WEBVTT - OpenAI Cannot Survive

0:00:03.160 --> 0:00:03.560
<v Speaker 1>Media.

0:00:04.720 --> 0:00:06.960
<v Speaker 2>Hello and welcome to Better Offline. I'm your host ed

0:00:07.000 --> 0:00:21.239
<v Speaker 2>Sir Tron. Now, before we go any further, I hate

0:00:21.280 --> 0:00:23.080
<v Speaker 2>to ask you to do this, but I need your help.

0:00:23.440 --> 0:00:25.520
<v Speaker 2>I'm up for this year's Webbe's for the Best Business

0:00:25.520 --> 0:00:27.760
<v Speaker 2>Podcast Award. I know it's a pain in the ars

0:00:27.760 --> 0:00:29.600
<v Speaker 2>to register, but can you sign up and vote for

0:00:29.640 --> 0:00:31.920
<v Speaker 2>Better Offline? Never won an award in my life, and

0:00:31.920 --> 0:00:33.480
<v Speaker 2>I have enough listeners that I think we can tip

0:00:33.479 --> 0:00:37.000
<v Speaker 2>the scales. But on with the show. The following two

0:00:37.000 --> 0:00:39.760
<v Speaker 2>part episode is the culmination of months of research, presenting

0:00:39.760 --> 0:00:41.920
<v Speaker 2>a case I'd been making in parts since July of

0:00:42.000 --> 0:00:45.559
<v Speaker 2>last year. Open AYE is a financial abomination, a thing

0:00:45.600 --> 0:00:48.120
<v Speaker 2>that should not be, an aberration, a symbol of rot

0:00:48.280 --> 0:00:51.239
<v Speaker 2>at the heart of Silicon Valley, a company that unrepentantly

0:00:51.280 --> 0:00:53.880
<v Speaker 2>and needlessly burns billions of dollars with no end in sight,

0:00:54.240 --> 0:00:57.480
<v Speaker 2>helmed by a creticist, dis ingenuous billionaire who continually lies

0:00:57.720 --> 0:00:59.760
<v Speaker 2>about what it is that it will do because he,

0:01:00.280 --> 0:01:02.920
<v Speaker 2>like I've been saying since last year, knows that Generative

0:01:02.920 --> 0:01:05.640
<v Speaker 2>AI can't do much more than it does today. What

0:01:05.720 --> 0:01:07.320
<v Speaker 2>I'm going to lay out for you is my case

0:01:07.360 --> 0:01:11.040
<v Speaker 2>the Open AI can't survive, that it's boadline impossible under

0:01:11.240 --> 0:01:14.120
<v Speaker 2>any of the current terms for this company to continue,

0:01:14.160 --> 0:01:16.080
<v Speaker 2>and that its demise will bring about the death of

0:01:16.120 --> 0:01:19.400
<v Speaker 2>a or significant harm of several other firms. A cottage

0:01:19.440 --> 0:01:23.240
<v Speaker 2>industry of despicable, billion dollar burn rate capitalistic monstrosities has

0:01:23.240 --> 0:01:26.280
<v Speaker 2>sprung up around this stupid fucking company in the hopes

0:01:26.280 --> 0:01:28.639
<v Speaker 2>of further inflating a bubble set to burst at any moment.

0:01:29.360 --> 0:01:31.880
<v Speaker 2>My discuss for the parties involved is unrelenting. This is

0:01:32.120 --> 0:01:34.200
<v Speaker 2>my discuss for a media industry that failed to even

0:01:34.200 --> 0:01:36.959
<v Speaker 2>attempt to tackle the subject matter I'm going to detail.

0:01:37.280 --> 0:01:39.080
<v Speaker 2>I believe at the end of these episodes you'll see

0:01:39.080 --> 0:01:41.600
<v Speaker 2>my point and at the very least agree that open

0:01:41.600 --> 0:01:45.440
<v Speaker 2>AI's current situation is totally untenable. If I'm right, open

0:01:45.480 --> 0:01:47.640
<v Speaker 2>ai will go down in history as an abdication of

0:01:47.720 --> 0:01:51.920
<v Speaker 2>due diligence, fiscal responsibility, in frankly common sense, both in

0:01:51.960 --> 0:01:54.280
<v Speaker 2>the venture capitalists and entities that propped it up and

0:01:54.320 --> 0:01:57.160
<v Speaker 2>a tech media that was more concerned with taking detailed

0:01:57.200 --> 0:01:59.760
<v Speaker 2>notes on my comings and goings than knowing their ass

0:01:59.760 --> 0:02:02.400
<v Speaker 2>from their ear hole. Those who have failed to hold

0:02:02.400 --> 0:02:05.200
<v Speaker 2>men like Sam Altman and darra Ama Day Accountable have

0:02:05.320 --> 0:02:09.639
<v Speaker 2>tacitly approved a financial and environmentally destructive movement that will

0:02:09.720 --> 0:02:12.840
<v Speaker 2>lead to the very little actually happening or changing in

0:02:12.840 --> 0:02:15.440
<v Speaker 2>the world other than damaging our power grid and the

0:02:15.440 --> 0:02:19.040
<v Speaker 2>theft of art from millions of people. These episodes will

0:02:19.080 --> 0:02:21.920
<v Speaker 2>be full of numbers and statements and very few declarations

0:02:22.000 --> 0:02:24.360
<v Speaker 2>or personal opinions, though probably a few swear words. If

0:02:24.400 --> 0:02:26.959
<v Speaker 2>I'm honest, I don't need to get that personal here.

0:02:27.200 --> 0:02:30.880
<v Speaker 2>The numbers in question, they're damaging, they're staggering, they're worrying,

0:02:30.880 --> 0:02:34.119
<v Speaker 2>and ultimately spell collapse. The truth is, we don't even

0:02:34.120 --> 0:02:36.960
<v Speaker 2>need to talk about tarifs for things to go sideways

0:02:36.960 --> 0:02:39.519
<v Speaker 2>for this industry. The price of a GPU could rise

0:02:39.520 --> 0:02:42.160
<v Speaker 2>one hundred percent, or it could have it really wouldn't

0:02:42.160 --> 0:02:44.799
<v Speaker 2>make much of a difference to open AI's chances of survival.

0:02:45.200 --> 0:02:48.400
<v Speaker 2>That's how bad the fundamentals are. And to illustrate that point,

0:02:48.400 --> 0:02:51.120
<v Speaker 2>I'm going to ask a number of relatively simple questions

0:02:51.120 --> 0:02:53.680
<v Speaker 2>over the next couple of episodes and make an attempt

0:02:53.720 --> 0:02:57.560
<v Speaker 2>to answer them. First, let's start with something simple, how

0:02:57.639 --> 0:03:02.080
<v Speaker 2>much cash does open ai have. At the start of April,

0:03:02.120 --> 0:03:05.280
<v Speaker 2>open Ai closed what was called the largest private tech

0:03:05.320 --> 0:03:09.720
<v Speaker 2>funding round in history, where it raised and astonishing forty

0:03:09.760 --> 0:03:12.200
<v Speaker 2>billion dollars. And the reason I'm saying this with the

0:03:12.240 --> 0:03:15.000
<v Speaker 2>sarcastic inflection is that open ai has only actually raised

0:03:15.040 --> 0:03:16.960
<v Speaker 2>ten billion dollars of the forty billion dollars, with the

0:03:16.960 --> 0:03:19.920
<v Speaker 2>rest arriving by the end of the year, and even

0:03:19.960 --> 0:03:22.000
<v Speaker 2>as I record this, I don't know if the money's

0:03:22.000 --> 0:03:25.880
<v Speaker 2>actually gone yet. I'll get it into that in a minute. Now.

0:03:25.919 --> 0:03:27.799
<v Speaker 2>A lot can happen in a year, and the remaining

0:03:28.200 --> 0:03:31.120
<v Speaker 2>thirty billion dollars twenty billion dollars of which will allegedly

0:03:31.160 --> 0:03:33.760
<v Speaker 2>be provided by soft Bank, is partially contingent on an

0:03:33.760 --> 0:03:36.640
<v Speaker 2>open AI's conversion from a nonprofit to a for profit

0:03:36.720 --> 0:03:38.920
<v Speaker 2>by the end of twenty twenty five, and if it fails,

0:03:38.960 --> 0:03:41.200
<v Speaker 2>soft Bank will only give open Ai a further twenty

0:03:41.200 --> 0:03:44.600
<v Speaker 2>billion dollars. I'll get into how fucking stupid this gets later.

0:03:45.480 --> 0:03:49.400
<v Speaker 2>The round also valued open Ai and astonishing three hundred

0:03:49.440 --> 0:03:51.960
<v Speaker 2>billion dollars. To put that in context, open Ai had

0:03:52.000 --> 0:03:54.560
<v Speaker 2>revenues of four billion dollars in twenty twenty four. This

0:03:54.640 --> 0:03:57.760
<v Speaker 2>deal values open Ai at seventy five times its revenue.

0:03:57.960 --> 0:04:00.560
<v Speaker 2>That's more than Tesla even at its most looted Chris Peak.

0:04:01.480 --> 0:04:04.120
<v Speaker 2>I also want to add that, as of writing this sentence,

0:04:04.160 --> 0:04:07.040
<v Speaker 2>this money is yet to arrive. Maybe it will arrive

0:04:07.080 --> 0:04:09.440
<v Speaker 2>by the time this is out. Maybe what really make

0:04:09.480 --> 0:04:12.560
<v Speaker 2>me look stupid. But based on soft Bank's filings, which

0:04:12.560 --> 0:04:15.040
<v Speaker 2>I'll link to in the spreadsheet for this episode, say

0:04:15.040 --> 0:04:17.679
<v Speaker 2>that the money will arrive mid April. It's April fourteenth

0:04:17.720 --> 0:04:20.000
<v Speaker 2>as I'm recording this, and that soft Bank would be

0:04:20.080 --> 0:04:23.920
<v Speaker 2>borrowing as much as ten billion dollars of the financing

0:04:23.920 --> 0:04:26.200
<v Speaker 2>for the round, with the option to syndicate, meaning you

0:04:26.200 --> 0:04:29.039
<v Speaker 2>bring in other investors the rest of them. For the

0:04:29.040 --> 0:04:31.359
<v Speaker 2>sake of argument, I'm going to assume the money actually arrives,

0:04:31.360 --> 0:04:36.080
<v Speaker 2>though filings also suggest that, and I'm quoting in certain circumstances,

0:04:36.080 --> 0:04:38.600
<v Speaker 2>the second thirty billion dollar tranch could arrive in early

0:04:38.640 --> 0:04:43.200
<v Speaker 2>twenty twenty six. This isn't great. It also seems that

0:04:43.240 --> 0:04:45.839
<v Speaker 2>soft banks ten billion dollar commitment is contingent on getting

0:04:45.839 --> 0:04:49.719
<v Speaker 2>a loan, as it says it's financed through borrowings from

0:04:50.160 --> 0:04:54.920
<v Speaker 2>Mizuho Bank Limited, among other financial institutions. Soft Banks had

0:04:54.920 --> 0:04:56.600
<v Speaker 2>plenty of loans in the past, so I think they're

0:04:56.640 --> 0:04:58.600
<v Speaker 2>going to get it. But I think this is one

0:04:58.640 --> 0:05:01.320
<v Speaker 2>of their biggest open a I also revealed it now

0:05:01.320 --> 0:05:04.799
<v Speaker 2>has twenty million paying subscribers and over five hundred million

0:05:04.800 --> 0:05:07.599
<v Speaker 2>weekly active users. If you're wondering why it doesn't talking

0:05:07.640 --> 0:05:10.440
<v Speaker 2>about monthly users, it's because they're likely much higher than

0:05:10.440 --> 0:05:13.600
<v Speaker 2>five hundred million, which would reveal exactly how poorly open

0:05:13.600 --> 0:05:17.120
<v Speaker 2>ai converts free chat GPT users to paying ones. The

0:05:17.200 --> 0:05:20.120
<v Speaker 2>Information reported back in January that open ai was generating

0:05:20.120 --> 0:05:22.560
<v Speaker 2>twenty five million dollars in revenue a month from its

0:05:22.560 --> 0:05:25.359
<v Speaker 2>two hundred dollars a month pro subscribers, and just so

0:05:25.400 --> 0:05:27.880
<v Speaker 2>we're clear, they lose money on every one of those two,

0:05:28.040 --> 0:05:30.640
<v Speaker 2>suggesting that they have around one hundred and twenty five

0:05:30.680 --> 0:05:34.560
<v Speaker 2>thousand Chat GPT pro subscribers, each losing the money somehow.

