WEBVTT - Deepseek Offers Lower-Cost AI Model 0:00:02.520 --> 0:00:07.000 Bloomberg Audio Studios, Podcasts, radio news. 0:00:08.640 --> 0:00:12.680 This is Bloomberg business Week Insight from the reporters and 0:00:12.800 --> 0:00:16.800 editors that bring you America's most trusted business magazine, plus 0:00:17.000 --> 0:00:20.799 global business, finance and tech news. The Bloomberg Business Week 0:00:20.840 --> 0:00:25.760 Podcast with Carol Masser and Tim Stenoveek on Bloomberg Radio. 0:00:27.440 --> 0:00:29.960 Our first guest, he's part of the family, Man Deep Sing. 0:00:30.160 --> 0:00:31.520 He doesn't know, but we're not going to let him 0:00:31.520 --> 0:00:33.720 go for sixty minutes because there's so much to talk about. 0:00:33.760 --> 0:00:36.400 When it comes to Deep Seek, we want to get 0:00:36.400 --> 0:00:39.680 into it. As you know, it's the Chinese artificial intelligence 0:00:39.760 --> 0:00:43.720 startup Deep Seek, their latest model AI model, sparking a 0:00:43.840 --> 0:00:46.800 one trillion dollar route in US and European technology stocks. 0:00:46.880 --> 0:00:50.879 We have been seeing this AI shot play around the 0:00:50.920 --> 0:00:54.760 globe on this Monday. A lot of questions though about 0:00:54.800 --> 0:00:57.000 maybe valuations here in the United States, but what it 0:00:57.040 --> 0:01:00.040 really means in the AI race. So let's get to it. 0:01:00.160 --> 0:01:02.920 Man Deep is Bloomberg Intelligence Senior Tech industry unless Man 0:01:02.920 --> 0:01:07.040 Deep seeing he's here in studio. You've been up early. 0:01:07.200 --> 0:01:09.199 We've been listening to you, both Tim and I throughout 0:01:09.240 --> 0:01:13.720 the day. Why is this all happening? Because for your world. 0:01:13.800 --> 0:01:16.479 You guys, this has been on your radar, But why 0:01:16.520 --> 0:01:17.800 is it all of a sudden on our radar and 0:01:17.840 --> 0:01:19.480 why should we care or shouldn't we care? 0:01:19.880 --> 0:01:23.840 Well, every company has been talking about how capex needs 0:01:23.880 --> 0:01:27.399 to go up because they believe in scaling laws. I 0:01:27.400 --> 0:01:31.280 think what this development has shown is the hardware that 0:01:31.319 --> 0:01:35.960 these companies have procured, they weren't using it efficiently and 0:01:36.040 --> 0:01:39.640 literally it's both on the training side as well as 0:01:39.680 --> 0:01:42.360 the inferencing of the models when it comes to using 0:01:42.360 --> 0:01:45.440 it in chatbots and so on. And that's where I 0:01:45.480 --> 0:01:48.880 think you have to ask yourself as an investor, are 0:01:48.880 --> 0:01:51.880 you ready to pay up for these companies when they are, 0:01:51.920 --> 0:01:54.400 you know, increasing their capex when we know they're not 0:01:54.560 --> 0:01:57.320 running it efficiently. So I think a lot of questions 0:01:57.360 --> 0:02:00.440 around just the efficiency of the hardware. I mean, the 0:02:00.440 --> 0:02:03.880 scaling will continue just because there is more demand to 0:02:03.920 --> 0:02:07.040 incorporate AI, but the fact that you know this model 0:02:07.120 --> 0:02:10.639 does it so much more efficiently, I think raised a 0:02:10.680 --> 0:02:11.320 lot of questions. 0:02:11.400 --> 0:02:15.760 Does it show that deep Sinck is actually ahead of 0:02:16.240 --> 0:02:20.119 open AI, of Anthropic, of Lama at Facebook? 0:02:20.800 --> 0:02:24.239 No, In fact, they are using the output of open 0:02:24.280 --> 0:02:28.560 AI and Anthropic and Lama to train their own models. 0:02:28.880 --> 0:02:32.480 So think about this, Deep Sea. Didn't you know, train 0:02:32.639 --> 0:02:36.000 this model from scratch the way open ai did. They 0:02:36.200 --> 0:02:41.400 used output from open AI's models to train its own model. 0:02:41.639 --> 0:02:44.120 So in effect, they are paying open ai, you know, 0:02:44.200 --> 0:02:46.880 for all the API calls. But then they didn't have 0:02:46.960 --> 0:02:49.400 to do things from scratch. And that's where you know, 0:02:49.480 --> 0:02:52.600 if you are a meta Llama who has invested billions 0:02:52.600 --> 0:02:56.000 of dollars in creating LAMA model, what was their incentive 0:02:56.080 --> 0:02:59.400 to open source it to begin with? Because open sourcing 0:02:59.480 --> 0:03:02.520 it made it easy for someone like deep Seek to 0:03:02.600 --> 0:03:06.320 build their own model. And that's the thing about open sources, 0:03:06.400 --> 0:03:09.440 you lose your IP. Someone else can come up with 0:03:09.520 --> 0:03:13.240 a better approach to do things more efficiently, and suddenly 0:03:13.280 --> 0:03:14.440 everyone runs to them. 0:03:14.639 --> 0:03:16.680 Man, Deep is this like the Apple Store. You don't 0:03:16.720 --> 0:03:18.240 have to be first, you just have to be better. 0:03:18.520 --> 0:03:20.880 Like I'm thinking about like creating a smartphone. 0:03:21.280 --> 0:03:24.160 I won't compare them to Apple yet, I mean, they'll 0:03:24.200 --> 0:03:26.799 have their problem when it comes to distributing this model. 0:03:26.840 --> 0:03:29.400 There's no way I can see, you know, Deep Seek 0:03:29.919 --> 0:03:32.960 getting to the adoption of open Ai, especially here in 0:03:33.000 --> 0:03:35.040 the US. But at the same time, you have to 0:03:35.040 --> 0:03:37.600 give them credit for you know, just being novel in 0:03:37.680 --> 0:03:40.960 terms of you know, the limited GPUs they had and 0:03:41.000 --> 0:03:42.360 how they ended up using them. 0:03:42.400 --> 0:03:47.280 Do you trust the information, the research, and do you 0:03:47.360 --> 0:03:49.960 trust that what they say they did they did do? 0:03:50.520 --> 0:03:53.720 Probably not. If they are saying six million dollars for 0:03:53.880 --> 0:03:57.440 training their entire model, I don't trust it because when 0:03:57.480 --> 0:04:00.680 I look at OpenAI costs for training their you know, 0:04:00.960 --> 0:04:03.600 GPT four model, it was close to three hundred million. 