WEBVTT - Monologue: AI's Demand Story Is A Lie

0:00:02.800 --> 0:00:08.160
<v Speaker 1>Zone Media hell On, Welcome to this week's Better Offline Monologue.

0:00:08.200 --> 0:00:18.640
<v Speaker 1>I'm your host ed Zi Trump. It's earning season and

0:00:18.680 --> 0:00:21.920
<v Speaker 1>then the quarterly drawn through the psychosis of financial analysts

0:00:21.960 --> 0:00:23.759
<v Speaker 1>that have worked for years to prop up an AI

0:00:23.800 --> 0:00:27.840
<v Speaker 1>industry with questionable returns, spending now over seven hundred and

0:00:27.840 --> 0:00:30.960
<v Speaker 1>fifty billion dollars in this year. I believe on capex

0:00:31.240 --> 0:00:34.919
<v Speaker 1>for well, no real return on investment. In fact, all

0:00:34.920 --> 0:00:37.519
<v Speaker 1>of them seemed to be losing money. I know, I know,

0:00:37.600 --> 0:00:39.120
<v Speaker 1>though some of you are going to say, but these

0:00:39.159 --> 0:00:41.960
<v Speaker 1>businesses had great quarters. But you'll nose is that said

0:00:42.040 --> 0:00:44.400
<v Speaker 1>quarterly growth does not actually seem to be coming from

0:00:44.440 --> 0:00:47.320
<v Speaker 1>AI when you read their results where they actually share

0:00:47.360 --> 0:00:51.880
<v Speaker 1>the AI roic return on investing capital. So, while Meta

0:00:51.920 --> 0:00:54.680
<v Speaker 1>and Google refused to actually explain their AI returns because

0:00:54.760 --> 0:00:57.680
<v Speaker 1>that would require admitting how bad they were, Microsoft revealed

0:00:57.680 --> 0:01:00.440
<v Speaker 1>that he had thirty seven billion dollars in AI revenue

0:01:00.520 --> 0:01:02.680
<v Speaker 1>run rate about three point oh eight billion dollars a

0:01:02.720 --> 0:01:05.640
<v Speaker 1>month or so, and Amazon have fifteen billion dollars annualized

0:01:05.720 --> 0:01:08.000
<v Speaker 1>or around one point two five billion dollars a month.

0:01:08.520 --> 0:01:12.480
<v Speaker 1>Do not be fooled. These are bad numbers based on

0:01:12.520 --> 0:01:15.119
<v Speaker 1>my own reporting, Open ai spent three point six four

0:01:15.200 --> 0:01:17.480
<v Speaker 1>eight billion dollars on inference in the third quarter of

0:01:17.520 --> 0:01:20.959
<v Speaker 1>twenty twenty five on Microsoft Azure, or around fourteen point

0:01:20.959 --> 0:01:23.920
<v Speaker 1>four billion dollars on an annualized basis. While I only

0:01:23.920 --> 0:01:26.880
<v Speaker 1>add half of the fourth quarter numbers, I estimate based

0:01:26.880 --> 0:01:29.559
<v Speaker 1>on them that open AI's annualized spent hit over eighteen

0:01:29.600 --> 0:01:32.080
<v Speaker 1>point five billion dollars or around four point six billion

0:01:32.080 --> 0:01:34.319
<v Speaker 1>dollars a quarter by the end of twenty twenty five,

0:01:34.560 --> 0:01:36.920
<v Speaker 1>and that's not accounting for things like Sorra two or

0:01:36.959 --> 0:01:40.440
<v Speaker 1>a launch of its code ex coding platform. In total,

0:01:40.560 --> 0:01:43.399
<v Speaker 1>this puts open ai and estimated thirteen billion dollars of

0:01:43.480 --> 0:01:47.160
<v Speaker 1>spend on Azure just on inference in twenty twenty five,

0:01:47.640 --> 0:01:52.520
<v Speaker 1>with billions of dollars more on training models. Now, Microsoft

0:01:52.560 --> 0:01:54.920
<v Speaker 1>also accounted for sixty seven percent of core we's five

0:01:54.920 --> 0:01:57.560
<v Speaker 1>point one five billion dollars in twenty twenty five revenue,

0:01:57.760 --> 0:02:01.720
<v Speaker 1>or around three point four five billion dollars, and as

0:02:02.000 --> 0:02:04.800
<v Speaker 1>all of that is used by open Ai, I really

0:02:04.840 --> 0:02:08.240
<v Speaker 1>think you need to think very deeply about well, how

0:02:08.320 --> 0:02:10.919
<v Speaker 1>much AI demand there really is outside of two very

0:02:11.000 --> 0:02:14.799
<v Speaker 1>large fail suns. I'm also confident that Corewave's computer is

0:02:14.919 --> 0:02:18.079
<v Speaker 1>used for training open AI's models, as Corewave's announcement related

0:02:18.080 --> 0:02:20.720
<v Speaker 1>to its direct deal with open ai specifically said it

0:02:20.760 --> 0:02:23.240
<v Speaker 1>was contracted and I quote to power the training of