0:05:34.960 --> 0:05:38.080
<v Speaker 2>Assuming the other nineteen million, eight hundred and seventy five

0:05:38.160 --> 0:05:41.000
<v Speaker 2>thousand users are paying twenty bucks a month, that puts

0:05:41.040 --> 0:05:43.280
<v Speaker 2>revenue about four hundred and twenty three million dollars a month,

0:05:43.400 --> 0:05:45.080
<v Speaker 2>or at least five billion dollars a year from chat

0:05:45.120 --> 0:05:48.640
<v Speaker 2>GPT subscriptions. This is what reporters mean when they say

0:05:48.760 --> 0:05:51.640
<v Speaker 2>annualized revenue. By the way, it's literally the month. The

0:05:51.680 --> 0:05:53.680
<v Speaker 2>monthly revenue that the money they're making in one month

0:05:53.720 --> 0:05:55.960
<v Speaker 2>multiplied by twelve, and you'll be surprised to hear that

0:05:56.000 --> 0:05:59.240
<v Speaker 2>people play silly buggers with that all the time. In March,

0:05:59.240 --> 0:06:02.120
<v Speaker 2>Bloomberg reports the open ai expects its revenue to triple

0:06:02.160 --> 0:06:04.719
<v Speaker 2>to twelve point seven billion dollars in twenty twenty five,

0:06:04.920 --> 0:06:07.440
<v Speaker 2>assuming a similar split of revenue to twenty twenty four.

0:06:07.560 --> 0:06:10.320
<v Speaker 2>This would require open ai to nearly double its annualized

0:06:10.320 --> 0:06:13.360
<v Speaker 2>subscription revenue from Q one twenty twenty five from five

0:06:13.360 --> 0:06:15.640
<v Speaker 2>billion dollars right now to around nine point twenty seven

0:06:15.680 --> 0:06:19.600
<v Speaker 2>billion dollars, and nearly kuldruple API revenue from twenty twenty

0:06:19.640 --> 0:06:22.559
<v Speaker 2>four's revenue of one billion, which includes Microsoft's twenty percent

0:06:22.600 --> 0:06:25.520
<v Speaker 2>payment for access to open AI's models, and that would

0:06:25.560 --> 0:06:27.400
<v Speaker 2>get them about three point four to three billion. We

0:06:27.440 --> 0:06:29.800
<v Speaker 2>all need to super worry about these numbers. And I

0:06:29.880 --> 0:06:33.359
<v Speaker 2>realized these are messy numbers. It's just unclear how open

0:06:33.400 --> 0:06:35.240
<v Speaker 2>ai intends to pull any of this off. It's an

0:06:35.240 --> 0:06:39.320
<v Speaker 2>incredible leap, and open AI's own plans don't exactly inspire confidence.

0:06:39.680 --> 0:06:43.120
<v Speaker 2>They're really good at getting free subscribers, they don't seem

0:06:43.160 --> 0:06:45.359
<v Speaker 2>to be able to get paying ones in quite the

0:06:45.400 --> 0:06:48.680
<v Speaker 2>same number, and also they even lose them money. Every

0:06:48.760 --> 0:06:50.560
<v Speaker 2>time I think about this company, I start feeling a

0:06:50.600 --> 0:06:53.440
<v Speaker 2>little crazy, and if I'm on it anyway, The Information

0:06:53.520 --> 0:06:56.039
<v Speaker 2>reported in February that if the open ai plan to

0:06:56.040 --> 0:06:58.360
<v Speaker 2>grow its revenue by making three billion dollars a year

0:06:58.400 --> 0:07:01.520
<v Speaker 2>selling agents with chat GBT subscriptions seven point nine billion

0:07:01.520 --> 0:07:03.720
<v Speaker 2>dollars and API cous one point eight billion dollars making

0:07:03.800 --> 0:07:07.600
<v Speaker 2>up the rest. This is, of course, what's the technical term, Oh,

0:07:07.640 --> 0:07:11.360
<v Speaker 2>it's bollocks. It's complete fucking bollocks. I'm sorry. Agents, by

0:07:11.400 --> 0:07:14.000
<v Speaker 2>the way, are AI chatbots that can do something like

0:07:14.080 --> 0:07:17.160
<v Speaker 2>interact with another program on the user's behalf. Open aiy's

0:07:17.200 --> 0:07:20.120
<v Speaker 2>agents can't even do the simplest tasks. And three billion

0:07:20.160 --> 0:07:22.440
<v Speaker 2>dollars of twelve point seven billion dollar figure appears to

0:07:22.480 --> 0:07:24.960
<v Speaker 2>be a commitment made by soft Bank to purchase three

0:07:25.000 --> 0:07:28.160
<v Speaker 2>billion dollars a year of Open Aiyes, tech, now, let's

0:07:28.480 --> 0:07:32.080
<v Speaker 2>pass out these numbers precisely. So incoming monthly revenue roughly

0:07:32.080 --> 0:07:34.280
<v Speaker 2>four hundred and twenty five million dollars give or take

0:07:34.800 --> 0:07:37.760
<v Speaker 2>theoretical revenue from soft Banks completely made up thing two

0:07:37.800 --> 0:07:40.440
<v Speaker 2>hundred and fifty million dollars a month. However, I really

0:07:40.440 --> 0:07:43.000
<v Speaker 2>can find no proof that soft Bank has begun to

0:07:43.000 --> 0:07:45.520
<v Speaker 2>make these payments, or indeed that it intends to make

0:07:45.560 --> 0:07:49.200
<v Speaker 2>them even how it intends to. Now, let's talk liquidity.

0:07:49.920 --> 0:07:52.320
<v Speaker 2>Open Ai has ten billion dollars that they yet to

0:07:52.360 --> 0:07:54.720
<v Speaker 2>receive a recording this, but let's assume they get it,

0:07:54.880 --> 0:07:56.840
<v Speaker 2>and it will be ten billion dollars seven point five

0:07:56.880 --> 0:07:59.760
<v Speaker 2>billion from soft Bank and a syndicate of investors including

0:07:59.760 --> 0:08:02.760
<v Speaker 2>my mic Krossoft and Cutchiu and a few others. Potentially.

0:08:03.520 --> 0:08:06.000
<v Speaker 2>There's also an indeterminate amount of remaining capital on the

0:08:06.000 --> 0:08:08.840
<v Speaker 2>four billion dollar credit facility provided by multiple banks back

0:08:08.880 --> 0:08:11.240
<v Speaker 2>in October twenty twenty four, and that was raised alongside

0:08:11.240 --> 0:08:13.000
<v Speaker 2>of funding ground that valued the company one hundred and

0:08:13.040 --> 0:08:16.480
<v Speaker 2>fifty seven billion dollars now. As a note, the October

0:08:16.480 --> 0:08:19.800
<v Speaker 2>announcement stated that open ai had access to over ten

0:08:19.840 --> 0:08:22.360
<v Speaker 2>billion dollars in liquidity, giving us a sense of how

0:08:22.360 --> 0:08:24.320
<v Speaker 2>fast it's burning through the cash it has on hand,

0:08:24.320 --> 0:08:26.560
<v Speaker 2>as that was pretty much all the money they'd raised

0:08:26.560 --> 0:08:29.400
<v Speaker 2>at the time. Based on reports, open Ai will not

0:08:29.440 --> 0:08:31.600
<v Speaker 2>have access to the rest of the forty billion dollars

0:08:32.000 --> 0:08:34.120
<v Speaker 2>that soft Bank is funding them with. Until the end

0:08:34.120 --> 0:08:35.880
<v Speaker 2>of the year. And it's unclear what part of the

0:08:35.960 --> 0:08:38.760
<v Speaker 2>end of the year that is, but SoftBank's filing says December.

0:08:39.080 --> 0:08:41.520
<v Speaker 2>Will it be December first? Will it be Christmas time?

0:08:41.800 --> 0:08:43.960
<v Speaker 2>Will Sam Mortman look on to the tree and Messiyoshi

0:08:43.960 --> 0:08:46.560
<v Speaker 2>san will be there? We'll find out, won't we. But

0:08:46.640 --> 0:08:49.120
<v Speaker 2>we can assume in this case that open ai likely has,

0:08:49.120 --> 0:08:52.240
<v Speaker 2>in the best case scenario, access to about roughly sixteen

0:08:52.240 --> 0:08:55.319
<v Speaker 2>billion dollars in liquidity at any given time. It's reasonable

0:08:55.400 --> 0:08:57.680
<v Speaker 2>to believe that open ai will raise more debt this

0:08:57.800 --> 0:09:01.040
<v Speaker 2>year despite this massive raise, and it does so to

0:09:01.080 --> 0:09:03.800
<v Speaker 2>the tune of around five or six billion dollars. Without it,

0:09:04.000 --> 0:09:06.880
<v Speaker 2>I'm genuinely not sure what they're going to do. And

0:09:07.040 --> 0:09:10.040
<v Speaker 2>as a reminder, kids, open ai loses money on every

0:09:10.080 --> 0:09:24.400
<v Speaker 2>single user, free or paying. Now here's my next question,

0:09:24.880 --> 0:09:28.600
<v Speaker 2>what are open AI's obligations? Now? When I put out

0:09:28.600 --> 0:09:30.839
<v Speaker 2>how does open ai survive and open aiy is a

0:09:30.880 --> 0:09:33.480
<v Speaker 2>bad business? I used reported information to explain how this

0:09:33.559 --> 0:09:36.960
<v Speaker 2>company was at its court unsustainable. I'll link to the

0:09:37.000 --> 0:09:39.240
<v Speaker 2>newsletters and the spreadsheet in the podcast too. But let's

0:09:39.240 --> 0:09:43.240
<v Speaker 2>refresh our memory, shall we. Okay, compute costs at least

0:09:43.240 --> 0:09:46.280
<v Speaker 2>thirteen billion dollars of which are going to Microsoft a

0:09:46.280 --> 0:09:48.400
<v Speaker 2>loan in twenty twenty five, and as much as five

0:09:48.559 --> 0:09:51.320
<v Speaker 2>hundred and ninety four million dollars to my favorite company,

0:09:51.400 --> 0:09:55.360
<v Speaker 2>core Wave. Core Weave's back. Every time I think I