0:04:03.960 --> 0:04:06.320 So we are talking about a fifty x gap. 0:04:06.520 --> 0:04:09.560 Now, even if Opening Eye wasn't very efficient, I don't 0:04:09.560 --> 0:04:12.480 buy the fifty x gap. So clearly, you know there 0:04:12.560 --> 0:04:15.560 is something that they are not counting when they you know, 0:04:15.640 --> 0:04:18.800 release the paper. But at the same time, they've give 0:04:18.880 --> 0:04:21.839 given us enough evidence to show that they are doing 0:04:21.920 --> 0:04:25.320 things much more efficiently on the hardware side because of 0:04:25.360 --> 0:04:27.280 the constraints they had Men Deep. 0:04:27.320 --> 0:04:29.080 I know you don't cover in video, and we're gonna 0:04:29.080 --> 0:04:30.960 be speaking to Ian King in just a few minutes, 0:04:31.040 --> 0:04:33.480 who's been covering in video in the ship space for years. 0:04:33.920 --> 0:04:36.279 But we can't ignore that in videos down close to 0:04:36.320 --> 0:04:38.400 eighteen percent right now. I mean it's shed almost a 0:04:38.520 --> 0:04:41.800 fifth of its value in just today. Yes, the run 0:04:41.880 --> 0:04:44.120 up over the last couple of years has been immense, 0:04:44.160 --> 0:04:45.960 but a day like today is just mind boggling. I 0:04:46.000 --> 0:04:47.400 think for a lot of people should in video in 0:04:47.440 --> 0:04:48.200 your view be worried. 0:04:49.040 --> 0:04:51.279 I mean, look, that's where a lot of the questions 0:04:51.320 --> 0:04:54.960 come is the cost for training the next model. So 0:04:55.160 --> 0:04:58.640 if deepseek is saying that the cost was six million, 0:04:59.480 --> 0:05:01.960 I think it should be really worried. But a lot 0:05:02.040 --> 0:05:04.839 of people are apprehensive of that six million number because 0:05:04.880 --> 0:05:07.800 they think it'll be higher. And it won't be three 0:05:07.839 --> 0:05:10.800 hundred million the way open Aies was, but it will 0:05:10.800 --> 0:05:15.000 be higher. And that's where I think open Ai or 0:05:15.520 --> 0:05:18.320 Entropic will have a very tough time selling to their 0:05:18.320 --> 0:05:20.760 investors that they need to build a one million GPU 0:05:20.800 --> 0:05:23.320 cluster to train the next version of their model. 0:05:23.440 --> 0:05:26.600 Mandi, does this show that even though the US has been. 0:05:28.120 --> 0:05:31.000 Pretty rough in terms of restricting advanced technology, going over 0:05:31.040 --> 0:05:33.919 to China, does this show that Chinese R and D 0:05:34.240 --> 0:05:36.800 and Chinese companies can be pretty scrappy and pretty smart 0:05:36.800 --> 0:05:39.560 and maybe figuring a go around and coming up with 0:05:39.600 --> 0:05:42.760 their own way of competing against Nvidia. 0:05:42.400 --> 0:05:42.839 And the like. 0:05:43.400 --> 0:05:45.920 Yeah, and look, I think this could be another way 0:05:45.960 --> 0:05:48.919 of them saying Look, we are open sourcing whatever we 0:05:48.960 --> 0:05:50.919 have learned so far. So they want to build a 0:05:50.960 --> 0:05:54.560 community because the export restrictions aren't going away, so they 0:05:54.560 --> 0:05:56.599 are not going to get the you know, black ball 0:05:56.680 --> 0:05:59.880 chips on paper and all the other allocation. And the 0:06:00.080 --> 0:06:02.560 thing about model training is it's continuous. It's not a 0:06:02.560 --> 0:06:05.360 one and done thing. So you have to continuously improve 0:06:05.400 --> 0:06:07.560 your model. The other labs will go ahead if they 0:06:07.560 --> 0:06:10.760 are not able to, you know, keep doing the training 0:06:10.839 --> 0:06:13.440 and all the other stuff that other labs are doing. 0:06:13.480 --> 0:06:16.520 So from that perspective, I think the whole aspect of 0:06:16.600 --> 0:06:20.320 community building is front and center here. Yeah, but I 0:06:20.400 --> 0:06:23.920 still think the cost parity is a very big concept here. 0:06:23.960 --> 0:06:25.560 Hey, before we let you go, Mandy, you do cover 0:06:25.640 --> 0:06:28.520 meta platforms. The company out on Friday, Mark Zuckerberg saying 0:06:28.560 --> 0:06:30.720 he was going to be spending more on Capex than 0:06:30.800 --> 0:06:35.040 previously announced. Then the stock moved higher on Friday. Stock 0:06:35.120 --> 0:06:38.680 was lower earlier in the session, but higher now. I mean, 0:06:39.160 --> 0:06:40.680 I look at a move like that today and that 0:06:40.760 --> 0:06:42.960 runs counterintuitive to kind of everything we're talking about. 0:06:43.279 --> 0:06:45.960 Look, there is a lot of confusion, but if there's 0:06:46.040 --> 0:06:49.080 one thing that happened today is this moment is good 0:06:49.120 --> 0:06:53.080 for open source, and meta Lama is an open source model, 0:06:53.200 --> 0:06:56.800 so if anything into open source does get a nod 0:06:56.920 --> 0:06:58.960 when it comes to you know what deep seek is 0:06:59.320 --> 0:06:59.960 touting here. 0:07:00.080 --> 0:07:01.880 But you do also think about if it's cheaper for 0:07:01.960 --> 0:07:05.840 companies to build out, then does everybody have to rely 0:07:05.920 --> 0:07:07.120 on some of the bigger players. 0:07:07.160 --> 0:07:07.520 Maybe not. 0:07:07.839 --> 0:07:11.080 Next time Meta Lama will collect a fee from deep 0:07:11.120 --> 0:07:13.360 seek for you know, using their model to train their 0:07:13.400 --> 0:07:15.440 own models, so that could be a positive. 