1 00:00:02,440 --> 00:00:03,199 Speaker 1: Zoe Media. 2 00:00:04,400 --> 00:00:05,360 Speaker 2: All Right, we're back. 3 00:00:05,440 --> 00:00:20,000 Speaker 3: I'm ed Zetron and this is better offline. This is 4 00:00:20,000 --> 00:00:22,280 Speaker 3: the fourth and final episode of this four part series 5 00:00:22,320 --> 00:00:26,480 Speaker 3: where I dissect in excruciating comprehensive detail the AI bubble 6 00:00:26,520 --> 00:00:28,840 Speaker 3: and tell you why I think its implosion is inevitable. 7 00:00:29,480 --> 00:00:31,520 Speaker 3: We're at the end, which is fitting because in this 8 00:00:31,600 --> 00:00:33,559 Speaker 3: episode we're going to talk about why and how this 9 00:00:33,760 --> 00:00:37,440 Speaker 3: entire sham dies again. If you jumping in now, start 10 00:00:37,440 --> 00:00:39,080 Speaker 3: from the beginning. We talk about a lot of things 11 00:00:39,120 --> 00:00:40,920 Speaker 3: and it all kind of threads together. I know, I 12 00:00:40,960 --> 00:00:42,720 Speaker 3: know you want to hear my explanation of how open 13 00:00:42,760 --> 00:00:45,080 Speaker 3: ai is fucked with a capital fuck. But we've got 14 00:00:45,120 --> 00:00:48,160 Speaker 3: to lay the groundwork to get there. Okay, got up, good, Good, 15 00:00:48,240 --> 00:00:51,279 Speaker 3: We're good. We're god, You're ready, We're good, You're got yep, yep. 16 00:00:51,200 --> 00:00:52,040 Speaker 2: Okay, we'll begin. 17 00:00:52,720 --> 00:00:55,200 Speaker 3: One of the comfortable lies that people tell themselves is 18 00:00:55,240 --> 00:00:57,200 Speaker 3: that the AI bubble is similar to the fiber boom 19 00:00:57,240 --> 00:00:59,279 Speaker 3: or the dot com boom or uber, or that we're 20 00:00:59,280 --> 00:01:01,800 Speaker 3: in the growth stay, that this is what software companies do. 21 00:01:01,880 --> 00:01:03,720 Speaker 3: They spend a bunch of money, then pull the profit lever. 22 00:01:04,720 --> 00:01:07,560 Speaker 3: The thing is, this is nothing like anything you've ever 23 00:01:07,600 --> 00:01:09,880 Speaker 3: seen before, because this is the dumbest shit the tech 24 00:01:09,920 --> 00:01:13,959 Speaker 3: industry has ever ever done. AI data centers are nothing 25 00:01:14,040 --> 00:01:16,679 Speaker 3: like fiber because there are very few actual use cases 26 00:01:16,680 --> 00:01:19,200 Speaker 3: for these GPUs outside of AI, and none of them 27 00:01:19,200 --> 00:01:22,840 Speaker 3: are remotely hyper scale revenue drivers. As I discussed a 28 00:01:22,840 --> 00:01:25,400 Speaker 3: month or so ago, data center development accounted for more 29 00:01:25,440 --> 00:01:28,520 Speaker 3: of America's GDP growth than all consumers spending combined in 30 00:01:28,560 --> 00:01:29,280 Speaker 3: the first half of. 31 00:01:29,240 --> 00:01:30,119 Speaker 2: Twenty twenty five. 32 00:01:30,360 --> 00:01:32,679 Speaker 3: And there really isn't any demand for AI in general, 33 00:01:32,760 --> 00:01:34,800 Speaker 3: let alone at the scale that these hundreds of billions 34 00:01:34,840 --> 00:01:37,960 Speaker 3: of dollars are being sunk into. The conservative estimate of 35 00:01:38,040 --> 00:01:41,080 Speaker 3: capital expenditures related to data centers is around four hundred 36 00:01:41,080 --> 00:01:43,760 Speaker 3: billion dollars, but given the fifty billion dollar in a 37 00:01:43,880 --> 00:01:46,240 Speaker 3: quarter in private equity invested, I'm going to guess it 38 00:01:46,240 --> 00:01:48,560 Speaker 3: breaks half a trillion, all to build capacity for an 39 00:01:48,560 --> 00:01:51,680 Speaker 3: industry yet to prove itself. And this whole in video 40 00:01:51,720 --> 00:01:54,680 Speaker 3: OpenAI one hundred billion dollar funding news should only fill 41 00:01:54,720 --> 00:01:57,360 Speaker 3: you full of dread. But also it isn't fucking finalized. 42 00:01:57,360 --> 00:02:02,240 Speaker 3: Stop reporting as if it's done, it's fucking anyway good. 43 00:02:02,680 --> 00:02:05,360 Speaker 3: According to CNBC and I quote the initial ten billion 44 00:02:05,400 --> 00:02:07,760 Speaker 3: dollar tranch is locked in at a five hundred billion 45 00:02:07,760 --> 00:02:10,200 Speaker 3: dollar valuation and expected to close within a month or 46 00:02:10,200 --> 00:02:13,359 Speaker 3: so once the transaction has been finalized, with successive ten 47 00:02:13,360 --> 00:02:16,240 Speaker 3: billion dollar rounds planned, each to be priced at the 48 00:02:16,240 --> 00:02:19,040 Speaker 3: company's then current valuation. Is new capacity comes online, what 49 00:02:19,080 --> 00:02:21,440 Speaker 3: a horrible deal. They can just raise whatever they can, 50 00:02:21,560 --> 00:02:24,640 Speaker 3: just go back to video. But let me just be clear. 51 00:02:24,919 --> 00:02:26,919 Speaker 3: Open Ai has written a lot of checks that neither 52 00:02:26,960 --> 00:02:28,720 Speaker 3: it nor anyone else can cash, and for it to 53 00:02:28,720 --> 00:02:31,720 Speaker 3: survive it needs to raise more than, honestly about five 54 00:02:31,800 --> 00:02:35,240 Speaker 3: hundred billion dollars and maybe another four hundred billion dollars 55 00:02:35,280 --> 00:02:38,800 Speaker 3: and other people paying for stuff. It's astonishing, really, and 56 00:02:38,880 --> 00:02:42,120 Speaker 3: no point is anyone asking how exactly open ai builds 57 00:02:42,120 --> 00:02:44,160 Speaker 3: the data centers to fill for of the GPUs to 58 00:02:44,160 --> 00:02:46,840 Speaker 3: get the rest of Nvidia's funding. In fact, I'm genuinely 59 00:02:46,840 --> 00:02:49,160 Speaker 3: shocked and a little disgusted by how poorly this story 60 00:02:49,160 --> 00:02:49,800 Speaker 3: has been told. 61 00:02:50,120 --> 00:02:51,120 Speaker 2: Let's go point by point. 62 00:02:52,040 --> 00:02:54,160 Speaker 3: Ten billion dollars is not enough for open ai to 63 00:02:54,160 --> 00:02:56,480 Speaker 3: build a data center. The one point one to one 64 00:02:56,480 --> 00:02:59,920 Speaker 3: point two Gigawa Gabiline Texas data center being built by 65 00:03:00,040 --> 00:03:02,960 Speaker 3: Oracle and Cruiser for open Ai required fifteen billion dollars 66 00:03:03,040 --> 00:03:05,720 Speaker 3: in debt, and that's not including the forty billion dollars 67 00:03:05,760 --> 00:03:08,079 Speaker 3: of chips needed to power it. Though a question whether 68 00:03:08,120 --> 00:03:10,120 Speaker 3: all of those are going into Abilene or somewhere else, 69 00:03:10,120 --> 00:03:13,640 Speaker 3: which you get to later. Open Ai can also not 70 00:03:13,720 --> 00:03:16,560 Speaker 3: afford to pay for everything it's promised. Open Ai will 71 00:03:16,560 --> 00:03:18,799 Speaker 3: burn they say one hundred and fifteen billion dollars in 72 00:03:18,800 --> 00:03:20,680 Speaker 3: the next four years ago into the information, but I 73 00:03:20,720 --> 00:03:23,560 Speaker 3: believe the company intentionally leaked those numbers before the announcement 74 00:03:23,560 --> 00:03:26,600 Speaker 3: of their three hundred billion dollar deal with Oracle, as 75 00:03:26,639 --> 00:03:28,359 Speaker 3: when you take into account the numbers share but the 76 00:03:28,440 --> 00:03:31,360 Speaker 3: information which involved them burning forty seven billion dollars in 77 00:03:31,360 --> 00:03:33,280 Speaker 