1 00:00:02,800 --> 00:00:03,560 Speaker 1: Z Media. 2 00:00:05,120 --> 00:00:08,039 Speaker 2: Hello on, Welcome to Better Offline. I'm your host ed 3 00:00:08,119 --> 00:00:23,120 Speaker 2: zetron and I am not a car. Yesterday, a business 4 00:00:23,160 --> 00:00:26,400 Speaker 2: inside of report that Uber COO Andrew McDonald had said, 5 00:00:26,440 --> 00:00:29,400 Speaker 2: and I quote, that its AI costs were becoming harder 6 00:00:29,440 --> 00:00:32,400 Speaker 2: to justify and that the link was not there between 7 00:00:32,440 --> 00:00:35,919 Speaker 2: spending money on AI tokens and creating more useful features. 8 00:00:36,800 --> 00:00:37,000 Speaker 3: Yeah. 9 00:00:37,200 --> 00:00:39,480 Speaker 2: Just gonna throw a bastball through a hoop real quick. 10 00:00:39,720 --> 00:00:42,920 Speaker 2: After three long years of hammering it home, I have 11 00:00:43,040 --> 00:00:46,519 Speaker 2: finally been proven right. AI's outputs and efficacy do not 12 00:00:46,640 --> 00:00:49,920 Speaker 2: match up with its ruinous costs. When organizations have to 13 00:00:49,920 --> 00:00:54,400 Speaker 2: pay the actual token costs of AI versus using subsidized subscriptions, 14 00:00:54,560 --> 00:00:56,959 Speaker 2: they're forced to measure the actual return on investment from 15 00:00:57,000 --> 00:01:00,920 Speaker 2: AI and are immediately balking at the results. Wheeling for mercy. 16 00:01:01,000 --> 00:01:03,760 Speaker 2: They're saying, honey, I can afford it now. To give 17 00:01:03,800 --> 00:01:07,400 Speaker 2: you some context, Anthropic only moved organizations to token based 18 00:01:07,400 --> 00:01:11,319 Speaker 2: billing sometime in Q one, twenty twenty six. This is 19 00:01:11,400 --> 00:01:13,720 Speaker 2: at most four months of having to pay the true 20 00:01:13,760 --> 00:01:16,800 Speaker 2: costs of their AI token burn and they're already squealing. 21 00:01:16,800 --> 00:01:19,640 Speaker 2: They're already begging for mercy. They're already saying, sir, no 22 00:01:19,760 --> 00:01:23,400 Speaker 2: more tokens. AI has a revenue ceiling and an economic 23 00:01:23,440 --> 00:01:26,640 Speaker 2: mismatch with its customer base. It's time to accept it. 24 00:01:28,360 --> 00:01:31,240 Speaker 2: Every time you've heard somebody say that AI is real, 25 00:01:31,480 --> 00:01:35,240 Speaker 2: it's here, and it's transformative, you've heard from somebody paying 26 00:01:35,240 --> 00:01:38,200 Speaker 2: a monthly subscription to a service that allows its customers 27 00:01:38,200 --> 00:01:40,920 Speaker 2: to burn anywhere from three dollars the thirteen dollars worth 28 00:01:40,959 --> 00:01:44,080 Speaker 2: of tokens for every dollar of their subscription. Even get 29 00:01:44,160 --> 00:01:47,559 Speaker 2: hub Copilot, which paid the model providers directly, was letting 30 00:01:47,600 --> 00:01:50,040 Speaker 2: people burn on a thirty nine dollar a month subscription 31 00:01:50,200 --> 00:01:54,200 Speaker 2: thirteen hundred to six thousand dollars in a month. Every 32 00:01:54,280 --> 00:01:58,040 Speaker 2: effervescent booster and captured business idiot editor crowing about the 33 00:01:58,040 --> 00:02:01,120 Speaker 2: power of AI has done so without ever really facing 34 00:02:01,120 --> 00:02:05,040 Speaker 2: its real costs or i think, even using it very much. However, 35 00:02:05,200 --> 00:02:07,920 Speaker 2: useful lms may seem to them as a facade for 36 00:02:07,960 --> 00:02:10,360 Speaker 2: a product that costs far too much money for outputs 37 00:02:10,480 --> 00:02:13,120 Speaker 2: that may or may not actually result in something functional 38 00:02:13,200 --> 00:02:17,240 Speaker 2: or helpful. When you're not paying for tokens, these mistakes 39 00:02:17,280 --> 00:02:21,520 Speaker 2: are easy to ignore. These subscriptions musk the ugly truth 40 00:02:21,560 --> 00:02:24,680 Speaker 2: of AI that you're paying on a per million token basis, 41 00:02:24,680 --> 00:02:26,799 Speaker 2: regardless of whether you get what you want, or even 42 00:02:26,840 --> 00:02:29,120 Speaker 2: if the model makes mistakes or creates more problems that 43 00:02:29,160 --> 00:02:32,240 Speaker 2: you then have to spend more tokens to fix. It's 44 00:02:32,280 --> 00:02:35,240 Speaker 2: a scam, it's a con. It never made sense. An 45 00:02:35,320 --> 00:02:38,400 Speaker 2: uber COO has given everybody permission to talk about the 46 