0:07:15.560 --> 0:07:17.640 I love this world that keeps us on our toes men, 0:07:17.680 --> 0:07:20.480 deep saying keeps us on the straight and narrow. He's 0:07:20.520 --> 0:07:23.720 Bloomberg Intelligence senior tech industry analyst here and has been 0:07:23.760 --> 0:07:25.600 all over this story. We want to focus a little 0:07:25.600 --> 0:07:28.080 bit more on semiconductors because we have seen that group 0:07:28.680 --> 0:07:32.560 really taking a beating. You've got almost all thirty names 0:07:32.560 --> 0:07:36.000 twenty eight out of thirty down in the Philadelphia Semiconductor 0:07:36.000 --> 0:07:38.440 Index down about ten percent, the index as a whole. 0:07:38.480 --> 0:07:42.040 Nvidios down seventeen percent. So ouch, what does this mean 0:07:42.080 --> 0:07:45.480 for Nvidia specifically? We get to now Bloomberg News US 0:07:45.480 --> 0:07:47.960 semiconductor and networking reporter Ian King. He's out there in 0:07:47.960 --> 0:07:51.040 San Francisco. I don't know Ian, how we've all been 0:07:51.120 --> 0:07:54.120 like talking about this non stop today you're the expert. 0:07:54.440 --> 0:07:57.800 How are you thinking about Nvidia and possible impact the 0:07:57.840 --> 0:07:59.000 reality of it all? 0:07:59.160 --> 0:07:59.800 How do you see it? 0:08:00.960 --> 0:08:04.040 I mean a lot depends upon how much reality there 0:08:04.080 --> 0:08:08.240 actually is behind this claim behind you know, it was 0:08:08.280 --> 0:08:14.080 a very well set out, well referenced research paper at 0:08:14.080 --> 0:08:17.040 this point, and the model has been open sourced and 0:08:17.040 --> 0:08:19.920 we'll see what people can do with it. The implication is, 0:08:19.960 --> 0:08:22.880 of course that things got cheaper, things got more simple, 0:08:23.360 --> 0:08:25.440 but we'll have to see that manifest itself in the 0:08:25.480 --> 0:08:28.840 real world. And then obviously you can absolutely bet that 0:08:29.200 --> 0:08:32.000 in video and a lot of other computer scientists that 0:08:32.080 --> 0:08:34.600 some of the most powerful companies in the world are 0:08:34.600 --> 0:08:37.800 looking at this right now working on this. For some 0:08:37.840 --> 0:08:40.760 of them this would be great news. For others it 0:08:40.760 --> 0:08:43.640 would be a disaster. So they'll be definitely testing out 0:08:43.679 --> 0:08:46.640 the veracity of this, and I think the actual where 0:08:46.640 --> 0:08:50.320 the truth lies will emerge with a bit more time. 0:08:50.920 --> 0:08:53.280 In What do you make of selloffs in other names 0:08:53.320 --> 0:08:58.160 today Broadcom, for example, Marvel, I mean everybody is getting 0:08:58.240 --> 0:09:00.240 hit in the sector. As Carol mentioned, twenty eight of 0:09:00.320 --> 0:09:01.920 thirty names lower. 0:09:01.679 --> 0:09:05.800 I mean, yeah, yeah, the logic is absolutely clear, right, 0:09:05.840 --> 0:09:09.120 I mean, if you are no longer a believer in 0:09:09.160 --> 0:09:14.280 this exponential growth in spending, in this exponential need for 0:09:14.920 --> 0:09:17.760 more and more computing, then you're going to sell Micron, 0:09:17.800 --> 0:09:20.720 You're going to sell broad Coom, you're going to sell Marvel, 0:09:20.800 --> 0:09:22.880 and of course you're going to sell in video or 0:09:23.040 --> 0:09:27.319 you know. The logic based upon an assumption, and that's 0:09:27.360 --> 0:09:30.079 what it is at this point is flawless. I mean, 0:09:30.120 --> 0:09:35.079 for example, Broadcom, they are the key ingredient in terms 0:09:35.080 --> 0:09:38.160 of these companies like Google that are now designing their 0:09:38.160 --> 0:09:41.440 own chips for this type of purpose. So if there's 0:09:41.480 --> 0:09:43.079 going to be less need for that kind of high 0:09:43.160 --> 0:09:45.480 end chip, then it makes sense to sell them. 0:09:45.800 --> 0:09:47.600 I mean, none of this happens in a vacuum, right, 0:09:47.640 --> 0:09:50.120 so I can only imagine the meetings that are happening 0:09:50.160 --> 0:09:53.080 at Nvidia today and all of the chip companies. So 0:09:53.400 --> 0:09:57.240 would you expect them to, you know, you know, tear 0:09:57.280 --> 0:10:01.520 apart the research from this and try to understand exactly 0:10:01.600 --> 0:10:03.760 what they did and then come back with their own response. 0:10:05.320 --> 0:10:07.560 You've got I mean, they've got to do that, right, 0:10:07.600 --> 0:10:09.600 I mean, the the you know, if this was a 0:10:09.640 --> 0:10:12.280 two or three percent move and a few market jitters. 0:10:12.320 --> 0:10:15.440 Then you know you probably as a management team you 0:10:15.480 --> 0:10:17.720 look into it fiercely, but you just keep your mouth shut. 0:10:18.160 --> 0:10:20.560 Or we're just about to go into earning season, right, 0:10:20.640 --> 0:10:24.120 So whether they like it or not, for example, Intel 0:10:24.360 --> 0:10:26.960 on Thursday are trying to get into this market, They're 0:10:26.960 --> 0:10:27.960 going to get asked about this. 0:10:28.040 --> 0:10:28.760 What have you got? 0:10:28.840 --> 0:10:31.280 How can you deal with this? Is this good for you? 0:10:31.360 --> 0:10:33.960 These are all questions that I think, you know, everybody's 0:10:33.960 --> 0:10:36.200 companies are going to be answering over the next two 0:10:36.280 --> 0:10:38.760 or three weeks, and they better have some good answers. 0:10:39.320 --> 0:10:41.640 And do folks you taught you speak to who are 0:10:41.840 --> 0:10:44.760 covering the industry, have they been waiting for a day 0:10:44.800 --> 0:10:46.280 like today? 0:10:46.320 --> 0:10:47.559 Like was this bound to happen? 