0:09:55.440 --> 0:09:58.040
<v Speaker 2>got away from core Weave, they pop out like a

0:09:58.120 --> 0:10:01.080
<v Speaker 2>sour turd coming back in the toilet. Anyway, it seems

0:10:01.080 --> 0:10:03.559
<v Speaker 2>from even a cursory glance that open AI's costs are

0:10:03.559 --> 0:10:07.040
<v Speaker 2>increasing dramatically. The information reported earlier in the year the

0:10:07.040 --> 0:10:10.280
<v Speaker 2>open ai projects to spend thirteen billion dollars on compute

0:10:10.320 --> 0:10:13.199
<v Speaker 2>with Microsoft the loan in twenty twenty five, nearly tripling

0:10:13.280 --> 0:10:15.400
<v Speaker 2>what it spent in total compute in twenty twenty four,

0:10:15.400 --> 0:10:17.960
<v Speaker 2>which was five billion dollars, with three billion for training

0:10:18.000 --> 0:10:20.719
<v Speaker 2>and two billion for running their models. This suggests that

0:10:20.760 --> 0:10:24.160
<v Speaker 2>open aiyes costs as skyrocketing, and that was before the

0:10:24.240 --> 0:10:26.880
<v Speaker 2>launch of their new image generator, which led to multiple

0:10:26.920 --> 0:10:29.520
<v Speaker 2>complaints from Sam Mormon about a lack of available GPUs,

0:10:29.800 --> 0:10:33.360
<v Speaker 2>leading to open Aiy's Terrible Little Man CEO saying that

0:10:33.400 --> 0:10:37.720
<v Speaker 2>he expected stuff to break and delays in new products. Nevertheless,

0:10:37.760 --> 0:10:39.480
<v Speaker 2>even if we assume open ai in factored in the

0:10:39.480 --> 0:10:42.400
<v Speaker 2>compute increases into its projections, it still expects to pay

0:10:42.440 --> 0:10:46.280
<v Speaker 2>that thirteen billion dollars to Microsoft a loan. This number, however,

0:10:46.280 --> 0:10:49.360
<v Speaker 2>doesn't include the twelve point nine billion dollar five year

0:10:49.400 --> 0:10:51.760
<v Speaker 2>long compute deal signed with core Weave, a deal that

0:10:51.800 --> 0:10:54.360
<v Speaker 2>was a result of Microsoft declining to pick up the

0:10:54.400 --> 0:10:57.760
<v Speaker 2>option to buy the compute themselves for open Ai. As

0:10:57.800 --> 0:11:00.840
<v Speaker 2>an aside from what I am here and this is

0:11:01.080 --> 0:11:04.840
<v Speaker 2>sources telling me most of the Microsoft compute was open Ai.

0:11:04.880 --> 0:11:08.240
<v Speaker 2>It's basically they put an open Ai sticker over Microsoft.

0:11:08.320 --> 0:11:11.120
<v Speaker 2>It's basically the same thing. Now, payments for this deal

0:11:11.480 --> 0:11:14.320
<v Speaker 2>with with open Ai and core Weave they start in

0:11:14.320 --> 0:11:18.600
<v Speaker 2>October twenty twenty five according to the information, and assuming

0:11:18.640 --> 0:11:22.000
<v Speaker 2>that it's even paid, or like evenly paid or even

0:11:22.040 --> 0:11:24.680
<v Speaker 2>like anyone gets the cash. But these contracts are weird,

0:11:24.720 --> 0:11:27.760
<v Speaker 2>you never know exactly how they're paid, this would still

0:11:27.880 --> 0:11:31.600
<v Speaker 2>amount to roughly two point three eight billion dollars a year. However,

0:11:31.600 --> 0:11:33.960
<v Speaker 2>for the sake of argument, let's consider the payments are

0:11:33.960 --> 0:11:36.199
<v Speaker 2>around one hundred and ninety eight million dollars a month.

0:11:36.559 --> 0:11:39.800
<v Speaker 2>Though there are scenarios such as say core weaves build

0:11:39.840 --> 0:11:42.200
<v Speaker 2>out partner not being able to build the data centers

0:11:42.240 --> 0:11:43.800
<v Speaker 2>or core weave not having the money to pay to

0:11:43.800 --> 0:11:46.400
<v Speaker 2>build them, where open ai might end up paying less.

0:11:46.440 --> 0:11:49.800
<v Speaker 2>And don't worry, I'll get to that later. Let's talk

0:11:49.800 --> 0:11:54.240
<v Speaker 2>about Stargate. You heard this, You've seen this. Stargate is

0:11:54.280 --> 0:11:57.040
<v Speaker 2>this data center project the open ai is going into.

0:11:57.120 --> 0:11:59.720
<v Speaker 2>And they've dedicated somewhere in the region of nineteen billion

0:11:59.720 --> 0:12:02.880
<v Speaker 2>dollars to this project, along with another nineteen billion dollars

0:12:02.920 --> 0:12:07.800
<v Speaker 2>provided by Software Bank at some indeterminate time. And there

0:12:07.800 --> 0:12:10.400
<v Speaker 2>are other partners too, like oor Coins. Not even obvious

0:12:10.440 --> 0:12:12.920
<v Speaker 2>what they're putting in either. One thing I can tell you, though,

0:12:13.000 --> 0:12:16.000
<v Speaker 2>is the Trump administration is not actually spending any money

0:12:16.000 --> 0:12:18.600
<v Speaker 2>on this anyone. I've had tons of people emailing me

0:12:18.679 --> 0:12:21.480
<v Speaker 2>being like, hey, yeah, Donald Trump was there. Yeah, Donald

0:12:21.520 --> 0:12:25.000
<v Speaker 2>Trump was there. Doesn't mean shit anyway. Based on reporting

0:12:25.000 --> 0:12:28.240
<v Speaker 2>from Bloomberg, open Ai plans to have sixty four thousand

0:12:28.240 --> 0:12:30.760
<v Speaker 2>Blackwell GPUs running by the end of twenty twenty six,

0:12:31.200 --> 0:12:33.720
<v Speaker 2>or roughly about three point eighty four billion dollars worth

0:12:33.760 --> 0:12:36.760
<v Speaker 2>of them. I should also note that Bloomberg said that

0:12:36.800 --> 0:12:39.360
<v Speaker 2>sixteen thousand of these chips would be operational by summer

0:12:39.360 --> 0:12:42.120
<v Speaker 2>twenty twenty five, though it's unclear if that will actually happen,

0:12:42.160 --> 0:12:44.559
<v Speaker 2>and as I'll get to later, I got some more

0:12:44.640 --> 0:12:48.400
<v Speaker 2>questions about what exactly is happening at Stargate or told.

0:12:48.600 --> 0:12:51.680
<v Speaker 2>Though it's unclear who actually pays what parts of Stargate,

0:12:51.679 --> 0:12:53.560
<v Speaker 2>it's safe to assume that open Ai will have to

0:12:53.600 --> 0:12:55.760
<v Speaker 2>at least put a billion dollars into a project that

0:12:55.840 --> 0:12:57.240
<v Speaker 2>is meant to be up and running by the end

0:12:57.280 --> 0:13:01.640
<v Speaker 2>of twenty twenty six, if not more. Now Stargate really

0:13:01.679 --> 0:13:04.640
<v Speaker 2>on has one like data center project under development is

0:13:04.679 --> 0:13:08.160
<v Speaker 2>in Abilene, Texas, and as I've mentioned, it's not really

0:13:08.160 --> 0:13:10.080
<v Speaker 2>clear how it's going though. Are recent piece from the

0:13:10.080 --> 0:13:13.720
<v Speaker 2>Information reported that it was currently empty and incomplete, and

0:13:13.720 --> 0:13:16.160
<v Speaker 2>if it stays that way, and I quote again, open

0:13:16.200 --> 0:13:18.080
<v Speaker 2>Ai could walk away from the deal, which would cost

0:13:18.120 --> 0:13:21.600
<v Speaker 2>Oracle billions of dollars. Though the article takes great pains

0:13:21.600 --> 0:13:23.840
<v Speaker 2>to assure the reader that that won't be likely. Even

0:13:23.880 --> 0:13:26.920
<v Speaker 2>an inkling of such a possibility is a bad fucking sign.

0:13:27.440 --> 0:13:30.120
<v Speaker 2>Business insiders reported on the site in Abilene calls it

0:13:30.160 --> 0:13:33.280
<v Speaker 2>a three point four billion dollar data center development, as

0:13:33.320 --> 0:13:35.520
<v Speaker 2>did the press release from the developer Crusoe, who will

0:13:35.520 --> 0:13:38.720
<v Speaker 2>get too later. Though these numbers don't include GPUs hardware

0:13:38.800 --> 0:13:41.880
<v Speaker 2>or the labour necessary to run them. Right now, Cruso is,

0:13:41.880 --> 0:13:44.719
<v Speaker 2>according to Business Insider, building six new data centers, each

0:13:44.760 --> 0:13:47.520
<v Speaker 2>with a minimum square footage, which will join the two

0:13:47.600 --> 0:13:50.720
<v Speaker 2>it is already constructing for Oracle. Oracle is signed, according

0:13:50.760 --> 0:13:53.000
<v Speaker 2>to the information, a fifteen year long lease with Cruiser

0:13:53.000 --> 0:13:55.440
<v Speaker 2>for its data centers, all of which will be rented

0:13:55.480 --> 0:13:58.880
<v Speaker 2>to open Ai. In any case, open AI's exposure could

0:13:58.920 --> 0:14:02.040
<v Speaker 2>be much, much, much higher than billion dollars. And I'll

0:14:02.080 --> 0:14:04.120
<v Speaker 2>explain in greater depth how I've reached the figure at

0:14:04.120 --> 0:14:06.640
<v Speaker 2>the women I get there. But nevertheless, it could be

0:14:06.720 --> 0:14:09.520
<v Speaker 2>much higher. If OpenAI has to contribute significantly to the

0:14:09.600 --> 0:14:13.760
<v Speaker 2>costs of associating with Stargate in general, it could cost

0:14:13.800 --> 0:14:15.839
<v Speaker 2>them a lot of money. Data centers aren't something you

0:14:15.880 --> 0:14:19.720
<v Speaker 2>can do. Funny money math with Data Center Dynamics reports

0:14:19.720 --> 0:14:21.560
<v Speaker 2>that the Abilene site is meant to have two hundred

0:14:21.600 --> 0:14:24.120
<v Speaker 2>megawatts of compute capacity in the first half of twenty

0:14:24.160 --> 0:14:25.920
<v Speaker 2>twenty five, and then as much as one point two

0:14:25.960 --> 0:14:28.680
<v Speaker 2>gigawats by mid twenty twenty six. To give you a

0:14:28.720 --> 0:14:31.120
<v Speaker 2>sense of the total cost of the project, Former Microsoft

0:14:31.240 --> 0:14:33.920
<v Speaker 2>VP of Energy Brian Janis said in January that it

0:14:33.960 --> 0:14:36.840
<v Speaker 2>costs twenty five million dollars a megawat or about twenty

0:14:36.840 --> 0:14:39.640
<v Speaker 2>five billion dollars a gigawe, meaning that the initial capital

0:14:39.680 --> 0:14:42.000
<v Speaker 2>expensures for Stargate to spin up its first two hundred

0:14:42.040 --> 0:14:45.760
<v Speaker 2>megawat data center will be around five billion dollars, spiraling

0:14:45.760 --> 0:14:48.280
<v Speaker 2>to thirty billion dollars or more for the entire project.