00:02:38,400 --> 00:02:41,600 Speaker 2: inherent economic mismatch of AI, and also revealed that AI 47 00:02:41,680 --> 00:02:44,480 Speaker 2: companies like Anthropic and open Ai have a ceiling far 48 00:02:44,560 --> 00:02:47,919 Speaker 2: below that which they need to justify their valuations. These 49 00:02:47,960 --> 00:02:50,880 Speaker 2: companies need to be making two hundred billion dollars a 50 00:02:50,960 --> 00:02:53,640 Speaker 2: year by twenty thirty or they cannot keep up with 51 00:02:53,680 --> 00:02:58,600 Speaker 2: their own costs. Now, anthropics rapid revenue growth is the 52 00:02:58,639 --> 00:03:01,520 Speaker 2: result of companies spending millillions or tens or even hundreds 53 00:03:01,560 --> 00:03:04,800 Speaker 2: of millions of dollars on its tokens. For its revenue 54 00:03:04,840 --> 00:03:08,600 Speaker 2: story to make sense, this revenue would have to be stable, replicable, 55 00:03:08,600 --> 00:03:12,799 Speaker 2: and sustainable. Instead, it appears that organizations have been burning 56 00:03:12,840 --> 00:03:16,160 Speaker 2: tokens without any real understanding of why, other than that 57 00:03:16,280 --> 00:03:19,680 Speaker 2: the need to do AI and LLLM coding is the future, 58 00:03:19,840 --> 00:03:22,880 Speaker 2: and all of their other dickhead business idiot friends are saying, oh, 59 00:03:22,919 --> 00:03:25,120 Speaker 2: I got, i'm let and everyone spend ten million dollars 60 00:03:25,160 --> 00:03:26,960 Speaker 2: a month on this shit because I have a fucking 61 00:03:27,000 --> 00:03:30,840 Speaker 2: piece of rebar in my skull. For this to make sense, 62 00:03:30,919 --> 00:03:34,280 Speaker 2: the majority of organizations would have to sustain and grow 63 00:03:34,560 --> 00:03:39,000 Speaker 2: a massive spend on AI tokens from anthropic and open AI. Instead, 64 00:03:39,000 --> 00:03:41,920 Speaker 2: it appears that this token bone was inherently experimental and 65 00:03:42,120 --> 00:03:45,120 Speaker 2: entirely disconnected from messy things like I don't know return 66 00:03:45,160 --> 00:03:48,720 Speaker 2: on investment. Uber already noted back in April that it 67 00:03:48,760 --> 00:03:51,440 Speaker 2: blew through its annual token budget in a few months, 68 00:03:51,440 --> 00:03:54,440 Speaker 2: and that conversation reported by Laura Bretton of the Information 69 00:03:54,560 --> 00:03:56,920 Speaker 2: for End of the Show clearly led to an internal 70 00:03:56,960 --> 00:03:58,840 Speaker 2: back and forth that will end in it cutting its 71 00:03:58,840 --> 00:04:02,440 Speaker 2: spend on AI models. I've now spoken to three different 72 00:04:02,480 --> 00:04:05,800 Speaker 2: large organizations that are all echoing similar anxieties about the 73 00:04:05,960 --> 00:04:10,360 Speaker 2: RII of AI. Nobody can actually tell why they're spending 74 00:04:10,400 --> 00:04:13,280 Speaker 2: this much money. Things aren't getting shipped fast, that software 75 00:04:13,320 --> 00:04:15,800 Speaker 2: isn't better, and the only people that seem excited about 76 00:04:15,840 --> 00:04:18,880 Speaker 2: it a business idy it's disconnected from production, and even 77 00:04:19,240 --> 00:04:22,279 Speaker 2: they are becoming cost conscious when faced with millions of 78 00:04:22,320 --> 00:04:26,719 Speaker 2: dollars of token bills anthropic and open ai cannot afford 79 00:04:26,760 --> 00:04:29,080 Speaker 2: for things to slow down, as they've both signed up 80 00:04:29,120 --> 00:04:33,840 Speaker 2: to over a trillion dollars of compute commitments across Google, Microsoft, Amazon, 81 00:04:33,960 --> 00:04:38,000 Speaker 2: Core Weave, Oracle and Cerebrus. Just to be clear as well, 82 00:04:38,160 --> 00:04:40,240 Speaker 2: open ai has to be making two hundred and eighty 83 00:04:40,240 --> 00:04:42,919 Speaker 2: four billion dollars by twenty thirty, and they need to 84 00:04:42,920 --> 00:04:46,440 Speaker 2: be profitable at that point too, otherwise Oracle cannot afford 85 00:04:46,520 --> 00:04:50,560 Speaker 2: its own bills. This is not hyperbole. This is quoting 86 00:04:50,680 --> 00:04:56,159 Speaker 2: back open AI's own projected revenues from their investor dex. 87 00:04:58,160 --> 00:05:00,599 Speaker 2: But in reality, it appears there's a limit to which 88 00:05:00,680 --> 00:05:03,520 Speaker 2: organizations can be abused and manipulated into believing that the 89 00:05:03,520 --> 