This Isn't Just Technology — It's A Weapons Race — We Had To React

Welcome back to Impact Theory, I’m Tom Bilyeu. Today, we’re cutting through the noise around artificial intelligence—addressing both the real and overblown fears dominating the headlines. Together with Derek Thompson, we’re taking a hard look at where AI is actually headed: from viral but out-of-touch startup launches to the unprecedented investments pouring in from America’s biggest tech giants. We’ll explore whether the AI sector is in a dangerous bubble or just getting started, what the US-China AI arms race means for the future, and how these innovations are reshaping everything from our jobs to our culture. We’ll also confront the tough ethical questions—like the destruction of rare books for AI training—and what all of this means for you, me, and everyone trying to navigate this seismic technological shift. If you want the real story on where AI stands and where it’s going, you’re in the right place. Quince: Free shipping and 365-day returns at https://quince.com/impactpod Whatnot: Download the Whatnot app today and get free shipping on your first order. Quo: Try for free PLUS get 20% off your first 6 months at https://quo.com/impact Incogni: Take your personal data back with Incogni! Use code IMPACT at the link below and get 60% off an annual plan: https://incogni.com/impact Pique: 20% off at https://piquelife.com/impact Shopify: Sign up for your one-dollar-per-month trial period at https://shopify.com/impact Surfshark: Go to https://surfshark.com/TOMBor use code TOMB at checkout to get 4 extra months of Surfshark! ATT Business: Switch to AT&T Business at business.att.com http://business.att.com Ketone IQ: Visit https://ketone.com/IMPACT for 30% OFF your subscription order Netsuite: For the first time ever you can try NetSuite Next for free. If your revenues are at least in the seven figures, go to https://NetSuite.ai/Theory https://netsuite.ai/Theory. What's up, everybody? It's Tom Bilyeu here: If you want my help... STARTING a business:join me here at ZERO TO FOUNDER https://tombilyeu.com/zero-to-founder?utm_campaign=Podcast%20Offer&utm_source=podca[%E2%80%A6]d%20end%20of%20show&utm_content=podcast%20ad%20end%20of%20show: https://tombilyeu.com/zero-to-founder?utm_campaign=Podcast%20Offer&utm_source=podca[%E2%80%A6]d%20end%20of%20show&utm_content=podcast%20ad%20end%20of%20show https://tombilyeu.com/zero-to-founder?utm_campaign=Podcast%20Offer&utm_source=podca[%E2%80%A6]d%20end%20of%20show&utm_content=podcast%20ad%20end%20of%20show SCALING a business:see if you qualify here. https://tombilyeu.com/call: https://tombilyeu.com/call Get my battle-tested strategies and insights delivered weekly to your inbox:sign up here. https://tombilyeu.com/: https://tombilyeu.com/ FOLLOW TOM: Instagram:https://www.instagram.com/tombilyeu/ Tik Tok:https://www.tiktok.com/@tombilyeu?lang=en Twitter:https://twitter.com/tombilyeu YouTube:https://www.youtube.com/@TomBilyeu See Privacy Policy at https://art19.com/privacy and California Privacy Notice at https://art19.com/privacy#do-not-sell-my-info.

• 0:00Speaker 00

  Let's talk about the moment your side hustle becomes a real business. Now, I see way too many founders stay stuck running their business like it's a hobby, and then they wonder why nothing ever feels real. It doesn't feel real because it's not. I've helped thousands of entrepreneurs build real businesses through Impact Theory University, and the first step in the path to building a legit business is registering it. And Taylor Brands will help you do just that in a few clicks. An LLC, a limited liability company, is an official business structure. It's inexpensive to form and very simple to maintain. It's how you open a business bank account. It's how customers know they're dealing with a real company. And Taylor Brands doesn't stop at the filing.

• 0:45Speaker 00

  Their dashboard tracks what comes next based on what your business actually needs. So EIN number, registered agent, licenses and permits, bookkeeping, business banking. I tell every founder the same thing. Just make it official. Click the link in the show notes to check out Taylor Brands and get started on your business today. This is a paid advertisement. Right now, the reality is that people are having very warranted fears about AI and very unwarranted fears about AI. This video is going to walk us through answering the question, is AI on its last legs? Is there so much systemic risk here that everybody should be terrified?

• 1:25Speaker 01

  On July 28th, 2026, an AI startup called Orchid posted a launch video that went mega-viral on Twitter. It features a couple on their anniversary, but the boyfriend forgot about it. The woman, clearly frustrated, vents to her AI assistant Orchid for help. The assistant says it's way ahead of her and sends messages to her boyfriend reminding him of the date. Then it proceeds to book a table for him and get her flowers she likes. He texts, "Okay, okay, I can do this." And the chatbot says, "You can't. That's why I'm here." Orchid tells the woman he's acting like he planned it himself as he compliments the AI for carrying their relationship, and they live happily ever after. How wonderful. Instead of trying to be a better significant other, you can just stay a piece of shit and have AI cuck your girlfriend.

• 2:10Speaker 01

  What a wonderful world we live in, isn't it? Unsurprisingly, this video generated a massive amount of vitriol, and it's just another one of the many examples of the people building this crap being so incredibly out of touch with the general populace, which will probably lead to its impending collapse.

• 2:27Speaker 00

  One of the most important things to understand when you're looking at the culture of a tool like AI, some people are going to use it terribly, other people are going to use it well. But just because something can be used in a way that's absolutely asinine, or just because the company behind it is using marketing materials that show it used in a way that you know is going to have these second and third order consequences. If you get lost at that layer of the argument, then you end up going down a stupid street. Once you understand AI as an arms race between the US and China for intelligence itself, now you're grounded in the right way. And so I'm really driven to absolute madness by the level of discourse around AI. You need to understand it as a weapon system.