0:10:49.240 --> 0:10:52.440 I mean, you know, investors do two things right. You 0:10:52.480 --> 0:10:56.800 have different types of investors, but basic pattern recognition is 0:10:56.840 --> 0:10:59.600 supposed to be the skill that everybody can bring to this, 0:10:59.679 --> 0:11:02.520 whether your technical or whether you're just a you know, 0:11:02.559 --> 0:11:05.600 a general investor. So I think a lot of the 0:11:05.640 --> 0:11:09.960 general investors will be like, let's let's not be on 0:11:10.040 --> 0:11:12.680 the wrong side of what happened in two thousand with 0:11:12.760 --> 0:11:15.719 all of the Internet sell off. Let's get out right now, right, 0:11:15.800 --> 0:11:19.840 maybe we're at the peak. More sort of technically savvy 0:11:19.880 --> 0:11:22.800 investors will probably take their time look at this and 0:11:22.840 --> 0:11:25.240 try to find out just how much truths there is 0:11:25.280 --> 0:11:28.240 there and just how many concrete implications there are. 0:11:28.559 --> 0:11:30.080 Well in terms of implications too. 0:11:30.400 --> 0:11:32.199 You know, a lot of what's been kicked around throughout 0:11:32.240 --> 0:11:35.040 the day is you know, I feel like TikTok, you know, 0:11:35.160 --> 0:11:38.319 version two point zero in other words, another Chinese company. 0:11:38.480 --> 0:11:40.319 You know, you want to know how much the government 0:11:40.360 --> 0:11:42.760 is involved and how much maybe you know, do they 0:11:42.800 --> 0:11:46.079 support the r indeed, siphius from the United States, the 0:11:46.120 --> 0:11:48.120 Committee on Foreign Investment in the United States, you know, 0:11:48.120 --> 0:11:49.640 are they going to weigh in on this whether or 0:11:49.640 --> 0:11:53.480 not you know, American companies or you know, can tap 0:11:53.559 --> 0:11:55.839 into it. I mean, it's a whole other round of this, 0:11:56.400 --> 0:11:56.920 it seems. 0:11:57.640 --> 0:11:57.880 Yeah. 0:11:58.000 --> 0:12:00.920 No, I clearly there's a lot of scrap rambling going on. 0:12:01.200 --> 0:12:03.960 And you know, the path that's happened so far in 0:12:04.000 --> 0:12:07.079 that particular area has been to restrict access to these 0:12:07.120 --> 0:12:11.040 high end chips. So guess what, the country that is 0:12:11.720 --> 0:12:14.240 restricted from these high end chips suddenly comes up with 0:12:14.559 --> 0:12:17.679 the magic bullet solution, which says, maybe we don't need 0:12:17.720 --> 0:12:19.280 as many of these high end chips, or maybe we 0:12:19.280 --> 0:12:22.600 don't need them at all. So there is an element 0:12:22.640 --> 0:12:25.920 of like, hmm, okay, is this really Is this the 0:12:26.000 --> 0:12:29.000 answer that they want us to be concerned about, or 0:12:29.120 --> 0:12:32.040 is this really the answer? Or have we given them 0:12:32.040 --> 0:12:34.440 so much incentive to find a way around it that 0:12:34.440 --> 0:12:36.960 they've poured all their energy into it being really smart 0:12:37.000 --> 0:12:39.720 and got around it. I mean, layers upon layers of 0:12:39.800 --> 0:12:43.960 things going on here, but in general, there are no 0:12:44.040 --> 0:12:46.800 magic bullets in at least the chip industry. It tends 0:12:46.880 --> 0:12:51.160 to be a pretty incremental industry where we see changes 0:12:51.160 --> 0:12:55.160 over time. Software can be more volatile and you know, 0:12:55.280 --> 0:12:59.360 is prone to more kind of instant changes in thinking. 0:12:59.400 --> 0:13:01.559 But even then, and you know, the change to a 0:13:01.600 --> 0:13:04.679 whole industry, the direction that's going in, it's a little 0:13:04.679 --> 0:13:07.080 bit hard to imagine that one morning we all wake 0:13:07.160 --> 0:13:10.920 up and go, oh dear right. But at the same time, 0:13:11.000 --> 0:13:15.479 if you're an investor, you can't afford to ignore the possibility. 0:13:15.520 --> 0:13:17.960 And if you're the CEO of in Video or the 0:13:17.960 --> 0:13:20.720 board of InVideo or the senior management of video, you 0:13:20.760 --> 0:13:23.160 cannot afford to ignore the possibility. 0:13:23.400 --> 0:13:26.520 Ian on a day like today, you know, we're certainly 0:13:26.559 --> 0:13:29.320 focused on the big names that are moving lower by 0:13:29.360 --> 0:13:34.720 double digits in video, certainly down Broadcom is lower today, 0:13:34.960 --> 0:13:37.120 Oracles down by fourteen percent. 0:13:37.679 --> 0:13:38.160 But from a. 0:13:38.080 --> 0:13:41.840 Hardware perspective and from Deep six perspective, what do you 0:13:41.920 --> 0:13:46.199 know about the hardware that Deep seek users to train 0:13:46.280 --> 0:13:46.559 its own? 0:13:46.679 --> 0:13:50.120 That's no, and that's that's a very good question. That's 0:13:50.240 --> 0:13:53.280 probably one that the US government are looking into, which is, 0:13:53.760 --> 0:13:55.760 you know, some of the analysis that we've seen has 0:13:55.800 --> 0:14:00.200 been like, look, yes, this model that they're citing that's 0:14:00.200 --> 0:14:02.640 actually a derivative of another model. How did they train 0:14:02.679 --> 0:14:07.000 that original model? So should we classify this derivative model 0:14:07.040 --> 0:14:11.440 that they're offering up as part of this larger equation 0:14:11.520 --> 0:14:16.120 which they must have developed with Western technology? And the answer, 0:14:16.160 --> 0:14:18.600 simple answer is we do not know. And there's going 0:14:18.679 --> 0:14:22.640 to be a lot of pressure on those that need 0:14:22.680 --> 0:14:25.440 to know and should know to find that out. And 0:14:25.480 --> 0:14:27.600 I think we're going to see a lot of reporting emerge. 