00:05:05,880 Speaker 2: future is here. And that limit is when they paid 90 00:05:05,920 --> 00:05:08,440 Speaker 2: millions of dollars a month for something that doesn't appear 91 00:05:08,480 --> 00:05:12,120 Speaker 2: to have a measurable return on investment. My friends, the 92 00:05:12,160 --> 00:05:14,920 Speaker 2: business idiots are losing because they never had a plan 93 00:05:15,000 --> 00:05:19,360 Speaker 2: to begin with. As I've said before, Ai is only 94 00:05:19,440 --> 00:05:22,360 Speaker 2: as real as its subsidies. Chat GPT is only free 95 00:05:22,400 --> 00:05:24,720 Speaker 2: to hundreds of billions of people because open ai is 96 00:05:24,760 --> 00:05:27,480 Speaker 2: able to raise hundreds of billions of dollars have Microsoft 97 00:05:27,520 --> 00:05:29,880 Speaker 2: and Google and Amazon and Core. We've built their own 98 00:05:29,920 --> 00:05:34,040 Speaker 2: infrastructure and basically do everything for them, much like Anthropic 99 00:05:34,160 --> 00:05:37,080 Speaker 2: is only able to subsidize its subscribers anywhere from eight 100 00:05:37,120 --> 00:05:39,520 Speaker 2: to thirteen and a half dollars for every dollar of 101 00:05:39,600 --> 00:05:43,159 Speaker 2: revenue because of the endless venture welfare they receive and 102 00:05:43,200 --> 00:05:48,400 Speaker 2: from hyperscalers too. The underlying economics suggests that no subscription 103 00:05:48,480 --> 00:05:51,560 Speaker 2: based AI service, let alone a free one, makes any 104 00:05:51,640 --> 00:05:54,320 Speaker 2: kind of financial sense, and the only reason that everybody 105 00:05:54,360 --> 00:05:57,000 Speaker 2: has had such unrestrained access to AI is because the 106 00:05:57,040 --> 00:05:59,159 Speaker 2: media and the market's approved it, and the people with 107 00:05:59,160 --> 00:06:01,360 Speaker 2: the money are deluded and disconnected from the process of 108 00:06:01,400 --> 00:06:05,679 Speaker 2: value creation on almost every imaginable level. Any statements around 109 00:06:05,720 --> 00:06:09,200 Speaker 2: Anthropic actually being profitable on inference of products of fantasy 110 00:06:09,240 --> 00:06:12,080 Speaker 2: and magical thinking the still copium for people that would 111 00:06:12,120 --> 00:06:14,800 Speaker 2: rather delude themselves into believing that any of this ever 112 00:06:14,839 --> 00:06:19,120 Speaker 2: made sense than face reality or do basic mathematics. Again, 113 00:06:19,400 --> 00:06:21,880 Speaker 2: the assumption is that companies would never just burn a 114 00:06:21,920 --> 00:06:24,240 Speaker 2: lot of money, but that too is catering to the 115 00:06:24,240 --> 00:06:27,400 Speaker 2: grade to myth of executive competence, something that nobody who 116 00:06:27,440 --> 00:06:30,320 Speaker 2: spends any amount of time around managers or CEOs would 117 00:06:30,320 --> 00:06:34,359 Speaker 2: ever believe get hub copilot let people burn thousands of 118 00:06:34,360 --> 00:06:36,800 Speaker 2: dollars on a thirty nine dollars a month subscription as 119 00:06:36,800 --> 00:06:40,520 Speaker 2: a means of expressing growth. I absolutely, one hundred percent 120 00:06:40,520 --> 00:06:43,640 Speaker 2: believe that both open ai and Anthropic are doing exactly 121 00:06:43,680 --> 00:06:46,200 Speaker 2: the same thing, and that neither of them has some 122 00:06:46,400 --> 00:06:49,480 Speaker 2: magical way of making inference cheap enough to justify letting 123 00:06:49,480 --> 00:06:52,080 Speaker 2: people burn thousands of dollars and one hundred or two 124 00:06:52,200 --> 00:06:55,400 Speaker 2: hundred dollars a month subscription. The whole profitable on inference 125 00:06:55,480 --> 00:06:58,039 Speaker 2: thing is a fucking myth, and to give them the 126 00:06:58,040 --> 00:07:00,000 Speaker 2: benefit of the doubt is to empower them to continue 127 00:07:00,120 --> 00:07:03,479 Speaker 2: You're raising money by conning their investors and the general public, 128 00:07:03,560 --> 00:07:06,440 Speaker 2: and to continue perpetuating an era of software that runs 129 00:07:06,520 --> 00:07:10,640 Speaker 2: contrary to what makes technology good. Their goal is simple 130 00:07:10,800 --> 00:07:12,640 Speaker 2: to ram as much of this through to as many 131 00:07:12,680 --> 00:07:14,560 Speaker 2: people as possible, to get them to spend as much 132 00:07:14,600 --> 00:07:18,440 Speaker 2: money as possible, until they work out a way to 133 00:07:18,480 --> 00:07:21,440 Speaker 2: make open AI or Anthropic or these endless data centers 134 00:07:21,480 --> 00:07:25,160 Speaker 2: into something approaching a real business. One of the greatest 135 00:07:25,200 --> 00:07:27,560 Speaker 2: mistakes we can make in our lives is to assume 136 00:07:27,600 --> 00:07:30,040 Speaker 2: that the rich and powerful have any idea what the 137 00:07:30,080 --> 00:07:33,240 Speaker 2: future holds, or that they have any grand plan or strategy. 