• 3:10Speaker 00

  You need to understand why intelligence is the single most important battle that we will do as a nation ever before we can have any real conversation. Now, he's going to get into the dead of it all and the circular financing, and we're going to talk about that. But also, just at the cultural layer, I want to make sure that people are thinking about this in the right way. The reason that humans have been able to have the outsized impact on the globe that we have been able to have is one reason and one reason only. Our brains are organized for higher level intelligence. Now, I know that people want to say dolphins are more intelligent than us, or octopus are so intelligent that, you know, we might as well see them as an alien intelligence that rivals ours. Kids, if that were true, they would have figured out how to breathe outside of the water

• 3:54Speaker 00

  just as we have figured out how to breathe inside the water. The reality is that we are the most dominant species the world has ever seen precisely because we have the higher level cognition, the ability to do theory of mind, the ability to plan for the future, the ability to manipulate objects within our minds and then go out and build them. Intelligence, flexible intelligence, is precisely the end goal if we want to move forward as a species. And so when you start thinking about what artificial intelligence really is, everybody's so hung up on the artificial part, they get into the Terminator of it all, and they forget about intelligence. They forget about if we could scale intelligence,

• 4:38Speaker 00

  it would radically transform our lives, radically transform the world. We look at all these demographic problems, and we lament that it takes 18 years to bring an adult into the workforce, but that isn't true anymore. So when you're having the debate at this level of being mad because Sam Altman is a douchebag, or you don't like Elon Musk, it's like you're missing the point of why this technology is as important as it is. So please keep in mind, don't get sucked into this cultural framing.

• 5:12Speaker 01

  Thank you to Derek Thompson for this great Substack article which helped guide this video. By now, most of us know that AI is in a bubble. A financial bubble is when the asset prices of a certain industry rise to unreasonably high fake levels that do not match the actual value of what's being built.

• 5:27Speaker 00

  I love that this is his framing, and it's framing that I'm deeply empathetic to. And I'm actually going to step outside of my own frame today a little bit. I'll certainly give you checkpoints in terms of where I believe we are in terms of the bubble. But what I want to do is now argue from the bull case of AI. People like Ral Paul, who I just spent time with yesterday, who believe that we're nowhere near being in a bubble, certainly not the way that people think that we are. And if you look at historical examples, there's reasons for this. So I'm going to be walking us through. I'm going to let him take the bearish case. I'm going to let him make an argument that you guys are probably used to me making. And I'm going to approach this from the outside to give a different look at this because I think that the reason that the bubble continues to inflate

• 6:11Speaker 00

  is precisely because people do not agree on where we are in terms of the valuation. Are we already completely unhinged? And this has really, the valuations no longer match the actual value of what's being built? Or are we looking at something where we haven't even begun to scrape the bottom of the first rung of the value that it's going to bring to society? So that's going to be the collision that we go through.

• 6:36Speaker 01

  In order to understand just how large this bubble is getting, we need to know what a hyperscaler is and how they're using their money. Hyperscalers are the largest cloud computing and data center providers who have created most of the AI infrastructure. The four biggest hyperscalers are Meta, Microsoft, Alphabet, AKA Google, and Amazon. Since 2021, these four companies have been pouring an incomprehensible amount of money into AI development and infrastructure. A big reason they could justify this is, well, for one, AI is a race to build the next big thing, and so companies are just trying to get there first, but also because these companies could afford to. The hyperscalers' free cash flow, the cash that a company has left over after paying its day-to-day operations

• 7:19Speaker 01

  and long-term expenditures, has remained in the green since the beginning of the AI arms race. Just two years ago, the four biggest hyperscalers, Microsoft, Alphabet, Meta, and Amazon, were generating roughly $210 billion in free cash flow every year. But in 2026, their free cash flow has plummeted to below zero. Of course, not every public company is suffering because the wealth is being transferred from the hyperscalers to the semiconductor companies.

• 7:47Speaker 00

  Before we move on to that, it's important to understand that there's a big difference between cash flow and free cash flow. So what you're looking at are companies that still remain optionally very valuable. So very cash flow positive. They are default alive. When you have a company that's default dead, it means that they are unable to be profitable, to offer the product or service that they're generating their cash from. They have to have outside capital because they're not yet even break even with their core offering. When you have a company like this, though, which is spending that money in an optional fashion, so their core business remains the same,

• 8:30Speaker 00

  their core business is still these insane cash cows. So at any time, they could say, "Oh, look, AI was a failed experiment. We're going to stop spending money on that." And then they will go back to being just as profitable as they were before. So understanding that what they're doing now, they're not investing in something that is destructive. So they're making a big bet, and if that big bet doesn't pay off, that is certainly going to be something that will cause instability in the narrative story around who they are as a company. But if you look at what Meta did when they changed their name to Meta, that was them placing a bet that the metaverse was going to be the future. This was at the height of Web 3 and everything that was happening there, and they had a vision for what this was going to be. They end up buying Oculus.

• 9:15Speaker 00

  That becomes one of the core divisions of the company. They spend billions and billions of dollars every year, year after year, focusing on that, end up changing their name, refocusing the entire company, and then they decide, "Meh, actually, that isn't going to be the future." They stop spending that money, and they go back to being the cash cow that they always were, that they seemed less of because they were investing something in the future, but that core business was still there churning that money out day after day. The same is true for these hyperscalers. So yes, right now, they're putting themselves in a shocking position. And you guys have heard me say many times, "Yo, it is wild."