0:14:48.960 --> 0:14:51.120
<v Speaker 2>The Information has reported that the site, which could be

0:14:51.120 --> 0:14:53.320
<v Speaker 2>and I quote potentially one of the biggest AAY identify

0:14:53.560 --> 0:14:56.040
<v Speaker 2>data centers, could cost fifty billion to one hundred billion

0:14:56.040 --> 0:15:00.280
<v Speaker 2>dollars in the coming years. Where's that fucking money coming from? Me?

0:15:00.320 --> 0:15:02.520
<v Speaker 2>Stick with the lower end of these cost estimates, it's

0:15:02.560 --> 0:15:04.520
<v Speaker 2>likely the open AI is on the hook for over

0:15:04.560 --> 0:15:07.200
<v Speaker 2>five billion dollars for the Abilene site. Based on the

0:15:07.280 --> 0:15:10.400
<v Speaker 2>nineteen billion dollars it's committed to the overall Stargate Data

0:15:10.400 --> 0:15:13.800
<v Speaker 2>center project. This expenditure won't come all at once and

0:15:13.840 --> 0:15:17.000
<v Speaker 2>will be spread across several years. Still, assuming even the

0:15:17.160 --> 0:15:19.640
<v Speaker 2>rosiest numbers, it's hard to see how open ai doesn't

0:15:19.640 --> 0:15:21.680
<v Speaker 2>have to pony up at least a billi in twenty

0:15:21.720 --> 0:15:24.200
<v Speaker 2>twenty five, and that's likely because the development at this

0:15:24.320 --> 0:15:26.480
<v Speaker 2>site is going to be heavily delayed by both tariffs,

0:15:26.720 --> 0:15:30.760
<v Speaker 2>labor shortages, and oracles as reported by Information. Well, they're

0:15:30.800 --> 0:15:33.000
<v Speaker 2>trusting and I this is the kind of quote you

0:15:33.000 --> 0:15:36.640
<v Speaker 2>really want to hear. They're trusting scrappy but unproven startups

0:15:36.640 --> 0:15:39.280
<v Speaker 2>to develop the project. Is that good? Is that who

0:15:39.320 --> 0:15:41.680
<v Speaker 2>you want doing this? I know when I get a

0:15:41.720 --> 0:15:44.720
<v Speaker 2>contractor in to fix something, my first thought is, is

0:15:44.760 --> 0:15:48.280
<v Speaker 2>this guy's scrappy? Anyway, Let's talk about the other costs.

0:15:48.440 --> 0:15:51.120
<v Speaker 2>They're at least three point five billion based on reporting

0:15:51.160 --> 0:15:53.480
<v Speaker 2>from the information from last year, open ai will spend

0:15:53.480 --> 0:15:56.520
<v Speaker 2>at least two point five billion dollars across salaries, data

0:15:56.600 --> 0:15:59.400
<v Speaker 2>referring to buying data from other companies, hosting another cost

0:15:59.440 --> 0:16:01.920
<v Speaker 2>of sales in sales and marketing, and then another billion

0:16:01.960 --> 0:16:05.080
<v Speaker 2>on what infrastructure open ai owns. I expect the latter

0:16:05.120 --> 0:16:07.960
<v Speaker 2>cost to balloon with open AI's investment in physical infrastructure

0:16:08.000 --> 0:16:11.640
<v Speaker 2>for stargame. Here's another bloody question. How does open ai

0:16:11.760 --> 0:16:15.480
<v Speaker 2>meet its obligations? Based on previous estimates, open ai spends

0:16:15.520 --> 0:16:17.560
<v Speaker 2>about two dollars and twenty five cents to make a buck.

0:16:17.920 --> 0:16:20.200
<v Speaker 2>At that rate, it's likely the open AI's cost in

0:16:20.280 --> 0:16:23.440
<v Speaker 2>its most rosy reed revenue projections of twelve point seven

0:16:23.440 --> 0:16:26.240
<v Speaker 2>billion dollars or at least twenty eight billion, meaning that

0:16:26.280 --> 0:16:28.960
<v Speaker 2>it's on course to burn at least fourteen billion dollars

0:16:29.000 --> 0:16:31.920
<v Speaker 2>in twenty twenty five a loan, assuming that open ai

0:16:32.120 --> 0:16:35.560
<v Speaker 2>has literally all of the money they had from last year.

0:16:35.640 --> 0:16:37.920
<v Speaker 2>They don't, but for the sake of argument, let's pretend

0:16:37.920 --> 0:16:40.440
<v Speaker 2>they have ten billion dollars, as well as ten billion

0:16:40.480 --> 0:16:43.440
<v Speaker 2>from soft Bank. It's still unclear how they pay for

0:16:43.520 --> 0:16:48.120
<v Speaker 2>everything now. While open ai likely has preferential payment structures

0:16:48.160 --> 0:16:51.160
<v Speaker 2>with all of their vendors, such as discounted rates with Microsoft,

0:16:51.240 --> 0:16:54.880
<v Speaker 2>Razi or cloud Services, it will eventually have to pay someone,

0:16:55.000 --> 0:16:57.720
<v Speaker 2>especially in the case of costs related to stargate, many

0:16:57.760 --> 0:17:00.880
<v Speaker 2>of which are upfront and involve physical things happening. In

0:17:00.920 --> 0:17:03.600
<v Speaker 2>the event that its costs a severe as reporting suggests,

0:17:03.640 --> 0:17:06.320
<v Speaker 2>open AI's revenue comes at a terrible cost and will

0:17:06.400 --> 0:17:09.280
<v Speaker 2>likely be immediately funneled directly into funding the obscene cost

0:17:09.320 --> 0:17:12.680
<v Speaker 2>behind inference and training models like GPT four point five,

0:17:12.880 --> 0:17:17.280
<v Speaker 2>which Sam Wuonton called a giant expensive model, and yeah,

0:17:17.320 --> 0:17:21.000
<v Speaker 2>nevertheless he pushed that to every single user. Worse. Still,

0:17:21.080 --> 0:17:24.560
<v Speaker 2>open ai has while delaying its next model, GPT five

0:17:24.800 --> 0:17:27.720
<v Speaker 2>promised to launch after all that its O three reasoning

0:17:27.760 --> 0:17:30.440
<v Speaker 2>model after saying it wouldn't do so, which is strange

0:17:30.480 --> 0:17:33.200
<v Speaker 2>because it turns up three is actually way more expensive

0:17:33.200 --> 0:17:35.960
<v Speaker 2>to run that people thought. With arc Price Foundation, a

0:17:36.000 --> 0:17:39.520
<v Speaker 2>nonprofit that makes the RKGI test for benchmarking models, because

0:17:39.560 --> 0:17:41.480
<v Speaker 2>real tests don't work on them, they need to make

0:17:41.520 --> 0:17:45.000
<v Speaker 2>something out. Well, they're estimating they will cost about thirty

0:17:45.040 --> 0:17:48.159
<v Speaker 2>grand a task, which is that's a lot of money.

0:17:49.280 --> 0:17:52.440
<v Speaker 2>And yes, they again it. Open Ai has a new

0:17:52.480 --> 0:17:54.760
<v Speaker 2>sugar Daddy in the form of SoftBank. But SoftBank has

0:17:54.800 --> 0:17:58.959
<v Speaker 2>to borrow money to meet its obligations for Stargate and

0:17:59.200 --> 0:18:03.439
<v Speaker 2>also open Ai, and this is leading to its financial

0:18:03.480 --> 0:18:07.240
<v Speaker 2>condition likely deteriorating. And that's S and P Global that

0:18:07.440 --> 0:18:10.320
<v Speaker 2>you know, you know, the S and P five those people,

0:18:10.920 --> 0:18:13.240
<v Speaker 2>those people said that that's not what you want to hear.

0:18:13.720 --> 0:18:15.680
<v Speaker 2>Let's look soft Bank. As of right now, soft Bank

0:18:15.720 --> 0:18:18.359
<v Speaker 2>has committed to the following at least thirty billion dollars

0:18:18.359 --> 0:18:20.720
<v Speaker 2>in funding, is part of open AI's recent forty billion

0:18:20.760 --> 0:18:24.080
<v Speaker 2>dollar funding round. Now. Soft Bank's filing surrounding open AI's

0:18:24.119 --> 0:18:26.720
<v Speaker 2>funding also suggests that they are ultimately on the hook

0:18:26.760 --> 0:18:30.280
<v Speaker 2>for the entire forty billion, but they can syndicate it.

0:18:30.359 --> 0:18:33.879
<v Speaker 2>Like I mentioned earlier, reporting suggests that the syndication will happen,

0:18:34.040 --> 0:18:36.280
<v Speaker 2>and it will happen with people at COUTE, Microsoft and

0:18:36.320 --> 0:18:40.000
<v Speaker 2>other investors. Now here's the funny part. If open ai

0:18:40.080 --> 0:18:42.320
<v Speaker 2>fails to convert to a for profit, that forty billion

0:18:42.320 --> 0:18:44.960
<v Speaker 2>dollars slashed down to a paltry thirty billion dollars, although

0:18:44.960 --> 0:18:48.000
<v Speaker 2>again soft Banks shares contingent upon whether it finds other

0:18:48.080 --> 0:18:51.080
<v Speaker 2>investors to join the deal. Now there's another three billion

0:18:51.119 --> 0:18:54.000
<v Speaker 2>dollars that soft Bank's promised to spend on open AI's tech,

0:18:54.560 --> 0:18:57.680
<v Speaker 2>and then nineteen billion dollars for the Stargate Data Center project,

0:18:57.680 --> 0:19:01.119
<v Speaker 2>which soft Bank is taking full financial responsibility for. The

0:19:01.119 --> 0:19:04.920
<v Speaker 2>information reports, and the total is either fifty two billion

0:19:05.000 --> 0:19:07.040
<v Speaker 2>dollars or sixty two billion dollars, with at least ten

0:19:07.080 --> 0:19:09.240
<v Speaker 2>billion dollars due by the end of twenty twenty five,

0:19:09.640 --> 0:19:12.679
<v Speaker 2>but more like twenty Like there's so much money. To

0:19:12.760 --> 0:19:16.679
<v Speaker 2>be clear, SoftBank had to borrow all of the ten

0:19:16.760 --> 0:19:19.120
<v Speaker 2>billion dollars. How the fuck are they meant to get

0:19:19.119 --> 0:19:22.399
<v Speaker 2>the thirty billion dollars? Now I kind of mentioned s

0:19:22.440 --> 0:19:24.240
<v Speaker 2>and P. But I want to really return to that now,

0:19:24.280 --> 0:19:27.320
<v Speaker 2>SoftBank's exposure to open Ai is materially harming the company,

0:19:27.359 --> 0:19:30.600
<v Speaker 2>and I'm quoting the Wall Street Journal here. Ratings agency

0:19:30.640 --> 0:19:33.560
<v Speaker 2>S and P Global said Tuesday that SoftBank's financial condition

0:19:33.600 --> 0:19:36.640
<v Speaker 2>will likely deteriorate as a result of the open Ai investment,

0:19:36.880 --> 0:19:39.280
<v Speaker 2>and that its plans to add debt could lead the

0:19:39.280 --> 0:19:43.560
<v Speaker 2>agency to consider downgrading SoftBank's ratings. While one might argue

0:19:43.560 --> 0:19:45.440
<v Speaker 2>that SoftBank has a good amount of cash, the journal

0:19:45.480 --> 0:19:48.000
<v Speaker 2>also adds that they're somewhat hamstrung in its use as

0:19:48.040 --> 0:19:51.800
<v Speaker 2>a result of CEO Massaoshi's son's dick shit reckless gambles.

