WEBVTT - Monologue: AI Can't Afford To Slow Down

0:00:02.800 --> 0:00:07.320
<v Speaker 1>Z media greetms and salutations. I m ed zatrom and

0:00:07.400 --> 0:00:10.320
<v Speaker 1>this is better off limee and you're better off line monologue.

0:00:17.960 --> 0:00:19.919
<v Speaker 2>It's been a funny week for me, went on.

0:00:20.000 --> 0:00:22.959
<v Speaker 1>Bloomberg did an interview cleanly covering the highlights of my

0:00:23.040 --> 0:00:26.120
<v Speaker 1>AI bubble thesis. Didn't say fuck shit or bugger even once.

0:00:26.400 --> 0:00:28.600
<v Speaker 1>A bunch of people loved it. And now everybody is,

0:00:28.640 --> 0:00:31.360
<v Speaker 1>for some reason mysteriously talking about the lack of return

0:00:31.400 --> 0:00:32.560
<v Speaker 1>of investment on AI.

0:00:33.960 --> 0:00:36.159
<v Speaker 2>That's because everything I said has come true.

0:00:36.520 --> 0:00:38.760
<v Speaker 1>Back in September twenty twenty four, I talked about the

0:00:38.840 --> 0:00:42.360
<v Speaker 1>subprime AI crisis and said that when companies adopting LLM

0:00:42.440 --> 0:00:45.360
<v Speaker 1>based services were forced to pay the actual real cost

0:00:45.400 --> 0:00:48.080
<v Speaker 1>of tokens, they'd start to freak the fuck out.

0:00:48.560 --> 0:00:50.519
<v Speaker 2>Now, clam you all, clemanth clammy.

0:00:50.560 --> 0:00:53.519
<v Speaker 1>Sam Ultman himself has added an unfortunate comment in the

0:00:53.560 --> 0:00:57.560
<v Speaker 1>mix purpose There's insider. Ormond said at a recent event

0:00:57.600 --> 0:01:00.200
<v Speaker 1>that ROI had now become a meme, quoting so but

0:01:00.240 --> 0:01:02.960
<v Speaker 1>he's saying, my company spent my entire twenty twenty six

0:01:03.000 --> 0:01:05.480
<v Speaker 1>budget in Q one, Can you make this more efficient?

0:01:05.720 --> 0:01:08.360
<v Speaker 1>Adding that the company was an open AI in this case,

0:01:08.400 --> 0:01:10.759
<v Speaker 1>I mean was exploring other ways to deliver more value

0:01:10.800 --> 0:01:13.679
<v Speaker 1>for less spend. I love how casual he says this too.

0:01:13.680 --> 0:01:16.600
<v Speaker 1>It's like, ah, yeah, you know, just people are blowing

0:01:16.640 --> 0:01:18.720
<v Speaker 1>their entire yearly budgets in a few months.

0:01:18.720 --> 0:01:20.560
<v Speaker 2>But it's fine.

0:01:20.920 --> 0:01:23.320
<v Speaker 1>He also said that at the beginning of twenty twenty six,

0:01:23.400 --> 0:01:25.720
<v Speaker 1>the issue was spend never came up, and people were

0:01:25.800 --> 0:01:28.400
<v Speaker 1>totally happy with the amount they were spending. But now

0:01:28.480 --> 0:01:32.720
<v Speaker 1>AI costs were and I quote a huge issue. Well

0:01:32.720 --> 0:01:35.560
<v Speaker 1>that's probably because open Ai and Anthropic moved their enterprise

0:01:35.600 --> 0:01:39.360
<v Speaker 1>customers from heavily subsidace monthly subscriptions to token based billing

0:01:39.360 --> 0:01:40.800
<v Speaker 1>in Q one twenty twenty six.

0:01:41.520 --> 0:01:44.319
<v Speaker 2>Hmmm, I wonder if it's that. It could be.

0:01:44.440 --> 0:01:47.680
<v Speaker 1>I guess it could be there. But put differently, nobody

0:01:47.680 --> 0:01:50.600
<v Speaker 1>actually paid what AI actually really cost, and now they're

0:01:50.640 --> 0:01:52.840
<v Speaker 1>having to and they're screaming like they stepped on a

0:01:52.840 --> 0:01:55.560
<v Speaker 1>mouse trap. Like Tom and Jerry, They're freaking the fuck out.

0:01:55.920 --> 0:01:59.160
<v Speaker 1>Journalis are freaking out too. Everyone's writing pieces that discussed

0:01:59.240 --> 0:02:01.960
<v Speaker 1>the debate around the return on investment on AI as

0:02:01.960 --> 0:02:04.680
<v Speaker 1>if it's a normal, insane thing that we're four years

0:02:04.720 --> 0:02:08.680
<v Speaker 1>in and still debating whether there's ROI debating whether this

0:02:08.760 --> 0:02:11.800
<v Speaker 1>is all actually worth it? Did anybody fucking check No?