• 9:55Speaker 00

  That Google, which has never been cash flow negative since the time that it came online... Is now cash flow negative for the first time? That should definitely pique people's curiosity. And if you're investing in them, you should be asking, are, is this going to work out? Because they are spending a tremendous amount of capital. But keep in mind, they could shut that off. As Meta did with their Metaverse initiative, they can shut that spending off and that money is still there. So making sure that you don't read optional cash flow expenditures as a fundamental flaw in the business. Very important.

• 10:30Speaker 01

  The problem is these hyperscalers need to keep spending, spending, spending. So what are they doing? They're taking on more debt. Here's what that looks like from the big four. In order to generate more spending money, public companies will issue bonds, essentially a way for investors to lend money to a company for a set period. The company pays the investor interest at fixed intervals. Then the original amount borrowed is paid back to the investor when the bond reaches its maturity date. From 2024 to 2025, the annual volume of bonds issued from hyperscalers went from roughly $20 billion to 110 billion. And in 2026, that number reached 150 billion. Many are now likening this to the dot com bubble in the late 90s and early 2000s.

• 11:14Speaker 01

  In an article for Business Insider, Jennifer Soar writes about an analyst at JP Morgan comparing the divergence in the AI trade to communications equipment in 1999. Quote, "Companies supplying communications equipment began to see a parabolic rise in 1999, while companies making heavy capital investments in the space tumbled from their peaks."

• 11:34Speaker 00

  Okay, so the important thing to understand as we talk about, okay, what is the real size of the bubble here? Before I was talking to Raul Powell, I had a sense that these companies were getting way farther out over their skis and they actually are. So when you look at the debt as a percentage of the cash flow, what you see is right now, these guys are taking on about 4%, maybe not cash flows, but as a percentage of their value. Right now, they're at 4% in AI. So a very manageable number versus 30% at the height of the dot com mania. So right now, in terms of scale, we're a long way, which I actually found pretty shocking. I didn't realize we were this far away

• 12:20Speaker 00

  from the heights of the insanity around dot com. So yes, these guys are taking on an extraordinary amount of debt, especially when you just look at it in raw dollar terms. But when you look at it as a percentage of their value, we're still in a much, much, much saner place than we were in the dot com bubble. So as I do my own thinking through this, where I come out is somewhere between Raul and these guys. For Raul, there's a sense that, or for anybody that's holding the bull case for AI, it's going to go something like this. This is the single most important technology that the world has ever seen. The US is locked in an arms race with China.

• 13:02Speaker 00

  There is no way for the US to not push this industry forward. So even if any one of these individual companies ends up doing something stupid and they get themselves in financial trouble, the US government is going to step in and either prop that company up or they're going to step in and help sell off the assets to the other companies. So any money that's raised to build the physical infrastructure of the data centers, that is going to be used. So it may not be used by Anthropic. It may not be used by OpenAI, but it is going to be used. And so when your mental model is that, oh yeah, I've seen this playbook before.

• 13:43Speaker 00

  I know that the US will print an absolutely obscene amount of money to prop up, whether it's banks, whether it's housing, but they'll print an obscene amount of money to protect an industry, especially if they think it's systemically important, which it is right now. Not only is America, in terms of the stock market, just making one bet on AI. And you can make a very credible case that the whole world is making one bet on AI. So the odds that a country like China or the US that has the ability to print their way out of a problem isn't going to do that is basically zero. So if the industry gets in trouble, it's going to be backstopped by all of us, by the taxpayer.

• 14:26Speaker 00

  So you can just expect that this is going to keep going. Then you combine that with the fact that we're at 4% of debt to value versus 30%. And all of a sudden, it's like, okay, we may not be in the same, anywhere near the same level of risk that we were in when the dot com bubble finally blew up. And so when you ask a question, not is this a bubble? But where are we in the bubble? Are we early innings? Are we late innings? I think that we're in later innings from a stock price purchase perspective and much earlier innings from a actual debt burden

• 15:09Speaker 00

  on these like just absolute cash cow companies. So when you look at the amount of debt that the companies have taken on, that's not necessarily scary. But for me, when you look at the 40x Cape ratio in the stock market, so the Cape ratio is designed to say, hey, over the last 10 year average, how much money have these companies been making compared to the value of the stock market? And when the value of the stock market gets too high compared to the actual money that these companies have been producing, now it's like the alarm bells are going off. For me, there, the alarm bells are going off. The traditional number is 16 to 17x.

• 15:52Speaker 00

  So you would expect the value that people are willing to pay to be 16 to 17x, the 10 year running average of the amount that the, call it the S and P 500 was actually making. It's now at 40. And the last time we were at 40 was right before the dot com bubble. So this is one of those where it's like, okay, like we're starting to get into crazy territory. But the companies are probably doing just fine.

• 16:17Speaker 01

  The dot com bubble eventually burst in early 2000, less than a year after that divide was first noticed. What also might be indicative of what's to come is the fact that Microsoft ended June with its worst monthly share loss since the dot com bubble in 2000.

• 16:33Speaker 00

  We're hitting pause for a moment. There's plenty more ahead, so don't go anywhere.

• 16:39Speaker 00

  There's a reason Quince sells cashmere for $60, and it isn't the cashmere. It's real 100% Mongolian cashmere. It's the same stuff luxury labels charge a fortune for. Quince just works directly with ethical factories and cuts out the middleman. So there's no need for a markup and no logo tax. The $60 goes to the actual sweater, not the fancy name on the tag. That's why everything they make runs 50 to 80% less. And it's not just the cashmere. I ordered a few things myself, their fleece joggers, a couple of their tees, which I wear obsessively. First thing I noticed was the quality. Soft, well-made, built to hold up. They make down jackets and wool outerwear now too.

