Monologue: The AI Industry Is Lying To You

This is an iHeart podcast. Guaranteed Human. Hello and welcome to this week's Better Offline monologue. I'm your host Ed Zittron. Now I'm turning 40 in a month or so, and at 40 years young, I'm old enough to remember as far back as December 11th, 2025, when Disney and OpenAI reached an agreement to bring beloved characters from across Disney's brands to Sora, OpenAI's video platform. As part of the deal, Disney would, and I quote, become a major customer of OpenAI, use its API to build new products, tools and experiences, as well as showing Sora videos in Disney Plus, and deploy chat GPT for its employees, as well as making a $1 billion equity investment in OpenAI. Except now Sora is dead, shot in the head by Clammy Sam Altman and OpenAI, because it cost $15 million or more to run, and had everybody making videos of Goofy saying, GORISH, HELLFUCK ANYTHING THAT MOVES, LIKE FRANK FROM BLUE VELVET. To be clear, OpenAI is shutting down both Sora, the app, and access to the Sora video model through its API, framing the changes, focusing the company on business and enterprise customers, which is a euphemism for a product that never had any real business model or product market fit, other than burning millions of dollars of compute today and violating copyright. I think it's also completely insane that nobody bothered to check whether Disney had ever invested, or even talked about investing, that billion dollars. I went and had a look, and there wasn't a damn thing in any of Disney's most recent quarterly or annual reports. I should have checked this myself, but it's wild that nobody else did either. There are business and tech reporters. There are people that do this full time. I... But I also want to take this moment to say something, for the bottom of my heart. I was right. I was fucking right. I said it back in October. I said it back in 2024. Sora was never a business. Sora was never replacing anything. going to work, everyone who said otherwise is a god damn rube and so many people were wrong. The app that CNBC said was challenging Hollywood and freaking out the movie industry and the Hollywood reporter would suggest could somehow challenge Pixar and was Sam Malt when successfully playing Hollywood and the fucking Ankhlo who should be ashamed of themselves said was open AI going to war with Hollywood as it shook the industry and deadline said made Hollywood soar and boardroom said was in a standoff with Hollywood and the LA Times said was deepening a battle between Hollywood and open AI and igniting a firestorm in Hollywood and Puck said had Hollywood panicking and Techno Lama said was the end of copyright as we know it and Slate said was a case of AI crushing Hollywood as it as we've known it and that I swear to god I wrote that right in the script but I'm keeping going. Anyway this thing is completely dead a little more than 5 months after everybody claimed it was changing everything. You all were wrong every single god damn wrong wrong wrong wrong wrong on the wrong a sandwich of wrong that you're shoving up your asshole I'm so fucking tired of this. I cannot express how angry I am. Because it's almost as if everybody making these proclamations was instinctually printing whatever marketing copy had been imagined by the AI labs to promote compute intensive vaporware and absolutely nobody is going to apologize to the people working in the entertainment industry for scaring the fuck out of them with ghost stories for no apparent reason. Every single person who blindly repeated that Sora existed and was changing everything should be forced to apologize to their readers. I cannot express the sheer amount of panic that's spread through every single part of the entertainment industry as a result of the specious poorly founded mythology spread by people that didn't give enough of a shit to understand what was actually going on. Sora 2 was always an act of desperation, an attempt to create a marketing cycle to prop up a tool that was burning as much as $15 million a day, the most of the mainstream media bought into because they believe everything open AI says and are willing to extrapolate the destruction of an entire industry from a fucking facade driven by a clammy man that lies for a living. I am tired of everybody scarfing down the slot from these companies. I'm tired of the majority of reporters not doing the hard work to actually understand what's going on. And good lord, do I not have another example? Well I do. You get what I mean. I spent the last week or so trying to work out how many actual data centers got built last year and found that despite somewhere between 190 and 240 gigawatts of supposed capacity being announced, based on my analysis only 3 gigawatts of IT load, which is the GPUs in the associated hardware that's being turned on, got built in America in 2025. To explain how bad this number 3 gigawatts of critical IT hardware is about $90 billion per Jerome Darling of TD Cohen, an Nvidia sold about $135 billion worth of GPUs and hardware to America in its last fiscal year. Based on the last quarter of sales, which is in the $62 billion range, 69% of that nice is America, it now takes about 6 months to install a single quarter's sales worth of GPUs and associated hardware. And if you think AI is the biggest and most hugest and most special boy, what's the fucking point of buying