138 00:07:34,000 --> 00:07:37,160 Speaker 2: It's very likely that Dario Amaday and Sam Mortman's plan 139 00:07:37,240 --> 00:07:39,480 Speaker 2: is to keep burning money until somebody works out for 140 00:07:39,680 --> 00:07:43,000 Speaker 2: them how to stop burning money. And in the interim, 141 00:07:43,040 --> 00:07:44,760 Speaker 2: their plan is to get as many users as they 142 00:07:44,800 --> 00:07:48,560 Speaker 2: can to keep raising money. That's it. That's the entire plan. 143 00:07:48,680 --> 00:07:51,360 Speaker 2: There is no secret thing. They're not building AGI. They 144 00:07:51,360 --> 00:07:53,920 Speaker 2: don't know how to do. Their AGI doesn't exist. They 145 00:07:53,920 --> 00:07:57,560 Speaker 2: don't have secretly super profitable inference. If they did, they 146 00:07:57,600 --> 00:08:00,400 Speaker 2: would only sell model access. That's all they would do, 147 00:08:00,440 --> 00:08:03,480 Speaker 2: and it would be profitable. Right. They don't have that. 148 00:08:03,560 --> 00:08:05,800 Speaker 2: They don't have a plan. There is no secret model 149 00:08:05,800 --> 00:08:09,080 Speaker 2: behind the scenes. Even mythos is a fucking myth. I've 150 00:08:09,080 --> 00:08:11,400 Speaker 2: been over that before. Put a link in the notes. 151 00:08:11,880 --> 00:08:15,680 Speaker 2: And similarly, Microsoft, Google, Meta and Amazon's plan is to 152 00:08:15,720 --> 00:08:17,520 Speaker 2: keep building data centers in the hope they'll have a 153 00:08:17,520 --> 00:08:19,679 Speaker 2: reason to use them by the time they're actually finished. 154 00:08:20,240 --> 00:08:23,440 Speaker 2: There is no other plan. They do not have a 155 00:08:23,480 --> 00:08:26,200 Speaker 2: secret invention coming. They do not have AGI in a 156 00:08:26,240 --> 00:08:28,680 Speaker 2: box in their office. They do not have anything. And 157 00:08:28,720 --> 00:08:31,120 Speaker 2: the reason they're spending so much money and shoving AI 158 00:08:31,160 --> 00:08:33,840 Speaker 2: into everything you use is because they have no fucking 159 00:08:33,840 --> 00:08:37,280 Speaker 2: clue what to do. This is why Warrio Amida makes 160 00:08:37,280 --> 00:08:40,440 Speaker 2: wild claims about AI replacing fifty percent of all white 161 00:08:40,440 --> 00:08:45,000 Speaker 2: collar workers, or Microsoft aiceo Mustafa Suliman claims all white 162 00:08:45,040 --> 00:08:48,160 Speaker 2: collar labor will be automated in eighteen months because the 163 00:08:48,200 --> 00:08:51,719 Speaker 2: products themselves, the actual products aren't impressive enough to win 164 00:08:51,800 --> 00:08:54,160 Speaker 2: you over or to justify the hundreds of billions of 165 00:08:54,200 --> 00:08:57,400 Speaker 2: dollars being sunk into AI. They say these things to 166 00:08:57,440 --> 00:08:59,720 Speaker 2: make you think that they have a scary or powerful 167 00:09:00,000 --> 00:09:04,080 Speaker 2: technology behind the scenes when they do not. And yes, God, yes, 168 00:09:04,480 --> 00:09:08,800 Speaker 2: does that include fucking mythos. I'm so tired of seeing 169 00:09:09,360 --> 00:09:12,079 Speaker 2: news stories about how banks are meeting up to talk 170 00:09:12,120 --> 00:09:15,880 Speaker 2: about the scary monster, the scary Dracula of Mythos. They 171 00:09:15,880 --> 00:09:18,640 Speaker 2: should have called it Dracula. I pretend to take this 172 00:09:18,760 --> 00:09:23,880 Speaker 2: shit seriously, and this forceful harassment grade incursion if AI 173 00:09:24,000 --> 00:09:26,679 Speaker 2: services into our daily lives is not a sign of 174 00:09:26,720 --> 00:09:29,360 Speaker 2: its power, but a gesture of the lack of confidence 175 00:09:29,400 --> 00:09:32,440 Speaker 2: and fear in the hearts of its progenitors. Good shit 176 00:09:32,600 --> 00:09:35,880 Speaker 2: sells by telling you why it's good. Dodgy shit cells 177 00:09:35,920 --> 