0:02:23.240 --> 0:02:26.960
<v Speaker 1>open AI's most advanced next generation models, and said capacity

0:02:27.000 --> 0:02:30.440
<v Speaker 1>per semaphore was only available because Microsoft declined to extend

0:02:30.480 --> 0:02:33.320
<v Speaker 1>its current agreement with core Weave to use that compute

0:02:33.360 --> 0:02:37.000
<v Speaker 1>for open Ai altogether. That puts open ai spend on

0:02:37.040 --> 0:02:40.680
<v Speaker 1>Microsoft services at over eighteen billion dollars in twenty twenty five,

0:02:40.760 --> 0:02:42.800
<v Speaker 1>and it's easy to see how that could grow to

0:02:42.880 --> 0:02:45.600
<v Speaker 1>over twenty four billion dollars on an analyzed basis in

0:02:45.600 --> 0:02:48.600
<v Speaker 1>the last quarter, or around two billion dollars a month.

0:02:49.360 --> 0:02:52.760
<v Speaker 1>Microsoft is open AI's primary cloud provider, and I estimate

0:02:52.800 --> 0:02:55.440
<v Speaker 1>the open Ai represents around seventy percent of its AI

0:02:55.520 --> 0:02:58.160
<v Speaker 1>revenue if not more, while taking up the majority of

0:02:58.160 --> 0:03:02.440
<v Speaker 1>its AI infrastructure. Otherwise, Microsoft's twenty million Copilot three sixty

0:03:02.480 --> 0:03:05.120
<v Speaker 1>five subscribers likely pay no more than seven billion dollars

0:03:05.120 --> 0:03:07.760
<v Speaker 1>a year. I'll get to my feelings about that in

0:03:07.800 --> 0:03:11.000
<v Speaker 1>a bit now. As far as Amazon goes, things get

0:03:11.000 --> 0:03:14.880
<v Speaker 1>a lot grimmar. In early April, per Reuter's Andy Jesse,

0:03:15.280 --> 0:03:19.040
<v Speaker 1>their CEO admitted that its cloud business's AI revenue run

0:03:19.120 --> 0:03:21.640
<v Speaker 1>rate was more than fifteen billion dollars in the first

0:03:21.680 --> 0:03:24.160
<v Speaker 1>quart of twenty twenty six, which translates to about one

0:03:24.160 --> 0:03:27.200
<v Speaker 1>point twenty five billion dollars in monthly revenue, or roughly

0:03:27.320 --> 0:03:29.520
<v Speaker 1>zero point four to one nine percent of the two

0:03:29.639 --> 0:03:32.160
<v Speaker 1>hundred and ninety eight point three billion dollars in capex

0:03:32.200 --> 0:03:35.400
<v Speaker 1>Amazon spent so far on the AI bubble, or around

0:03:35.400 --> 0:03:37.800
<v Speaker 1>twenty five percent of the five billion dollars Amazon just

0:03:37.840 --> 0:03:43.120
<v Speaker 1>invested in Anthropic last week. I think it's also reasonable

0:03:43.160 --> 0:03:45.280
<v Speaker 1>to assume that a large part, if not a majority,

0:03:45.320 --> 0:03:48.960
<v Speaker 1>of that revenue comes from Anthropic. Per my reporting last year,

0:03:49.080 --> 0:03:52.080
<v Speaker 1>Anthropic spent five hundred and eighteen point nine million dollars

0:03:52.080 --> 0:03:55.640
<v Speaker 1>on Amazon Web services in September twenty twenty five, at

0:03:55.680 --> 0:03:57.640
<v Speaker 1>a time when it had around seven billion dollars in

0:03:57.680 --> 0:04:00.520
<v Speaker 1>annualized revenue, a figure that increased by five hundred percent

0:04:00.520 --> 0:04:03.160
<v Speaker 1>if you believe Anthropic, which I don't know if I do,

0:04:03.320 --> 0:04:06.360
<v Speaker 1>to thirty billion dollars in annualized revenue since as of April,

0:04:07.320 --> 0:04:10.200
<v Speaker 1>five hundred and eighteen point nine million dollars is about

0:04:10.200 --> 0:04:12.640
<v Speaker 1>six point two billion dollars in annualized spend, and I

0:04:12.640 --> 0:04:14.760
<v Speaker 1>think it's fair to assume that its spend will have

0:04:14.840 --> 0:04:18.279
<v Speaker 1>at least doubled to twelve billion dollars in annualized AI

0:04:18.320 --> 0:04:21.960
<v Speaker 1>revenue for Amazon, or around eighty percent of the AI revenue.