0:14:27.600 --> 0:14:30.120 We're going to see the government looking into this to 0:14:30.200 --> 0:14:32.720 find out what did they have access to, should they 0:14:32.720 --> 0:14:35.360 have had access to it, if what they had access 0:14:35.440 --> 0:14:39.200 to was in fact more efficient computing than what say, 0:14:39.200 --> 0:14:42.520 Microsoft is putting in place. Then wow, here we go. Right, 0:14:42.560 --> 0:14:45.920 this is a breakthrough so many layers to this onion, 0:14:46.160 --> 0:14:48.320 and that's not a very useful conclusion, but I think 0:14:48.320 --> 0:14:51.120 it's the best that we can offer at this particular point. 0:14:51.160 --> 0:14:52.880 A lot of questions still out there. Ian, Thank you 0:14:52.960 --> 0:14:54.520 so much, so appreciated. Of course. 0:14:54.520 --> 0:14:57.479 Our Bloomberg News US Semiconductor and Networking reporter. 0:15:00.240 --> 0:15:03.920 Is the Bloomberg Business Week Podcast. Listen live each weekday 0:15:03.960 --> 0:15:07.200 starting at two pm Eastern on Applecarplay and Android Auto 0:15:07.320 --> 0:15:10.320 with the Bloomberg Business App. You can also listen live 0:15:10.360 --> 0:15:13.960 on Amazon Alexa from our flagship New York station, Just 0:15:14.000 --> 0:15:16.520 say Alexa Play Bloomberg eleven thirty. 0:15:17.360 --> 0:15:19.920 It feels like the last week or so, topped off 0:15:19.920 --> 0:15:22.080 by today's news on Deep Seek, has put the AI 0:15:22.160 --> 0:15:24.680 story once again tim front and center in a super 0:15:24.680 --> 0:15:25.040 big way. 0:15:25.120 --> 0:15:25.320 Yeah. 0:15:25.320 --> 0:15:27.600 Remember just last week, President Trump any group of tech 0:15:27.640 --> 0:15:31.120 companies SoftBank, Oracle and open Ai initially planning to invest 0:15:31.160 --> 0:15:34.680 one hundred billion dollars to build US based infrastructure, including 0:15:34.760 --> 0:15:38.160 data centers for open Ai, the project we know called Stargate. 0:15:38.560 --> 0:15:41.760 On that though, one very high profile tech and administration 0:15:41.880 --> 0:15:44.200 voice who's not all in on it It's something that 0:15:44.240 --> 0:15:47.840 Bloomberg BusinessWeek columnist Max Chaffkin gets into in his latest 0:15:47.880 --> 0:15:50.520 story and in the latest Bloomberg BusinessWeek daily newsletter. You 0:15:50.560 --> 0:15:52.840 can read Max's business Week story on the Bloomberg and 0:15:52.840 --> 0:15:56.000 at Bloomberg dot com slash BusinessWeek. Max the co host 0:15:56.280 --> 0:15:58.680 of the Lining podcast, also the author of the concheriont 0:15:58.680 --> 0:16:01.560 Peter Tiel in Silicon Valley Pursuit of Power. Max, We're 0:16:01.560 --> 0:16:05.040 gonna talk about elon Stargate in just a second, but 0:16:05.760 --> 0:16:08.280 deep Seek is where I want to start because it's 0:16:08.320 --> 0:16:10.800 what's hitting the tech names today. Well, I don't want 0:16:10.800 --> 0:16:13.400 to say all tech names. It's what's hitting you know, Nvidia, 0:16:13.760 --> 0:16:15.680 Broadcoms and of the other chip names in a big 0:16:15.720 --> 0:16:16.120 way today. 0:16:16.120 --> 0:16:18.240 How are you looking at absolutely well, you know, as 0:16:18.280 --> 0:16:21.840 you're hinting at with this discussion of stargate, until basically 0:16:22.000 --> 0:16:25.280 this weekend, you know, AI infrastructure was seen as. 0:16:25.200 --> 0:16:25.960 The end all be all. 0:16:25.960 --> 0:16:28.280 You know, they were talking about not not one hundred billion, 0:16:28.360 --> 0:16:31.600 five hundred billion, right and at times Sam Altman has 0:16:31.680 --> 0:16:34.720 said he's gonna he's gonna need seven trillion dollars in 0:16:35.560 --> 0:16:38.240 you know, funding for AI infrastructure, and there's been this 0:16:38.480 --> 0:16:43.400 sense especially among Google, Open Eye, a bunch of these 0:16:43.480 --> 0:16:46.040 big players, that this is gonna be the thing that 0:16:46.120 --> 0:16:49.240 like only hyperscalers are going to be able to compete 0:16:49.280 --> 0:16:51.600 in that whoever has the most infrastructure, even whoever has 0:16:51.640 --> 0:16:53.800 the most power. Right, We've seen a big run up 0:16:53.840 --> 0:16:56.680 in nuclear stocks and things like that, so so, and 0:16:56.720 --> 0:16:58.760 this sort of threw all that out the window because 0:16:59.080 --> 0:17:01.960 if we're to believe what deep seek is saying, this 0:17:02.160 --> 0:17:05.280 was done with a much much lighter weight, uh you know, 0:17:05.320 --> 0:17:08.560 amount of usage of chips and it and it actually 0:17:08.600 --> 0:17:12.160 is running a lot more efficiently efficiently. So that suggests, 0:17:12.240 --> 0:17:15.000 right that that these this crazy run up we've seen 0:17:15.320 --> 0:17:19.080 in demand for GPUs and so on might be overheated. 0:17:19.080 --> 0:17:21.440 And we we had other reasons to think that there 0:17:21.520 --> 0:17:24.040 was perhaps some sort of bubble or a little bit 0:17:24.040 --> 0:17:26.680 of frothiness in this market. And and I do think 0:17:26.720 --> 0:17:28.640 if it hadn't been this, it might have been something else. 0:17:28.720 --> 0:17:28.840 Right. 0:17:28.880 --> 0:17:32.000 I think investors were and are primed. 0:17:31.600 --> 0:17:35.440 To see, you know, to see to see potential flaws 0:17:35.480 --> 0:17:37.480 and with the with the business model that so many 0:17:37.480 --> 0:17:37.920 have bought into. 0:17:38.080 --> 0:17:40.439 So not necessarily an end of like the AI spend, 0:17:40.440 --> 0:17:42.040 but just kind of a twist and turn in the story. 0:17:42.080 --> 0:17:44.800 And there's still more to be known, right Max, in 0:17:44.880 --> 0:17:48.160 terms of the reliability of the information we're hearing from 0:17:48.240 --> 0:17:49.520 them and the reality of it. 0:17:49.720 --> 0:17:52.320 Yeah, there's so much we don't know here, and again 0:17:53.000 --> 0:17:57.719 it's definitely worth I think approaching Deep Seek's claims skeptically 0:17:57.960 --> 0:18:00.879 and also understanding that there if they were able to 0:18:00.920 --> 0:18:04.119 develop a much more inexpensive model that could help parts 0:18:04.119 --> 0:18:07.360 of the AI market. It doesn't necessarily mean that all 0:18:07.440 --> 0:18:10.720 of this goes away or something, but it may. 0:18:10.400 --> 0:18:12.960 Make it more accessible, right, that's less expensive. 