00:09:38,520 Speaker 2: by tricking and scaring you and taking advantage of business 178 00:09:38,520 --> 00:09:40,319 Speaker 2: idiots who think that using an L and M to 179 00:09:40,400 --> 00:09:42,680 Speaker 2: type emails and spending twelve hours a day on Twitter 180 00:09:42,760 --> 00:09:46,480 Speaker 2: constitutes work. I believe that the vast majority of these 181 00:09:46,559 --> 00:09:49,640 Speaker 2: data centers go unuse and or unfinished, and the most 182 00:09:49,640 --> 00:09:53,240 Speaker 2: AI startups will die once the venture capital subsidies dry up. 183 00:09:54,360 --> 00:09:57,199 Speaker 2: I believe that neither open AI nor anthropic have a future, 184 00:09:57,360 --> 00:09:59,600 Speaker 2: and that their revenues are only made possible through venture 185 00:09:59,600 --> 00:10:03,320 Speaker 2: subsidyce for startups using their models, and the experimental revenue 186 00:10:03,320 --> 00:10:05,520 Speaker 2: of business idiots that don't really know why they're doing 187 00:10:05,559 --> 00:10:09,880 Speaker 2: AI in the first place. AI demand remains a result 188 00:10:09,920 --> 00:10:12,960 Speaker 2: of societal psychosis and a weakness in those who are 189 00:10:12,960 --> 00:10:16,680 Speaker 2: meant to scrutinize the untrustworthy. It's unraveling will be framed 190 00:10:16,679 --> 00:10:19,760 Speaker 2: as impossible to see coming because nobody in power bothered 191 00:10:19,760 --> 00:10:32,240 Speaker 2: to fucking look and look. I've seen some of you 192 00:10:32,280 --> 00:10:35,199 Speaker 2: on the Reddit. I've had some emails from you. You're scared. 193 00:10:35,679 --> 00:10:38,240 Speaker 2: I get him. It's easy to feel hopeless. We're at 194 00:10:38,240 --> 00:10:41,080 Speaker 2: a point where agreed and the shamelessness and the stupidity 195 00:10:41,120 --> 00:10:44,680 Speaker 2: is at its fever pitch. We've reached a time too 196 00:10:44,760 --> 00:10:47,400 Speaker 2: when the maskers started to slip and the c suite 197 00:10:47,440 --> 00:10:50,920 Speaker 2: imbecile classes unabashed and about its loathing of people, as 198 00:10:51,000 --> 00:10:53,000 Speaker 2: was the case when the CEO of a UK bank, 199 00:10:53,160 --> 00:10:56,080 Speaker 2: Bill Winters, talked about how those at risk of losing 200 00:10:56,120 --> 00:10:59,839 Speaker 2: their jobs to AI are lower value human capital the 201 00:11:00,080 --> 00:11:02,240 Speaker 2: same event where he said that the company would likely 202 00:11:02,400 --> 00:11:05,040 Speaker 2: shared nearly eight thousand roles in the coming years due 203 00:11:05,080 --> 00:11:09,000 Speaker 2: to AI. Winters would later apologize for his choice of words, 204 00:11:09,040 --> 00:11:11,320 Speaker 2: although to be clear, he was being absolutely honest when 205 00:11:11,320 --> 00:11:14,200 Speaker 2: he made the remarks. That's what he believes, that's what 206 00:11:14,280 --> 00:11:16,360 Speaker 2: gets him up in the morning, that's what gets him hard. 207 00:11:16,720 --> 00:11:19,160 Speaker 2: Bill Winton's loves him. Now it must be clear for 208 00:11:19,240 --> 00:11:21,480 Speaker 2: legal reasons that was a joke. I don't know about 209 00:11:21,480 --> 00:11:26,600 Speaker 2: Bill Winter's erection. I won't guess it at any further anyway, 210 00:11:27,000 --> 00:11:30,480 Speaker 2: Move it on, though. Everything feels rough because the AI 211 00:11:30,559 --> 00:11:34,960 Speaker 2: industry is equal parts desperate and overconfident. AI executives believe 212 00:11:35,000 --> 00:11:37,480 Speaker 2: that they can cram through enough promises of money into 213 00:11:37,520 --> 00:11:40,199 Speaker 2: the system that the system would rather cannibalize itself than 214 00:11:40,240 --> 00:11:44,760 Speaker 2: admit it made a mistake. Sundarbishai, Andy Jase, Larry Ellison, 215 00:11:44,840 --> 00:11:48,760 Speaker 2: Elon Musk Morg Zuckerberg and Sachinadella will gladly annihilate hundreds 216 00:11:48,760 --> 00:11:52,000 Speaker 2: of billions of dollars to avoid facing the inevitable, but 217 00:11:52,160 --> 00:11:56,600 Speaker 2: once they do, it'll be gruesome. At the same time. 218 00:11:56,679 --> 00:11:59,800 Speaker 2: The things that they need to happen, actual profitability, actual 219 00:11:59,800 --> 00:12:02,240 Speaker 2: time and investment, actual tangible proof that this is a 220 00:12:02,280 --> 00:12:04,800 Speaker 2: real thing rather than something