0:04:23.200 --> 0:04:27.120
<v Speaker 1>I also want to address this continual claim that capacity

0:04:27.160 --> 0:04:31.400
<v Speaker 1>constraints in AI are proof of some incredible demand. In reality,

0:04:31.440 --> 0:04:33.680
<v Speaker 1>the reason that there's so little available compute is that

0:04:33.720 --> 0:04:35.679
<v Speaker 1>the majority of it is taken up by either open

0:04:35.680 --> 0:04:38.840
<v Speaker 1>ai Andthropic or hyperscalers running their own services with the

0:04:38.920 --> 0:04:42.119
<v Speaker 1>dregs of what's left behind. Think of it like seeing

0:04:42.160 --> 0:04:44.880
<v Speaker 1>a very very large man on the bus taking up

0:04:45.320 --> 0:04:49.000
<v Speaker 1>four seats. There are absolutely capacity issues on this bus,

0:04:49.000 --> 0:04:53.400
<v Speaker 1>but they're not a result of everybody wanting to ride it. Now. Epokai,

0:04:53.560 --> 0:04:56.720
<v Speaker 1>who I have had my differences with, estimate and I'm

0:04:56.800 --> 0:04:58.800
<v Speaker 1>using these numbers because there's fuck all out there as

0:04:58.880 --> 0:05:02.240
<v Speaker 1>far as compute capacity goes. They estimate the Microsoft had

0:05:02.279 --> 0:05:05.279
<v Speaker 1>around two point oh two gigawatts of compute capacity available

0:05:05.320 --> 0:05:07.600
<v Speaker 1>at the end of twenty twenty five, a time when

0:05:07.680 --> 0:05:10.040
<v Speaker 1>open ai said it had access to around one point

0:05:10.120 --> 0:05:13.719
<v Speaker 1>nine gigawats of compute. As Microsoft is open AI's primary

0:05:13.720 --> 0:05:15.920
<v Speaker 1>cloud provider, and Oracle had yet to stand up more

0:05:15.960 --> 0:05:19.200
<v Speaker 1>than one hundred megawatts of stargate abeline. I think it's

0:05:19.240 --> 0:05:21.560
<v Speaker 1>safe to assume that open Ai takes up eighty to

0:05:21.720 --> 0:05:27.360
<v Speaker 1>ninety percent of Microsoft's available capacity. Similarly, Epoch estimates that

0:05:27.600 --> 0:05:30.279
<v Speaker 1>by end of twenty twenty five, Amazon had access to

0:05:30.279 --> 0:05:33.080
<v Speaker 1>an estimated one point sixty seven gigawatts of capacity, and

0:05:33.240 --> 0:05:36.800
<v Speaker 1>I estimate over seventy five percent it's taken up by anthropic.

0:05:37.400 --> 0:05:41.120
<v Speaker 1>Amazon's anthropic dedicated project Rainier Data Center accounted for five

0:05:41.200 --> 0:05:43.359
<v Speaker 1>hundred megawats by the end of that year, around a

0:05:43.400 --> 0:05:46.360
<v Speaker 1>third of that capacity, and considering anthropic spend, I think

0:05:46.400 --> 0:05:48.560
<v Speaker 1>it's reasonable to believe it's taking up the vast majority

0:05:48.600 --> 0:05:51.320
<v Speaker 1>of what's available. Of course, there are capacity issues. You're

0:05:51.360 --> 0:05:55.400
<v Speaker 1>selling all the capacity to one fucking company. EPOC also

0:05:55.720 --> 0:05:58.720
<v Speaker 1>estimates the Google had around two point ninety five gigawats

0:05:59.240 --> 0:06:02.720
<v Speaker 1>uh remark figure that also realized on literally doubling its

0:06:02.720 --> 0:06:05.719
<v Speaker 1>capacity since the beginning of twenty twenty five. Nevertheless, it

0:06:05.920 --> 0:06:08.240
<v Speaker 1>likely uses a lot of its GPUs for anthropic too,

0:06:08.279 --> 0:06:10.280
<v Speaker 1>but it's hard to estimate how much as we have

0:06:10.400 --> 0:06:13.640
<v Speaker 1>very little insight into anthropic spent on Google Cloud outside

0:06:13.640 --> 0:06:15.840
<v Speaker 1>of its promise to use and I quote after one

0:06:15.880 --> 0:06:18.559
<v Speaker 1>million TPUs with over a gigawad of capacity coming online

0:06:18.600 --> 0:06:20.680
<v Speaker 1>in twenty twenty six, from the end of last year

0:06:21.000 --> 0:06:22.680
<v Speaker 1>and a month ago when it said it would use

0:06:22.720 --> 0:06:26.159
<v Speaker 1>and I quote multiple gigawats at next generation TPU capacity

0:06:26.320 --> 0:06:29.880
<v Speaker 1>starting in twenty twenty seven. I also think it's fair

0:06:29.920 --> 0:06:32.480
<v Speaker 1>to assume that large swards of Google's compute infrastructure is

0:06:32.480 --> 0:06:34.680
<v Speaker 1>taken up by Google Gemini being shoved into literally every

0:06:34.720 --> 0:06:38.039
<v Speaker 1>single part of Google's products. It's everywhere. It's like a

0:06:38.080 --> 0:06:42.400
<v Speaker 1>fucking weed. It's horrible. Google also refuses to break out

0:06:42.400 --> 0:06:45.559
<v Speaker 1>its AI revenues, but considering it owns around thirteen percent

0:06:45.600 --> 0:06:47.599
<v Speaker 1>of anthropic I also think it's fair to assume a

0:06:47.680 --> 0:06:52.200
<v Speaker 1>large capacity amount goes to them. All Right, you know what,

0:06:53.560 --> 0:06:56.920
<v Speaker 1>I tried to be nice. I tried to keep my cool,

0:06:57.000 --> 0:06:59.160
<v Speaker 1>but this whole situation is beginning to piss me off.