0:19:51.920 --> 0:19:54.600
<v Speaker 2>Again quoting from the Wall Street Journal, SoftBank had a

0:19:54.600 --> 0:19:57.320
<v Speaker 2>decent buffer of thirty one billion dollars of cash as

0:19:57.359 --> 0:19:59.600
<v Speaker 2>of December thirty first, but the company is also pledged

0:19:59.600 --> 0:20:01.440
<v Speaker 2>to hold mone much of that in reserve to quell

0:20:01.480 --> 0:20:04.479
<v Speaker 2>worried investors. Soft Bank is committed to not borrow more

0:20:04.520 --> 0:20:06.359
<v Speaker 2>than twenty five percent of the value of its holdings,

0:20:06.359 --> 0:20:08.560
<v Speaker 2>which means it will likely need to sell some of

0:20:08.600 --> 0:20:10.520
<v Speaker 2>the other parts of its empire to pay the rest

0:20:10.520 --> 0:20:14.320
<v Speaker 2>of the open Ai deal. Worse still, it seems that

0:20:14.400 --> 0:20:17.760
<v Speaker 2>as mentioned before that SoftBank will be financing the entirety

0:20:17.840 --> 0:20:20.280
<v Speaker 2>of the first ten billion seven point five billion if

0:20:20.280 --> 0:20:23.360
<v Speaker 2>they're able to syndicate it. As a result, soft Bank

0:20:23.440 --> 0:20:25.000
<v Speaker 2>is likely going to have to sell off parts of

0:20:25.040 --> 0:20:28.240
<v Speaker 2>their actually valuable holdings in companies like Alababa or arm

0:20:28.359 --> 0:20:30.840
<v Speaker 2>or worse still, parts of their ailing investments from a

0:20:30.960 --> 0:20:34.400
<v Speaker 2>vision fund, resulting in a material loss and its underwater deals.

0:20:35.200 --> 0:20:39.560
<v Speaker 2>This is an untenable strategy, and I'd like to explain why. First.

0:20:39.680 --> 0:20:42.440
<v Speaker 2>Open Ai needs at least forty billion dollars a year

0:20:42.480 --> 0:20:45.679
<v Speaker 2>to survive and its costs are only increasing. While we

0:20:45.760 --> 0:20:48.600
<v Speaker 2>don't have much transparency into open AI's actual day to

0:20:48.680 --> 0:20:51.720
<v Speaker 2>day finances, we can make educated guesses that its costs

0:20:51.720 --> 0:20:54.240
<v Speaker 2>are increasing based on the amount of capital it's raising.

0:20:54.560 --> 0:20:57.200
<v Speaker 2>If open AI's costs were flat or only mildly increasing,

0:20:57.280 --> 0:20:59.720
<v Speaker 2>we'd expect to see raises roughly the same size as

0:20:59.720 --> 0:21:01.680
<v Speaker 2>the pre vis one six point six billions something in

0:21:01.760 --> 0:21:04.680
<v Speaker 2>that range. A forty billion dollar raise is nearly six

0:21:04.800 --> 0:21:08.640
<v Speaker 2>times the previous funding round. Admittedly, multiples like that aren't

0:21:08.680 --> 0:21:11.679
<v Speaker 2>particularly unusual. If a company raises three hundred grand in

0:21:11.680 --> 0:21:13.960
<v Speaker 2>the pre seed round and a three million dollar series

0:21:14.000 --> 0:21:17.600
<v Speaker 2>a round in funding. That's a tenfold increase. But we're

0:21:17.600 --> 0:21:20.040
<v Speaker 2>not talking about hundreds of thousands of dollars or even

0:21:20.119 --> 0:21:23.360
<v Speaker 2>millions of dollars. We're talking about billions of dollars. If

0:21:23.400 --> 0:21:26.040
<v Speaker 2>open AI's funding round with soft Bank goes as planned,

0:21:26.200 --> 0:21:29.280
<v Speaker 2>it will raise the equivalent of the entire GDP of Estonia,

0:21:29.520 --> 0:21:32.200
<v Speaker 2>a fairly wealthy country itself that's also a member of

0:21:32.280 --> 0:21:34.879
<v Speaker 2>NATO and the European Union. That alone should give you

0:21:34.920 --> 0:21:38.680
<v Speaker 2>a sense of how fucking stupid this all is. Stupid, sure,

0:21:38.760 --> 0:21:42.639
<v Speaker 2>but undoubtedly necessary. Per the information, open ai plans to

0:21:42.720 --> 0:21:45.040
<v Speaker 2>spend as much as twenty eight billion dollars in compute

0:21:45.040 --> 0:21:47.840
<v Speaker 2>on Microsoft dosy or cloud in twenty twenty eight. Over

0:21:47.880 --> 0:21:50.560
<v Speaker 2>a third of open AI's revenue, per the same article,

0:21:50.760 --> 0:21:55.000
<v Speaker 2>will come from SoftBank's alleged spend. It's reasonable to believe

0:21:55.040 --> 0:21:57.200
<v Speaker 2>that open ai will, as a result, need to raise

0:21:57.240 --> 0:21:59.880
<v Speaker 2>in excess of forty billion dollars a year, though it's

0:21:59.880 --> 0:22:01.840
<v Speaker 2>reasonable to believe that it will need to raise more

0:22:01.880 --> 0:22:05.320
<v Speaker 2>like fifty billion a year until they reach profitability you

0:22:05.359 --> 0:22:09.240
<v Speaker 2>know never And now this actually has a reason. It's

0:22:09.320 --> 0:22:11.199
<v Speaker 2>due to the growing cost of doing business as well

0:22:11.240 --> 0:22:13.879
<v Speaker 2>as the various infrastructure commitments they've made, both in the

0:22:13.920 --> 0:22:16.680
<v Speaker 2>terms of Stargate as well as deals with third party

0:22:16.680 --> 0:22:20.960
<v Speaker 2>suppliers like core Weavan, indeed, Microsoft. Open Ai CEO Sam

0:22:20.960 --> 0:22:23.960
<v Speaker 2>Mortman's statement around costs also suggests that they're going up

0:22:24.080 --> 0:22:27.120
<v Speaker 2>quite fiercely. In late February, he claimed that open ai

0:22:27.359 --> 0:22:29.960
<v Speaker 2>was out of GPUs. While this suggests that there is

0:22:30.280 --> 0:22:33.360
<v Speaker 2>demand for some products like it's horrifying him as generating

0:22:33.520 --> 0:22:37.359
<v Speaker 2>tech that has made abominations that insult Miyazaki and I

0:22:37.480 --> 0:22:40.080
<v Speaker 2>will hate them for it forever and thus made them

0:22:40.119 --> 0:22:42.280
<v Speaker 2>go viral in much it also means that to meet

0:22:42.320 --> 0:22:46.080
<v Speaker 2>the demand of the horrible abominations, open Ai needs to

0:22:46.080 --> 0:22:48.480
<v Speaker 2>spend more, and at the risk of repeating myself, that

0:22:48.560 --> 0:22:53.760
<v Speaker 2>demand doesn't necessarily translate into revenue or profitability. Now. Second,

0:22:53.920 --> 0:22:56.960
<v Speaker 2>soft Bank cannot fund open ai long term. I must

0:22:57.040 --> 0:22:59.480
<v Speaker 2>be clear. Open AI's costs are projected to be three

0:22:59.600 --> 0:23:04.119
<v Speaker 2>hundred twenty billion dollars over the next five years. SoftBank

0:23:04.160 --> 0:23:07.320
<v Speaker 2>has to overcome significant challenges to fund both open ai

0:23:07.480 --> 0:23:09.720
<v Speaker 2>and it's part of Stargate, and when I say fund,

0:23:09.760 --> 0:23:13.160
<v Speaker 2>I mean fund the current state of both projects, assuming

0:23:13.240 --> 0:23:17.720
<v Speaker 2>no further obligations or complications. The Information reports the open

0:23:17.760 --> 0:23:20.160
<v Speaker 2>Ai forecast that will spend as I mentioned, twenty eight

0:23:20.200 --> 0:23:22.360
<v Speaker 2>billion dollars on compute with Microsoft a loan in twenty

0:23:22.400 --> 0:23:25.359
<v Speaker 2>twenty eight, and I apologize for repeating myself, these numbers

0:23:25.400 --> 0:23:28.400
<v Speaker 2>fucking matter. The same arcle also reports that open ai

0:23:28.480 --> 0:23:30.400
<v Speaker 2>would turn profitable by the end of the decade after

0:23:30.440 --> 0:23:34.240
<v Speaker 2>building out Stargate, suggesting that open AI's operating expend ditches

0:23:34.280 --> 0:23:37.880
<v Speaker 2>will grow exponentially year over year. According to the Information,

0:23:37.920 --> 0:23:41.280
<v Speaker 2>it expects its costs to surpass three hundred and twenty

0:23:41.320 --> 0:23:43.639
<v Speaker 2>billion dollars between twenty twenty five and twenty thirty, with

0:23:43.720 --> 0:23:46.160
<v Speaker 2>half of that going towards funding model training and development.

0:23:47.400 --> 0:23:50.159
<v Speaker 2>How the fuck does building stargate make them profitable? I

0:23:50.200 --> 0:23:53.919
<v Speaker 2>really can't anyway, it won't. It's the same shit. But

0:23:54.000 --> 0:23:56.800
<v Speaker 2>SoftBank has had to and will continue having to go

0:23:56.880 --> 0:24:00.280
<v Speaker 2>to remarkable lengths to fund open AI's current forty billion

0:24:00.320 --> 0:24:02.919
<v Speaker 2>dollar round, lengths so significant that it may lead to

0:24:02.960 --> 0:24:06.080
<v Speaker 2>their credit rating being further downgraded. As I mentioned, even

0:24:06.119 --> 0:24:09.879
<v Speaker 2>if we assume the best case scenario, open ai successfully

0:24:09.920 --> 0:24:11.760
<v Speaker 2>converts to a for profit entity by the end of

0:24:11.760 --> 0:24:14.560
<v Speaker 2>twenty twenty five and gets that full thirty billion dollars.

0:24:15.200 --> 0:24:18.199
<v Speaker 2>It seems unlikely, if not impossible, for it to continue

0:24:18.280 --> 0:24:21.600
<v Speaker 2>raising the amount of capital they need to to continue operations.

0:24:22.040 --> 0:24:24.359
<v Speaker 2>As I've argued in multiple news letters and podcasts, there

0:24:24.359 --> 0:24:26.200
<v Speaker 2>are only a few entities that can provide the kind

0:24:26.200 --> 0:24:28.840
<v Speaker 2>of funding that open AI needs. These include big tech

0:24:28.880 --> 0:24:31.960
<v Speaker 2>focused investment firms like soft Bank, sovereign wealth funds like

0:24:31.960 --> 0:24:34.359
<v Speaker 2>those of Saudi Arabia and the UAE, and perhaps the

0:24:34.440 --> 0:24:37.560
<v Speaker 2>largest tech companies. I also want to be clear that

0:24:37.920 --> 0:24:40.600
<v Speaker 2>I keep getting messages being like the gatherment could do

0:24:40.720 --> 0:24:42.679
<v Speaker 2>it daily, couldn't it get me word? If the government

0:24:42.680 --> 0:24:44.840
<v Speaker 2>did get it's not going to happen. Stop fucking asking me.