they all have to actively 221 00:12:04,840 --> 00:12:08,920 Speaker 2: conspire to keep alive. Well, they're not happening. Each week 222 00:12:08,960 --> 00:12:11,760 Speaker 2: we hear about new AI megaprojects that will dominate our 223 00:12:11,760 --> 00:12:15,040 Speaker 2: countryside with blinding lights, endless noise, and few belging gas 224 00:12:15,080 --> 00:12:18,200 Speaker 2: turbines at such a scale that it feels impossible it 225 00:12:18,200 --> 00:12:21,640 Speaker 2: could ever stop. The system is absolutely going to try 226 00:12:21,640 --> 00:12:24,760 Speaker 2: and exhaust itself to keep it going. The government bought 227 00:12:24,840 --> 00:12:28,480 Speaker 2: nine billion dollars of Blackwell GPUs, which may seem like 228 00:12:28,600 --> 00:12:30,440 Speaker 2: it's a too big to fail, but what it actually 229 00:12:30,520 --> 00:12:32,560 Speaker 2: is is just the way to keep in Vidia's plates 230 00:12:32,600 --> 00:12:35,520 Speaker 2: spinning for another quarter. Nine billion dollars isn't going to 231 00:12:35,559 --> 00:12:38,800 Speaker 2: be enough. In truth, the amount of money that in 232 00:12:38,840 --> 00:12:41,200 Speaker 2: Nvidia needs to keep this going is so extreme that 233 00:12:41,240 --> 00:12:42,880 Speaker 2: it is now a test of how long the debt 234 00:12:42,880 --> 00:12:46,920 Speaker 2: markets and the hyperscalers can keep sustaining the situation, the hype, 235 00:12:46,920 --> 00:12:50,480 Speaker 2: and the industry itself. A trillion dollars in annual revenue 236 00:12:50,520 --> 00:12:53,200 Speaker 2: is necessary by the end of twenty twenty eight for Invidia, 237 00:12:53,240 --> 00:12:56,720 Speaker 2: which would require over thirty gigawots of actual operational data 238 00:12:56,760 --> 00:12:58,920 Speaker 2: center capacity to be built by then at a time 239 00:12:58,960 --> 00:13:01,040 Speaker 2: when only five gig was It most appears to be 240 00:13:01,120 --> 00:13:04,760 Speaker 2: under construction, and under construction could mean anything from actual 241 00:13:04,800 --> 00:13:09,440 Speaker 2: building happening to a scaffolding yard. Nevertheless, even the sweatiest, 242 00:13:09,520 --> 00:13:12,640 Speaker 2: least trustworthy boosters have begun to sneak in statements about 243 00:13:12,640 --> 00:13:15,199 Speaker 2: how we're probably not in a bubble, or yeah, it's 244 00:13:15,240 --> 00:13:17,679 Speaker 2: a bubble, but it's a good bubble. Jeff Bezos, when 245 00:13:17,679 --> 00:13:19,920 Speaker 2: asked about the AI bubble, said that you shouldn't worry 246 00:13:19,960 --> 00:13:25,079 Speaker 2: about it, which, well, well, Jeff, that that doesn't really 247 00:13:25,120 --> 00:13:27,720 Speaker 2: help anyone. Jeff, I mean, you probably don't need to 248 00:13:27,720 --> 00:13:31,240 Speaker 2: worry about it. You're rich, twap. Anyway, none of this 249 00:13:31,400 --> 00:13:33,319 Speaker 2: is to say that the mood is good. The vibes 250 00:13:33,320 --> 00:13:37,040 Speaker 2: are disastrous. Everybody's exhausted. Those who love AI vibrate with 251 00:13:37,080 --> 00:13:40,320 Speaker 2: a strange soullessness, constantly talking about the incredible power of 252 00:13:40,360 --> 00:13:43,080 Speaker 2: AI without ever showing what it did, or perhaps what 253 00:13:43,160 --> 00:13:48,080 Speaker 2: all that supposed saved time got them. What are you 254 00:13:48,120 --> 00:13:50,560 Speaker 2: doing with the spare time? I don't see any chuckle 255 00:13:50,600 --> 00:13:53,720 Speaker 2: folks providing anything useful to the word. I don't see 256 00:13:53,720 --> 00:13:57,880 Speaker 2: you living lives of leisure. I just see you posting online. 257 00:13:58,360 --> 00:14:00,400 Speaker 2: I can do that without the power of a because 258 00:14:00,400 --> 00:14:03,960 Speaker 2: I have a working brain. What's your fucking excuse, good lord. 259 00:14:04,920 --> 00:14:07,200 Speaker 2: But one thing I'll add though, that adds to the 260 00:14:07,280 --> 00:14:09,560 Speaker 2: malaise is that everyone I talk to to a lot 261 00:14:09,559 --> 00:14:12,800 Speaker 2: of people at hyperscales, they're all fucking miserable. It's terrible there. 