Monologue: NVIDIA's No IT Loads Refused Cash Dump

0:02Speaker 1:Zone Media, Hello and welcome to this week's Better Offline Monologue. I'm your host ed Zeitron. I know some of you are going to say, Ed, why are we hearing a monologue on a Wednesday? I don't like change. I'm screaming, I'm screaming and crying. And the answer is We're going to have a full in studio episode running Friday with Mainstage, Chloe Radcliffe and ed Onnguiso Junior. They're back, folks, and those of you in the New York metro should also get tickets to see closed new stage showed sheet at Joe's Pub. Link will be in the episode. Notes. It's actually fantastic. It's genuinely one of the best things I've ever seen, and I generally don't like going outside anyway. Man, let's talk about this fucking in Video deal. Earlier in the week, various media outlets reported that In Video was lining up five hundred billion dollars in financing for AI infrastructure in partnership with Apollo, Blackstone, black Rock, Brookfield Asset Management and Goldman Sachs, KKR, Copra, Command, Gargamel, Grew and The Minions. Mortimer mcmire from Commander Keane and the ghosts of Margaret Thatcher and Ronald Reagan. Okay, well, technically partnership

1:11Speaker 1:and lined up don't really reflect the truth of the matter, and neither there's five hundred billion dollars for that matter. What actually happened was in video, signed memorandums of understanding, which are not worth the paper they're printed on, with successet managers to maybe build some data centers at some point, using money they might raise. The Wall Street Journal also reports in a paragraph that was since removed from the story that the financing would involve different vehicles rather than one large collaboration, meaning that they'll probably just raise a bunch of data center deals separately in exactly the same way they've already been doing it. If they already funded across the board, They're in core Weave, They're in every goddamn deal. Every single person I've just named, other than obviously the joke Ones, is involved with core Weave or an AI data center of some sort. Jesus not in any of the stories either. In any case, all of this is entirely theoretical, because a memorandum of understanding from video is worthless. In September last year, and Video CEO Jensen Huang, along with open Ai executives Clammy Sam Moltman

2:09Speaker 1:and Greg Brockman, went on CNBC to talk about how it was investing one hundred billion dollars in open Ai to build ten gigawatts of compute capacity and that said deal was monumental in size, even though the deal was based on a memorandum of understanding. In the end, Jan would claim to Taiwanese journalist that this was and I quote never a commitment and that in Video was invited to invest one hundred billion dollars and they were happy and honored as a result. In the end, In Video invested thirty billion dollars in open Ai six months after the announcement of the original deal, and open Ai built zero gigawats of data center capacity as a result. Pobody's an effect. You know, just these things happened, Things moved crazy.

2:48Speaker 1:You know, seven seventy billion dollars didn't happen, Ten gigawatts of compute capacity didn't get built, nothing happened, just completely fucking forced. They did an interview on CNBC, just like they did with this deal. I feel like I'm going insane anyway, on Monday night, one went on Twitter and delivered a seemingly AI written screed or maybe he's just a shitty writer about this theoretical partnership, adding that in video would and I quote, provide a residual value support mechanism for up to twenty five percent of an opportunity, as says carefully on a project by project basis. And if you're wondering what that means, that makes two of us.

3:23Speaker 1:That was very clearly collect CRES, the CFO and the Legal Council just being like Jensen, you will fucking run these You will run this goddamn sentence. I will. I'll get the club Jensen. Anyway, if I had the guess that twenty five percent refers to in video guaranteeing a certain amount of compute revenue for the company building the project, as in like, they will guarantee that, say, if open AI I can't afford to pay, they'll cover the costs up to twenty five percent, which would in turn allow this developer to raise billions of dollars to buy in video GPUs. This is actually very common. It's how Calways raised a bunch of debt. They go to the various financiers and they say, hey, look, I've got a contract.

3:59Speaker 1:I got contract from a customer. Sometimes that customer is in video, by the way, and the bankers go, oh shit, you got a guaranteed contract. Huh, I'll give you the money for this. Hey, yeah, you look like you're running out of money and we're going to need to do nine percent interest. But fuck me up, Grandpa. I should also add that none of this is actually real until actual contracts assigned and actual money is raised, and that Jensen Wong's word is effectively worthless. At this point. In the Twitter post, Huang also added two questions to I imagine alleviate concerns. He responded to the question of whether this was circular financing by saying that this initiative was built to address that concern, bringing independent, long term institutional capital into the AI infrastructure market, before adding that the demand is real and it comes from Frontier AI Labs, followed by a series of other companies that do not have significant demand for AI compute, ending by saying that this was the beginning of an open capital market for AI infrastructure, which means nothing. Just to be clear, this

4:57Speaker 1:is absolutely circular financing. Nvidia is helping companies raise debt, offering a guarantee on that debt to make sure that the debt is issued, also that Nvidia can be paid using the proceeds from that debt, which in Vidia is guaranteeing. It's customers are companies that can quite literally not afford to buy its products, mostly because they lack an affected business model or the cash flows to do so. I'll add that it's also laughable to read someone say that the demand is real as they attempt to raise half a trillion dollars explicitly to create demand. Yet my favorite palm was Jensen's last question, where is the return on investment?