0:06:59.200 --> 0:07:02.880
<v Speaker 1>I'm tired out of hearing people online may do this

0:07:03.000 --> 0:07:05.680
<v Speaker 1>victory lap thing about how AI demand is there when

0:07:05.720 --> 0:07:09.360
<v Speaker 1>it isn't. And when I started these monologues, I was

0:07:09.360 --> 0:07:11.440
<v Speaker 1>told I'd be allowed to rent, and I haven't really

0:07:11.520 --> 0:07:13.520
<v Speaker 1>let myself off the chain, so I think it's time

0:07:13.560 --> 0:07:17.600
<v Speaker 1>to do so. The story of massive AI demand is

0:07:17.640 --> 0:07:21.440
<v Speaker 1>a lie and nearly a trillion dollars, if not more, annihilated.

0:07:21.480 --> 0:07:24.520
<v Speaker 1>To create the largest circle joke of all time. Venture

0:07:24.600 --> 0:07:27.760
<v Speaker 1>capitalists and hyperscalers feed money to open ai and Anthropic

0:07:27.840 --> 0:07:30.400
<v Speaker 1>so that vengre capitalists can feed money to startups to

0:07:30.480 --> 0:07:33.920
<v Speaker 1>feed money back to open Ai and Anthropic to feed

0:07:33.960 --> 0:07:37.080
<v Speaker 1>back into hyperscalers for GPU compute, at which point they

0:07:37.080 --> 0:07:40.800
<v Speaker 1>turn to Jensen, Hogovinvidia and buy more GPUs. While it

0:07:40.880 --> 0:07:43.880
<v Speaker 1>might seem tempting to credit these men as geniuses for

0:07:43.960 --> 0:07:48.640
<v Speaker 1>creating companies specifically to feed themselves revenue, well, but to

0:07:48.720 --> 0:07:51.120
<v Speaker 1>keep up with this k fabe of doing AI to

0:07:51.160 --> 0:07:53.800
<v Speaker 1>substantiate the build out means that they've had to massively

0:07:53.840 --> 0:07:55.920
<v Speaker 1>over commit to the bit even though the only two

0:07:56.040 --> 0:07:59.720
<v Speaker 1>meaningful businesses in AI appear to be Anthropic and Open Ai,

0:07:59.800 --> 0:08:03.600
<v Speaker 1>and that's only because they're effectively intellectual honeypots for the

0:08:03.760 --> 0:08:07.000
<v Speaker 1>entire industry. Outside of those two, the only other competitive

0:08:07.040 --> 0:08:10.600
<v Speaker 1>AI businesses are those of Amazon, Microsoft, and Google, two

0:08:10.680 --> 0:08:13.240
<v Speaker 1>of which have an annualized revenue of around six percent

0:08:13.280 --> 0:08:17.440
<v Speaker 1>of their capital expenditures. So far, Google's AI business is

0:08:17.480 --> 0:08:19.840
<v Speaker 1>so booming that it refuses to break it out. And

0:08:19.840 --> 0:08:22.760
<v Speaker 1>while it nebulously claims that AI is creating growth, it's

0:08:22.800 --> 0:08:25.000
<v Speaker 1>not really clear how, and they're vague about it because

0:08:25.000 --> 0:08:27.160
<v Speaker 1>analysts and the media are ready to swallow the narrative

0:08:27.160 --> 0:08:29.680
<v Speaker 1>as long as number go up. That's why Google doesn't

0:08:29.680 --> 0:08:33.480
<v Speaker 1>break out their AI revenues by the fucking way. Amazon

0:08:33.520 --> 0:08:35.800
<v Speaker 1>and Microsoft had their hands forced by the markets after

0:08:35.840 --> 0:08:39.040
<v Speaker 1>their stocks tumbled and fucked up by sharing their AI revenues.

0:08:39.480 --> 0:08:42.400
<v Speaker 1>Amazon's two hundred and ninety eight point three billion dollars

0:08:42.400 --> 0:08:45.560
<v Speaker 1>in capex has successfully created an AI business that and

0:08:45.600 --> 0:08:47.760
<v Speaker 1>more than a quarter of a way to a trillion dollars,

0:08:48.120 --> 0:08:50.800
<v Speaker 1>has successfully managed to make one point two five billion

0:08:50.840 --> 0:08:54.920
<v Speaker 1>dollars a month. That's fucking pathetic. If we had investors

0:08:54.920 --> 0:08:58.480
<v Speaker 1>with IQs above room temperature, they'd run Anty Jassey out

0:08:58.480 --> 0:09:02.480
<v Speaker 1>of Arlington like fucking But this is the AI bubble

0:09:02.600 --> 0:09:06.480
<v Speaker 1>where everybody is wrong, because once everybody admits what's actually happening,

0:09:06.600 --> 0:09:09.000
<v Speaker 1>they're gonna have to admit they all sounded insane for years.

