It's not the traditional dot bubble, right? Because back then there was companies going public getting crazy valuations and people are buying them and the stock would go up, you know, 50% 100% with companies that had no revenue, no traffic, no nothing. and you'd go get a cab back then and people would be talking about them. And you don't you don't see that at all today. So, it's not a bubble that's going to impact most people in the room, right? Or most people um across the US, but it could just destroy a lot of VCs and a lot of funds and a lot of PE, right? Because they're going all in.

Cuban draws a precise structural distinction between the dot-com bubble and the current AI bubble, arguing the damage will be concentrated among institutional investors rather than the general public.

What's happening now is the market leaders, Google, etc., Meta, they're borrowing hundreds of million, billions of dollars. And there's already a private credit problem right now, right? So you you you just layer on private credit like um Al Capital getting all the um refunds and then you you know, you have these huge companies that have cash flow, but there's, you know, they're they're spending all their cash flow on capex and then they're borrowing on top of that. bonds, right? That's planning for perfection. And we're building these data centers and you know, if there's a price performance curve on AI that minimizes the power requirements, there's going to be a lot of data centers that are going to be turned into pickleball courts.

Cuban flags a compounding financial risk — free-cash-flow-eating capex plus new debt — and colorfully predicts data center overbuilding could parallel the dark-fiber glut of the early 2000s.

There's going to be breakthroughs in technological breakthroughs as well. Just like we saw fiber back in the day, it was all about putting in fiber. Then it went from 1 GB fiber to 10 to 100 gigabyte and then there wasn't a fiber problem anymore. There wasn't a bandwidth problem anymore. And just sitting there, right? And how is it not going to be the case that we don't get the same price performance improvements on the AI side and on the data center side?

The dark-fiber analogy is a historically grounded warning that the current AI infrastructure buildout may become stranded assets if efficiency gains outpace demand growth.

In this particular bubble, because it's so driven by private capital, there should be a lot more companies going public, not at the SpaceX level for not open AI, not anthropic, but the hundred million dollar um IPO. Because if AI does what AI knows, we all know what it'll do in terms of disruption. Then you want to have um some sort of um currency that allows you to buy all those companies. Well, if you don't have that currency, the stock is currency, you're going to have to go out and raise money to do it. And if anything happens, and that money is not cheap, that money is really, really expensive.

Cuban makes the strategic case that AI-era companies should go public earlier to use stock as acquisition currency before legacy businesses become cheap targets.

I'm like trying to tell my um my portfolio companies, go public, motheruckers. Go public. They just don't think that's the right thing to do. Well, and M&A is, you know, not to make this political, but you know, when you're an entrepreneur, you have to play the game on the field. M&A was off the table. Lena Khan had a certain perspective, which was we have to be like precogs in Minority Report. We have to predict who's going to be a monopoly in the future and stop them.

Cuban bluntly argues that the FTC's precautionary anti-merger stance under Lena Khan functionally denied startups an exit path and acquisition currency for years.

If AI does what we know it will do the companies that are disruptive don't have to be enormous companies. So you don't want to have to always raise cash to go out there. Because like back in the day with broadcast.com we bought like five companies just for stock and then there was always a bigger fish to come along.

Cuban illustrates that serial acquisition via stock — not cash fundraising — is the playbook for AI disruptors who want to scale without diluting themselves through private rounds.

You know I collared my stock because how rich do I need to be, you know, and if you if you're able to change your life. If I work for any SpaceX, any of them, whatever, I'd be like, you know, somebody put together a collar for me. Yeah. You know, because I just need to be protected, save part for my upside, but just cover my downside. No, it didn't. So, I had to actually create an index of internet stocks. I had Goldman Sachs create an index of internet stocks that I thought sucked. And then I shorted that index.

Cuban's vivid account of improvising a protective collar on his Yahoo shares is a masterclass in asymmetric risk management that every employee holding illiquid private-company stock should understand.

AI is a lot harder to implement than anybody expected, right? You know you can do an agent pretty straightforward right you can prompt away you know we cheated on tests we you know cheated at work you know did projects improved productivity 100x, right? All easy peasy. And we just assumed, you know, at the enterprise it'd be just as easy. Yeah. And it's terrifying. And not terrifying for employees because Dario and everybody saying 50% of white collar people are going to lose their jobs. Here we are two years later, they said within two years and you know, employment still growing, people are hiring, we need more AI literate people, right? So CEOs have no clue what is going on. None whatsoever.

Cuban punctures both the hype that AI would rapidly eliminate white-collar jobs and the assumption that consumer-grade AI ease of use translates to enterprise implementation.

If AI, we're talking about, you know, AGI and we're talking about taking over the world. If you can't you if you need to have forward deployed engineers that tells you all you need to know about AI because by definition you should just be able to ask AI to do what I need you to do and ask and tell me how to implement it right yet here you have Microsoft hiring 6,000 people right you have anthropic open AI all saying they're going to deploy to these companies which tells you AI is hard and then you have um Alex Karp from Palunteer freaking out saying how can you give your alpha to these companies when in my opinion he's just saying they're doing what we're doing right we're we're all about forward deployed engineers at Palunteer and they're doing the same thing.