0:02:12.520 --> 0:02:15.160
<v Speaker 1>And it must be clear that these companies cannot, under

0:02:15.200 --> 0:02:19.360
<v Speaker 1>any circumstances slow down. Both open Ai and Anthropic have

0:02:19.400 --> 0:02:21.480
<v Speaker 1>sold investors on a vision that both of them will

0:02:21.480 --> 0:02:24.880
<v Speaker 1>make around two hundred billion dollars in annual revenue.

0:02:24.600 --> 0:02:26.320
<v Speaker 2>Each by the end of the decade.

0:02:27.280 --> 0:02:30.560
<v Speaker 1>Those data centers are being built under the belief that

0:02:30.639 --> 0:02:34.320
<v Speaker 1>they and probably multiple other Anthropic and open Ai sized

0:02:34.360 --> 0:02:38.480
<v Speaker 1>companies will ab exist and be have the money and

0:02:38.560 --> 0:02:41.480
<v Speaker 1>see be able to spend that money on compute. And

0:02:41.560 --> 0:02:43.799
<v Speaker 1>both of them have made a combined one point one

0:02:43.960 --> 0:02:50.120
<v Speaker 1>trillion dollars in compute commitments across Google, Microsoft, Amazon, Oracle, Core, Weave, Cerebrus,

0:02:50.160 --> 0:02:53.640
<v Speaker 1>and even fucking SpaceX. Both companies lose billions and billions

0:02:53.680 --> 0:02:55.680
<v Speaker 1>of dollars. I'd probably sell like tens of billions of

0:02:55.720 --> 0:02:58.320
<v Speaker 1>dollars if I really add the guess. Their only hope

0:02:58.360 --> 0:03:00.720
<v Speaker 1>is to keep growing to such an astoni level that

0:03:00.800 --> 0:03:04.040
<v Speaker 1>midwit investors in the recently lead poison keep feeding them

0:03:04.080 --> 0:03:07.360
<v Speaker 1>cash under the auspices of some sort of theoretical profitability.

0:03:07.840 --> 0:03:10.040
<v Speaker 1>The only way to keep that cong going is to

0:03:10.120 --> 0:03:12.679
<v Speaker 1>keep on growing, and it's become very obvious that both

0:03:12.720 --> 0:03:15.120
<v Speaker 1>companies are bonking their heads against the ceiling of their

0:03:15.120 --> 0:03:18.600
<v Speaker 1>customer spend to make matters worse. Earlier in the week,

0:03:18.720 --> 0:03:22.000
<v Speaker 1>Sam Altman talked to CNBC's David Faber on the subject.

0:03:22.200 --> 0:03:24.520
<v Speaker 1>When asked about these spending problems, he said that he

0:03:24.600 --> 0:03:27.480
<v Speaker 1>assumed that the industry will figure that out pretty quickly.

0:03:29.040 --> 0:03:34.120
<v Speaker 1>Samuel Credenza, methusilla, formaldehyde. Altman, you just raised one hundred

0:03:34.160 --> 0:03:38.040
<v Speaker 1>and twenty billion fucking dollars. You are the AI industry.

0:03:38.360 --> 0:03:40.200
<v Speaker 1>You are the one that's meant to work this out.

0:03:40.440 --> 0:03:43.920
<v Speaker 1>That's your job. You can't be four years into this no,

0:03:44.080 --> 0:03:47.360
<v Speaker 1>it loads refused cash dump and still shrugging your shoulders

0:03:47.400 --> 0:03:49.760
<v Speaker 1>like the Kevin James mean and saying, I'm just a

0:03:49.760 --> 0:03:52.000
<v Speaker 1>little guy. I can't do everything. I'm a birthday boy.

0:03:52.040 --> 0:03:54.080
<v Speaker 1>I'm a little birthday boy. Give me one hundred billion

0:03:54.120 --> 0:03:57.240
<v Speaker 1>dollars because it's my birthday. You cannot do that, Sam Altman,

0:03:57.800 --> 0:04:01.680
<v Speaker 1>you can't do that. That's that's crazy, man. And it's

0:04:01.800 --> 0:04:04.480
<v Speaker 1>very clear that open AI is the quote drill cooked

0:04:04.520 --> 0:04:07.720
<v Speaker 1>on the pit Memphis style. The Q one twenty twenty

0:04:07.720 --> 0:04:10.680
<v Speaker 1>six non gap operating margin was negative one hundred and

0:04:10.720 --> 0:04:13.640
<v Speaker 1>twenty two percent. They lose billions and billions and billions

0:04:13.680 --> 0:04:16.480
<v Speaker 1>of dollars a year, and chat GPT's growth is stalled

0:04:16.520 --> 0:04:19.520
<v Speaker 1>at around nine hundred and five million weekly active users.