3: twenty twenty eight a year when it open ai is 78 00:03:33,280 --> 00:03:36,120 Speaker 3: meant to make seventy billion dollars in payments to compute 79 00:03:36,160 --> 00:03:38,840 Speaker 3: to Oracle, not for twenty eight billion dollars to Microsoft 80 00:03:38,920 --> 00:03:41,240 Speaker 3: or any other cost. There's no space for the three 81 00:03:41,320 --> 00:03:44,320 Speaker 3: hundred billion dollars to go, and there's actually another one 82 00:03:44,400 --> 00:03:47,280 Speaker 3: hundred billion dollars they're promising for backup compute as well. 83 00:03:47,600 --> 00:03:49,680 Speaker 3: On top of all of this, open ai is still 84 00:03:49,720 --> 00:03:52,200 Speaker 3: yet to convert to a for profit entity, and must 85 00:03:52,240 --> 00:03:54,360 Speaker 3: do so by the end of twenty twenty five, or 86 00:03:54,360 --> 00:03:57,440 Speaker 3: they'll lose twenty billion dollars in funding from SoftBank. A 87 00:03:57,480 --> 00:04:00,119 Speaker 3: few weeks ago, Microsoft and open Ai co published an 88 00:04:00,120 --> 00:04:03,320 Speaker 3: announcement that said they'd signed a non binding memorandum of 89 00:04:03,400 --> 00:04:06,000 Speaker 3: understanding for the next phase of their partnership, which the 90 00:04:06,040 --> 00:04:08,840 Speaker 3: media quickly took to mean the deal was done. No 91 00:04:09,000 --> 00:04:11,920 Speaker 3: deal has been done. This is an announcement of nothing. 92 00:04:12,840 --> 00:04:15,000 Speaker 3: On top of all of this, venture capital might run 93 00:04:15,040 --> 00:04:17,760 Speaker 3: out at the current rate of investment in around six quarters, 94 00:04:17,760 --> 00:04:20,719 Speaker 3: and open ai needs more investment than ever. The Information 95 00:04:20,839 --> 00:04:24,520 Speaker 3: recently published some worrying data from venture capitalist John Sokoda 96 00:04:24,560 --> 00:04:26,600 Speaker 3: that a Today's clip the industry would run out of 97 00:04:26,600 --> 00:04:29,240 Speaker 3: money in six quarters, adding that the money wouldn't run 98 00:04:29,279 --> 00:04:30,960 Speaker 3: out until the end of twenty twenty eight if it 99 00:04:31,000 --> 00:04:31,840 Speaker 3: wasn't for open. 100 00:04:31,600 --> 00:04:32,560 Speaker 2: Ai and Anthropic. 101 00:04:33,000 --> 00:04:35,760 Speaker 3: In a very real sense, open ai threatens the future 102 00:04:35,760 --> 00:04:39,240 Speaker 3: of available capital for the tech industry. Well, the good 103 00:04:39,240 --> 00:04:41,240 Speaker 3: news is that they're going to make lots of money, right. 104 00:04:41,520 --> 00:04:44,119 Speaker 3: I mean, open Ai leaked that they'll make two hundred 105 00:04:44,200 --> 00:04:48,400 Speaker 3: billion dollars in four years time. I mean, if you're concussed, 106 00:04:48,720 --> 00:04:50,839 Speaker 3: I can see how you believe that. Let's go, I'm 107 00:04:50,839 --> 00:04:52,680 Speaker 3: going to read you a bunch of numbers. I know 108 00:04:52,720 --> 00:04:55,040 Speaker 3: it's kind of going to be annoying to hear this. 109 00:04:56,200 --> 00:04:57,760 Speaker 3: For the next thirty second, I'm going to just throw 110 00:04:57,800 --> 00:04:59,800 Speaker 3: figures at you, but pay attention because it's important to 111 00:04:59,800 --> 00:05:02,240 Speaker 3: really us how insane this is. So. In twenty twenty five, 112 00:05:02,320 --> 00:05:05,120 Speaker 3: open ai projects to make thirteen billion dollars and have 113 00:05:05,200 --> 00:05:08,159 Speaker 3: free cash flow of negative nine billion dollars. In twenty 114 00:05:08,160 --> 00:05:10,920 Speaker 3: twenty six, open ai will make twenty nine billion dollars 115 00:05:10,960 --> 00:05:13,440 Speaker 3: or thirty billion dollars, but have free cash flow of 116 00:05:13,480 --> 00:05:17,360 Speaker 3: negative seventeen billion. In twenty twenty seven, open ai will 117 00:05:17,360 --> 00:05:20,200 Speaker 3: make sixty billion dollars but have negative cash flow of 118 00:05:20,560 --> 00:05:23,600 Speaker 3: negative thirty five billion dollars. And in twenty twenty eight, 119 00:05:23,640 --> 00:05:25,640 Speaker 3: open ai will make one hundred billion dollars but have 120 00:05:25,720 --> 00:05:28,520 Speaker 3: free cash flow of negative forty seven billion dollars. In 121 00:05:28,560 --> 00:05:30,680 Speaker 3: twenty twenty nine, open ai will make one hundred and 122 00:05:30,720 --> 00:05:32,880 Speaker 3: forty five billion dollars but a free cash flow of 123 00:05:32,960 --> 00:05:36,440 Speaker 3: negative eight billion dollars. Somehow, I just want to reiterate 124 00:05:36,480 --> 00:05:39,640 Speaker 3: there open AI's projection. See it reduce its negative cash 125 00:05:39,640 --> 00:05:42,120 Speaker 3: flow by thirty nine billion dollars or one and a 126 00:05:42,120 --> 00:05:44,800 Speaker 3: half times three m's revenue from the last financial year 127 00:05:44,920 --> 00:05:47,200 Speaker 3: in a single year, while also increasing their revenue by 128 00:05:47,200 --> 00:05:49,960 Speaker 3: a similar amount, give or take a few billion. Who's counting, 129 00:05:50,279 --> 00:05:52,920 Speaker 3: certainly not Sarah Fryer, the CFO of open Ai. 130 00:05:53,440 --> 00:05:54,719 Speaker 2: It's also fucking stupid. 131 00:05:54,720 --> 00:05:57,960 Speaker 3: But don't worry. I know, I know some of you 132 00:05:58,120 --> 00:06:00,760 Speaker 3: have been worried. You see big strong men with tears. 133 00:06:00,839 --> 00:06:03,520 Speaker 3: Now is going to be saying cirsa open Ai. They're 134 00:06:03,520 --> 00:06:05,120 Speaker 3: not gonna They're going to have enough money. 135 00:06:05,160 --> 00:06:06,320 Speaker 2: They're gonna die. But don't worry. 136 00:06:06,360 --> 00:06:09,040 Speaker 3: In twenty thirty, open ai will make two hundred billion 137 00:06:09,040 --> 00:06:12,000 Speaker 3: dollars and somehow have positive free cash flow of thirty 138 00:06:12,040 --> 00:06:14,920 Speaker 3: eight billion dollars. It's just that easy. I teach you 139 00:06:14,960 --> 00:06:17,479 Speaker 3: this in business school. It's just that easy. 140 00:06:17,480 --> 00:06:19,160 Speaker 2: How they're gonna make it. They're gonna make it. They'll 141 00:06:19,200 --> 00:06:19,719 Speaker 2: be fine. 142 00:06:20,000 --> 00:06:22,320 Speaker 3: Open ai is current reported burn is one hundred and 143 00:06:22,360 --> 00:06:24,760 Speaker 3: fifteen billion dollars through twenty thirty, which means there's no 144 00:06:24,800 --> 00:06:27,279 Speaker 3: way that these projections that were leaked included three hundred 145 00:06:27,320 --> 00:06:30,240 Speaker 3: billion dollars in compute costs. Even when you factor in 146 00:06:30,279 --> 00:06:33,120 Speaker 3: the revenue, there's no space in the projections to absorb 147 00:06:33,160 --> 00:06:34,840 Speaker 3: the Oracle money, and from what I can tell, by 148 00:06:34,880 --> 00:06:37,600 Speaker 3: twenty twenty nine, open AI will burned up have burned 149 00:06:37,680 --> 00:06:40,280 Speaker 3: upwards of two hundred ninety billion dollars, assuming it suffves 150 00:06:40,360 --> 00:06:43,039 Speaker 3: that long, which I do not believe it will. Don't 151 00:06:43,040 --> 00:06:46,039 Speaker 3: worry though, open ai is about to make some crazy money. 