0:09:09.520 --> 0:09:14.040
<v Speaker 1>One point two five billion dollars a month after nearly

0:09:14.120 --> 0:09:17.760
<v Speaker 1>three hundred billion dollars of Capex. Unbelievable, what a joke.

0:09:18.559 --> 0:09:21.600
<v Speaker 1>And like I said, more than seventy five percent, if

0:09:21.640 --> 0:09:24.079
<v Speaker 1>not eighty percent or more of that revenue is from

0:09:24.200 --> 0:09:27.320
<v Speaker 1>fucking anthropic who Amazon just agreed to fund up to

0:09:27.360 --> 0:09:32.920
<v Speaker 1>another twenty five billion dollars. At some point, it's just Microsoft, Google,

0:09:32.960 --> 0:09:35.680
<v Speaker 1>and Amazon handing themselves money or handing it to Nvidia

0:09:35.760 --> 0:09:38.840
<v Speaker 1>to justify what they've burned over eight hundred billion fucking

0:09:38.880 --> 0:09:41.959
<v Speaker 1>dollars creating some of the most wretched businesses of all time.

0:09:42.760 --> 0:09:46.840
<v Speaker 1>I'm sorry, Wow, Satcha. You managed to get up to

0:09:47.000 --> 0:09:52.520
<v Speaker 1>twenty million paying Microsoft three sixty five copilot subscriptions six

0:09:52.640 --> 0:09:56.080
<v Speaker 1>hundred million dollars a month in revenue, not profit, and

0:09:56.160 --> 0:09:58.920
<v Speaker 1>all it took was you investing more than thirteen billion

0:09:58.960 --> 0:10:02.480
<v Speaker 1>dollars in open forcing large language models into every one

0:10:02.480 --> 0:10:03.920
<v Speaker 1>of your products in a way that brought us on

0:10:03.960 --> 0:10:07.520
<v Speaker 1>harassment and about what two hundred and forty something billion

0:10:07.559 --> 0:10:10.000
<v Speaker 1>dollars in Capex, as well as laying off tens of

0:10:10.040 --> 0:10:13.160
<v Speaker 1>thousands of people and savaging the Xbox brand. What the

0:10:13.280 --> 0:10:26.920
<v Speaker 1>fucking shit, man, You should be ashamed of yourself. Open

0:10:26.960 --> 0:10:30.720
<v Speaker 1>AI is in and of itself a kind of psychosis generator.

0:10:31.400 --> 0:10:33.120
<v Speaker 1>It was the first thing in a long time that

0:10:33.160 --> 0:10:35.240
<v Speaker 1>felt like a new thing since the iPhone for the

0:10:35.280 --> 0:10:38.439
<v Speaker 1>people that entirely obsess over growth. It was the panacea

0:10:38.480 --> 0:10:41.040
<v Speaker 1>for the entire tech industry, creating a new way for

0:10:41.080 --> 0:10:43.880
<v Speaker 1>business idiots to spend money on infrastructure, a new thing

0:10:44.000 --> 0:10:46.800
<v Speaker 1>for consultants to scam people, with a new series of

0:10:46.800 --> 0:10:49.120
<v Speaker 1>things to be an expert in as a journalist, the blogger,

0:10:49.320 --> 0:10:51.800
<v Speaker 1>or just a scam artist, all wrapped up in something

0:10:51.840 --> 0:10:55.360
<v Speaker 1>that could also be a consumer product and enterprise software products,

0:10:55.559 --> 0:10:57.680
<v Speaker 1>and a new kind of API to attach to other

0:10:57.840 --> 0:11:02.320
<v Speaker 1>enterprise software products and then arch more money. In theory,

0:11:02.520 --> 0:11:06.839
<v Speaker 1>open AI's success would lift everything at once, hardware, software

0:11:06.880 --> 0:11:10.680
<v Speaker 1>and even adjacent fields. It promised to both democratize access

0:11:10.720 --> 0:11:14.720
<v Speaker 1>to creating software while also heavily reinforcing existing power structures

0:11:14.720 --> 0:11:17.240
<v Speaker 1>to the point that every dollar inevitably ended up in

0:11:17.280 --> 0:11:21.400
<v Speaker 1>the Magnificent seven's pocket or asshole. The problem is that

0:11:21.440 --> 0:11:24.319
<v Speaker 1>the system needed to actually work one day. It needed

0:11:24.360 --> 0:11:27.600
<v Speaker 1>to eventually make more money than it cost. Every single

0:11:27.720 --> 0:11:30.800
<v Speaker 1>one of these companies is talking about AI NonStop, and

0:11:30.840 --> 0:11:33.240
<v Speaker 1>not one of them can show you a profit. The

0:11:33.320 --> 0:11:35.440
<v Speaker 1>only thing they can do is tell Liz of omission

0:11:35.480 --> 0:11:38.200
<v Speaker 1>by saying AI helped boost everything, and when you ask

0:11:38.320 --> 0:11:41.520
<v Speaker 1>for specifics, well, the results are either tepid or so

0:11:41.640 --> 0:11:44.880
<v Speaker 1>secretive you'd think they were hiding a dead body. The