0:04:20.800 --> 0:04:23.160
<v Speaker 1>We are approaching the end of this farce, and these

0:04:23.200 --> 0:04:26.800
<v Speaker 1>companies cannot afford to slow down a little, even a little,

0:04:27.040 --> 0:04:30.040
<v Speaker 1>not a touch and as Goldman's sacks. As Jim Cavello

0:04:30.120 --> 0:04:33.080
<v Speaker 1>said in a recent podcast, and I love this quote

0:04:33.120 --> 0:04:35.320
<v Speaker 1>because it's Jim is very smart and he's been ahead

0:04:35.320 --> 0:04:35.479
<v Speaker 1>of this.

0:04:35.600 --> 0:04:36.480
<v Speaker 2>But it's so simple.

0:04:37.200 --> 0:04:39.640
<v Speaker 1>At some point, you've got to make money as you

0:04:39.720 --> 0:04:42.120
<v Speaker 1>make investments in a business, so you can generate returns.

0:04:42.120 --> 0:04:45.880
<v Speaker 1>As people are just quoting that, like, damn, brother, shit,

0:04:45.960 --> 0:04:47.520
<v Speaker 1>why did not think of that?

0:04:48.120 --> 0:04:49.280
<v Speaker 2>Hey, you invest money?

0:04:49.279 --> 0:04:51.760
<v Speaker 1>And I thought you just invested money in businesses because

0:04:51.800 --> 0:04:54.440
<v Speaker 1>it felt good, because it made it made you you

0:04:54.520 --> 0:04:56.800
<v Speaker 1>feel good inside, it made you feel warm and fuzzy.

0:04:56.839 --> 0:05:00.640
<v Speaker 1>I didn't know, what do you mean a return on investment?

0:05:00.720 --> 0:05:03.240
<v Speaker 1>What do you mean she's talking about put money in

0:05:03.279 --> 0:05:04.880
<v Speaker 1>this for a reason you create?

0:05:05.120 --> 0:05:06.599
<v Speaker 2>Jim Cabello, you so crazy.

0:05:07.800 --> 0:05:10.279
<v Speaker 1>But four years in, companies are spending more than ever

0:05:10.320 --> 0:05:12.719
<v Speaker 1>on a product without a clear return on investment.

0:05:12.800 --> 0:05:14.120
<v Speaker 2>It's obvious at this point.

0:05:14.279 --> 0:05:18.479
<v Speaker 1>However, many use cases the mildly mold poisoned people in

0:05:18.560 --> 0:05:21.280
<v Speaker 1>AI dancing around about. No one can seem to actually

0:05:21.360 --> 0:05:25.440
<v Speaker 1>like say what it means or why it matters or indeed,

0:05:25.960 --> 0:05:28.480
<v Speaker 1>even if all of these people claim they're writing all

0:05:28.520 --> 0:05:31.520
<v Speaker 1>their code with AI, what that actually changes. Doesn't make

0:05:31.520 --> 0:05:35.040
<v Speaker 1>the software better, doesn't I'm so close to Tucker Carlson voice.

0:05:35.040 --> 0:05:38.599
<v Speaker 1>Doesn't make the software better, doesn't seem to make anything better.

0:05:39.040 --> 0:05:42.920
<v Speaker 1>It just seems to make our apps worse. But folks,

0:05:42.960 --> 0:05:44.880
<v Speaker 1>we're reaching the end of this era, and I am

0:05:44.920 --> 0:05:48.200
<v Speaker 1>currently working on a massive story, bigger than both of

0:05:48.240 --> 0:05:51.240
<v Speaker 1>my huge exclusives from last year, and one you will

0:05:51.240 --> 0:05:54.080
<v Speaker 1>not want to miss. I am really excited about this.

0:05:54.120 --> 0:05:56.800
<v Speaker 1>Technically there's too but I know one's going to come in.

0:05:57.000 --> 0:06:01.320
<v Speaker 1>One's pretty big, one's ginormous. But before then, I'm taking

0:06:01.320 --> 0:06:03.760
<v Speaker 1>a week off. I'm getting my wisdom teeth out. I

0:06:03.760 --> 0:06:06.600
<v Speaker 1>do need a break, so I'm going to rest up.

0:06:06.680 --> 0:06:09.040
<v Speaker 1>You should rest up too, and keep on laughing at

0:06:09.080 --> 0:06:12.440
<v Speaker 1>these oaths burning billions of dollars on software that doesn't

0:06:12.440 --> 0:06:13.640
<v Speaker 1>have a return on investment.

0:06:14.400 --> 0:06:15.119
<v Speaker 2>Zechra nowt