0:24:45.040 --> 0:24:47.720
<v Speaker 2>It's just not going to happen. The government's not sending

0:24:47.800 --> 0:24:50.480
<v Speaker 2>forty billion dollars to Sam Altman. They're not going to

0:24:50.520 --> 0:24:53.400
<v Speaker 2>do it. I will apologize personally to each and every

0:24:53.480 --> 0:24:55.280
<v Speaker 2>one of you if I'm wrong, but they will need

0:24:55.320 --> 0:24:58.359
<v Speaker 2>to send forty billion dollars cash to open ai. But

0:24:58.440 --> 0:25:01.160
<v Speaker 2>these entities can meet open ai needs for some time,

0:25:01.200 --> 0:25:05.119
<v Speaker 2>but not all the time. It's not realistic to expect

0:25:05.119 --> 0:25:07.639
<v Speaker 2>soft Bank or Microsoft or the Soald's Oracle or whoever

0:25:07.680 --> 0:25:10.640
<v Speaker 2>to provide forty billion dollars every year for the foreseeable future.

0:25:11.000 --> 0:25:15.040
<v Speaker 2>And just to be clear, this is what open Ai needs. Eventually,

0:25:15.040 --> 0:25:17.800
<v Speaker 2>even the Souds have to have a break. And I

0:25:17.800 --> 0:25:19.560
<v Speaker 2>don't know if you remember from a previous episode, but

0:25:19.560 --> 0:25:21.840
<v Speaker 2>Masayoshi's son gets a lot of his money from them,

0:25:22.040 --> 0:25:24.680
<v Speaker 2>and they're not happy with him. In fact, Masaoshi's son

0:25:24.720 --> 0:25:27.080
<v Speaker 2>has said a few years ago that he owed Muhammad

0:25:27.080 --> 0:25:31.080
<v Speaker 2>bill and Salmon. Not a great guy to owe anyway.

0:25:31.119 --> 0:25:33.160
<v Speaker 2>This really is especially true for soft Bank by the way.

0:25:33.200 --> 0:25:36.000
<v Speaker 2>They're bruised. They're battered after several rough years, including a

0:25:36.000 --> 0:25:39.800
<v Speaker 2>failed multi billion dollar investment in we Work. Based on

0:25:39.840 --> 0:25:41.639
<v Speaker 2>its current promise to not borrow more than twenty five

0:25:41.680 --> 0:25:44.320
<v Speaker 2>percent of its holdings, it's near impossible that SoftBank will

0:25:44.359 --> 0:25:46.920
<v Speaker 2>be able to continue to fund open Ai at this rate,

0:25:47.400 --> 0:25:50.399
<v Speaker 2>and forty billion dollars a year may not actually be enough.

0:25:51.520 --> 0:25:54.640
<v Speaker 2>Based on soft Bank's last reported equity value of its holdings,

0:25:54.720 --> 0:25:57.400
<v Speaker 2>they have about two hundred and twenty nine billion dollars

0:25:57.440 --> 0:25:59.920
<v Speaker 2>of stuff, meaning that they can borrow just over fifty

0:26:00.080 --> 0:26:03.600
<v Speaker 2>seven billion while remaining compliant with these guidelines. In any case,

0:26:03.640 --> 0:26:07.000
<v Speaker 2>it's unclear how soft Bank can fund open Ai, but

0:26:07.040 --> 0:26:24.040
<v Speaker 2>it's far clearer that nobody else is willing to. Now

0:26:24.440 --> 0:26:26.920
<v Speaker 2>we're going to move off the questions for a minute

0:26:26.960 --> 0:26:29.680
<v Speaker 2>because I just want to get into some problems. I've

0:26:29.720 --> 0:26:32.280
<v Speaker 2>popped because it turns out that open ai they got

0:26:32.320 --> 0:26:36.399
<v Speaker 2>doodoo ass. That's a professional finance term. I did just

0:26:36.400 --> 0:26:38.520
<v Speaker 2>say financi and we're just going to keep that. Now.

0:26:38.600 --> 0:26:41.119
<v Speaker 2>Open Ai has started running into capacity issues, and this

0:26:41.240 --> 0:26:44.119
<v Speaker 2>is a real problem, and it suggests material instability in

0:26:44.160 --> 0:26:47.600
<v Speaker 2>their business or infrastructure, and it's really not clear how

0:26:47.640 --> 0:26:51.479
<v Speaker 2>open ai expands further. Let me explain it's important to

0:26:51.480 --> 0:26:53.720
<v Speaker 2>note that open ai does not really have any of

0:26:53.720 --> 0:26:56.840
<v Speaker 2>its own compute infrastructure. The majority of its compute is

0:26:56.880 --> 0:27:00.679
<v Speaker 2>provided by Microsoft, though as mentioned previously, open now has

0:27:00.680 --> 0:27:02.880
<v Speaker 2>a deal with core Weave to take over the capacity

0:27:02.880 --> 0:27:05.200
<v Speaker 2>that Microsoft was going to have twelve billion dollars or

0:27:05.240 --> 0:27:08.400
<v Speaker 2>so of capacity in the future. Anyway, in the last

0:27:08.520 --> 0:27:10.560
<v Speaker 2>ninety days, Sam Altman has complained about the lack of

0:27:10.640 --> 0:27:13.919
<v Speaker 2>GPUs and pressure on open ai servers many times. In

0:27:13.960 --> 0:27:16.640
<v Speaker 2>my newsletter published a few days ago, I was did

0:27:16.720 --> 0:27:20.320
<v Speaker 2>six such examples. Should you be curious, these statements in

0:27:20.359 --> 0:27:23.000
<v Speaker 2>a bubble seem either harmless or like open AI's growth

0:27:23.080 --> 0:27:25.600
<v Speaker 2>is skyrocketing, the latter of which might indeed be true,

0:27:25.600 --> 0:27:27.760
<v Speaker 2>but botzil for a company that burns money with every

0:27:27.760 --> 0:27:31.240
<v Speaker 2>single user, any mention of rate limits or performance issues

0:27:31.240 --> 0:27:34.399
<v Speaker 2>suggests that open ai is having significant capacity issues, and

0:27:34.480 --> 0:27:36.679
<v Speaker 2>at this point it's unclear what further capacity it can

0:27:36.720 --> 0:27:40.439
<v Speaker 2>actually expand to outside of that currently available. Sam Wilton's

0:27:40.440 --> 0:27:44.600
<v Speaker 2>complaining about melting GPUs, You've got the lead from Sourer saying, yeah,

0:27:44.640 --> 0:27:47.160
<v Speaker 2>we're gonna have some problems showing you stuff for a minute,

0:27:47.359 --> 0:27:49.520
<v Speaker 2>just because like, yeah, we're melting GPUs. There was a

0:27:49.560 --> 0:27:52.720
<v Speaker 2>whole thing about how you should expect delays on product

0:27:52.800 --> 0:27:56.720
<v Speaker 2>launches and service problems. None of it's really good, And

0:27:56.920 --> 0:27:59.560
<v Speaker 2>like I said, it isn't really obvious how open ai

0:27:59.680 --> 0:28:03.000
<v Speaker 2>is going to expand much further. Remember Microsoft has now

0:28:03.040 --> 0:28:06.160
<v Speaker 2>pulled as much as two gigawatts of data center projects,

0:28:06.280 --> 0:28:09.240
<v Speaker 2>walked away from a billion dollar data center development in

0:28:09.280 --> 0:28:11.840
<v Speaker 2>Ohio and declined the option the one I just mentioned

0:28:11.920 --> 0:28:14.360
<v Speaker 2>on twelve billion dollars a compute from Corewave that open

0:28:14.400 --> 0:28:17.439
<v Speaker 2>ai had to pick up, Meaning that open ai may

0:28:17.480 --> 0:28:20.200
<v Speaker 2>be pushing up against the limits of what is physically available.

0:28:20.720 --> 0:28:24.120
<v Speaker 2>While The total available capacity of GPUs at many providers

0:28:24.160 --> 0:28:26.400
<v Speaker 2>like Lambda and Crusoe is unknown, and indeed, I don't

0:28:26.400 --> 0:28:28.919
<v Speaker 2>know if Cruso has a single data center at this point.

0:28:29.119 --> 0:28:31.879
<v Speaker 2>We know that Corewave has upon approximately three hundred and

0:28:31.960 --> 0:28:35.720
<v Speaker 2>sixty megawats available compared to microsoft six point five to

0:28:35.760 --> 0:28:38.960
<v Speaker 2>seven point five gigawads, a large chunk of which I

0:28:39.000 --> 0:28:42.360
<v Speaker 2>think powers open Ai. If open ai is running into

0:28:42.360 --> 0:28:44.880
<v Speaker 2>capacity issues, it could be one of the following. They

0:28:44.880 --> 0:28:46.800
<v Speaker 2>could be running up against the limit of what Microsoft

0:28:46.800 --> 0:28:49.000
<v Speaker 2>has available or is willing to offer the company. The

0:28:49.000 --> 0:28:51.680
<v Speaker 2>Information reported in October twenty twenty four the open Ai

0:28:51.880 --> 0:28:55.040
<v Speaker 2>was frustrated with Microsoft and said it wasn't moving fast

0:28:55.120 --> 0:28:58.200
<v Speaker 2>enough to supply open ai with servers and now. It

0:28:58.200 --> 0:29:00.920
<v Speaker 2>could also be that while open ai is cony is sufficient,

0:29:00.920 --> 0:29:03.719
<v Speaker 2>it does not have the resources available to easily handle

0:29:03.760 --> 0:29:06.320
<v Speaker 2>bursts in user growth in the stable manner. Per The

0:29:06.360 --> 0:29:09.880
<v Speaker 2>Information's reporting, Microsoft promised open Ai three hundred thousand in

0:29:09.960 --> 0:29:12.560
<v Speaker 2>Vidio GB two hundred Blackwell chips. By the end of

0:29:12.560 --> 0:29:15.600
<v Speaker 2>twenty twenty five, are roughly eighteen billion dollars worth of them.

0:29:15.800 --> 0:29:18.720
<v Speaker 2>It's unclear if this has changed since Microsoft allowed open

0:29:18.760 --> 0:29:21.160
<v Speaker 2>Ai to seek other compute from other companies in late

0:29:21.240 --> 0:29:24.200
<v Speaker 2>January twenty twenty five. I also don't believe that open

0:29:24.200 --> 0:29:27.640
<v Speaker 2>ai has any other viable options for existing compute infrastructure

0:29:27.640 --> 0:29:31.320
<v Speaker 2>outside of Microsoft. Corweep's current data centers mostly feature in

0:29:31.400 --> 0:29:34.840
<v Speaker 2>Vidia's aging Hopper GPUs, and while it could and likely

0:29:35.000 --> 0:29:38.400
<v Speaker 2>is retrofitting its current infrastructure with Blackwell chips, doing so

0:29:38.520 --> 0:29:41.920
<v Speaker 2>isn't easy or cheap. Blackwell chips require far more powerful

0:29:41.960 --> 0:29:44.600
<v Speaker 2>cooling and server infrastructure to make them run smoothly, a

0:29:44.600 --> 0:29:46.400
<v Speaker 2>problem which led to a delay in their delivery to

0:29:46.440 --> 0:29:49.240
<v Speaker 2>most customers according to the information. And even if Corey

0:29:49.480 --> 0:29:52.400
<v Speaker 2>was able to replace every last Hopper GPU with a Blackwell,

0:29:52.440 --> 0:29:55.280
<v Speaker 2>and they won't, it still wouldn't match what open ai

0:29:55.440 --> 0:29:58.440
<v Speaker 2>needs to expand one might argue that it simply needs

0:29:58.440 --> 0:30:00.800
<v Speaker 2>to wait for the construction of the Stargus data center,

0:30:00.960 --> 0:30:02.560
<v Speaker 2>or for core we have to finish the gigawa or

0:30:02.600 --> 0:30:05.800
<v Speaker 2>so of construction it's working on. I want to be

0:30:05.880 --> 0:30:08.480
<v Speaker 2>clear how impossible that is. I need to be clear.