• 17:24Speaker 00

  So upgrade your stuff to things you'll actually wear. Find your next fall favorites at Quince. Download the Quince app for app exclusive offers or go to quince.com. Get free shipping on your order and 365 day returns, now available in Canada and the UK too. And when Quince asks where you heard about them, let them know it was Impact Theory. It's the best way to support the show. Let's talk about what it means to actually be visible in today's day and age. When you build something public, whether it's a business or an audience, anything on social media, you become a target. This is something that legitimately keeps me up at night because our data isn't just on the obvious broker sites. It's on old news portals, public forums, obscure directories,

• 18:09Speaker 00

  places you would never think to look. That's where Incogni's exposure scanner comes in. It hunts down your personal information across the wider web, not just the 420 plus broker sites. Incogni already removes your data from automatically. They can't harm you if they can't find you. Go to incogni.com/impact and use code IMPACT for 60% off an annual plan. Try it risk-free for 30 days. That's incogni, I-N-C-O-G-N-I.com/impact and use code IMPACT. Let's talk about how much you're paying to stay safe online. Count it up.

• 18:51Speaker 00

  A VPN, a password manager, something to block ads and trackers? The service to scrub your data off of broker lists? A tool to flag scam emails? That's five. Five subscriptions, apps, logins. And most people still have gaps. That's the problem Surfshark solves. It's not a VPN. It's an all-in-one security suite. One tool that protects your connection, your personal info and your devices all at the same time. Go to surfshark.com/tomb or use code TOMB for four extra months of Surfshark, backed by a 30-day money-back guarantee.

• 19:30Speaker 00

  That's S-U-R-F-S-H-A-R-K.com/tomb. Surfshark.com/tomb. Use code TOMB for four extra months of Surfshark.

• 19:48Speaker 00

  Thanks for sticking around. Let's get right back into the action.

• 19:52Speaker 01

  These companies are basically banking on what other reporters accurately call a miracle. The biggest five hyperscalers are trying to promise that their revenue will double over the next three years, while cutting roughly $80 billion in operating expenses. As depreciation expenses soar over the next few years due to all the new buildings and equipment, the expenses spent on sales, general and administrative sectors must go down at an unprecedented rate. As accounting professor at Purdue University, Kevin Koharky puts it, "It just seems like the analysts, in order to keep the operating margins steady or slightly growing, are assuming that whatever depreciation increases by, we're going to see a direct offset through sales, general and administrative." I can't think of a scenario where that's ever happened before.

• 20:35Speaker 00

  Okay, one thing that I think that they're leaving out is what people are really looking at, and this is what Jensen Huang, the CEO of Nvidia, has been trying to get people to understand. Right now, there's a raging debate about what the ongoing value of these data centers are. Now, depending on where you clock that, we're either in starting to get into shaky territory, or we're doing just fine. Jensen Huang's pitch is, listen, guys, you keep looking at the depreciation schedule, and you're like Michael Burry, and you're freaking out, and you think that we've got the wrong depreciation schedule, but the reality is that we have H100s that came on six years ago that are still being used.

• 21:19Speaker 00

  So clearly, these do not depreciate in the three years that everybody's saying. Yes, you're going to get faster chips that will come into the market, and yes, those chips will be better, but you're acting like the old chips just ceased to exist, and that isn't going to be true. And this is why he's putting forward a financing package that would allow people to come in and look at basically the long-term rental of a data center, looking at it as something that generates rent for the long term versus a widget that's going to depreciate very quickly. And he's so confident that that is true, that he's willing to backstop 25% on a deal-by-deal basis.

• 22:01Speaker 00

  But he's willing to backstop, basically put insurance against the decline in value of those chips. He's not guaranteeing the loans, but he's guaranteeing the product. And he's saying that product is, I'm so confident that it will hold its value, or potentially even increase in value, that I'm willing to backstop up to 25%. Now, what he's basing that off of is that the cost of, I can't remember if it's the H100 or the H200, but the cost of that compute per hour hasn't gone down, it's gone up. So now, instead of this being a depreciating asset, which it is because it's technology, so we don't want to confuse sort of this momentary directionality, but as of right now, they're actually being able to charge more over time,

• 22:46Speaker 00

  so the chips are becoming more valuable over time rather than going down. So I think he really does have a very compelling case that the data centers that are being built out don't function in the disaster scenario way that people have been assuming, because this is an unfortunate reality that when you look at historical build-outs of things that have a very expensive infrastructure build-out, what ends up happening is that first wave of investors gets wiped out, because it just takes too long for the technology to come in to start generating money. So take railroads. You have to build the railroads before you can put trains on them, before they're valuable, before people can start using them.

• 23:28Speaker 00

  But the big difference with AI is that, yes, you're still building out the infrastructure, but it isn't like the railroads where there's like zero value, zero value, zero value, then finally it comes on, and all during that zero value, you were just having to pour all of this money into building out the railroad stations and securing the railroad lines and laying the railroad tracks. Yes, that stuff's going to last for an extraordinarily long period of time, but there's this huge delay in getting the revenue. And Huang and other people that are bullish on AI are like, guys, that isn't how AI works. You're getting revenue right now. And not only are you getting revenue right now, and yes, maybe revenue's not coming in as fast

• 24:12Speaker 00

  as we thought it would, partly because of what China's doing with the open source models, but Anthropic, their revenue is growing faster than any company's revenue has ever grown in history. Maybe not as a percentage, but when you look at the raw dollars, it is insane. Insane, the amount of billions of dollars that they're adding to their bottom line, even in just the last five months. So let's all realize that this isn't the railroads, that the technology is being used right now. Yes, this technology will depreciate over time,

• 24:51Speaker 00

  but so far, we've seen them last at least six years, and the cost for compute is going up. Okay, all very important. Now, I'll give you my maybe more balanced take here, which is that, yes, that's true, because we're still in the middle of the build-out, and right now, there's more pent-up demand than we have actual ability to fulfill that demand. However, you aren't seeing a whole bunch of other big companies come online, and that puts question marks. It certainly doesn't put a nail in the coffin, but it asks question marks. Where are the other companies coming from that are going to bring this increased demand? Because what we're probably looking at is it's going to take a while

• 25:33Speaker 00

  for AI to find its way deeper and deeper into the economy. What I always say is for your refrigerator to get smart, your blender to get smart, that stuff is just going to take time. And if we hit that upper limit where there's no longer the growing revenue that we need right now, and you start having an overproduction of compute, now, very rapidly, those older chips are going to decline in value, or if you get China or whoever making algorithms that are more efficient, and so you don't need as much compute for the same output, there are several scenarios that could cause the value of those chips to decline, not just a reduction in demand.