0:11:44.920 --> 0:11:48.199
<v Speaker 1>only reason Google, Amazon and Microsoft are being tolerated that

0:11:48.240 --> 0:11:51.400
<v Speaker 1>their current excess is because their non AI segments continue

0:11:51.440 --> 0:11:55.040
<v Speaker 1>to grow through endless price increases and in Sitification and

0:11:55.080 --> 0:11:57.560
<v Speaker 1>its other business units, by which I mean Open AI

0:11:57.640 --> 0:11:59.720
<v Speaker 1>and Anthropic are yet to run out of money to

0:11:59.760 --> 0:12:03.839
<v Speaker 1>feed them. Sorry, By the way, I just I don't

0:12:03.840 --> 0:12:06.440
<v Speaker 1>know what Meta's doing. I haven't talked about very much

0:12:06.480 --> 0:12:09.120
<v Speaker 1>because I don't think Meta knows what Meta is doing.

0:12:09.679 --> 0:12:11.560
<v Speaker 1>Every so often they bury effect in one of their

0:12:11.640 --> 0:12:14.840
<v Speaker 1>blogs about they saw a three percent increase in something

0:12:14.880 --> 0:12:17.720
<v Speaker 1>related to AI. Then they promised to burn another one

0:12:17.800 --> 0:12:20.240
<v Speaker 1>hundred and seventy billion dollars in a year, and again

0:12:20.280 --> 0:12:23.880
<v Speaker 1>we can't really tell why. They also lost another four

0:12:23.920 --> 0:12:26.920
<v Speaker 1>billion dollars on reality labs in the quarter. By the way,

0:12:27.080 --> 0:12:29.640
<v Speaker 1>there should be a legitimate criminal inquiry into where this

0:12:29.679 --> 0:12:32.320
<v Speaker 1>money is going. This is very strange. I think they're

0:12:32.320 --> 0:12:34.920
<v Speaker 1>at like eighty eighty six billion dollars and all we

0:12:35.000 --> 0:12:39.439
<v Speaker 1>have to show for it is the metaverse and pervert glasses. Meanwhile,

0:12:39.520 --> 0:12:44.079
<v Speaker 1>SpaceX and mister Musk is rushing to have the strangest

0:12:44.080 --> 0:12:47.360
<v Speaker 1>and largest IPO of all time. All those daily stories

0:12:47.440 --> 0:12:49.720
<v Speaker 1>leak about billions of dollars of losses and whatever the

0:12:49.760 --> 0:12:53.000
<v Speaker 1>fuck that deal with Cursor is. Apparently SpaceX will buy

0:12:53.000 --> 0:12:55.240
<v Speaker 1>it for sixty billion dollars or pay it ten billion

0:12:55.280 --> 0:12:58.000
<v Speaker 1>dollars to fuck off. I think what actually happens is

0:12:58.000 --> 0:13:00.600
<v Speaker 1>a secret third thing. SpaceX five it for a bit,

0:13:00.720 --> 0:13:03.240
<v Speaker 1>then there's a falling out between Musk and Michael Truell

0:13:03.320 --> 0:13:06.439
<v Speaker 1>of Cursor, and then the company either rushes another acquisition

0:13:06.520 --> 0:13:10.600
<v Speaker 1>or just dies. Remember Elon Musk killed Curs's funding round

0:13:10.640 --> 0:13:13.000
<v Speaker 1>that was in place by doing this acquisition, and he

0:13:13.080 --> 0:13:17.440
<v Speaker 1>can't buy this company before SpaceX goes public. Elon Musk

0:13:17.480 --> 0:13:19.680
<v Speaker 1>took fucking open Ai to court. Do you think he'll

0:13:19.720 --> 0:13:23.960
<v Speaker 1>care about killing Cursor? Who's gonna sue them the remains?

0:13:24.520 --> 0:13:27.000
<v Speaker 1>It's just a fucking rapper company. I'm so fucking tired

0:13:27.040 --> 0:13:31.120
<v Speaker 1>of it. Anyway, that lawsuit with open Ai. It's a

0:13:31.160 --> 0:13:34.000
<v Speaker 1>real alien versus predator situation, And if I'm honest, I

0:13:34.080 --> 0:13:36.360
<v Speaker 1>found the whole thing a little boring a duo of

0:13:36.440 --> 0:13:38.679
<v Speaker 1>Dullard's shoulder, budging each other to see who can run

0:13:38.679 --> 0:13:40.959
<v Speaker 1>a company that neither of them can really describe because

0:13:40.960 --> 0:13:43.559
<v Speaker 1>neither of them do anything other than pontificate and take

0:13:43.600 --> 0:13:46.880
<v Speaker 1>credit for other people's work. Two dumb bitches going exactly

0:13:47.880 --> 0:13:50.120
<v Speaker 1>if this breaks open AI, I'll be surprised. But if

0:13:50.120 --> 0:13:53.040
<v Speaker 1>it does, it would be extremely fitting that fucking Elon

0:13:53.200 --> 0:13:56.640
<v Speaker 1>Musk would accidentally destroy the AI industry, like mister Bean

0:13:57.040 --> 0:14:00.280
<v Speaker 1>sitting down on a button that launches a nuke here

0:14:00.320 --> 0:14:04.520
<v Speaker 1>would be very funny. I'm just not giving it much hope. Nevertheless,

0:14:04.559 --> 0:14:08.120
<v Speaker 1>this entire industry is only made possible by the global

0:14:08.240 --> 0:14:11.839
<v Speaker 1>kfabe circular economy of taking every single sign is good

0:14:11.920 --> 0:14:16.080
<v Speaker 1>for AI and ignoring every single possible glaring warning sign

0:14:16.080 --> 0:14:19.800
<v Speaker 1>and the hopes they'll go away. That is the entire strategy.