5:34Speaker 1:His answer was, I shit you not that the return is the usefulness of AI. Fuck off man, fuck you for shut the fuck up. Fuck fucking motherfucker. That is not what return on investment refers to. And you know it. Apparently the usefulness of AI is not significant enough to provide cash for your customers to pay you. Outside of financial wizardry. That would make Anti Fastau blush. If there were actual diverse demand for Invidia's GPUs, commit nswer it with analystic expectations, you wouldn't have to do these bizarre, painfully circular deals that exist only to inflate your fucking revenues a further prop up the existence of unprofitable AI labs.

6:09Speaker 1:You know this, Jensen, You're full of shit. Anyway, If Jensen Huo actually listens that, I would be really surprised. If you know, Jensen want to play this for him, I really am just quite excited for it to eat. He's never gonna hear it. He has no idea. Who am who cares though? Anyway? Anyway, this deal is about as real as Stargate and will work out much the same way. For those of you who forget Trump, Altman, Ellison and Massioshi Son announced in January twenty twenty five that they would be building five hundred billion dollars worth of AI infrastructure, even though no actual money was raised, no partnership was created, and genuinely no LLC was ever formed.

6:47Speaker 1:In the end, Stargate was a name that got put on any open AI data center, no matter how it was funded, and mostly referred to Oracle's massive build out for the least viable customer of all time. It was just a chance for Trump to get up their post inauguration and go, We're gonna build the biggest buildings. It's my beautiful friend Larry, and then nasty friend Sammy is a nasty clammy little man, And yeah, I could do that for an hour. I'm just gonna stop myself. In Jensen's case, I expect further data center deals to get funded by these asset managers that get referred to in brief as part of the five hundred billion dollar partnership within Video, and I expect the media to dutifully say that too. They will help ratify this as they did with Stargate, even though not a single data center being built is part of a Stargate initiative of any kind, because one does not exist. Even though I should be clear, Jensen WLL did these deals and nothing has functionally changed.

7:37Speaker 1:And Video has already been offering preferred pricing and guarantee compute spend to influence providers. It already discussed backstopping open AI data centers, it already backstopped one of Corweav's leases, and it will continue to do whatever it can to keep growth going and to sell more GPUs. Now, if you wondering why they're doing all this craziness, it's because consensus analyst estimates for the revenue for the next three years are utterly fucking insane. And Video made about two hundred and sixteen billion dollars in fiscal year twenty twenty six, which ended on January twenty fifth, twenty twenty six, frustrating, it's really annoying, and analyst expectations have it at three hundred and ninety three point seven billion dollars in fiscal year twenty twenty seven, a little under doubling its revenue year over year. In fiscal year twenty twenty eight, expectations are five hundred and sixty five point seven billion dollars, and in fiscal year twenty twenty nine, and Video is expected to make six hundred and ninety four billion dollars.

8:31Speaker 1:Over ninety percent of its revenues come from selling GPUs and other data center hardware, so nothing else is going to make up the shortfall here, And I realized that's a lot of numbers, so I'll put it real simple in Video is expected to sell over one point six trillion dollars worth of GPUs by January twenty twenty nine, and the vast majority of its customers, including hyperscalers are having to take out hundreds of billions of dollars worth of debt to pay for it, if they can even raise it. As I discussed last week, seventy percent of Microsoft, Google, and amaz On's AI revenues come from open AI and Anthropic, meaning that the demand does not exist that remotely the scale necessary to justify this AI data center capex Jensen Hwang has no choice but to engage in the darkest arts of circular financing to keep up with investor expectations.

9:15Speaker 1:But honestly, I think everyone's kind of running on empty. Anthropic is apparently having to reassure investors about the dangers of open source competition as it rushes towards the September or October IPO adding that it's going to do stuff in biology to deal with the negative perspective that people might have on AI. Fucking laughable. Open Ai just lost COO brad Lightcap a day after completing a seven billion dollar internal stock buyback, which is peculiar. When a company does an employee share sale, which is when insiders like people who worked there for years, sell their stock, they usually sell to external investors rather than by the company's cash buying it. In fact, I looked around and I really struggled to find examples of this. Even SpaceX found dupes to sell it to, and every other up and ai internal share sale involves selling to Thrive Capital or SoftBank. It' is Masayoshi Son not quite as horny? Is he not into it anymore? He's not getting his rocks off? Is this not the golden goose anymore? Masayosi Son? What's going on?

10:14Speaker 1:Massa Yoshi Son? Why didn't you invest more? And I realized by the way this share sale was at the eight hundred and fifty two billion dollar valuation from March, so maybe they didn't want to double down. I don't know if you think this company's actually going to go public. It feels strange for you to sell at this point, unless, of course you don't. But yeah, things are about to get a little desperate. The five hundred billion dollar deal didn't move the markets much, with Nvidio down about half a percent about an hour before market close, which means investors are no longer impressed by Jensen Wong's fantastical promises.

10:48Speaker 1:And it worked really well last year. Out of last year, you could just say Yeah, we're gonna sell a gazillion GPUs to this guy, and the markets would give you five ten percent of the day. Not so much anymore. The CON's not work him, and I'll be here to sheap with you through wherever the con goes next and however insane it becomes as a result. Thanks us ever for listening.