262 00:14:12,840 --> 00:14:15,720 Speaker 2: The business idiots are running the show, and they're both saying, 263 00:14:15,760 --> 00:14:18,280 Speaker 2: spend more money on AI, but also don't spend quite 264 00:14:18,280 --> 00:14:20,280 Speaker 2: as much money on AI because it costs a lot. 265 00:14:20,880 --> 00:14:23,120 Speaker 2: Nobody really knows why they're doing this other than the 266 00:14:23,200 --> 00:14:27,000 Speaker 2: fact it's what everyone else is doing. And basically every 267 00:14:27,160 --> 00:14:30,400 Speaker 2: person in every job has had somebody intimate that they're 268 00:14:30,440 --> 00:14:32,560 Speaker 2: going to lose their job to a computer every time 269 00:14:32,560 --> 00:14:35,120 Speaker 2: they open a newspaper or use a website, and every 270 00:14:35,160 --> 00:14:37,640 Speaker 2: app has some sort of desperate, vulgar pop up about 271 00:14:37,680 --> 00:14:40,760 Speaker 2: a feature that will generate some bullshit officscating the features 272 00:14:40,800 --> 00:14:42,560 Speaker 2: you actually want to use in favor of those that 273 00:14:42,680 --> 00:14:46,320 Speaker 2: might lose the company money, because the company has to 274 00:14:46,360 --> 00:14:48,640 Speaker 2: prove to the people that invested in the company that 275 00:14:48,680 --> 00:14:52,120 Speaker 2: they're futuristic and have the AI integrations, even though no 276 00:14:52,120 --> 00:14:56,479 Speaker 2: one likes them and they only seem to lose money. Alternatively, 277 00:14:56,520 --> 00:14:59,160 Speaker 2: their CEO either has my old or severe AI sside 278 00:14:59,200 --> 00:15:01,560 Speaker 2: coosis to the point they have decided to violate your 279 00:15:01,640 --> 00:15:05,800 Speaker 2: user experience because AI is a non consensual technology at 280 00:15:05,840 --> 00:15:09,680 Speaker 2: its heart, this is what it is. Nobody asked for this, 281 00:15:10,080 --> 00:15:13,520 Speaker 2: no one wants this. We've had it forced upon us. 282 00:15:13,880 --> 00:15:16,760 Speaker 2: Even the people excited by it would rather do something else. 283 00:15:17,440 --> 00:15:20,200 Speaker 2: And all of that excitement doesn't seem to amount to anything. 284 00:15:21,120 --> 00:15:23,680 Speaker 2: And they all seem desperate, and they seem anxious and 285 00:15:23,880 --> 00:15:28,560 Speaker 2: scared and vulgar in the way they attack their detractors. 286 00:15:29,280 --> 00:15:33,800 Speaker 2: And that's because they're losing. They all know it. They're desperate, 287 00:15:33,840 --> 00:15:37,160 Speaker 2: they're scared. They're scared that the thing that they mortgage 288 00:15:37,200 --> 00:15:41,400 Speaker 2: their companies and careers and reputations on is all falling apart. 289 00:15:43,280 --> 00:15:46,040 Speaker 2: And it seems that there are nearly as many announcements 290 00:15:46,040 --> 00:15:49,120 Speaker 2: of new large data center developments as there are cancelations 291 00:15:49,120 --> 00:15:53,000 Speaker 2: of said data center projects. While hyperscolars can dismiss that 292 00:15:53,120 --> 00:15:56,200 Speaker 2: as a simple reallocation of capital and nothing to worry about, 293 00:15:56,440 --> 00:15:59,280 Speaker 2: it's hard to ignore the growing backlash against these facilities 294 00:15:59,320 --> 00:16:02,040 Speaker 2: from locals and the success that these locals have had, 295 00:16:02,560 --> 00:16:06,760 Speaker 2: especially in blocking multiple different developments or setting up actual 296 00:16:07,440 --> 00:16:11,080 Speaker 2: what county wide and statewide moratoriums depending on where you look. 297 00:16:11,960 --> 00:16:15,160 Speaker 2: And it gets worse. Anthropic had to conspire with Elon 298 00:16:15,360 --> 00:16:18,680 Speaker 2: fucking Musk to conjure up a single profitable quarter with 299 00:16:18,800 --> 00:16:22,200 Speaker 2: SpaceX deliberately discounting two months of compute costs as a 300 00:16:22,240 --> 00:16:25,160 Speaker 2: means of swindling the media and its investors. One last time, 301 00:16:25,520 --> 00:16:28,320 Speaker 2: I swear to God, the object permanence in the valley 302 00:16:28,440 --> 00:16:32,120 Speaker 2: is not so good, folks. But in response, open Ai 303 00:16:32,240 --> 00:16:34,400 Speaker 2: either leaked or had leaked that it had a negative 304 00:16:34,440 --> 00:16:37,600 Speaker 2: one hundred and twenty two percent non gap operating margin 305 00:16:37,640 --> 00:16:40,480 Speaker 2: in Q one twenty twenty six and chat GPT growth 306 00:16:40,480 --> 00:16:44,960 Speaker 2: that's stalled. Anthropic is either the single most successful grifter 307 00:16:45,080 --> 00:16:46,920 Speaker 2: of all