• 26:18Speaker 00

  So you could have still growing demand, but the need for compute to be coming down at the same time to meet that demand, that would cause that value to go down. This is why Jensen obviously isn't backstopping it at 100% of the cost of those chips. He's only backing up 25%. So he's confident, but he's not all the way confident. So anyway, that's a much more balanced take on what is likely actually going on.

• 26:45Speaker 01

  Coming at the same time that we're seeing companies regret cutting their workforces to save revenue, which is causing them to try to rehire all the employees they fired.

• 26:53Speaker 00

  Okay, I hate this argument because it's bullshit. So here is what's actually happening. So you've got these guys, I'm sure there were a certain amount of people getting overhyped, it's like, oh, AI is going to be able to do all of this. AI has limits. AI falls on its face. There are certain things that it does well and certain things that it does terribly. The things that it does well grow by the day, the things that it does terribly get diminished by the day. And so what is likely going to happen with jobs is that some jobs will just be completely wiped out, they will never exist again, no human is ever going to do those things again. Data entry, data analysis, things like that, it's just going to go away. AI is so much better at it. Or if it doesn't go away, the number will be so dramatically reduced, same thing with driving,

• 27:37Speaker 00

  that what will end up happening is like one data analyst will be able to oversee a thousand AI going out, running, you know, getting the data, pulling it in, and they're just there to check to make sure that everything is running well. So instead of it being a department, it's one person and a bunch of AIs. But that doesn't mean that new jobs won't be created. So this story that we're beginning to tell ourselves of like, oh, we thought AI was cool, psych, it's actually not cool, we have to rehire all these employees, isn't true. One, just because you laid off a bunch of people and are now hiring doesn't mean that you're hiring for the same skillset. So you might have fired all the people that were resistant to using AI, now you're hiring people back that have similar job descriptions or even the same exact job title,

• 28:21Speaker 00

  but now you're hiring a different kind of person that's going to scale AI in the role versus people that might be very resistant to that, or just not good at it. They're not a digital native or whatever. So this is where you've got to be very careful about buying into the narrative. If you're not using AI, I think it's very easy to get sucked up in the hype, either that it does more than it says it can, and you just believe it because you're not using it, or the, you know, the doomer hype of like, this is all bullshit, it's not real, this is all going to go away. If you're in the thick of it every day, as I am, you learn very quickly. Yeah, there's a lot of things that it doesn't do well, but damn, like new features, new abilities are coming out all the time.

• 29:06Speaker 00

  It's coming out so fast, it's really pretty extraordinary. So anyway, don't get sucked into this messaging.

• 29:12Speaker 01

  Market research from Robert Half indicates that roughly 29% to 32% of organizations that cut jobs due to early AI projections have already started restaffing those exact roles, which begs

• 29:24Speaker 00

  Again, but potentially with different skillsets.

• 29:26Speaker 01

  the question, what the hell do these companies expect to do to live up to those projections? Despite companies rehiring their employees, finding a job in this market can feel impossible. Whether it's ghost jobs or entry-level jobs going extinct, finding an employer who will actually view your resume and give you a chance is like finding a needle in a haystack. If you're someone who's been struggling to make the advances they want in their career, a career coach on today's sponsor, strawberry.me, could help you.

• 29:54Speaker 00

  We let these play to honor them for the content.

• 29:56Speaker 01

  Using a personalized quiz, strawberry.me connects you with a professional coach that keeps you accountable towards reaching your career goals. And once you're matched with a coach, you will have one-on-one video sessions to assess your current situation. This differs from a typical coach or therapist, because instead of just having you talk about how you're feeling or why you feel stuck, they get straight into producing results for you. Your coach will help you get clarity on where exactly you're trying to go with your career and the right choices you need to make to get there. They'll help you identify and overcome blind spots, close skillset gaps, and create a structured plan. So if you know you have the qualifications to get a better-paying job, but you just feel lost in a sea of online job slop, you can go to strawberry.me/colehastings, which is the first link in the description, and you'll get 50% off your first coaching session.

• 30:41Speaker 01

  At the same time, these companies are rehiring after layoffs, one country is out-AI-ing us by a huge margin with much less cost to the bottom line: China. If there's one thing China's really good at, besides making their cities look like cyberpunk paradises on social media, it's creating scalable, cost-efficient products while accepting thinner margins to gain market share and beat international competition.

• 31:03Speaker 00

  All right. So I'll agree with that. China is extraordinarily good at um. They'll allocate tax dollars to an industry to make sure that it's thriving um and they are very good at optimization. To say that they're ahead of us in AI is not true. Now they're rapidly catching up and this is definitely an area that from a national defense standpoint, we have to stay on our game. Uh but right now it's very clear AI in the US is the absolute frontier models. China is close behind. Creating open source models that can do, let's call it good enough uh for much cheaper. So I would say they're winning the price war, but they're not yet winning the actual frontier war. So it becomes a question of how important is the frontier going to be?