0:14:20.200 --> 0:14:22.560
<v Speaker 1>That's all it is. It's just people being like, well,

0:14:22.560 --> 0:14:26.960
<v Speaker 1>the bad things don't matter. They do matter. Last week,

0:14:27.080 --> 0:14:30.680
<v Speaker 1>Microsoft said it's shifting gehub copilot to token based billing,

0:14:31.160 --> 0:14:33.240
<v Speaker 1>meaning that people will have to pay the actual cost

0:14:33.280 --> 0:14:36.240
<v Speaker 1>of their models. This is effectively killing the product as

0:14:36.240 --> 0:14:39.360
<v Speaker 1>people know it and invalidates every single story about its growth,

0:14:39.400 --> 0:14:42.320
<v Speaker 1>revenue or otherwise ever written. I don't see anyone on

0:14:42.400 --> 0:14:44.200
<v Speaker 1>writing those. By the way, no one gives a fuck.

0:14:44.760 --> 0:14:46.840
<v Speaker 1>And to give you some context about the scale of

0:14:46.880 --> 0:14:50.840
<v Speaker 1>how big GitHub copilot was or is until they stop

0:14:50.920 --> 0:14:53.880
<v Speaker 1>subsidizing people's compute because right now when you use GitHub Copilot,

0:14:54.360 --> 0:14:57.200
<v Speaker 1>Microsoft's just paying the models for you with them anyway,

0:14:57.640 --> 0:15:00.960
<v Speaker 1>gehub Copilot is the second largest customer Ofanthropics models and

0:15:01.120 --> 0:15:03.200
<v Speaker 1>was only that large because it was subsidized in the

0:15:03.200 --> 0:15:06.320
<v Speaker 1>compute spend of its customers. A lot of that money

0:15:06.400 --> 0:15:10.960
<v Speaker 1>spend Onlanthropic. This is bad, and really that's the only

0:15:11.000 --> 0:15:13.520
<v Speaker 1>way to build any kind of AI business. Google and

0:15:13.560 --> 0:15:16.600
<v Speaker 1>Amazon realize that AI revenues are contingent on the continued

0:15:16.640 --> 0:15:20.040
<v Speaker 1>survival of Anthropic, and Amazon and Microsoft's revenues are contingent

0:15:20.080 --> 0:15:23.400
<v Speaker 1>on open AI and Anthropic. They know that if these

0:15:23.400 --> 0:15:25.800
<v Speaker 1>companies die, they're going to lose billions of dollars in revenue.

0:15:25.800 --> 0:15:28.680
<v Speaker 1>But also they have to compete with them for fear

0:15:28.720 --> 0:15:32.119
<v Speaker 1>that they'll be seen as falling behind their horrible progeny.

0:15:32.480 --> 0:15:35.560
<v Speaker 1>And thus they're incinerating their brands and endlessly pontificating about

0:15:35.560 --> 0:15:38.280
<v Speaker 1>the power of AI or spending nearly a trillion dollars

0:15:38.320 --> 0:15:41.400
<v Speaker 1>on Capex, almost entirely to make sure their competition, which

0:15:41.440 --> 0:15:44.720
<v Speaker 1>is also their customer, which is also the welfare recipient,

0:15:44.840 --> 0:15:47.320
<v Speaker 1>doesn't die because if they die, they lose a bunch

0:15:47.320 --> 0:15:49.120
<v Speaker 1>of revenue. But they need to spend way more money

0:15:49.160 --> 0:15:54.080
<v Speaker 1>than they'll ever make to justify that. Yeah, it's a

0:15:54.080 --> 0:15:57.520
<v Speaker 1>mess and a mistake, and eventually one of these companies

0:15:57.640 --> 0:16:00.480
<v Speaker 1>is going to grow tired of it. Microsoft was already

0:16:00.560 --> 0:16:04.400
<v Speaker 1>billions under the analyst estimates for Capex. They're moving to

0:16:04.440 --> 0:16:07.920
<v Speaker 1>token based billing on GitHub Copilot. They claim to invest

0:16:07.960 --> 0:16:10.920
<v Speaker 1>an Anthropic in February, but didn't mention it in their

0:16:10.960 --> 0:16:13.520
<v Speaker 1>ten Q their earnings in any way, shape or form,

0:16:13.680 --> 0:16:18.360
<v Speaker 1>which means Nomani was sent. At some point, these fuck

0:16:18.480 --> 0:16:20.520
<v Speaker 1>nuts are going to be forced to reckon with what

0:16:20.600 --> 0:16:24.960
<v Speaker 1>they're doing. Until then we'll have increasingly more frenzied and

0:16:25.000 --> 0:16:29.360
<v Speaker 1>ejaculatory statements about AI demand that fail to match with reality.