time or speed running a con where it 308 00:16:46,960 --> 00:16:49,360 Speaker 2: fudget together numbers to raise endless amounts of money to 309 00:16:49,440 --> 00:16:53,360 Speaker 2: keep its billion or trillion dollar burn going. These are 310 00:16:53,360 --> 00:16:56,440 Speaker 2: not the actions of honest, sustainable companies that will exist 311 00:16:56,440 --> 00:16:59,320 Speaker 2: in the future. I believe that we are on course 312 00:16:59,360 --> 00:17:02,320 Speaker 2: for a truly horrible crash, the likes of which may 313 00:17:02,480 --> 00:17:06,159 Speaker 2: rewrite the venture capital industry and mortally wound one or 314 00:17:06,200 --> 00:17:09,720 Speaker 2: more hyperscalers, as well as fundamentally divide society on so 315 00:17:09,800 --> 00:17:12,440 Speaker 2: many levels into those that fall for this and those 316 00:17:12,480 --> 00:17:15,879 Speaker 2: that did not. This will, in the short term be 317 00:17:16,000 --> 00:17:19,040 Speaker 2: absolutely fucking horrible for our markets in our wider economy 318 00:17:19,119 --> 00:17:21,439 Speaker 2: is a result of the time bomb of private credit 319 00:17:21,560 --> 00:17:25,119 Speaker 2: and private equity and their failed investments in software and 320 00:17:25,160 --> 00:17:28,879 Speaker 2: their failing investments in data centers. In the long term, 321 00:17:29,000 --> 00:17:30,800 Speaker 2: I see it as that they live moment for many 322 00:17:30,800 --> 00:17:33,720 Speaker 2: millions of secret imbeciles and credits in our mist and 323 00:17:33,800 --> 00:17:36,200 Speaker 2: I don't think it'll be easy to wash the stench off. 324 00:17:36,200 --> 00:17:38,760 Speaker 2: For those that really pledge themselves to the graveyard smash 325 00:17:38,760 --> 00:17:43,240 Speaker 2: of AI, we will win long term, what they are 326 00:17:43,240 --> 00:17:46,240 Speaker 2: doing is not working. The future will not be without pain, 327 00:17:46,320 --> 00:17:48,560 Speaker 2: nor will it be easy or pleasant or something I 328 00:17:48,560 --> 00:17:51,120 Speaker 2: will relish in. But in the long term, I think 329 00:17:51,119 --> 00:17:53,760 Speaker 2: this is a moment where the greater business idiot incursion 330 00:17:54,000 --> 00:17:56,400 Speaker 2: faces a reckoning with a reality it believed it could 331 00:17:56,440 --> 00:17:59,760 Speaker 2: change through sheer force of will. These people don't know 332 00:17:59,800 --> 00:18:02,080 Speaker 2: how to build things that work anymore, and that's the 333 00:18:02,080 --> 00:18:04,639 Speaker 2: only thing they can do is spend money in fire people. 334 00:18:05,280 --> 00:18:08,080 Speaker 2: They believe in nothing other than growth, and one cannot 335 00:18:08,080 --> 00:18:11,679 Speaker 2: exist on belief in hyperlone, at least not forever. And 336 00:18:11,760 --> 00:18:14,399 Speaker 2: I can't wait to watch what happens when it collapses. 337 00:18:22,760 --> 00:18:24,440 Speaker 2: Thank you for listening to Better Offline. 338 00:18:24,560 --> 00:18:27,000 Speaker 3: The editor and composer of the Better Offline theme song 339 00:18:27,080 --> 00:18:29,720 Speaker 3: is Matasowski. You can check out more of his music 340 00:18:29,760 --> 00:18:33,400 Speaker 3: and audio projects at Matasowski dot com, M A T 341 00:18:33,400 --> 00:18:34,000 Speaker 3: T O. 342 00:18:34,200 --> 00:18:37,679 Speaker 2: S O w Ski dot com. 343 00:18:37,760 --> 00:18:40,080 Speaker 3: You can email me at easy at Better offline dot 344 00:18:40,119 --> 00:18:42,320 Speaker 3: com or visit Better Offline dot com to find more 345 00:18:42,359 --> 00:18:45,760 Speaker 3: podcast links and of course, my newsletter. I also really 346 00:18:45,760 --> 00:18:48,080 Speaker 3: recommend you go to chat dot Where's youreaed dot at 347 00:18:48,080 --> 00:18:50,520 Speaker 3: to visit the discord, and go to our slash Better 348 00:18:50,560 --> 00:18:53,720 Speaker 3: Offline to check out our reddit. Thank you so much 349 00:18:53,760 --> 00:18:54,280 Speaker 3: for listening. 350 00:18:55,119 --> 00:18:57,760 Speaker 1: Better Offline is a production of cool Zone Media. For 351 00:18:57,920 --> 00:19:00,560 Speaker 1: more from cool Zone Media, visit our web site cool 352 00:19:00,640 --> 00:19:03,960 Speaker 1: Zonemedia dot com or check us out on the iHeartRadio app, 353 00:19:04,040 --> 00:19:06,440 Speaker 1: Apple Podcasts, or wherever you get your podcasts.