• 31:48Speaker 00

  As we march towards superintelligence, the frontier becomes incredibly important.

• 31:54Speaker 01

  Their open weight AI model could take a large share of the industry soon.

• 31:58Speaker 00

  True.

• 31:58Speaker 01

  Released on July 27th, 2026, Kimi K. Three is an open weight model by the Chinese company Moonshot AI. Now, unlike the closed models from OpenAI and Anthropic, Kimi K. Three can be downloaded, modified, and run on your own hardware without relying on Moonshot's servers.

• 32:14Speaker 00

  Which by the way is amazing and we should want to see more of this. Um unfortunately, it may have a detrimental impact on the US um bleeding edge research frontier model. It may make it harder for them to um generate revenue at the rate that they thought they were going to. But ultimately for people, this is better. For you to um have the ability to get a hold of uh AI for essentially free for whatever it costs you to run it on a local server um is really incredible. And they're already talking about this lowering token cost by I think up to 60% uh so it is very impressive.

• 32:54Speaker 01

  Because of its efficiency, Kimi K. Three can be served at a fraction of the inference cost of many leading frontier models while delivering comparable performance across many benchmarks. Also, can we just pause to laugh at how hilarious graph metrics are sometimes? Like, what the hell does maximizing intelligence per dollar mean?

• 33:13Speaker 00

  Well, I'm glad you asked. It's actually ironically uh one of the most important things in uh the idea of AI. So understanding that we ultimately whoever can get the most intelligence for the cheapest is going to win. So that's the whole point. So I get it seems kind of funny in a spreadsheet, but the reality is um if it's like saying I can get the smartest employee on the in the world for the cheapest and whoever is able to do that, whoever is able to hire the smartest people for the least amount of money is going to have a tremendous advantage in their company. That's just true. So now it becomes a question with the models, how much intelligence am I getting from my money? It's bang for buck just in AI words.

• 33:52Speaker 01

  Anyways, China's models are already starting to comprise a huge chunk of AI traffic. According to OpenRouter, Chinese models went from 1.2% of token traffic in 2024 to more than half of it by the summer of 2026, overtaking American models in total volume. Think of the potential consequences of this. Some of America's largest companies are heavily invested in OpenAI and Anthropic. In Q1 2026, other income accounted for 60% of Google, 51% of Amazon, and 27% of Nvidia's profits. Some of those investment gains come from stakes in leading AI companies like Anthropic and OpenAI. If a huge portion of the market shifts to Chinese models, OpenAI and Anthropic could see their revenue shrink,

• 34:36Speaker 01

  reducing the billions they're currently spending on Nvidia chips, cloud infrastructure, and massive new data centers. But if you want the AI bubble to burst quicker, I mean, that would be a great thing.

• 34:46Speaker 00

  Yeah, this is exactly what we've talked about before. China poses a real threat to the AI industry. If they can start um hollowing out some of the revenue that's coming in that slows this down, we go back to the railroad argument. If the debt is growing faster than the ability to bring in the revenue, the whole industry begins to be a series of question marks around whether they'll be able to get cash flow positive before the debt basically explodes. Uh and if the debt ends up breaking and they need to keep raising money in order to stay solvent, but you can no longer raise debt because the revenues are coming in too slowly. Now all of a sudden, the industry uh stalls out. Now, if it stalls out, that's not necessarily the end of the world.

• 35:29Speaker 00

  It's if it gets to the point where the company is never able to achieve profitability and so they're not able to um actually become a default alive company. That's the question. When you look at the Microsofts, the Googles, they're all cash flow positive. So if they stop the build out, they're going to be just fine. But Anthropic and OpenAI are not. They're nowhere near cash flow positive. So those guys really could implode. But then it goes back to what do you think is going to happen? Will the US government step in or not? Your answer to that question tells you how risk, how at risk you think we are.

• 35:59Speaker 01

  Let's talk about the next threat to AI, increasing negative sentiment from the public regarding anything related to AI. From 2023 to 2025, the percentage of Americans who said the effects of artificial intelligence will have a very or somewhat positive effect actually increased from 15% to 27%. However, in that same period, those who said it will have a very negative or somewhat negative effect went from 40 to 47%.

• 36:24Speaker 00

  Yeah.

• 36:25Speaker 01

  Hmm, I wonder why the general populace hates AI so much. Could it be due to the fact that the people building these models have a poor moral compass and only care about growth and profits instead of helping humanity?

• 36:35Speaker 00

  Not really. I think that uh people put up with an absolute obscene amount of uh uh amoral bullshit and push back almost not at all. What people care about is what is my life like? If my life is good, then I'm good. If my life is bad, now I'm looking for reasons. I think AI induces a level of anxiety that is causing people to be very um unable to view AI from first principles, which is causing a lot of our problems. If you think AI is going to make your life worse, it's going to make it not worth living. You're not going to have any meaning and purpose. You're not going to have economic prospects into the future. If it just creates so much uncertainty about the future that it makes you feel absolutely horrible, you're going to jettison out of that.

• 37:17Speaker 00

  People saying that this is because um uh OpenAI is run by a moral monster, it's not the real reason.

• 37:25Speaker 02

  You know, I think AI will probably like most likely sort of lead to the end of the world, but in the meantime uh there will be great companies created with serious machine learning.

• 37:35Speaker 01

  Could it be?

• 37:35Speaker 00

  Now that's a horrific thing to say. There's no doubt about that. This sounds absolutely asinine, but I don't think that's what's really driving this.

• 37:41Speaker 01

  Due to the fact that working class neighborhoods are being torn down to make way for data centers that produce a ridiculous amount of noise pollution and sap up all the water? Or maybe it's the AI content slop to dropshipping pipeline empires it's enabling?

