WEBVTT - Single Best Idea with Tom Keene: Mandeep Singh & Anurag Rana

0:00:02.520 --> 0:00:14.360
<v Speaker 1>Bloomberg Audio Studios, podcasts, radio news, single best idea.

0:00:14.400 --> 0:00:17.960
<v Speaker 2>And to give you a reality into how we do this,

0:00:18.160 --> 0:00:22.120
<v Speaker 2>How Emily House, Gonzol and Rachel Worspan and I invented

0:00:22.120 --> 0:00:26.480
<v Speaker 2>this over the years with wonderful, wonderful people helping out,

0:00:26.560 --> 0:00:30.600
<v Speaker 2>Ken Pruitt, David Gerrol, Michael McKee, John Pharaoh and all.

0:00:31.800 --> 0:00:36.239
<v Speaker 2>The way we invent it is to go find voices.

0:00:36.920 --> 0:00:41.200
<v Speaker 2>And last night Sunday evening in New York, a concerted

0:00:41.280 --> 0:00:44.960
<v Speaker 2>effort was just put in and it's just text messages

0:00:45.040 --> 0:00:49.040
<v Speaker 2>back and forth people watching the football game and all that.

0:00:49.640 --> 0:00:51.800
<v Speaker 2>But it's just named. There's not a lot of discussion,

0:00:51.840 --> 0:00:54.320
<v Speaker 2>there's not a lot of meetings and all that. It's

0:00:54.440 --> 0:00:58.960
<v Speaker 2>just go get Joe Wisenthal. We did. Joe was brilliant

0:00:59.320 --> 0:01:03.880
<v Speaker 2>at odd LA on using deep seek. We said, go

0:01:03.960 --> 0:01:10.520
<v Speaker 2>get Stacy Raskin at Bernstein. He's in Los Angeles Chemical

0:01:10.560 --> 0:01:15.880
<v Speaker 2>Engineering MIT. He's definitive on this AI stuff. He was

0:01:15.959 --> 0:01:19.319
<v Speaker 2>spellbinding from Bernstein getting up in his case at I

0:01:19.319 --> 0:01:24.000
<v Speaker 2>think the five thirty six am hour in LA. But

0:01:24.120 --> 0:01:27.720
<v Speaker 2>really the first email I sent yesterday and it wasn't

0:01:27.760 --> 0:01:29.920
<v Speaker 2>just me, it was a whole team. We all agreed.

0:01:30.000 --> 0:01:33.959
<v Speaker 2>Man Deep Sing and Inner Ragrana at Bloomberg Intelligence have

0:01:34.080 --> 0:01:38.440
<v Speaker 2>a holistic view of this uproar in technology from this

0:01:38.560 --> 0:01:43.280
<v Speaker 2>Chinese company, Deep Seek like nobody first. Man Deep Sing

0:01:43.800 --> 0:01:46.000
<v Speaker 2>on the impact of this AI shock.

0:01:46.560 --> 0:01:50.880
<v Speaker 3>Clearly, I think the investor expectations were different, which is

0:01:50.880 --> 0:01:53.960
<v Speaker 3>why you see this sort of stock reaction. So, yes,

0:01:54.080 --> 0:01:56.600
<v Speaker 3>we have Moore's law and we have other laws in

0:01:56.680 --> 0:01:59.840
<v Speaker 3>terms of figuring out how the cost curve may look

0:01:59.880 --> 0:02:03.000
<v Speaker 3>like over time. But clearly this is a step change

0:02:03.040 --> 0:02:05.520
<v Speaker 3>in terms of you know, what you can do with

0:02:05.680 --> 0:02:08.919
<v Speaker 3>the existing chips that are out there, as well as

0:02:09.000 --> 0:02:12.480
<v Speaker 3>what you will need going forward. So I think this

0:02:12.760 --> 0:02:16.400
<v Speaker 3>is a big moment in terms of just the efficiency

0:02:16.440 --> 0:02:18.480
<v Speaker 3>that we can see with these chips and the kind

0:02:18.520 --> 0:02:22.160
<v Speaker 3>of scale that is required for the next versions of

0:02:22.520 --> 0:02:25.399
<v Speaker 3>these foundational models. But there is no doubt that this

0:02:25.639 --> 0:02:29.799
<v Speaker 3>makes generative AI more accessible to software companies as well

0:02:29.800 --> 0:02:31.160
<v Speaker 3>as you know, overall users.

0:02:31.400 --> 0:02:33.880
<v Speaker 2>Men Deep saying there, one thing came up today. I

0:02:33.880 --> 0:02:36.120
<v Speaker 2>didn't even have time on the show to talk about it.