152 00:06:46,080 --> 00:06:49,440 Speaker 3: I say in no way being totally sarcastic. I'm gonna 153 00:06:49,440 --> 00:06:53,839 Speaker 3: now read you some of the projections that CFO Sarah 154 00:06:53,880 --> 00:06:58,800 Speaker 3: Fryer signed off on. This is a professional chief financial officer. Again, 155 00:06:58,839 --> 00:07:02,359 Speaker 3: more numbers, but pay attention, Pay attention. This is I 156 00:07:02,400 --> 00:07:04,520 Speaker 3: want to give you, because these are numbers you just heard, 157 00:07:04,560 --> 00:07:07,560 Speaker 3: but I want to give you some comparison points. So 158 00:07:07,600 --> 00:07:09,920 Speaker 3: in twenty twenty six, open AI will make, according to 159 00:07:09,960 --> 00:07:12,760 Speaker 3: the projections, or these are the projections, thirty billion dollars 160 00:07:12,800 --> 00:07:15,320 Speaker 3: in revenue, or just under thirty the thirty four billion 161 00:07:15,360 --> 00:07:18,120 Speaker 3: dollars that Salesforce made in twenty twenty four. In twenty 162 00:07:18,120 --> 00:07:20,960 Speaker 3: twenty seven, open AI will allegedly make sixty billion dollars 163 00:07:21,000 --> 00:07:23,240 Speaker 3: or two point six billion, more than the fifty seven 164 00:07:23,280 --> 00:07:25,800 Speaker 3: point four billion dollars the Oracle made in its fiscal 165 00:07:25,840 --> 00:07:28,440 Speaker 3: year twenty twenty five, or eight point four billion dollars 166 00:07:28,480 --> 00:07:31,280 Speaker 3: more than Broadcom made in twenty twenty four. In twenty 167 00:07:31,280 --> 00:07:33,520 Speaker 3: twenty eight, this is crazy. Open ai will make one 168 00:07:33,560 --> 00:07:36,920 Speaker 3: hundred billion dollars two point three billion dollars more than 169 00:07:36,960 --> 00:07:40,000 Speaker 3: Tesla made in twenty twenty four, or eleven point eleven 170 00:07:40,040 --> 00:07:42,800 Speaker 3: point sixty six billion dollars more than TSMC, the single 171 00:07:42,880 --> 00:07:44,880 Speaker 3: largest chip manufacturer in the world. 172 00:07:44,720 --> 00:07:45,920 Speaker 2: Made in twenty twenty four. 173 00:07:46,480 --> 00:07:49,640 Speaker 3: In twenty twenty nine, things get crazier because open ai 174 00:07:49,720 --> 00:07:51,720 Speaker 3: will then make one hundred and fifty four billion dollars 175 00:07:51,800 --> 00:07:53,760 Speaker 3: or fourteen point five billion dollars more than the one 176 00:07:53,840 --> 00:07:55,960 Speaker 3: hundred and thirty point five billion dollars that in Video 177 00:07:56,000 --> 00:07:59,000 Speaker 3: made and its fiscal year twenty twenty five. But in 178 00:07:59,040 --> 00:08:01,440 Speaker 3: twenty thirty, this is when they really blow the doors off. 179 00:08:01,480 --> 00:08:04,120 Speaker 3: They will make two hundred billion dollars or thirty five 180 00:08:04,160 --> 00:08:06,120 Speaker 3: point five billion dollars in the one hundred and sixty 181 00:08:06,160 --> 00:08:09,280 Speaker 3: four point five billion that Meta made in twenty twenty four. 182 00:08:09,720 --> 00:08:12,280 Speaker 3: Just so we are clear, open ai intends to ten 183 00:08:12,320 --> 00:08:14,840 Speaker 3: exits revenue in the space of four years, selling software 184 00:08:14,880 --> 00:08:17,920 Speaker 3: and access to its models in an industry with about 185 00:08:17,960 --> 00:08:21,120 Speaker 3: sixty billion dollars of revenue. In twenty twenty five, How 186 00:08:21,200 --> 00:08:24,240 Speaker 3: will it do this? Open ai does not say. I 187 00:08:24,400 --> 00:08:27,160 Speaker 3: don't know open ai CFO Sarah Fryer, but I do 188 00:08:27,240 --> 00:08:29,200 Speaker 3: know that signing off on these numbers is at the 189 00:08:29,320 --> 00:08:33,599 Speaker 3: very least ethically questionable. But putting aside the ridiculousness of 190 00:08:33,640 --> 00:08:37,000 Speaker 3: open AI's deals or its funding requirements, Fryar has wilfully 191 00:08:37,080 --> 00:08:39,839 Speaker 3: allowed Samulman and open ai to state goals that defy 192 00:08:39,920 --> 00:08:43,239 Speaker 3: reality or good sense, all to take advantage of investors 193 00:08:43,360 --> 00:08:46,800 Speaker 3: and public markets that have completely lost the plot. I'm 194 00:08:46,840 --> 00:08:49,760 Speaker 3: actually you know what, I'm actually going to be blunter. 195 00:08:50,200 --> 00:08:53,040 Speaker 3: Open ai has signed multiple different deals and contracts for 196 00:08:53,080 --> 00:08:55,800 Speaker 3: amounts it cannot afford to pay that it cannot hope 197 00:08:55,800 --> 00:08:58,240 Speaker 3: to raise the money to pay for that defy the 198 00:08:58,280 --> 00:09:01,840 Speaker 3: amounts of venture capital and private capital available, or to 199 00:09:01,880 --> 00:09:05,640 Speaker 3: sustain a company that will burn half a trillion dollars 200 00:09:05,679 --> 00:09:08,800 Speaker 3: in the next four years based on their own expenditures, 201 00:09:08,880 --> 00:09:11,360 Speaker 3: and they have no path of the profitability of any kind. 202 00:09:12,240 --> 00:09:14,400 Speaker 3: But what about that Invidia deal? And one hundred billion 203 00:09:14,440 --> 00:09:16,280 Speaker 3: dollars is a lot of money, right, it is? But 204 00:09:16,320 --> 00:09:20,040 Speaker 3: the announcement is bullshit. I know you may have read otherwise, 205 00:09:20,040 --> 00:09:22,680 Speaker 3: but in Vidia didn't give open Ai a hundred billion dollars. 206 00:09:22,720 --> 00:09:25,080 Speaker 3: In fact, ninety billion dollars of that is contingent on 207 00:09:25,160 --> 00:09:28,000 Speaker 3: open Ai building roughly one hundred and twenty five billion 208 00:09:28,040 --> 00:09:30,520 Speaker 3: dollars worth of data centers with two hundred billion dollars 209 00:09:30,520 --> 00:09:33,439 Speaker 3: worth of GPUs inside them. And those data centers will 210 00:09:33,440 --> 00:09:35,520 Speaker 3: have to be built faster than anyone could possibly build 211 00:09:35,520 --> 00:09:37,680 Speaker 3: something of that scale. And there's no evidence that open 212 00:09:37,720 --> 00:09:40,280 Speaker 3: ai knows where or when these facilities will. 213 00:09:40,080 --> 00:09:41,760 Speaker 2: Be built or who will build them. 214 00:09:42,160 --> 00:09:45,400 Speaker 3: Important detail as well, in Vidia's data centers have nothing 215 00:09:45,440 --> 00:09:47,680 Speaker 3: to do with stargate. So when you read that open 216 00:09:47,720 --> 00:09:50,920 Speaker 3: ai has seven gigawatts of stargate whatever, it doesn't exist. 217 00:09:50,960 --> 00:09:54,679 Speaker 3: Who cares? That's nothing to do with Nvidia. Oracle is 218 00:09:54,720 --> 00:09:57,120 Speaker 3: not building these data centers. The ten gigawatts that they 219 00:09:57,160 --> 00:10:10,800 Speaker 3: need to build with video are something completely separate. But 220 00:10:10,840 --> 00:10:13,079 Speaker 3: you know what, let's take a step back with something 221 00:10:13,080 --> 00:10:17,640 Speaker 3: really simple. Data centers take forever to build. As I've 222 00:10:17,640 --> 00:10:20,560 Speaker 3: said previously, based on current reports, it's taking Oracle and 223 00:10:20,600 --> 00:10:22,000 Speaker 3: Cruiser around two point five. 224 00:10:21,920 --> 00:10:22,559 Speaker 2: Years per giga. 225 00:10:22,559 --> 00:10:25,680 Speaker 3: What of data circumpans city and nowhere in these reports 226 00:10:25,679 --> 00:10:28,040 Speaker 3: there's one reporter take a fucking second to say, hey, 227 00:10:28,080 --> 00:10:29,520 Speaker 3: what data centers are you talking about? 