0:18:12.600 --> 0:18:15.439 Yeah, it definitely, but it could cause investors to like 0:18:15.520 --> 0:18:17.960 dramatically revalue like some of their assumptions. 0:18:18.000 --> 0:18:19.879 How does it change the narrative around what China is 0:18:19.920 --> 0:18:21.840 able to do when it comes to AI right now 0:18:21.960 --> 0:18:24.520 and sort of the tech race that's happening there. The 0:18:24.600 --> 0:18:26.199 thing is, if they weren't supposed you know, there are 0:18:26.400 --> 0:18:28.159 these chips, A lot of these chips are not supposed 0:18:28.160 --> 0:18:30.480 to go to certain parts of the world. 0:18:30.680 --> 0:18:34.119 Right, So that so there's something potentially troubling there. If 0:18:34.160 --> 0:18:37.920 you're if you're looking at the attempts to restrict certain 0:18:38.000 --> 0:18:41.640 US technology to China, it seems like either those restrictions 0:18:41.640 --> 0:18:44.840 aren't working that well, maybe some GPUs found their way 0:18:44.880 --> 0:18:50.600 to China, or or that the Chinese software industry has 0:18:50.600 --> 0:18:54.240 found ways around it, which would also be potentially problematic, 0:18:54.280 --> 0:18:57.840 I suppose from a sort of US national security perspective, 0:18:57.960 --> 0:19:00.400 at least from the point of view of the China 0:19:00.440 --> 0:19:02.800 Hawks who have been sort of warning about this, and 0:19:02.840 --> 0:19:05.560 that includes many of the Silicon Valley companies like remember, 0:19:05.600 --> 0:19:09.520 one of the big reasons to justify this, like massive 0:19:09.640 --> 0:19:12.680 spend on AI And one of the reasons that many 0:19:12.720 --> 0:19:15.919 of these folks, including Sam Altman, sees this as a 0:19:15.920 --> 0:19:18.120 thing that government should be involved in, is because they 0:19:18.119 --> 0:19:21.600 have been arguing that without this, without spending huge sums 0:19:21.640 --> 0:19:25.760 of money on say on power or investing in infrastructure 0:19:25.760 --> 0:19:28.159 in terms of data centers, that we would fall behind 0:19:28.800 --> 0:19:32.399 behind the Chinese tech startups. And that may you know, 0:19:32.440 --> 0:19:34.800 if you sort of extrapolate from this very small piece 0:19:34.800 --> 0:19:37.600 of news, you might say, Okay, well it's happening anyway. 0:19:37.760 --> 0:19:40.160 So that sort of throws everything into a bit of disarray. 0:19:40.240 --> 0:19:43.440 All right, So let's get to stargate. Love the name, 0:19:43.520 --> 0:19:47.240 love the name Elon Musk. I find it interesting. I 0:19:47.240 --> 0:19:48.440 think we all found it interesting. 0:19:48.480 --> 0:19:50.879 That here's this big announcement right that comes out of 0:19:51.200 --> 0:19:54.959 the Trump administration last week, and Elon not so in 0:19:55.040 --> 0:19:55.480 on it? 0:19:55.560 --> 0:19:57.640 Is he no, not in on it at all. 0:19:57.680 --> 0:20:00.439 You know, it wouldn't be surprising if if you had 0:20:00.440 --> 0:20:03.119 this big announcement with Donald Trump and the CEOs of 0:20:03.119 --> 0:20:07.120 some major companies talking about huge numbers, right they were saying, 0:20:07.119 --> 0:20:09.920 we're going to invest one hundred billion dollars immediately five 0:20:09.960 --> 0:20:12.080 hundred billion with the next four years, would not have 0:20:12.080 --> 0:20:15.560 been surprising if some people said this sounds like a 0:20:15.560 --> 0:20:17.840 political stunt, I'll believe it when I see it. It's 0:20:17.960 --> 0:20:20.960 what made this thing that happened last week so surprising 0:20:21.040 --> 0:20:23.880 is that the person saying that was a senior advisor 0:20:23.960 --> 0:20:26.200 to Donald Trump, and it was Elon Musk. 0:20:26.160 --> 0:20:29.520 Tweeting over and over and over again last week. 0:20:29.560 --> 0:20:31.840 I mean, the last time I checked, twenty times, he's 0:20:31.840 --> 0:20:34.760 tweeted a few more times. You know, the usual mix 0:20:34.800 --> 0:20:37.920 of i'd say, off color jokes and put downs. You know, 0:20:37.960 --> 0:20:40.800 he's called Sam Altman, who was of course on stage 0:20:40.840 --> 0:20:44.400 with Donald Trump, a swindler. He suggested the announcement was fake, 0:20:44.760 --> 0:20:47.960 and Trump has sort of responded by saying, hey, you 0:20:47.960 --> 0:20:50.480 know Elon Musk has a dispute with one of the 0:20:50.520 --> 0:20:54.080 people in this in this deal, which is true. 0:20:54.160 --> 0:20:56.000 Sam Altman and Elon Musk have. 0:20:56.080 --> 0:20:58.200 Been feuding more or less in public for the last 0:20:58.280 --> 0:21:01.000 couple of years, So we're sort of seeing I'd say, 0:21:01.520 --> 0:21:04.120 you know, maybe one of the downsides to having somebody 0:21:04.119 --> 0:21:08.200 who is as deep in the business world also involved 0:21:08.240 --> 0:21:09.879 as a advisor to the president. 0:21:10.000 --> 0:21:11.920 I mean, is Elon Musk at this point saying, Hey, 0:21:11.920 --> 0:21:14.160 I gave you some giant bags of money to help 0:21:14.200 --> 0:21:14.919 you win the election. 0:21:15.960 --> 0:21:16.720 What were you thinking? 0:21:16.920 --> 0:21:19.320 And now you're doing this? Does this risk driving a 0:21:19.359 --> 0:21:20.440 wedge between the two of them? 0:21:20.840 --> 0:21:23.119 I think for both Musk and Trump. 