0:02:36.840 --> 0:02:43.480
<v Speaker 2>Multiple times today people spouted about Jevins paradox. William Stanley

0:02:43.560 --> 0:02:46.240
<v Speaker 2>Jevons is one of my heroes. He in the middle

0:02:46.240 --> 0:02:49.120
<v Speaker 2>of the nineteenth century, and he has a claim for

0:02:49.240 --> 0:02:53.360
<v Speaker 2>like sun spot theory, figuring out what sunspots mean. He

0:02:53.440 --> 0:02:57.320
<v Speaker 2>did things in the coal industry, very colectic, very applied

0:02:57.360 --> 0:03:01.760
<v Speaker 2>economics in the nineteenth century. William Stanley Jevons is the

0:03:01.840 --> 0:03:06.880
<v Speaker 2>person who dragged ancient economics into the modern age. There's

0:03:06.880 --> 0:03:11.840
<v Speaker 2>another guy named Walrus as well, but William Stanley Jevons

0:03:11.960 --> 0:03:18.760
<v Speaker 2>is the one who brought the calculus into static economics.

0:03:18.880 --> 0:03:24.080
<v Speaker 2>In a generalization, Jevins took the static economic analysis of

0:03:24.120 --> 0:03:28.760
<v Speaker 2>before over to a more dynamic economic analysis that was

0:03:28.800 --> 0:03:31.480
<v Speaker 2>picked up by Marshall and others at the turn of

0:03:31.480 --> 0:03:34.800
<v Speaker 2>the century. There's Jevons paradox, which I'm not going to

0:03:34.840 --> 0:03:37.080
<v Speaker 2>go into, but to make a long story short, you're

0:03:37.080 --> 0:03:39.120
<v Speaker 2>going to hear a lot about it, and it's just

0:03:39.160 --> 0:03:44.960
<v Speaker 2>about the system of technology's impact on the entire macroeconomic structure.

0:03:45.640 --> 0:03:47.880
<v Speaker 2>Andro Rana knows that he doesn't need to be lectured

0:03:47.880 --> 0:03:51.360
<v Speaker 2>by me on Jevins paradox. Aner Rana expert on the

0:03:51.440 --> 0:03:54.400
<v Speaker 2>cloud here on this uproar in tech.

0:03:55.120 --> 0:03:56.800
<v Speaker 4>So one of the things, if you would say, is

0:03:56.840 --> 0:04:01.880
<v Speaker 4>if this AI adoption could which is absolutely in infancy

0:04:01.960 --> 0:04:03.760
<v Speaker 4>right now, The only thing we have seen is in

0:04:03.800 --> 0:04:07.240
<v Speaker 4>the consumer world. If enterprises start to put this in

0:04:07.280 --> 0:04:11.080
<v Speaker 4>their core applications, then they would need a lot more

0:04:11.480 --> 0:04:14.120
<v Speaker 4>you could say, firepower to run it. Most of the

0:04:14.120 --> 0:04:16.680
<v Speaker 4>things that we have discussed is people would not want

0:04:16.720 --> 0:04:19.760
<v Speaker 4>to do this on their own premise. They're bigger companies well,

0:04:19.960 --> 0:04:22.240
<v Speaker 4>but most of them will go to a handful of

0:04:22.279 --> 0:04:26.039
<v Speaker 4>cloud providers and that would be Microsoft, Aws, Google and

0:04:26.080 --> 0:04:29.760
<v Speaker 4>then Oracle. So for all these companies to increase their

0:04:29.800 --> 0:04:33.200
<v Speaker 4>capacity now they are only seeing the orders and based

0:04:33.200 --> 0:04:36.400
<v Speaker 4>on that they are expanding their data center footwork.

0:04:36.800 --> 0:04:39.400
<v Speaker 2>Enter Agrana we had time with both enter Agrana and

0:04:40.200 --> 0:04:42.600
<v Speaker 2>man Deep saying to try to get forward to the

0:04:42.640 --> 0:04:46.599
<v Speaker 2>earning season upon us. That'll be our focus, I think

0:04:47.240 --> 0:04:50.560
<v Speaker 2>in the later this week, maybe tune into Bells of

0:04:50.640 --> 0:04:53.839
<v Speaker 2>Power with Joe Matthew and Kaylee Lyons because we hardly

0:04:53.880 --> 0:04:59.320
<v Speaker 2>talked about politics today. In international relations, thank you Damian Sasaur,

0:05:00.000 --> 0:05:04.120
<v Speaker 2>some currency chat, but all in all, just an extraordinary

0:05:04.200 --> 0:05:07.360
<v Speaker 2>day into the earning season, which we'll cover this week.

0:05:07.839 --> 0:05:10.880
<v Speaker 2>On your commute across the nation, Apple CarPlay, Android Auto

0:05:11.600 --> 0:05:14.520
<v Speaker 2>on Sirius and XM, and of course good morning on

0:05:14.600 --> 0:05:18.279
<v Speaker 2>a warmer Boston, down to Washington and New York City.

0:05:18.320 --> 0:05:21.480
<v Speaker 2>Bloomberg eleventh th toe to Oho. We're out on YouTube.

0:05:21.680 --> 0:05:26.560
<v Speaker 2>Subscribe to Bloomberg podcasts. This is a single best idea.