0:30:08.560 --> 0:30:10.920
<v Speaker 2>One of my least favorite responses to my work as

0:30:10.920 --> 0:30:13.440
<v Speaker 2>people saying they'll just build more data centers. They just

0:30:13.480 --> 0:30:14.920
<v Speaker 2>go and build them inn They're just going to build

0:30:14.960 --> 0:30:17.640
<v Speaker 2>them right now. They won't. You can't just data centers

0:30:17.680 --> 0:30:20.680
<v Speaker 2>don't just grow from the fucking ground. As I've argued

0:30:20.680 --> 0:30:23.120
<v Speaker 2>in the past, I have serious concerns over the viability

0:30:23.120 --> 0:30:26.160
<v Speaker 2>of core We've ever completing. It's alleged contracted one point

0:30:26.200 --> 0:30:29.440
<v Speaker 2>three gigawatts of capacity. Based on calculations, it'll have to

0:30:29.440 --> 0:30:32.080
<v Speaker 2>spend in excess of thirty nine billion dollars to build it.

0:30:32.080 --> 0:30:34.960
<v Speaker 2>It's unclear how that will happen, and it doesn't have

0:30:35.040 --> 0:30:38.360
<v Speaker 2>the money to do so, like it actually does not

0:30:38.400 --> 0:30:41.160
<v Speaker 2>have the cash. I'll get into that later, but they

0:30:41.200 --> 0:30:44.680
<v Speaker 2>don't have the money. However, even if I were to

0:30:44.800 --> 0:30:47.959
<v Speaker 2>humor this idea, it's impossible that any of this project

0:30:48.040 --> 0:30:49.640
<v Speaker 2>is done by the end of twenty twenty five, and

0:30:49.680 --> 0:30:52.560
<v Speaker 2>I'd argue even in twenty twenty six. I can find

0:30:52.560 --> 0:30:54.800
<v Speaker 2>no commitments to any time scale, other than the fact

0:30:54.800 --> 0:30:57.040
<v Speaker 2>that open Ai will allegedly start paying core Weave in

0:30:57.080 --> 0:31:00.080
<v Speaker 2>October per the information, which could very well be using

0:31:00.120 --> 0:31:03.960
<v Speaker 2>their current capacity. I also can't find any evidence that Cruso,

0:31:04.040 --> 0:31:06.760
<v Speaker 2>the company building the Stargate data center in Texas, has

0:31:06.840 --> 0:31:10.520
<v Speaker 2>any compute anywhere else. Lander, a GPU compute company that

0:31:10.640 --> 0:31:12.840
<v Speaker 2>raised three hundred and twenty million dollars earlier in the year,

0:31:13.160 --> 0:31:15.840
<v Speaker 2>and according to Data Center Dynamics, operates out of co

0:31:15.960 --> 0:31:19.200
<v Speaker 2>location data centers in San Francisco, California, and Allen, Texas.

0:31:19.320 --> 0:31:21.120
<v Speaker 2>It's backed by Moorner eight hundred and twenty million in

0:31:21.200 --> 0:31:24.640
<v Speaker 2>funds raised this year. All of that just kind of

0:31:25.000 --> 0:31:27.040
<v Speaker 2>doesn't say that they have a data center at all.

0:31:27.040 --> 0:31:30.480
<v Speaker 2>A colocation. That means you're in someone else's buddy. You

0:31:30.520 --> 0:31:33.640
<v Speaker 2>don't own a house, you're ranting one. And just to

0:31:33.640 --> 0:31:36.640
<v Speaker 2>be clear, open AI's ability to scale is entirely contingent

0:31:36.640 --> 0:31:39.239
<v Speaker 2>on the availability of whatever data center providers it has

0:31:39.280 --> 0:31:42.640
<v Speaker 2>relationships with, and all of their growths coming from these

0:31:42.640 --> 0:31:46.600
<v Speaker 2>to these two companies. I'll get there, don't worry. But

0:31:47.040 --> 0:31:50.480
<v Speaker 2>every time I kind of say this stuff out loud,

0:31:50.520 --> 0:31:54.200
<v Speaker 2>I feel my soul slightly stripped from me. I feel

0:31:54.200 --> 0:31:57.360
<v Speaker 2>like I feel like I'm in hell because this is

0:31:57.400 --> 0:32:00.480
<v Speaker 2>an insane thing. If you or I win to someone

0:32:00.720 --> 0:32:03.160
<v Speaker 2>went like, hey, look, I'm going to lose five billion dollars,

0:32:03.200 --> 0:32:05.240
<v Speaker 2>but I promise you God will come out of the computer.

0:32:05.960 --> 0:32:09.560
<v Speaker 2>They institutionalize you or me, they definitely would instalize me.

0:32:11.000 --> 0:32:13.760
<v Speaker 2>It's just fucking stupid, But in any case, this means

0:32:13.760 --> 0:32:16.160
<v Speaker 2>that open AI's only real choice for GPUs is either

0:32:16.240 --> 0:32:19.800
<v Speaker 2>Core Weave or Microsoft. And while it's hard to calculate precisely,

0:32:19.880 --> 0:32:23.960
<v Speaker 2>open AI's best case scenario is that sixteen thousand GPUs

0:32:24.000 --> 0:32:25.880
<v Speaker 2>come online in the summer of twenty twenty five as

0:32:25.880 --> 0:32:28.200
<v Speaker 2>part of the Stargate Data Center project. And that's a

0:32:28.280 --> 0:32:30.440
<v Speaker 2>drop in the bucket compared to the three hundred thousand

0:32:30.480 --> 0:32:34.120
<v Speaker 2>of the fucking things that Microsoft have previously promised. Now,

0:32:34.160 --> 0:32:39.240
<v Speaker 2>one last thing, any capacity or expansion issues that happen

0:32:39.320 --> 0:32:42.760
<v Speaker 2>with open Ai will needcap this company, open ai is,

0:32:42.800 --> 0:32:45.680
<v Speaker 2>regardless of how you or I may feel about generative Ai,

0:32:46.240 --> 0:32:49.080
<v Speaker 2>one of the fastest growing companies of all time. It

0:32:49.120 --> 0:32:52.440
<v Speaker 2>currently has, according to its own statements, over five hundred

0:32:52.520 --> 0:32:56.400
<v Speaker 2>million weekly active users. Putting aside that each user is unprofitable.

0:32:56.680 --> 0:32:59.640
<v Speaker 2>Such remarkable growth, especially as it's partially as a result

0:32:59.680 --> 0:33:03.400
<v Speaker 2>of its extremely resourced intensive image generator, is a massive,

0:33:03.840 --> 0:33:07.640
<v Speaker 2>horrifying strain on their infrastructure. The vast majority of open

0:33:07.680 --> 0:33:11.080
<v Speaker 2>aiyes users are free customers using chat GPT with only,

0:33:11.160 --> 0:33:14.080
<v Speaker 2>like I mentioned earlier, twenty million paying subscribers, and the

0:33:14.160 --> 0:33:16.720
<v Speaker 2>vast majority of them on the cheapest twenty bucks plan.

0:33:17.280 --> 0:33:20.680
<v Speaker 2>Open Aiyes, services, even in the case of image generation,

0:33:21.240 --> 0:33:24.480
<v Speaker 2>are relatively commoditized, meaning that users can, if they really care,

0:33:24.520 --> 0:33:27.480
<v Speaker 2>go and use any number of other large language model services.

0:33:27.800 --> 0:33:30.840
<v Speaker 2>They could use bing Oh, they could use stable Diffusion,

0:33:30.840 --> 0:33:33.400
<v Speaker 2>they could even use GROC if they really I don't

0:33:33.480 --> 0:33:37.240
<v Speaker 2>like saying it, but free users they're a burden on

0:33:37.280 --> 0:33:40.560
<v Speaker 2>the company, especially with such piss poor conversion rates, losing

0:33:40.840 --> 0:33:43.280
<v Speaker 2>money with each prompt, which is by the way, also

0:33:43.880 --> 0:33:46.760
<v Speaker 2>case with paying subscribers, and the remarkable popularity of its

0:33:46.760 --> 0:33:51.840
<v Speaker 2>horrible image generator, it only threatens to bring more burdens

0:33:51.840 --> 0:33:54.360
<v Speaker 2>than one off customers that will generate a few abominable

0:33:54.400 --> 0:33:58.000
<v Speaker 2>studio ghibli pictures of Garfield with giant knockers and then

0:33:58.040 --> 0:34:02.200
<v Speaker 2>never return. If AI's growth continues at this rate, it

0:34:02.280 --> 0:34:04.800
<v Speaker 2>will run into capacity issues, and it does not have

0:34:04.880 --> 0:34:07.240
<v Speaker 2>much room to expand. Well, we don't know how much

0:34:07.280 --> 0:34:10.040
<v Speaker 2>capacity they're taking up with Microsoft, or indeed whether Microsoft

0:34:10.040 --> 0:34:13.040
<v Speaker 2>is approaching capacity or otherwise limiting them. We do know

0:34:13.080 --> 0:34:15.480
<v Speaker 2>that open ai has seen reason to beg for access

0:34:15.480 --> 0:34:18.640
<v Speaker 2>for more GPUs. In simpler terms, even if open ai

0:34:18.800 --> 0:34:21.319
<v Speaker 2>wasn't running out of money, even If open ai wasn't

0:34:21.440 --> 0:34:25.240
<v Speaker 2>un horrifyingly unprofitable, it also may not have enough GPUs

0:34:25.280 --> 0:34:28.560
<v Speaker 2>to continue providing its services in a reliable manner. If

0:34:28.560 --> 0:34:30.799
<v Speaker 2>that's the case, there really isn't that much that can

0:34:30.840 --> 0:34:33.719
<v Speaker 2>be done other than significantly limiting free users activity on

0:34:33.760 --> 0:34:36.720
<v Speaker 2>the platform, which is open AI's primary mechanism for revenue

0:34:36.760 --> 0:34:40.280
<v Speaker 2>growth and customer acquisition. Limiting activity or changing the economics

0:34:40.280 --> 0:34:43.480
<v Speaker 2>behind its paid product. And this is quoting sam Orman.

0:34:44.200 --> 0:34:46.640
<v Speaker 2>They could potentially find some way to let people pay

0:34:47.040 --> 0:34:49.919
<v Speaker 2>for compute they want to use more dynamically, that's not good.