228 00:10:29,559 --> 00:10:30,640 Speaker 2: Hey, didn't Sam alm. 229 00:10:30,440 --> 00:10:32,280 Speaker 3: And say back in July he was building ten gigawats 230 00:10:32,280 --> 00:10:35,320 Speaker 3: of data center capacity with Oracle. But wait now Oracle 231 00:10:35,360 --> 00:10:37,600 Speaker 3: and open Ai have done another announcement that says they're 232 00:10:37,679 --> 00:10:40,240 Speaker 3: only doing seven gigawats, but they already had a schedule 233 00:10:40,240 --> 00:10:44,640 Speaker 3: on ten gigawatts. Like I said, totally different ten gigawats 234 00:10:44,640 --> 00:10:47,400 Speaker 3: to the one within video. But you know what, got 235 00:10:47,440 --> 00:10:49,800 Speaker 3: a few hundred billion dollars of data centers. Now you're 236 00:10:49,800 --> 00:10:52,880 Speaker 3: breaking real money anyway. I cannot be clear enough how 237 00:10:52,960 --> 00:10:54,880 Speaker 3: unlikely it is that the first giga what of in 238 00:10:54,960 --> 00:10:57,120 Speaker 3: video systems will be deployed in the second half of 239 00:10:57,200 --> 00:10:59,760 Speaker 3: twenty twenty six, as in Video has stated with their 240 00:10:59,760 --> 00:11:02,360 Speaker 3: deal with open Ai. And that's as if the land 241 00:11:02,440 --> 00:11:05,120 Speaker 3: has been bought and got permits and the construction has started. 242 00:11:05,360 --> 00:11:06,400 Speaker 3: None of this has happened. 243 00:11:06,760 --> 00:11:09,840 Speaker 2: But you know, let's get really specific on costs. 244 00:11:10,440 --> 00:11:13,080 Speaker 3: Kruso's one point two gigawatts of compute for open Ai 245 00:11:13,240 --> 00:11:15,599 Speaker 3: is a fifteen billion dollar joint venture, which means a 246 00:11:15,640 --> 00:11:18,800 Speaker 3: gigawat of compute runs about twelve point five billion dollars 247 00:11:18,800 --> 00:11:21,960 Speaker 3: worth of construction, Abilene's eight buildings are meant to hold 248 00:11:21,960 --> 00:11:24,959 Speaker 3: fifty thousand in video GB two hundred GPUs and their 249 00:11:25,000 --> 00:11:28,240 Speaker 3: associated networking infrastructure. So let's say a Gigawa is around 250 00:11:28,280 --> 00:11:31,040 Speaker 3: three hundred and thirty three thousand Blackwell GPUs, even though 251 00:11:31,040 --> 00:11:33,360 Speaker 3: the GB two hundreds are technically two of them put together. 252 00:11:33,840 --> 00:11:35,599 Speaker 3: Though this math is a little funky due to in 253 00:11:35,720 --> 00:11:38,160 Speaker 3: video promising to install its new Reuben GPUs in these 254 00:11:38,240 --> 00:11:41,120 Speaker 3: theoretical data centers. That means these data centers will require 255 00:11:41,120 --> 00:11:42,640 Speaker 3: a little under two hundred billion. 256 00:11:42,440 --> 00:11:43,559 Speaker 2: Dollars worth of GPUs. 257 00:11:43,880 --> 00:11:45,920 Speaker 3: By my maths, that's about three hundred and twenty five 258 00:11:45,960 --> 00:11:47,880 Speaker 3: billion dollars, but in videos saying it could be half 259 00:11:47,920 --> 00:11:50,040 Speaker 3: a trillion. But you know, who gives a shit, right, 260 00:11:50,600 --> 00:11:55,160 Speaker 3: numbers don't matter. You've got Jensen Wong telling CNBC that 261 00:11:55,320 --> 00:11:57,840 Speaker 3: one giga what costs fifty sixty billion dollars. You've got 262 00:11:57,840 --> 00:12:00,760 Speaker 3: Altman claiming the same thing to Bloomberg. But my mass 263 00:12:00,880 --> 00:12:04,640 Speaker 3: is something completely fucking different, But no one seems. 264 00:12:04,360 --> 00:12:05,319 Speaker 2: To want to bother. 265 00:12:06,360 --> 00:12:09,679 Speaker 3: You know. It's only like the places I'm quoting are 266 00:12:09,840 --> 00:12:12,520 Speaker 3: leading business outlets with teams of thousands of people, with 267 00:12:12,559 --> 00:12:15,680 Speaker 3: incredible infrastructure where they could sit down and say, well, 268 00:12:15,920 --> 00:12:19,040 Speaker 3: you know it costs this much to build stargate ABERI, 269 00:12:19,080 --> 00:12:21,920 Speaker 3: and why don't we work out actually what this costs? No, no, no, 270 00:12:22,080 --> 00:12:25,000 Speaker 3: just eat the info slop yum yum yum. You know 271 00:12:25,320 --> 00:12:28,120 Speaker 3: it doesn't exhaust me. It doesn't make me cynical and 272 00:12:28,200 --> 00:12:31,640 Speaker 3: frustrated with people who could do a good job. Indeed, 273 00:12:31,640 --> 00:12:34,000 Speaker 3: I've seen them do a good job, but they just 274 00:12:34,040 --> 00:12:35,920 Speaker 3: seem to not want to do one with this, and 275 00:12:36,320 --> 00:12:39,920 Speaker 3: it genuinely bothers me on a day to day basis. Okay, 276 00:12:40,120 --> 00:12:43,280 Speaker 3: I realize I'm just ranting at this point, but I'm 277 00:12:43,320 --> 00:12:45,280 Speaker 3: gonna be honest. I'm just I'm so tired of this. 278 00:12:47,800 --> 00:12:50,400 Speaker 3: As a reminder as well, Open Eyes agreed to spend 279 00:12:50,400 --> 00:12:52,600 Speaker 3: three hundred billion dollars on compute with Oracle. But at 280 00:12:52,600 --> 00:12:56,120 Speaker 3: this point, as you can tell, numbers basically don't have meaning. 281 00:12:56,760 --> 00:12:59,800 Speaker 3: Numbers have stopped meaning anything. If you believe these numbers, 282 00:13:00,280 --> 00:13:03,440 Speaker 3: you're a fantasist or a liar, or you just don't 283 00:13:03,440 --> 00:13:05,120 Speaker 3: want to think about them too much, because when you 284 00:13:05,160 --> 00:13:08,200 Speaker 3: think about them, you start to realize how unrealistic it is. 285 00:13:08,559 --> 00:13:10,960 Speaker 3: And even if you remain steadfast in your belief of 286 00:13:11,040 --> 00:13:14,800 Speaker 3: the transformative potential of large language models. Eventually, there comes 287 00:13:14,800 --> 00:13:17,040 Speaker 3: a point where reality must prevail and you accept that 288 00:13:17,080 --> 00:13:18,680 Speaker 3: we're in a bubble, whether you like it or not. 289 00:13:20,080 --> 00:13:22,480 Speaker 3: What we're witnessing is one of the most egregious wastes 290 00:13:22,520 --> 00:13:25,520 Speaker 3: of capital in history, sold by career charlatans, with their 291 00:13:25,520 --> 00:13:28,320 Speaker 3: reputations laundered by a tech and business media afraid to 292 00:13:28,320 --> 00:13:30,959 Speaker 3: criticize the powerful an analyst that don't seem to want 293 00:13:31,000 --> 00:13:34,079 Speaker 3: to tell their investors the truth. There are no historic 294 00:13:34,160 --> 00:13:38,240 Speaker 3: comparisons here. Even Britain's abominable eighteen hundreds railway bubble, which 295 00:13:38,280 --> 00:13:42,040 Speaker 3: absorbed half of the country's national income, created valuable infrastructure 296 00:13:42,040 --> 00:13:44,080 Speaker 3: for trains, a vehicle that can move people to and 297 00:13:44,120 --> 00:13:47,920 Speaker 3: from places. So the train doesn't randomly just go backwards 298 00:13:47,920 --> 00:13:49,800 Speaker 3: when you make it go forwards. It doesn't become a 299 00:13:49,840 --> 00:13:53,440 Speaker 3: bus or a dog at random. It doesn't hallucinate doors 300 00:13:53,440 --> 00:13:57,000 Speaker 3: that open and close at random. Trains work and GPUs 301 00:13:57,040 --> 00:13:59,560 Speaker 3: are not trains, nor are they cars or even CPUs, 302 00:14:00,080 --> 00:14:02,240 Speaker 3: not adaptable to many other kinds of work, nor are 303 00:14:02,240 --> 00:14:04,760 Speaker 3: they the infrastructure of the future of tech, because they're 304 00:14:04,800 --> 00:14:07,160 Speaker 3: already quite old and with everybody focused on buying them, 305 00:14:07,200 --> 00:14:09,600 Speaker 3: you'd absolutely see one other use case by now that 306 00:14:09,679 --> 00:14:14,800 Speaker 3: actually mattered. GPUs are expensive, power hungry, environmentally destructive, and 307 00:14:14,840 --> 00:14:17,800 Speaker 3: require their own kinds of cooling and server infrastructure, making 308 00:14:17,800 --> 00:14:22,120 Speaker 3: every GPU data center an environmental and fiscal bubble onto themselves. 