industry that needs hundreds of billions of dollars a year in revenue. The unbuilt data centers will likely lead to a fire sale of GPUs, which will eventually lead to nothing happening. The un, the partially finished ones will get capped off. Probably they'll find a way to turn them into user buildings and sell off the land. I think that hyperscalers will probably end up screwing multiple vendors because they can. Microsoft is one of the largest legal departments in the world. They will get out of any agreement they can. The real question at the end of this is who is left holding the bag? Is Nvidia, does Nvidia have cancellation terms? Can customers cancel on Nvidia? Can Micron, SK Hynix, Samsung, can their customers cancel on them? Who can cancel on who? Who can Broadcom's customers cancel? Can Google cancel TPU orders? Because it really comes down to if they can, you're going to see massive guidance drawbacks across the board. You're going to see a bunch of inventory getting written off as happened in 2023 with Micron. You're going to see hyperscalers probably have to take some sort of impairment. And I think you see someone's head getting taken off. I think you see one of the C, maybe Sachin Nadella. Amy Hood's been watching him for a while. Amy Hood take over the CFO of Microsoft. She could do it. You could see a change at Google, but Sundar is pretty well liked by Sergey and Larry who still had board seats. I believe you'll never see Meta move. Andy Jasse as well is always a good choice to fire. I think like the era of Jasse, because he was one of the early AWS people, was meant to be the big era of cloud computing. Amazon Web Services growth slowed for a bit. If AI isn't the thing, they built an entire data center in Indiana for a customer that will die. You might see Anthropic. I think you see Anthropic absorbed into Google or Amazon or both, like a weird consortium. I think you see OpenAI absorbed into Microsoft. Real question is whether they go public. If they do, well, it will be investors' problems too. And it's just, it really sucks. It sucks terribly because none of this had to happen. They could have stopped this in 2024. What I've been saying was true then. It's still true now. These things don't have ROI. They can do some code generation, but it's not a real business. It's not consistent enough to charge people for. And yeah, these are not real businesses. This isn't a real industry.
>> Well, Editron, thanks for taking the time.
>> Thanks for having me.
>> If you enjoyed today's episode and you want to hear more of the tech report, please consider liking and subscribing. Also, you can get episodes of the tech report wherever you get your podcasts.