0:34:50.400 --> 0:34:52.800
<v Speaker 2>Almans come up with some other ideas, like an idea

0:34:52.840 --> 0:34:55.840
<v Speaker 2>for paid plans on March fourth, where twenty bucks a

0:34:55.880 --> 0:34:58.680
<v Speaker 2>month goes to credits which you can use across features

0:34:58.719 --> 0:35:01.319
<v Speaker 2>like deep research, oh one at four point five sore

0:35:01.360 --> 0:35:03.680
<v Speaker 2>and so on, with no fixed limits per feature, and

0:35:03.840 --> 0:35:05.760
<v Speaker 2>you choose what you want. If you run out of credits,

0:35:05.760 --> 0:35:07.600
<v Speaker 2>you can buy more. I just want to be clear

0:35:07.640 --> 0:35:10.000
<v Speaker 2>that that is a terrible fucking deal. We have no

0:35:10.120 --> 0:35:12.240
<v Speaker 2>idea what the credits would be, and it would definitely

0:35:12.239 --> 0:35:14.359
<v Speaker 2>be rigged so that you would have to buy more.

0:35:14.920 --> 0:35:18.360
<v Speaker 2>He's also brought up things like mentioning losing two hundred

0:35:18.400 --> 0:35:20.799
<v Speaker 2>bucks a month on the pro subscription, but he's a

0:35:20.840 --> 0:35:23.799
<v Speaker 2>funny one. Buried in an article from the Information from

0:35:23.800 --> 0:35:26.319
<v Speaker 2>March fifth is a comment that suggests that open ai

0:35:26.440 --> 0:35:29.640
<v Speaker 2>is considering measures like changing its pricing model entirely, with

0:35:29.719 --> 0:35:32.960
<v Speaker 2>Altman reportedly telling developers in London in February twenty twenty

0:35:33.000 --> 0:35:35.279
<v Speaker 2>five the open aiy is prime to judge twenty or

0:35:35.320 --> 0:35:38.279
<v Speaker 2>thirty percent of pro customers a higher price because of

0:35:38.320 --> 0:35:41.520
<v Speaker 2>how many research queries they're doing. But he suggested in

0:35:41.680 --> 0:35:43.960
<v Speaker 2>a la carte or pays you go approach when it

0:35:43.960 --> 0:35:46.279
<v Speaker 2>comes to agents, though we have to judge more than

0:35:46.280 --> 0:35:49.319
<v Speaker 2>two hundred dollars a month. Fucking just this fucking guy.

0:35:50.239 --> 0:35:52.480
<v Speaker 2>The problem is that with all of these measures, even

0:35:52.520 --> 0:35:54.760
<v Speaker 2>if they succeeded in generating more money for the company,

0:35:54.880 --> 0:35:58.319
<v Speaker 2>also need to reduce the burden on open aiyes available infrastructure.

0:35:58.719 --> 0:36:03.040
<v Speaker 2>Remember remember data centers can take three to six years

0:36:03.080 --> 0:36:05.759
<v Speaker 2>to build, and even with stargates accelerated, and I'd argue

0:36:05.800 --> 0:36:09.400
<v Speaker 2>on realistic timelines, open ai isn't unlocking a tenth of

0:36:09.440 --> 0:36:12.880
<v Speaker 2>the promised compute that Microsoft gave them. Three hundred thousand

0:36:12.920 --> 0:36:16.400
<v Speaker 2>GPUs is a lot sixteen thousand. Really not, so what

0:36:16.560 --> 0:36:19.799
<v Speaker 2>might these capacities look like? What are the consequences? You

0:36:19.840 --> 0:36:23.440
<v Speaker 2>love payal horses. I love payal horses. Let's get riding.

0:36:24.080 --> 0:36:27.160
<v Speaker 2>Though downtime might be an obvious choice. Capacity issues that

0:36:27.200 --> 0:36:29.799
<v Speaker 2>open ai will likely manifest in hard limits on what

0:36:29.880 --> 0:36:34.000
<v Speaker 2>free users can do, some of which I've documented previously. Nevertheless,

0:36:34.040 --> 0:36:36.920
<v Speaker 2>I believe the real pay all horses of capacity issues

0:36:36.920 --> 0:36:40.000
<v Speaker 2>come from arbitrary limits on any given user group, meaning

0:36:40.040 --> 0:36:42.880
<v Speaker 2>both free users and paid users. Some limits on what

0:36:42.920 --> 0:36:44.719
<v Speaker 2>a user can do, a reduction in the number of

0:36:44.760 --> 0:36:48.440
<v Speaker 2>generations of images for paid users, any introduction of peak hours,

0:36:48.680 --> 0:36:50.960
<v Speaker 2>or any increases in prices are assigned that open ai

0:36:51.080 --> 0:36:53.920
<v Speaker 2>is running out of GPUs, which it's already said as

0:36:54.000 --> 0:36:57.960
<v Speaker 2>happening publicly. However, the really obvious thing, the real obvious

0:36:58.040 --> 0:37:01.400
<v Speaker 2>pale horse, would be service degradation delays in generations of

0:37:01.440 --> 0:37:04.520
<v Speaker 2>any kind, five hundred status code errors, or chat GBT

0:37:05.000 --> 0:37:08.080
<v Speaker 2>just failing to produce an answer. Open ai has up

0:37:08.160 --> 0:37:11.719
<v Speaker 2>until this point had fairly impressive uptime. Still, it's if

0:37:11.719 --> 0:37:13.839
<v Speaker 2>it's running up against the wall this streak will end.

0:37:15.239 --> 0:37:17.920
<v Speaker 2>The consequences depend on how often these issues occur. And

0:37:17.960 --> 0:37:20.840
<v Speaker 2>to whom they occur. If free users face service degradation,

0:37:20.920 --> 0:37:23.239
<v Speaker 2>they'll bounce off the product, as their US is likely

0:37:23.280 --> 0:37:25.719
<v Speaker 2>far more fleeting than the paid user, which will begin

0:37:25.800 --> 0:37:30.000
<v Speaker 2>to erode open AI's growth. Ironically, rapid and especially unprecedented

0:37:30.040 --> 0:37:32.800
<v Speaker 2>growth in one of open AI's competitors like Higher or

0:37:32.840 --> 0:37:35.560
<v Speaker 2>Anthropic could also represent a pale horse for open Ai,

0:37:35.680 --> 0:37:38.880
<v Speaker 2>though based on the monthly active users I've seen from Anthropic,

0:37:38.920 --> 0:37:41.160
<v Speaker 2>I don't think that's going to be a problem now.

0:37:41.200 --> 0:37:45.120
<v Speaker 2>If paid users face service degradation, it will likely cause

0:37:45.120 --> 0:37:47.120
<v Speaker 2>the most harm to the company is while paid users

0:37:47.200 --> 0:37:49.200
<v Speaker 2>still lose open Ai money, in the end, at least

0:37:49.200 --> 0:37:52.840
<v Speaker 2>they receive some money. Open Ai is effectively one choice

0:37:52.840 --> 0:37:56.400
<v Speaker 2>here get more GPUs from Microsoft, and really its future

0:37:56.440 --> 0:37:59.560
<v Speaker 2>depends heavily on both Microsoft's generosity and there being enough

0:37:59.600 --> 0:38:01.719
<v Speaker 2>of them time when Microsoft is pulled back from over

0:38:01.760 --> 0:38:05.520
<v Speaker 2>two gigabats of data centers, specifically according to td COD

0:38:05.760 --> 0:38:10.560
<v Speaker 2>because of it moving away from providing compute to open Ai. Now, admittedly,

0:38:10.600 --> 0:38:13.600
<v Speaker 2>open Ai has previously spent more on training models than inference,

0:38:13.640 --> 0:38:15.880
<v Speaker 2>and that's the actual running of them, and the company

0:38:15.920 --> 0:38:18.360
<v Speaker 2>might be able to smooth downtime issues by shifting capacity.

0:38:18.400 --> 0:38:20.000
<v Speaker 2>This would of course have a knock on effect on

0:38:20.040 --> 0:38:22.160
<v Speaker 2>its ability to develop new models, and the company is

0:38:22.200 --> 0:38:25.160
<v Speaker 2>already losing ground, particularly when it comes to Chinese rivals

0:38:25.200 --> 0:38:28.000
<v Speaker 2>like Deepseek. Now, I know this has been a long

0:38:28.040 --> 0:38:30.320
<v Speaker 2>episode and the fact is I'm not even close to finish.

0:38:30.480 --> 0:38:33.440
<v Speaker 2>I have some more tough, tough questions and tough problems

0:38:33.440 --> 0:38:36.080
<v Speaker 2>for open Ai. Tune in to the next episode to

0:38:36.120 --> 0:38:38.440
<v Speaker 2>hear them. I realized this has been a lot, and

0:38:38.480 --> 0:38:40.400
<v Speaker 2>I know you're very patient with me. You let me

0:38:40.440 --> 0:38:43.799
<v Speaker 2>read the long scripts. But this stuff is important. It's

0:38:43.840 --> 0:38:46.080
<v Speaker 2>important to me, and I think you're going to find

0:38:46.120 --> 0:38:50.120
<v Speaker 2>it important too. And I'll have a nice sexy conclusion

0:38:50.120 --> 0:38:51.960
<v Speaker 2>to this at the end of the next one. But

0:38:52.239 --> 0:38:54.720
<v Speaker 2>before I go, please vote for me in the fucking Webbies.

0:38:54.760 --> 0:38:56.799
<v Speaker 2>I want to win. It's my pin post on Blue

0:38:56.800 --> 0:38:59.040
<v Speaker 2>Sky and to wear in my handles eadsit Tran dot

0:38:59.080 --> 0:39:00.600
<v Speaker 2>Com and Blue Sky. Please he vote for me and

0:39:00.640 --> 0:39:11.759
<v Speaker 2>the Webbies. Anyway, catch you next episode. Thank you for

0:39:11.800 --> 0:39:14.480
<v Speaker 2>listening to Better Offline. The editor and composer of the

0:39:14.480 --> 0:39:17.600
<v Speaker 2>Better Offline theme song is Matasowski. You can check out

0:39:17.600 --> 0:39:21.279
<v Speaker 2>more of his music and audio projects at Matasowski dot com,

0:39:21.440 --> 0:39:24.760
<v Speaker 2>M A T T O S O W s ki

0:39:25.160 --> 0:39:28.040
<v Speaker 2>dot com. You can email me at easy at better

0:39:28.080 --> 0:39:30.480
<v Speaker 2>offline dot com or visit better offline dot com to

0:39:30.560 --> 0:39:33.400
<v Speaker 2>find more podcast links and of course my newsletter. I

0:39:33.440 --> 0:39:36.160
<v Speaker 2>also really recommend you go to chat dot Where's Youreed

0:39:36.239 --> 0:39:38.480
<v Speaker 2>dot at to visit the discord, and go to our

0:39:38.560 --> 0:39:41.880
<v Speaker 2>slash Better Offline to check out our reddit. Thank you

0:39:41.920 --> 0:39:42.880
<v Speaker 2>so much for listening.

0:39:43.719 --> 0:39:46.400
<v Speaker 1>Better Offline is a production of cool Zone Media. For

0:39:46.520 --> 0:39:49.719
<v Speaker 1>more from cool Zone Media, visit our website cool Zonemedia

0:39:49.760 --> 0:39:52.600
<v Speaker 1>dot com, or check us out on the iHeartRadio app,

0:39:52.640 --> 0:40:15.680
<v Speaker 1>Apple Podcasts, or wherever you get your podcasts.