0:16:30.560 --> 0:16:32.520
<v Speaker 1>I truly think that it's going to be like this,

0:16:32.600 --> 0:16:36.720
<v Speaker 1>if not crazier, until one day when the music suddenly stops.

0:16:37.480 --> 0:16:40.600
<v Speaker 1>Somebody's going to blink. Somebody is going to take a

0:16:40.600 --> 0:16:43.200
<v Speaker 1>step back and give everybody else permission to stop too.

0:16:43.960 --> 0:16:48.240
<v Speaker 1>Maybe a perplexity lovable rip or cognition dies. Maybe Microsoft

0:16:48.240 --> 0:16:51.440
<v Speaker 1>shifting GitHub copilots to token based billing on June first

0:16:51.640 --> 0:16:55.960
<v Speaker 1>inspires others like Anthropic to follow suit. Maybe AI tooken

0:16:56.000 --> 0:16:59.560
<v Speaker 1>austerity begins at Microsoft, Meta or any other large company

0:16:59.600 --> 0:17:03.600
<v Speaker 1>burning ten plus percent of their head count on AI tokens.

0:17:04.320 --> 0:17:07.000
<v Speaker 1>Maybe in video fails to inspire the markets in just

0:17:07.040 --> 0:17:09.920
<v Speaker 1>the right way, Or the fact that the data centers

0:17:09.920 --> 0:17:12.520
<v Speaker 1>are not opening fast enough to have fully digested twenty

0:17:12.560 --> 0:17:15.960
<v Speaker 1>twenty five's GPUs finally catches up with the economic mismatch.

0:17:16.040 --> 0:17:20.080
<v Speaker 1>The Jensen wog always beats and raises expectations, and that

0:17:20.160 --> 0:17:24.240
<v Speaker 1>really is the strangest thing. At the current rate of sales,

0:17:24.320 --> 0:17:26.760
<v Speaker 1>it's taking more than six months to install the single

0:17:26.880 --> 0:17:30.680
<v Speaker 1>quarter's worth of GPU sales. At this point, it's obvious

0:17:30.720 --> 0:17:33.280
<v Speaker 1>that there are warehouses of these fucking things. It just

0:17:33.359 --> 0:17:36.240
<v Speaker 1>isn't obvious whether they're in ones owned by Hyperscalers or

0:17:36.280 --> 0:17:39.919
<v Speaker 1>the Taiwanese odm's original design manufacturers like Quantum Computing and

0:17:39.960 --> 0:17:44.639
<v Speaker 1>fox con that build their servers, and last quarter it

0:17:44.680 --> 0:17:48.320
<v Speaker 1>looked like Quantu Computing inventories were growing by billions of dollars.

0:17:48.960 --> 0:17:53.000
<v Speaker 1>Is that good anyway? None of this makes sense. It

0:17:53.040 --> 0:17:55.520
<v Speaker 1>hasn't from the beginning. It's the largest bubble in history

0:17:55.560 --> 0:17:58.159
<v Speaker 1>and has reached such an intellectual and financial scale that

0:17:58.240 --> 0:18:00.639
<v Speaker 1>many have taken signs on it in a way that

0:18:00.680 --> 0:18:03.080
<v Speaker 1>will be completely impossible to walk back if they're wrong.

0:18:03.480 --> 0:18:06.399
<v Speaker 1>As things deteriorate, expect them to cling to their mythologies

0:18:06.440 --> 0:18:09.520
<v Speaker 1>tighter and become more and more agitated and make vulgar

0:18:09.600 --> 0:18:13.679
<v Speaker 1>threats and weird comments that don't really make sense but

0:18:13.760 --> 0:18:17.959
<v Speaker 1>are nevertheless quite verbose. And really, we've never seen anything

0:18:18.000 --> 0:18:21.080
<v Speaker 1>like this in our lives. You realize that Open Eye

0:18:21.119 --> 0:18:24.560
<v Speaker 1>doesn't make any sense right. It can't survive without eternal subsidies.

0:18:25.119 --> 0:18:29.439
<v Speaker 1>None of these businesses can. Every single AI business you

0:18:29.600 --> 0:18:33.120
<v Speaker 1>see is unprofitable, and none of them have a path

0:18:33.200 --> 0:18:37.919
<v Speaker 1>to break even, let alone sustainability. And to make matters worse,

0:18:38.240 --> 0:18:40.359
<v Speaker 1>it is the only way to create demand was to

0:18:40.400 --> 0:18:45.520
<v Speaker 1>literally fund it yourself. These men believe they've created perpetual energy.

0:18:46.000 --> 0:18:48.080
<v Speaker 1>What they've actually done is shit their pants and set

0:18:48.119 --> 0:18:51.040
<v Speaker 1>their houses on fire. Anyway, catch you next week.