Cuban uses the proliferation of human forward-deployment teams as empirical evidence against AGI-level capability claims, turning Big Tech's own hiring against its marketing.

But try going into cloud or or chat GPT and saying okay I just want you to do this search for Jason Calacanis investments and I want you to do report and I want you to email it to me every week. It can't do it. And then it says, would you like for me to do an an agent that gives you alerts and then you say, Sure. And then you have to know how to program because it either gives you a JSON file or it gives you, you know, code and it comes out slop. Yeah. And then you have to correct, right? You have to reiterate. But AI is not going to, you know, take away 50% of the jobs. There's just so much it can't do that regular people need it to do.

A concrete, relatable failure case that exposes the gap between AI's perceived capability and what non-technical users can actually accomplish with it today.

And when you have a narrow data set like code or legal, tax, it's magic and it's just math, right? Basic data, right? But when you wanted to do when normal people anywhere in the world want to use it for normal stuff, great. You start a business, great. But if you want to start getting advanced, it's like figuring out PowerPoint used to be or Excel, right? You always have to have your Excel expert, you know, or your PowerPoint in your company. Yeah. Or whoever that you knew, right? or there would be companies built um I even invested in like a company that slides share that all they did was have PowerPoint templates that you could download and redo. AI is not even that advanced for, you know, once you get to the second level. And so that creates so much opportunity for anybody to walk into a small, medium size, even large business and say, Hey, I understand AI.

Cuban reframes AI literacy as the new Excel expertise — a skill gap that creates a massive human consulting and service opportunity, not just a displacement threat.

But even better, if you're thinking as an entrepreneur to start a business, those things will eventually break. Agents get bored, right? and they drift because as the underlying um large language model starts to change, the way it was originally programmed doesn't match what the the large language model turned into, right? You see what I'm saying? And so you and you even wrote something about this, right? Where it's taking more people to manage all this stuff.

Cuban identifies 'agent drift' — model updates silently breaking deployed automations — as an underappreciated maintenance cost that creates a new category of business opportunity.

I've got lovable, I've got synthesia.io that was the first investor in 10 years ago, maybe that that are just that's just killing it, right? I've got AMI, which is Yakun's world model. Yeah, we haven't even talked about world models versus LLM and Transformers, right? Because everything we do is built on text and pictures. Nothing's going to be text and pictures in 10 years, right? And but what's going to drive it? And like people say, well, AI is so smart. It's going to change everything. I'm like, Okay, if you tell AI, if you show AI a video of a two-year-old on a high chair with a sippy cup, the a the two-year-old knows if you push the sippy cup over the edge, mom's going to come running and the AI's the kid's going to start laughing. AI's got no clue what's going to happen.

Cuban uses a simple toddler thought experiment to illustrate the profound gap between LLM pattern matching and genuine physical-world intuition — the core unsolved problem for AGI.

Would you ra if you're at a corner blindfolded and you have to cross the street, would you rather have a video? Would you rather have your phone with AI or would you rather have a CNI dog? Yeah, dog. I'm taking the dog every time. Right? Yeah, it's not ready, right? Yeah, it's not ready. So, just think about how far we have to go.

The guide-dog test is a memorable, jargon-free benchmark for real-world AI readiness that cuts through capability hype instantly.

The thing that I think will save us as a world more than anything else in terms of um information availability and reducing the information asymmetry as it applies to politics are large language models because large language models have to be as literate or and literal and honest as they possibly can. Truth seeking. Yeah. Truth is a better way to put it. Yeah. So they they have to seek truth otherwise you'll lose trust in them and the last thing you know claude open AI needs is well you know they're lying their ass off about this.

Cuban makes the optimistic counter-argument to social-media pessimism: LLMs are structurally incentivized toward truth in a way that engagement-driven algorithms are not.

You're not going to get rid of social media but I think people as they become more uncertain with their politics are going to go more and more and more to large language models and say, who should I vote for? And the large language model is going to come back and say, well, what do you think about this? What are your interests in this and da da da? And they'll give you honest answers. Yeah. And if you ask it, hey, what's a reasonable immigration policy? It'll give you a reasonable immigration policy.

Cuban predicts LLMs will become a key civic decision-making tool, offering a Socratic dialogue model that could counteract tribal algorithmic reinforcement.

95% of medicine is guessing, you know, because you can't. There's no way a doctor can memorize all the new that happens every single day. You just can't. But a doctor using these tools having, you know, the empathy and the ability to communicate and the ability to see. Right. Because again, there's you can't see. You know, you could have doctor, you know, why am I bleeding? You got a blood looks like you got a gunshot wound, right? AI is not going to tell you that, Dwayne.

Cuban articulates the durable human comparative advantage in medicine — physical presence, empathy, and sensory observation — while acknowledging AI's real value as a knowledge augmentation tool.