0:21:23.200 --> 0:21:25.280 I mean, there has been over the last couple of 0:21:25.280 --> 0:21:27.960 months people who have been hot on high alert for 0:21:28.119 --> 0:21:30.840 signs that this relationship might be might be breaking up. 0:21:31.200 --> 0:21:33.359 And I would say that it doesn't seem like this 0:21:33.440 --> 0:21:35.879 has done it either. You know, you know, Trump has 0:21:35.880 --> 0:21:38.800 cracked jokes before, We've seen these tweets. There have been 0:21:38.840 --> 0:21:40.879 there was a dispute over the H one B visa 0:21:40.960 --> 0:21:43.399 thing both Musk. 0:21:43.200 --> 0:21:46.560 And maybe that led to the divorce between Ramaswamy and Musk. 0:21:46.640 --> 0:21:48.520 Yeah, yeah, at least according to some reports. You know, 0:21:48.560 --> 0:21:51.080 it's possible there are other reasons as well. But but 0:21:51.160 --> 0:21:54.040 what I'll say is it seems like Musk and Trump 0:21:54.359 --> 0:21:58.439 are still working closely together. I'd say Trump's comments about 0:21:58.440 --> 0:22:02.240 this to me suggested that he was untroubled by Musks, 0:22:02.560 --> 0:22:05.359 you know, apparently kind of trying to undermine or or 0:22:05.440 --> 0:22:07.880 undermine some of his partners in this deal. And I'd 0:22:07.920 --> 0:22:10.159 say part of that is that, you know, there's not 0:22:10.440 --> 0:22:14.240 necessarily government money going into this. In the Stargate thing, 0:22:14.280 --> 0:22:17.720 it really was basically a media event for Donald Trump 0:22:17.760 --> 0:22:20.200 to announce a big business deal so that he can 0:22:20.240 --> 0:22:23.879 claim credit for, you know, helping to spur the economy. 0:22:23.920 --> 0:22:26.639 It's kind of like a vibes thing. And I think 0:22:26.720 --> 0:22:29.840 because of that, I'm not sure it really hurts Trump 0:22:29.880 --> 0:22:33.000 all that much to see Elon Musk dissing Sam Altman 0:22:33.119 --> 0:22:33.600 on Twitter. 0:22:33.680 --> 0:22:36.320 Well, you write in your story, Max, you said Musk 0:22:36.359 --> 0:22:38.720 has always treated his conflicts of interest like features rather 0:22:38.760 --> 0:22:41.600 than bugs, seeming to run his normally independent companies with 0:22:41.720 --> 0:22:44.560 their nominally independent shareholder basis as if they were a 0:22:44.560 --> 0:22:47.720 single conglomerate. You talk about employees being lent out, sharing 0:22:47.760 --> 0:22:49.320 of resources and stuff. 0:22:49.200 --> 0:22:51.720 So conflict he's kind of all in on this too. 0:22:52.080 --> 0:22:55.520 Well, yeah, absolutely, this is a big part of First 0:22:55.560 --> 0:22:59.280 of all, competing in this area is a big part. 0:22:59.080 --> 0:23:01.439 Of Elon Musk's big plan for. 0:23:01.320 --> 0:23:04.600 His companies, not just Tesla, which of course is developing 0:23:05.440 --> 0:23:08.480 you know, advanced robotics and is trying to build self 0:23:08.560 --> 0:23:12.200 driving cars, but also x Ai, his his chatbot competitor, 0:23:12.400 --> 0:23:15.240 so he is very much playing in this market. He 0:23:15.359 --> 0:23:20.000 is also competing to buy these Nvidia these high end 0:23:20.000 --> 0:23:23.760 in Nvidia GPUs. He's building giant data centers, so he 0:23:23.920 --> 0:23:26.200 very much has a dog in this fight, and that's 0:23:26.240 --> 0:23:30.480 why he's doing this. And I think Trump's attitude seems 0:23:30.480 --> 0:23:32.879 to be that's all out in the open, as long 0:23:32.880 --> 0:23:35.960 as people understand that what's behind this is a business dispute. 0:23:36.000 --> 0:23:39.119 Again, he said, well, one of the things is. 0:23:38.720 --> 0:23:40.480 Is he hates one of the guys in the deal, 0:23:40.720 --> 0:23:46.280 an obvious reference to Elon Sewings and but then Trump said, 0:23:46.280 --> 0:23:47.439 you know, I hate people too. 0:23:47.359 --> 0:23:49.359 So we all have a problem. 0:23:49.480 --> 0:23:49.760 Child. 0:23:50.160 --> 0:23:52.600 I wonder you know what what Donald Trump has been 0:23:52.640 --> 0:23:54.720 able to do in recent months has been pretty incredible 0:23:54.720 --> 0:23:56.200 when it comes to the tech industry. I mean a 0:23:56.200 --> 0:23:58.040 week ago, Carol and I were sitting here looking at 0:23:58.040 --> 0:24:01.360 the stage and during the inauguration, and you saw titans 0:24:01.359 --> 0:24:04.440 of industry who were not at previous inaugurations there all 0:24:04.480 --> 0:24:08.080 collectively getting behind the president. I mean, he's he is 0:24:08.119 --> 0:24:11.160 this larger than life figure who perhaps can bring people 0:24:11.240 --> 0:24:15.600 together who don't like each other, maybe because he offers 0:24:15.640 --> 0:24:17.480 them something that they want. 0:24:18.119 --> 0:24:21.360 Yeah, well, Trump probably more so, at least than than 0:24:21.760 --> 0:24:24.919 foreign President. Biden approaches the presidency in a way that 0:24:24.960 --> 0:24:27.000 I think would be fair to describe as a little 0:24:27.000 --> 0:24:27.800 bit more transactional. 0:24:27.880 --> 0:24:28.000 Right. 0:24:28.040 --> 0:24:31.800 He's willing to bend his policies to make deals. It's 0:24:31.800 --> 0:24:34.280 the whole thing that he's about, and in certain ways, 0:24:34.359 --> 0:24:37.080 right that works well for CEOs. 0:24:37.160 --> 0:24:39.200 I also think there's a lot going on there. 0:24:39.240 --> 0:24:41.720 There was a lot of built up anger in Silicon 0:24:41.800 --> 0:24:45.800 Valley over some of Biden's policies over anti trust for instance, 0:24:46.200 --> 0:24:49.119 and you have a lot of opportunism right now. There's 0:24:49.119 --> 0:24:52.719 a sense that they have leverage because by showing up 0:24:53.080 --> 0:24:56.280 in these Trump things, they they're giving Trump some additional 0:24:56.280 --> 0:24:58.600 credibility and they're trying to use that to get you know, 0:24:58.640 --> 0:25:01.840 what they can, which is, you know, favorable policies, favorable 0:25:01.880 --> 0:25:02.600 and trust policy. 