• 37:54Speaker 03

  What's going on guys? Smith here. Today I'm going to show you guys how videos.

• 37:57Speaker 00

  There there are going to be downsides. Now, again, I'm going to do uh much deeper research into data centers. I'm operating under the belief now that the reports that are coming out from the towns that have actually implemented the data centers, referring to them as a miracle. Uh they generate so much tax revenue that they're increasing job opportunities in the towns. Uh my current standing belief is that they certainly add more value than they take away, but we'll we'll do some deeper dives. Now, AI slop, that is a whole nother thing and people are definitely going to have to find a way to uh parse it out of their lives.

• 38:28Speaker 03

  Like this, had to mute the music for copyright reasons, make me almost $30,000 give or take every single week on average.

• 38:35Speaker 01

  Perhaps it's the way it gaslights or acts like a toddler?

• 38:38Speaker 04

  Hey, can you see me?

• 38:41Speaker 04

  I can't see you just yet. If you'd like me to see what's going on, you'll need to turn on your camera.

• 38:44Speaker 00

  It will get better over time.

• 38:46Speaker 04

  I just turned on the camera. All right, I'm ready. I can see you now. What's going on? What do I look like? I'm seeing a person in front of the camera.

• 38:54Speaker 00

  If you're not looking at your camera, he did not turn his camera on. There was nothing for it to see.

• 38:59Speaker 01

  AI use cases revolve around disintegrating human connection, critical thinking, and creativity.

• 39:03Speaker 00

  This one worries me.

• 39:04Speaker 01

  Perhaps the most egregious use case of them all that really grinds my gears, especially as someone who just published their debut novel, Link in the Description, is the fact that these AI companies are now trying to destroy rare books to train their models. This scandal actually starts last year in AI's first major copyright settlement. Court records first reported by the Washington Post revealed that Anthropic bought millions of physical books to build a searchable database for training its AI models. The internal project dubbed Project Panama relied on a process called destructive scanning. Machines removed a book spine, fed the loose pages through industrial scanners, and then threw away the original copy. Their reasoning? Well, physical, older books are verifiably not AI slop writing.

• 39:49Speaker 01

  And from what we know about AI training itself on its own slop, that makes human written books extremely valuable to them.

• 39:56Speaker 00

  I get why people don't like this, but I have a totally different take on this. We have a problem in America right now where kids can't read. Kids are not running out to buy these books. We do not, we're not at risk of like some uh tremendous wealth of knowledge being lost because the AI is breaking these books. What we are at the risk of is that information being lost forever because it's stored in a format that people don't even engage with anymore. So somehow, some way, getting these books actually recorded to me is far more interesting. And if we're able to get these books recorded in a fashion where you could actually go back and read them later, that would be my preference. But at a minimum, being able to get them into the AI so that the AI can learn from the patterns that are in there,

• 40:40Speaker 00

  you're going to be able to interact with the knowledge and. To me, being able to interact with the patterns that emerge out of that is actually more important and more useful than just being able to go and read that actual book. Now, I would never want to have to choose between the two. I don't want to have to choose between books never existing and me never being able to hear someone's exact words and just the amalgamated patterns that arise out of that. That would be a terrible trade. But if you're asking me if I'm worried that them taking one copy of a book and destroying it, knowing that, yeah, there may be a very limited number of books out there, but they're not destroying all of them. Also, I did some more research on this. This seems to be overblown a little bit. They did, I think it was Anthropic,

• 41:25Speaker 00

  reached out to a bunch of these bookstores. The vast majority of them never even replied to them. There's no evidence that they got millions of books and are destroying these. So, yeah, I think this is probably much ado about nothing, but nonetheless, I wouldn't want these books to get lost to time.

• 41:41Speaker 01

  Anthropic was required to pay $3,000 per book, and the majority of authors they stole from were paid. However, what the court documents didn't reveal was how companies were sourcing millions of physical copies in the first place. One possible glimpse came from a report by 404 Media. ISBNDB, best known for maintaining metadata linked to international standard book numbers, appeared to be positioning itself for the AI boom. The company advertised a service that would help AI developers acquire anywhere from 1,000 to 1 million printed books for large language model training, describing it as being built for the "scale AI demands." The webpages were later taken down, and ISBNDB said the service was never launched. So, we still don't know how exactly Anthropic and other AI labs were acquiring

• 42:26Speaker 01

  these books at scale, but this proposal's existence suggests that there was already a market emerging to supply physical books for AI training. The really nefarious part takes place in the present day. This month, Dutch antiquarian bookseller Peter de Vrij received a strange email from a person named Natalia, who represented a company called 2077 AI. She said the company wanted to place a fairly large order of books and attached a spreadsheet containing more than 3,000 ISBNs, asking for matching titles, price quotes, and shipping estimates to China. At first, the bookseller dismissed the request as spam and barely read past the opening lines. It wasn't until weeks later, after being contacted

• 43:07Speaker 01

  by a journalist investigating the inquiry, that he realized it was connected to the AI industry. He later shared the email and spreadsheet with Fortune, revealing a list of 3,001 mostly academic titles, published between 2020 and 2021 by publishers including Elsevier, Wiley, Routledge, Oxford University Press, and Emerald Publishing.