309 00:14:22,520 --> 00:14:25,360 Speaker 3: And whereas the Victorian train and infrastructure still exists in 310 00:14:25,400 --> 00:14:27,520 Speaker 3: the UK, though it has been upgraded over the years, 311 00:14:27,680 --> 00:14:30,760 Speaker 3: a GPU has a limited useful lifespan. These are cards 312 00:14:30,800 --> 00:14:34,120 Speaker 3: that can and will break after a period of extended usage, 313 00:14:34,200 --> 00:14:36,280 Speaker 3: whether that period is five years or later, though I 314 00:14:36,320 --> 00:14:39,800 Speaker 3: hear it's one to three, and they'll inevitably be superseded 315 00:14:39,840 --> 00:14:42,200 Speaker 3: by something better and more powerful, meaning that the resale 316 00:14:42,240 --> 00:14:44,280 Speaker 3: value of that GPU will only go down with a 317 00:14:44,280 --> 00:14:47,400 Speaker 3: price depreciation. That's the kin to a new car, and 318 00:14:47,440 --> 00:14:49,280 Speaker 3: then a used car, and then a car that you 319 00:14:49,360 --> 00:14:53,160 Speaker 3: kick the door over every so often. I am telling you, 320 00:14:53,440 --> 00:14:55,600 Speaker 3: as I have been telling you for years, again and 321 00:14:55,640 --> 00:14:58,000 Speaker 3: again and again, that the demand is not there for 322 00:14:58,040 --> 00:15:01,280 Speaker 3: generative AI, and the demand is never ever arriving. The 323 00:15:01,320 --> 00:15:03,920 Speaker 3: only reason anyone humors any of this crap is the 324 00:15:04,000 --> 00:15:07,400 Speaker 3: endless hoarding of GPUs to build capacity for a revolution. 325 00:15:07,040 --> 00:15:08,240 Speaker 2: That will never arrive. 326 00:15:08,880 --> 00:15:11,200 Speaker 3: Well, that and open Ai, a company that's built and 327 00:15:11,240 --> 00:15:15,520 Speaker 3: sold on lies about chat GPT's capabilities. Chat GPT's popularity 328 00:15:15,520 --> 00:15:18,680 Speaker 3: in open hunger for endless amounts of compute have created 329 00:15:18,720 --> 00:15:20,600 Speaker 3: the illusion of demand due to the sheer amount of 330 00:15:20,640 --> 00:15:23,600 Speaker 3: capacity required to keep their services operational. Also that they 331 00:15:23,640 --> 00:15:25,760 Speaker 3: can burn around eight billion dollars or more in twenty 332 00:15:25,800 --> 00:15:28,600 Speaker 3: twenty five and hundreds of billions of dollars more by 333 00:15:28,960 --> 00:15:32,400 Speaker 3: twenty thirty if they make him, which they won't. The 334 00:15:32,560 --> 00:15:35,120 Speaker 3: Nvidia dealers are fast an obvious attempt by the largest 335 00:15:35,120 --> 00:15:37,280 Speaker 3: company on the American stock market to prop up the 336 00:15:37,320 --> 00:15:40,760 Speaker 3: one significant revenue generator in the entire industry, knowing that 337 00:15:40,840 --> 00:15:42,800 Speaker 3: time is running out for it to create new avenues 338 00:15:42,840 --> 00:15:46,320 Speaker 3: for eternal growth. I'd argue that Nvidia's deal also shows 339 00:15:46,320 --> 00:15:48,880 Speaker 3: the complete content that these companies have for the media. 340 00:15:49,040 --> 00:15:51,520 Speaker 3: There are no details about how this deal works beyond 341 00:15:51,520 --> 00:15:54,200 Speaker 3: the initial ten billion dollars. There's no land purchase, no 342 00:15:54,280 --> 00:15:57,440 Speaker 3: data center construction started, yet the media slurps it down 343 00:15:57,480 --> 00:16:01,200 Speaker 3: without a second thought. I am but one man, and 344 00:16:01,240 --> 00:16:05,160 Speaker 3: I'm fucking peculiar. I did not learn financial analysis in school, 345 00:16:05,160 --> 00:16:06,760 Speaker 3: but I appear to be one of the few people 346 00:16:06,800 --> 00:16:09,480 Speaker 3: doing even the most basic analysis of these deals. And 347 00:16:09,480 --> 00:16:11,440 Speaker 3: while I'm having a great time doing so, I'm are 348 00:16:11,520 --> 00:16:14,320 Speaker 3: so extremely frustrated about how little effort is being put 349 00:16:14,320 --> 00:16:16,200 Speaker 3: into praying apart these deals by other people. 350 00:16:17,040 --> 00:16:19,720 Speaker 2: I realize how ridiculous all that this sounds. I get it. 351 00:16:19,760 --> 00:16:22,120 Speaker 3: There's so much money being promised, that so many people 352 00:16:22,160 --> 00:16:24,600 Speaker 3: market rallies built off the back of these massive deals, 353 00:16:24,640 --> 00:16:26,680 Speaker 3: and I get that the assumption that this is that 354 00:16:26,760 --> 00:16:29,160 Speaker 3: this much money can't be wrong, that this many people 355 00:16:29,160 --> 00:16:31,600 Speaker 3: wouldn't just say stuff without intending to follow through, or 356 00:16:31,600 --> 00:16:34,240 Speaker 3: without considering whether their company could afford it. I know 357 00:16:34,360 --> 00:16:36,560 Speaker 3: it's hard to conceive that these hundreds of billions of 358 00:16:36,600 --> 00:16:39,160 Speaker 3: dollars could be invested in something for no apparent reason, 359 00:16:39,520 --> 00:16:42,480 Speaker 3: but it's happening, right goddamn now, in front of your eyes. 360 00:16:42,640 --> 00:16:44,880 Speaker 3: And I am going to be merciless on anyone who 361 00:16:44,880 --> 00:16:47,800 Speaker 3: attempts to write, how could we possibly have seen this coming? 