these GPUs to 4 years in advance? Because even if it takes 6 months to install a single quarter's worth of sales, from what I've found it's actually taking a lot longer in some cases. Jensen Huang is announcing a new GPU every year. By the time you install your Blackwells they'll already have Vera Rubin. By the time they have Vera Rubin they'll have something else. I don't know, Daphne from Scooby Doo? They're gonna go… No. The only cool thing that Nvidia does is choosing the names of people. Anyway, with every new generation of these GPUs comes new power requirements and rack sizes. While the first generation of Vera Rubin, which is the next gen coming out very soon, will fit into the current Oberon Nvidia racks, Nvidia is already teasing its kyber racks which will require even more power, even more liquid cooling, which will mean those ouch, ouch, ouch, those old stinky dirty Blackwells, you're gonna have to throw them right in the trash. No more Blackwell for me, thanks. They've got Vera Rubin in town. Ewww, you're gonna make me use Blackwell? How gross! Ewww, Jensen put those stinky Blackwells in the toilet! Ewww, you couldn't fit in there far too large, I've tried. Yet things only get worse when you look at the actual construction in process. Per site land climate, despite over 190 gigawatts of global capacity that's supposedly in process, with 16 gigawatts that's meant to come online this year, only 5 gigawatts of it is actually under construction. And under construction can mean everything from a single steel beam to a nearly finished build out. Walk around London sometime, you should just look at building projects in London or New York, look at how quickly or not quickly those get built. Well, I don't know, go look around Stargate and all the data centers there. One of the infrastructure people from OpenAI was boasting about the Wisconsin data center the other day and he just had a single steel beam. Not even fucking with you, it was just a one beam and there were like 4 guys working on the beam and another guy walking up to it. I love builders. But anyway, in my premium this week, just hit my microphone, just going to keep going. I also dug into most of the major data center projects that have been announced and in basically every case found that they had either barely broken ground, hadn't actually picked the land yet or had stalled entirely. As was the case with Fermi's supposed 11 gigawatt data center project in Amarillo, Texas, which abruptly dismissed workers in February after moving forward without permitting or funding necessary to actually build the fucking thing. That's a public company by the way Fermi. This is Rick Scott. God. It's like the Avengers of assholes. Despite the many, many statements that AI is inevitable or unstoppable, the reality is that the things aren't getting built, billions of dollars aren't getting sent, industries are left undisrupted and the only people actually benefiting are grifters venture capitalists and banks providing debt to non-existent data center projects. Like I believe by the end of this era to quote Justario, one him and Kikashi, the lone gunman, these guys are two of the best out there and their pseudonymous accounts, this era will be remembered as the largest waste of capital in history. Most data centers aren't getting built and those that are won't be ready before 2027 at best. Most AI startups make piddly amounts of revenue and burn hundreds of millions of dollars to make single digit millions of dollars of revenue and constantly piss off their users by changing rate limits cause every single one is subsidized. Every single one is spending a couple of dollars or more per dollar of revenue and nothing is changing and nothing is improving. Fuck off with your whole inference is profitable thing by the fucking way. To paraphrase my good friend Casey Kagawa who made this point to me over Signal fairly recently, either inference isn't profitable or API calls are just a big, they're the biggest rip off ever that they could be charging even less than that. Indeed, if inference was so profitable, why are they charging a couple of bucks for a million dollars worth of input tokens, they should be charging pennies. Why are they charging so much if it's so profitable? Take that Aetherdits. Anyway, and for the most part the media has failed to notice all this stuff. I really must be clear how shocking this is. I really have read all over the shop. I still see people making the Amazon web services lost a lot of money argument. That's about $38 billion in over a decade that was spent on Amazon web services. So walk that one back will you. I don't know. I still see the same talking points peddled and it's just ridiculous. Big numbers get quoted without a second thought unrealistic or fantastical timelines of published ad verbatim. I'm talking about that. What 10, 20 billion dollar nebius deal with Meta that just data center doesn't exist. Haven't even raised the money. And scrutiny appears by the way to only be reserved for haters or skeptics or those who say fuck too much on a podcast. Well the scrutiny machine over here is just getting started and you know, you just know that I will be absolutely insufferable when I'm finally proven right. Thank you all for listening. Should have some fun stuff next week. I'm still working on. I love you all. This is an I heart podcast. Guaranteed human.