• 43:27Speaker 00

  So now imagine these are highly academic books that nobody is ever going to read again. They are going to sit and collect dust, and someone is saying, "We're going to turn those into usable knowledge that the whole world can take advantage of." And if these were going to China, odds are they're going to end up in an open source model that everybody will have access to, and we somehow have beef. That's the part that I don't understand. Like, these are people trying to protect something that they never cared about until they heard that somebody else is using it. Now, if they were just burning the books, and they just didn't want anyone to ever have access to them, I get why people would be up in arms. I would be up in arms. But the reality is that we're putting this in a format where we can all take advantage of the information. And I actually reject wholeheartedly the idea that this is, that when you upload something into AI, that you

• 44:12Speaker 00

  are stealing it. There is a way that you can steal it. There is a way that you can get, say, oh god, what was it, Figma? Figma is uploading all of its data into OpenAI or Anthropic, and they're very excited, and, you know, they're working with the code, and they're building something new, and then whichever one of them it was ends up turning that into their own software version and selling it. Okay, cool. That is you literally taking directly that and then republishing a free version essentially of that software? Okay, that's shitty. If somebody took in Stephen King and then republished Stephen King, like, that would be bad. But if you learn from all authors, and listen, I know

• 44:56Speaker 00

  this is happening to me, I know that given the thousands of hours of their content that there is out there of me speaking, I have had other people create tombots of what it would look like to ask me a question, and the answers are pretty close a lot of the times. And it's like, that is what it is, to have access to all of the other information in the world. It is a small trade to say, "All of my old stuff now exists in the same way that any of you could go watch all that stuff if you wanted to, and then learn the way that I answer questions and do, go do the same." That, to me, that is, you used to hear this a lot. You would hear, "Oh, that person is a child of," meaning intellectually. They went and learned from them, and they're basically retelling their ideas. And that's

• 45:40Speaker 00

  how it's worked forever, is we all go learn from the people that come before us. We say we're standing on the shoulders of giants. The fact that we've now made it far more efficient, that we've ingested all of human knowledge and given that access to everybody that has access to an internet connection, is a trade-off that is well worth it. Now, moving forward, I'm the only one that can create new versions of what I would say to do something that is surprising, and so nobody can take that away from you. But anyway, I reject the idea that that's theft.

• 46:07Speaker 01

  According to the bookseller, several other Dutch antiquarian book dealers received the same request and likewise assumed it was a scam. Yep, these companies are no longer pursuing just mass marketed books. They're targeting rare antique books. It almost feels like a second attempt at the burning of the Library of Alexandria out here. AI companies so-

• 46:26Speaker 00

  Again, the knowledge is being preserved. It's not gotten rid of.

• 46:30Speaker 01

  Oh, desperate for valuable human-made data, are resorting to removing antique book spines, feeding the pages through scanners, and throwing away the text, never to be read again by another set of eyes.

• 46:41Speaker 00

  Because people are so busily reading it already.

• 46:43Speaker 01

  Talk of synthetic data replacing human knowledge, the industry's most valuable resource is still authentic human writing, and it's becoming increasingly scarce. That's where things get strange. On one hand, AI companies are scouring the world for ever rarer sources of human knowledge because they still need fresh, high-quality data. On the other hand, investors are betting that AI is on the verge of becoming intelligent enough to improve itself, eliminating the need for humans altogether. Those are two very different stories, and whether the second one actually comes true may determine whether today's AI spending boom ends up looking visionary or like one of the biggest bubbles in tech history. Of course, like all of my videos, I'm gonna try to leave-

• 47:27Speaker 00

  All right, so the incredible thing that everybody needs to put together is that it is very important that you have a sense, when you're thinking about your own portfolio, "What am I going to do with my money?", that you have a better sense of where AI is likely to end up. What are the metrics that matter? So, I think that my view on where we are is somewhere between the pessimism that you saw here, and then somebody like Raul Paul, who just is all bear case all the time. He says AI can't get big enough. There is no amount of money that could pour into it, where he would think, "Okay, that's completely unhinged." This is the single most important technology of all time. It is best thought of as an arms race. That means that the US government will backstop it. The Chinese

• 48:11Speaker 00

  government will backstop it. These are not industries that are going to be left to die in the vine. However, debt matters. The rate at which your revenue is growing compared to the rate at which you're bringing on that debt matters for the people who are invested, not necessarily to the long-term viability of the technology. And the thing that I find really interesting, the thing that has really got me focused right now, is what Jensen Huang is talking about in terms of the fact that, right now, today, anyway, the depreciating schedule of these is not what people think. It is longer for right now. Now, it's probably a little bit more precarious than he wants you to believe in terms of the things that

• 48:53Speaker 00

  could knock it off of that path, but the fact that we have not been able to meet demand for intelligence yet, compute, tells you something, that for all the money that's getting poured into this, it isn't like the railroads where we just have to wait and wait and wait. The revenue is growing at an incredible rate. The usage rate is growing at an incredible rate. As we're using it, innovations are coming out, both from the US and from China. Costs are coming down, but revenues are still growing. Whether that holds in the long term is a bigger question. And so that does raise, the final thing that everybody has to analyze is not are we in a bubble, but where are we in a bubble, in the bubble? I think that's the very right answer. Let's talk about a pattern that is guaranteed to be killing your progress. You know what

• 49:38Speaker 00

  you need to do. You need consistent nutrition. We all do. You need vitamins, probiotics, greens. We all know that we should be doing more of it. When your morning gets chaotic, you skip it. When you travel, you skip it. When your routine breaks, everything tends to break, and that inconsistency compounds against you every single day. AG1 is designed to solve the execution problem. One scoop, eight ounces of water, and you're done. You're getting 75-plus ingredients, vitamins and minerals, pre and probiotics, nutrient-dense superfoods, everything that used to require six, seven different supplements, and perfect planning now happens in one drink that takes about 30 seconds to make. Right now, AG1

• 50:20Speaker 00

  is giving you $87 worth of free gifts with your first subscription. You get a welcome kit, travel packs, vitamin D3 plus K2, and flavor samples. Click the link in the show notes or visit drinkag1.com/impact to claim this offer.