362 00:16:48,880 --> 00:16:52,360 Speaker 3: Generator AI has never been reliable, has always been unprofitable, 363 00:16:52,400 --> 00:16:55,200 Speaker 3: and has always been unsustainable. And I've been saying so 364 00:16:55,240 --> 00:16:58,520 Speaker 3: since at least February twenty twenty four. These economics have 365 00:16:58,600 --> 00:17:01,600 Speaker 3: never made sense. Something I've said repeatedly since April twenty 366 00:17:01,600 --> 00:17:03,880 Speaker 3: twenty four, and when I wrote how does open Ai 367 00:17:03,960 --> 00:17:06,320 Speaker 3: Survive in July twenty twenty four, I had multiple people 368 00:17:06,359 --> 00:17:10,280 Speaker 3: suggest I was being an alarmist. Here's some alarmism for you. 369 00:17:10,760 --> 00:17:12,719 Speaker 3: The longer it takes for open ai to die, the 370 00:17:12,720 --> 00:17:15,600 Speaker 3: more damage it will cause once it does. As I 371 00:17:15,680 --> 00:17:18,120 Speaker 3: talked about earlier in this series, venture capital could run 372 00:17:18,119 --> 00:17:20,480 Speaker 3: out in six quarters, with investor and research of John 373 00:17:20,520 --> 00:17:23,240 Speaker 3: Socoda estimating that there will be only around one hundred 374 00:17:23,240 --> 00:17:25,960 Speaker 3: and sixty four billion dollars of dry powder, which is 375 00:17:25,960 --> 00:17:28,320 Speaker 3: available capital, by the way, in USVC firms, by the 376 00:17:28,400 --> 00:17:31,880 Speaker 3: end of twenty twenty five. In July, the French Tech 377 00:17:31,960 --> 00:17:35,040 Speaker 3: journal reported using Pitchbook DAYA, the global venture capital deal 378 00:17:35,040 --> 00:17:38,320 Speaker 3: activity reached its lowest first half total since twenty eighteen, 379 00:17:38,520 --> 00:17:40,879 Speaker 3: with one hundred and thirty nine point four billion dollars 380 00:17:40,880 --> 00:17:42,960 Speaker 3: in deal value in the first half of twenty twenty five, 381 00:17:43,200 --> 00:17:45,240 Speaker 3: down from one hundred and eighty three point four billion 382 00:17:45,240 --> 00:17:47,439 Speaker 3: in the first half of twenty twenty four, meaning that 383 00:17:47,480 --> 00:17:50,040 Speaker 3: any further expansion or demands for venge capital from open 384 00:17:50,119 --> 00:17:52,120 Speaker 3: ai will likely sap dwindling. 385 00:17:51,720 --> 00:17:53,600 Speaker 2: Funds available for other startups. 386 00:17:54,240 --> 00:17:56,000 Speaker 3: Things get worse when you narrow things to the US 387 00:17:56,119 --> 00:17:59,080 Speaker 3: venture capital In a piece from April Erstin Young report, 388 00:17:59,160 --> 00:18:01,760 Speaker 3: the VC backed in in US companies say eighty billion 389 00:18:01,760 --> 00:18:04,399 Speaker 3: dollars in Q one twenty twenty five, but one forty 390 00:18:04,400 --> 00:18:06,600 Speaker 3: billion dollar deal account for half of the investment. 391 00:18:06,680 --> 00:18:08,280 Speaker 2: Open Ai is forty billion. 392 00:18:08,000 --> 00:18:11,440 Speaker 3: Dollar deal, of which only ten billion dollar has actually closed, 393 00:18:11,520 --> 00:18:15,199 Speaker 3: and that didn't happen until fucking June. Without the imaginary 394 00:18:15,240 --> 00:18:17,719 Speaker 3: money from open Ai, US venture capital would have declined 395 00:18:17,720 --> 00:18:20,560 Speaker 3: by thirty six percent. I should also add another eight 396 00:18:20,640 --> 00:18:22,600 Speaker 3: point three billion dollars was added on the top of that, 397 00:18:22,680 --> 00:18:26,280 Speaker 3: but still that's less than forty dollars, folks. The longer 398 00:18:26,280 --> 00:18:28,320 Speaker 3: the open Ai survives, the longer it will sap the 399 00:18:28,359 --> 00:18:30,880 Speaker 3: remaining billions from the tech ecosystem, and I expect its 400 00:18:30,920 --> 00:18:34,200 Speaker 3: tendrils to extend to private credit soon too. The three 401 00:18:34,320 --> 00:18:36,359 Speaker 3: hundred and twenty five billion dollars it needs just to 402 00:18:36,359 --> 00:18:39,520 Speaker 3: fulfill its Nvidia contract, albeit over four years, is an 403 00:18:39,520 --> 00:18:42,399 Speaker 3: egregious sum that I believe exceeds the available private capital 404 00:18:42,440 --> 00:18:45,880 Speaker 3: in the world. And I went and actually looked, assuming 405 00:18:45,920 --> 00:18:48,000 Speaker 3: that all of this capital is currently available, the top 406 00:18:48,040 --> 00:18:50,000 Speaker 3: ten private equity films in the world have around four 407 00:18:50,080 --> 00:18:53,840 Speaker 3: hundred and seventy seven billion dollars of available capital. A reminder, 408 00:18:53,960 --> 00:18:56,040 Speaker 3: four hundred and seventy seven billion dollars is a lot 409 00:18:56,080 --> 00:18:58,520 Speaker 3: less than open AI's total commitments over the next few years. 410 00:18:58,520 --> 00:19:01,720 Speaker 3: In fact, I put a premium newsletter out just before 411 00:19:01,760 --> 00:19:04,320 Speaker 3: recording this, so I didn't totally update the script. But 412 00:19:04,720 --> 00:19:07,239 Speaker 3: open ai needs a trillion dollars. They need about five 413 00:19:07,320 --> 00:19:09,840 Speaker 3: hundred billion for operations and about four hundred and fifty 414 00:19:09,840 --> 00:19:12,679 Speaker 3: billion for all the data centers they promised and the 415 00:19:12,680 --> 00:19:14,280 Speaker 3: GPUs in them. 416 00:19:14,440 --> 00:19:16,400 Speaker 2: It's not really good, it's not good at all. 417 00:19:16,600 --> 00:19:18,280 Speaker 3: But if we include the cash at hand at the 418 00:19:18,280 --> 00:19:21,320 Speaker 3: biggest investment banks, this brings us up to By the way, 419 00:19:21,359 --> 00:19:23,000 Speaker 3: if we as seen that all of these funds are 420 00:19:23,000 --> 00:19:25,600 Speaker 3: available about five hundred and sixty two billion dollars in 421 00:19:25,640 --> 00:19:27,800 Speaker 3: capital and one hundred and sixty four billion dollars a 422 00:19:27,880 --> 00:19:30,360 Speaker 3: venture capital in the US to spend, and that's all 423 00:19:30,400 --> 00:19:32,400 Speaker 3: meant to go in more places than just open AI, 424 00:19:33,119 --> 00:19:36,760 Speaker 3: that's not enough. If we combind all the available private 425 00:19:36,760 --> 00:19:38,679 Speaker 3: equity firms in the top ten and all of the 426 00:19:38,720 --> 00:19:42,560 Speaker 3: available USVC, that's not enough to pay for all the 427 00:19:42,600 --> 00:19:47,639 Speaker 3: bills that the open aiy has said, it's we're in 428 00:19:47,680 --> 00:19:50,120 Speaker 3: a fucking bubble man. Come on, sure, I also get 429 00:19:50,119 --> 00:19:52,440 Speaker 3: that there's more than ten private equity firms and there's 430 00:19:52,480 --> 00:19:54,640 Speaker 3: more venture money in that they could raise more money 431 00:19:54,680 --> 00:19:57,080 Speaker 3: from lpiece, but what's the left exactly in our one 432 00:19:57,119 --> 00:19:59,520 Speaker 3: hundred and fifty two hundred and three hundred billion. They 433 00:19:59,560 --> 00:20:01,560 Speaker 3: need that, they need all of that. Hand that over 434 00:20:01,600 --> 00:20:04,960 Speaker 3: to Clammy Sami right now. Clammy needs to give that 435 00:20:05,040 --> 00:20:08,520 Speaker 3: money to Jensen. Jensen need money. Number go up, you asshole, 436 00:20:08,600 --> 00:20:25,160 Speaker 3: give money to Jensen Wong. Now you see open ai 437 00:20:25,320 --> 00:20:27,359 Speaker 3: needs to buy those GPUs, and it needs to build 438 00:20:27,359 --> 00:20:29,760 Speaker 3: those data centers, and it needs to pay it's thousands 439 00:20:29,800 --> 00:20:32,560 Speaker 3: of staff and marketing and sales costs too well. Open 440 00:20:32,600 --> 00:20:34,480 Speaker 3: Ai likely wouldn't be the ones raising the money for 441 00:20:34,520 --> 00:20:36,680 Speaker 3: the data centers, and honestly, I'm not sure who would. 442 00:20:36,680 --> 00:20:39,080 Speaker 3: At this point, somebody is going to need to build 443 00:20:39,119 --> 00:20:42,480 Speaker 3: twenty fucking gigawatts of data centers if we're going to 444 00:20:42,480 --> 00:20:45,960 Speaker 3: believe Oracle and Nvidia. You may argue that venture funds 445 00:20:45,960 --> 00:20:48,280 Speaker 3: and private equity funds can raise more, and you're right, 446 00:20:48,320 --> 00:20:50,639 Speaker 3: but at this point, there have been few meaningful acquisitions 447 00:20:50,640 --> 00:20:53,359 Speaker 3: of AI companies and zero exits from the billions of 448 00:20:53,400 --> 00:20:55,639 Speaker 3: dollars put into data centers. How are they meant to 449 00:20:55,760 --> 00:20:58,040 Speaker 3: justify doing this other than me need money? 450 00:20:58,080 --> 00:20:58,359 Speaker 2: Please? 451 00:20:58,960 --> 00:21:02,119 Speaker 3: Even open aims it's in its own announcement about new 452 00:21:02,160 --> 00:21:04,840 Speaker 3: stargate sites that this will be a four hundred billion 453 00:21:04,880 --> 00:21:07,919 Speaker 3: dollar investment over three years. Where the fuck's the money 454 00:21:07,920 --> 00:21:11,200 Speaker 3: coming from? Is open ai really going to absorb massive 455 00:21:11,240 --> 00:21:14,480 Speaker 3: junks of all available private credit and venture capital and 456 00:21:14,960 --> 00:21:17,760 Speaker 3: capex from major tech companies. 457 00:21:17,400 --> 00:21:18,480 Speaker 2: For the next few years? 458 00:21:19,359 --> 00:21:23,639 Speaker 3: And oh my fucking god, Oh my god, stop saying 459 00:21:23,720 --> 00:21:26,639 Speaker 3: the US government will bail this out. I'm tired of 460 00:21:26,680 --> 00:21:29,160 Speaker 3: hearing it. I get the things as scary right now. 461 00:21:29,240 --> 00:21:32,520 Speaker 3: I'm scared too, We're all fucking scared. But I get 462 00:21:32,600 --> 00:21:33,560 Speaker 3: jumping to doomerism. 463 00:21:33,640 --> 00:21:36,959 Speaker 2: But stop. The US government actually really can't afford to 464 00:21:37,000 --> 00:21:39,160 Speaker 2: do this. Open AI needs about. 465 00:21:38,840 --> 00:21:41,520 Speaker 3: A trillion dollars in the next four years. And while 466 00:21:41,560 --> 00:21:44,240 Speaker 3: the US government has spent equivalent sums in the past 467 00:21:44,240 --> 00:21:47,000 Speaker 3: to support private businesses, there's about four hundred and forty 468 00:21:47,000 --> 00:21:50,080 Speaker 3: billion dollars dispersed during the Great Recessions tart program, and 469 00:21:50,119 --> 00:21:52,439 Speaker 3: that was when the Treasury brought toxic assets from investment 470 00:21:52,520 --> 00:21:53,919 Speaker 3: banks to stop them from imploding. 471 00:21:54,040 --> 00:21:56,399 Speaker 2: Like Layman Brothers that one. 472 00:21:56,560 --> 00:21:58,680 Speaker 3: It's hard to imagine any case where open ai is 473 00:21:58,720 --> 00:22:01,040 Speaker 3: seen as vital to the global financial system and the 474 00:22:01,080 --> 00:22:04,080 Speaker 3: economic health of the US as the actual banking system. 475 00:22:04,119 --> 00:22:06,840 Speaker 3: And even if they were, that's more money than they 476 00:22:06,840 --> 00:22:10,040 Speaker 3: can easily mobilize, even for the US government. And sure, 477 00:22:10,080 --> 00:22:12,840 Speaker 3: we spent around a trillion dollars if we're being specific, 478 00:22:12,960 --> 00:22:15,440 Speaker 3: nine hundred and fifty three billion dollars on the paycheck 479 00:22:15,480 --> 00:22:18,080 Speaker 3: Protection program during the COVID era. That was to keep 480 00:22:18,080 --> 00:22:20,080 Speaker 3: people employed at a time when the economy outside of 481 00:22:20,119 --> 00:22:22,840 Speaker 3: Zoom and Walmart had, for all intents and purposes, ceased 482 00:22:22,880 --> 00:22:25,560 Speaker 3: to exist. There was an urgency that doesn't apply here. 483 00:22:26,400 --> 00:22:30,000 Speaker 3: If open ai goes tits up, soft Bank, Dragonnaer and 484 00:22:30,040 --> 00:22:32,560 Speaker 3: a few others lose a bit of money. Nothing new there, 485 00:22:32,760 --> 00:22:34,960 Speaker 3: and Sacha Nadella has to explain why he spent tens 486 00:22:34,960 --> 00:22:37,520 Speaker 3: of billions, actually hundreds of billions of dollars on data 487 00:22:37,520 --> 00:22:42,320 Speaker 3: centers fills with GPUs. But otherwise there's you can't really 488 00:22:42,320 --> 00:22:45,080 Speaker 3: bail this out. Are we going to nationalize in video? 489 00:22:45,280 --> 00:22:47,080 Speaker 3: Are we going to create a top fund just for 490 00:22:47,200 --> 00:22:50,720 Speaker 3: open ai? If you really believe that will happen. I 491 00:22:50,800 --> 00:22:52,240 Speaker 3: really must put a smile on your face. 492 00:22:52,440 --> 00:22:52,919 Speaker 2: It can't. 493 00:22:53,480 --> 00:22:58,080 Speaker 3: Even under the most crony capitalism scenario. This is more 494 00:22:58,080 --> 00:23:01,200 Speaker 3: money than actually really exists in the world. And yes, 495 00:23:01,320 --> 00:23:05,560 Speaker 3: while there will be and have been already disastrous economic consequences, 496 00:23:05,800 --> 00:23:08,439 Speaker 3: they won't be as systematically catastrophic as that of a 497 00:23:08,480 --> 00:23:11,800 Speaker 3: pandemic or a global financial crisis. To be clear, It'll 498 00:23:11,840 --> 00:23:14,679 Speaker 3: be bad, but not as bad. And there's also the 499 00:23:14,680 --> 00:23:17,199 Speaker 3: problem of a moral hazard. If the government steps in, 500 00:23:17,240 --> 00:23:19,919 Speaker 3: what's the stop big tech chasing its next frootless rainbow, 501 00:23:19,960 --> 00:23:22,399 Speaker 3: and of course the optics. If people resented bailing out 502 00:23:22,440 --> 00:23:24,880 Speaker 3: the banks after they acted like Gambler's and lost, how 503 00:23:24,920 --> 00:23:29,160 Speaker 3: will they feel bailing out fucking Sam Altman and Jensen Huang. Seriously, 504 00:23:30,080 --> 00:23:32,880 Speaker 3: no matter when you've got a conservative or Democrat government, 505 00:23:33,320 --> 00:23:35,520 Speaker 3: this is not going to be an easy thing to sell. 506 00:23:35,680 --> 00:23:37,879 Speaker 3: It's going to be hard to get the consensus, and 507 00:23:37,960 --> 00:23:42,080 Speaker 3: indeed every tech company will be kind of pissed. I 508 00:23:42,119 --> 00:23:47,359 Speaker 3: think that you miss you underestimate how upset companies that 509 00:23:47,440 --> 00:23:50,200 Speaker 3: have died would be, and indeed companies outside of tech 510 00:23:50,240 --> 00:23:52,720 Speaker 3: two this would be more than a scandal it would 511 00:23:52,760 --> 00:23:57,160 Speaker 3: be genuinely bad to do, even for the silliest government 512 00:23:57,200 --> 00:23:59,359 Speaker 3: in the world. I don't think it's happening, and even 513 00:23:59,400 --> 00:24:03,080 Speaker 3: if they try, there's not enough money. 514 00:24:03,080 --> 00:24:05,440 Speaker 2: Breathe, they can't get you. You're doing a podcast. 515 00:24:06,320 --> 00:24:08,280 Speaker 3: Look, this has been a slog and I appreciate you 516 00:24:08,280 --> 00:24:10,199 Speaker 3: sticking with me this long, but the significance of the 517 00:24:10,200 --> 00:24:14,240 Speaker 3: bubble requires this depth. There's little demand, little real money, 518 00:24:14,240 --> 00:24:17,320 Speaker 3: and little reason to continue, and the sheer lack of responsibility, 519 00:24:17,400 --> 00:24:19,960 Speaker 3: along with the willingness to kneel before the powerful, feels 520 00:24:20,000 --> 00:24:23,000 Speaker 3: me full of angry bio. I understand many journalists are 521 00:24:23,040 --> 00:24:24,520 Speaker 3: not in a position where they can just write this 522 00:24:24,560 --> 00:24:27,560 Speaker 3: shit sounds stupid, but we've entered a deeply stupid era, 523 00:24:27,800 --> 00:24:30,040 Speaker 3: and by continuing to perpetuate the myth of AI, the 524 00:24:30,080 --> 00:24:33,080 Speaker 3: media guarantees that retail investors and regular people's four O 525 00:24:33,160 --> 00:24:35,879 Speaker 3: one case will suffer. It is now inevitable that this 526 00:24:35,920 --> 00:24:38,639 Speaker 3: bubble bursts. Deutsche Bank has said that the AI boom 527 00:24:38,680 --> 00:24:42,040 Speaker 3: is unsustainable outside of tech spending remaining parabolic, which it 528 00:24:42,040 --> 00:24:44,959 Speaker 3: says is highly unlikely, and Bain Capital has said that 529 00:24:45,240 --> 00:24:48,480 Speaker 3: two trillion in new revenue is needed to fund AI scaling, 530 00:24:48,680 --> 00:24:50,960 Speaker 3: and even that math is completely fucked as it talks 531 00:24:50,960 --> 00:24:54,320 Speaker 3: about AI related savings, and I quote even if companies 532 00:24:54,359 --> 00:24:57,040 Speaker 3: in the US shifted all of their on premises, it 533 00:24:57,280 --> 00:25:00,000 Speaker 3: budgets the cloud and reinvested the savings from applying AI 534 00:25:00,040 --> 00:25:02,440 Speaker 3: in sales, marketing, customs support, and r D into capital 535 00:25:02,440 --> 00:25:04,960 Speaker 3: spending on new AI data centers, the amount would still 536 00:25:04,960 --> 00:25:07,560 Speaker 3: fall short of revenue, the revenue needed to fund the 537 00:25:07,600 --> 00:25:10,359 Speaker 3: full investment as AI's compute demands grow more than twice 538 00:25:10,359 --> 00:25:11,200 Speaker 3: the rate of Moore's law. 539 00:25:11,280 --> 00:25:11,800 Speaker 2: Ban notes. 540 00:25:12,520 --> 00:25:15,520 Speaker 3: Even when stared in the face by a ridiculous idea 541 00:25:15,600 --> 00:25:18,280 Speaker 3: two trillion dollars of new revenue in a global software 542 00:25:18,320 --> 00:25:21,240 Speaker 3: market and this is all software that's expected to be 543 00:25:21,280 --> 00:25:23,960 Speaker 3: around eight hundred and seventeen billion dollars in twenty twenty five, 544 00:25:24,200 --> 00:25:27,160 Speaker 3: Bain still ininks out some nonsense about the savings from 545 00:25:27,160 --> 00:25:31,040 Speaker 3: applying AI. Yet another myth perpetuated. I seem to plicate 546 00:25:31,119 --> 00:25:35,840 Speaker 3: the fucking moron sinking billions into this Every single vibe 547 00:25:35,840 --> 00:25:38,120 Speaker 3: Coding is the future, the power of AI and AI 548 00:25:38,240 --> 00:25:41,480 Speaker 3: job lot story written perpetuates a myth that will only 549 00:25:41,560 --> 00:25:43,480 Speaker 3: lead to more regular people getting hurt. 550 00:25:43,280 --> 00:25:44,320 Speaker 2: When the bubble bursts. 551 00:25:44,680 --> 00:25:47,480 Speaker 3: Every article written about open AI or in video or 552 00:25:47,520 --> 00:25:50,879 Speaker 3: oracle that does not explicitly state that the money doesn't exist, 553 00:25:50,960 --> 00:25:53,199 Speaker 3: that the revenues that are impossible, that one of the 554 00:25:53,240 --> 00:25:55,720 Speaker 3: companies involved burns billions of dollars and has no path 555 00:25:55,760 --> 00:26:00,199 Speaker 3: to profitability is an act of irresponsibility, and irresponsible make 556 00:26:00,240 --> 00:26:05,280 Speaker 3: belief and mythos. Look, I'm nobody, I'm not a financier, 557 00:26:05,480 --> 00:26:08,160 Speaker 3: I'm not anybody special. I write a lot, I read 558 00:26:08,160 --> 00:26:10,280 Speaker 3: a lot, and can do the most basic maths in 559 00:26:10,320 --> 00:26:13,080 Speaker 3: the world. I'm not trying to be anything other than myself, 560 00:26:13,119 --> 00:26:15,200 Speaker 3: nor do I have an agenda other than the fact 561 00:26:15,200 --> 00:26:16,920 Speaker 3: that I like doing this and I hate how. 562 00:26:16,840 --> 00:26:18,480 Speaker 2: This story is being told. 563 00:26:18,640 --> 00:26:21,560 Speaker 3: I never expected my newsletter to get big, nor did 564 00:26:21,600 --> 00:26:23,639 Speaker 3: I expect this podcast, nor did I expect to be 565 00:26:23,680 --> 00:26:27,120 Speaker 3: in the Financial Times. But here I am, and I'm 566 00:26:27,240 --> 00:26:30,040 Speaker 3: working my ass off, and I am pissed off at 567 00:26:30,080 --> 00:26:33,359 Speaker 3: where this has got to. I'm tired that things remain 568 00:26:33,640 --> 00:26:37,440 Speaker 3: honestly mythical, that we still see as things get more 569 00:26:37,480 --> 00:26:41,680 Speaker 3: and more ridiculous, that the headlines get more ridiculous too, 570 00:26:42,040 --> 00:26:44,960 Speaker 3: that there's questioning and people are saying, oh, the narrative 571 00:26:45,000 --> 00:26:48,640 Speaker 3: is shifting, sure, but they're still repeating increasingly dumb shit. 572 00:26:49,920 --> 00:26:51,720 Speaker 3: And I also believe that the way to stop this 573 00:26:51,800 --> 00:26:53,960 Speaker 3: happening again is to have a thorough and well sourced 574 00:26:54,000 --> 00:26:57,399 Speaker 3: explanation of everything that happens, ripping down the narratives as 575 00:26:57,400 --> 00:26:59,760 Speaker 3: they're spun, and making it clear who benefits from them 576 00:27:00,000 --> 00:27:02,960 Speaker 3: and how on why they're choosing to do so. When 577 00:27:03,000 --> 00:27:05,720 Speaker 3: things collapse, we need to be clear about how many 578 00:27:05,760 --> 00:27:08,199 Speaker 3: times people chose to look the other way or to 579 00:27:08,240 --> 00:27:11,320 Speaker 3: find good faith ways to interpret bad faith announcements and leaks. 580 00:27:12,440 --> 00:27:14,439 Speaker 2: So how could we have seen this coming? I don't know, 581 00:27:15,080 --> 00:27:16,280 Speaker 2: did anybody. 582 00:27:15,840 --> 00:27:22,159 Speaker 3: Try to fucking look? 583 00:27:25,440 --> 00:27:27,119 Speaker 2: Thank you for listening to Better Offline. 584 00:27:27,240 --> 00:27:29,640 Speaker 4: The editor and composer of the Better Offline theme song 585 00:27:29,720 --> 00:27:32,399 Speaker 4: is Matasowski. You can check out more of his music 586 00:27:32,400 --> 00:27:36,040 Speaker 4: and audio projects at Mattasowski dot com, m A T 587 00:27:36,040 --> 00:27:36,639 Speaker 4: T O. 588 00:27:36,880 --> 00:27:40,359 Speaker 2: S O W s ki dot com. 589 00:27:40,440 --> 00:27:43,040 Speaker 4: You can email me at easy at Better offline dot com, 590 00:27:43,160 --> 00:27:45,480 Speaker 4: or visit Better Offline dot com to find more podcast 591 00:27:45,520 --> 00:27:48,880 Speaker 4: links and of course my newsletter. I also really recommend 592 00:27:48,880 --> 00:27:50,800 Speaker 4: you go to chat dot Where's youreaed dot at to 593 00:27:50,880 --> 00:27:53,400 Speaker 4: visit the discord and go to our slash Better off 594 00:27:53,440 --> 00:27:56,400 Speaker 4: Line to check out our reddit thank you so much 595 00:27:56,440 --> 00:27:56,879 Speaker 4: for listening. 596 00:27:57,760 --> 00:28:00,480 Speaker 1: Better Offline is a production of Calls on Me. For 597 00:28:00,600 --> 00:28:04,639 Speaker 1: more from cool Zone Media, visit our website Coolzonemedia dot com, 598 00:28:04,720 --> 00:28:07,600 Speaker 1: or check us out on the iHeartRadio app, Apple Podcasts, 599 00:28:07,680 --> 00:28:09,159 Speaker 1: or wherever you get your podcasts.