0:25:02.600 --> 0:25:04.399 We're going to talk about that in the next couple 0:25:04.400 --> 0:25:05.000 of hours. 0:25:05.080 --> 0:25:06.680 Hey Max, do you want to go back to where 0:25:06.680 --> 0:25:09.280 we started and deep Seek because we're getting some breaking 0:25:09.320 --> 0:25:11.560 news and video is out saying that deep seek is 0:25:11.640 --> 0:25:15.760 quote excellent AI advancement and says that deep Seek's work 0:25:16.200 --> 0:25:20.360 quote fully export control compliant and says that inference requires 0:25:20.359 --> 0:25:23.359 significant numbers of its GPUs. This is an email statement 0:25:23.680 --> 0:25:26.959 on the deep seek AI model. They're they're coming out 0:25:27.160 --> 0:25:29.760 and saying talking about it. I'm wondering how you view that. 0:25:30.320 --> 0:25:32.920 I mean, I think it's first of all, everybody's talking 0:25:32.920 --> 0:25:35.880 about it. It seems like it seems like they that 0:25:35.960 --> 0:25:38.359 they have to and they also I think in that 0:25:38.440 --> 0:25:40.720 statement and I haven't looked at I'm just hearing it now, 0:25:41.000 --> 0:25:44.760 it seems like an attempt to address suggestions that perhaps 0:25:44.800 --> 0:25:48.280 deep seat Seek had access to you know, improper access 0:25:48.280 --> 0:25:49.560 to GPUs or something like that. 0:25:49.640 --> 0:25:52.080 It sounds like they are saying, no, that's not the case. 0:25:52.720 --> 0:25:55.760 Again, it definitely seems like this would be a business 0:25:55.880 --> 0:25:58.160 challenge for for Nvidio if you were if you were. 0:25:58.080 --> 0:26:00.800 Able to do the kinds of hatbot. 0:26:00.920 --> 0:26:04.080 You know, work much much more efficiently with about the 0:26:04.200 --> 0:26:07.720 same sort of level of quality that's gonna make these 0:26:07.800 --> 0:26:10.920 GPUs perhaps not as valuable. 0:26:11.080 --> 0:26:13.879 But again a lot. There's a lot here that we 0:26:13.920 --> 0:26:14.320 don't know. 0:26:14.400 --> 0:26:16.920 And you also have seen people on the other side 0:26:16.920 --> 0:26:19.119 of this saying, actually, this is going to be great 0:26:19.200 --> 0:26:22.400 because it's gonna make these chatbots so much cheaper. We're 0:26:22.400 --> 0:26:24.520 gonna see tons of them. Essentially, we're gonna make it 0:26:24.600 --> 0:26:27.240 up on volume. It's hard to know how to evaluate that, 0:26:27.560 --> 0:26:29.720 and we don't know that much about how this was 0:26:29.720 --> 0:26:33.040 put together or really just how good the quality looks. 0:26:33.240 --> 0:26:35.399 People have been playing with it only for a short amount. 0:26:35.520 --> 0:26:39.800 Wait you want details. Excilly you great stuff, Max, Thank 0:26:39.840 --> 0:26:42.680 you so much. Bloomberg Business Week commnist Max Chafkin is 0:26:42.720 --> 0:26:44.080 also co host of the Elon Inc. 0:26:44.160 --> 0:26:44.640 Podcast. 0:26:44.680 --> 0:26:45.640 You could check that out. 0:26:45.960 --> 0:26:47.880 It's on the Bloomberg and also at Bloomberg dot com 0:26:47.880 --> 0:26:51.280 slash Elon Inc. Max is also author of the contrarian 0:26:51.320 --> 0:26:54.240 Peter Teel and Silicon Values Pursuit of Power. 0:26:56.040 --> 0:26:59.719 This is the Bloomberg Business Week podcast. Listen live each 0:26:59.760 --> 0:27:02.840 week day starting at two pm Eastern on applecar Play 0:27:02.920 --> 0:27:05.760 and Android Auto with the Bloomberg Business app. You can 0:27:05.800 --> 0:27:08.959 also listen live on Amazon Alexa from our flagship New 0:27:09.040 --> 0:27:12.639 York station. Just say Alexa played Bloomberg eleven thirty. 0:27:13.920 --> 0:27:16.760 Hey, I guess say reading in this morning, Morgan Stanley 0:27:16.760 --> 0:27:20.040 continues to recommend that investors hold long five year treasuries 0:27:20.040 --> 0:27:24.080 ahead of two potentially devish market catalysts. And those catalysts 0:27:24.119 --> 0:27:27.359 are the January FED meeting, So not this week's meeting 0:27:27.720 --> 0:27:29.880 and then this month's job report. Meantime, you've got jp 0:27:30.000 --> 0:27:33.159 Morgan stay staying along the front end of the treasury 0:27:33.160 --> 0:27:36.160 curve well Bimo positioning for two quarter point FED cuts 0:27:36.160 --> 0:27:38.280 this year. So while we are focused so much tim 0:27:38.320 --> 0:27:40.840 on tech and AI news this morning, let's not forget 0:27:41.119 --> 0:27:43.120 we at the first FED decision of twenty twenty five, 0:27:43.119 --> 0:27:44.560 That decision out on Wednesday. 0:27:44.640 --> 0:27:46.040 Okay, with more on that, With. 0:27:46.080 --> 0:27:46.480 More on that. 0:27:46.560 --> 0:27:49.160 In the studio, we got Ian Wyatt's senior vice president 0:27:49.240 --> 0:27:52.000 director of Economics at Huntington Commercial Bank. It's a division 0:27:52.359 --> 0:27:55.480 of the publicly held roughly twenty five billion dollar market 0:27:55.520 --> 0:27:59.280 cap Huntington Bank. Shares Ian, good to see you. Look 0:27:59.320 --> 0:28:01.280 before we jump been to exactly what we're going to 0:28:01.359 --> 0:28:05.000 hear from Fed sho Jpowell, I would imagine he might 0:28:05.040 --> 0:28:07.760 get some questions around AI and the market sell off 0:28:07.760 --> 0:28:10.560 this week. Whether or not he answers those questions, that 0:28:10.680 --> 0:28:13.040 certainly remains to be seen. I would figure out a 0:28:13.040 --> 0:28:14.600 way to ask him a question around that, How do 0:28:14.640 --> 0:28:16.879 you think, how are you looking at what's happening in 0:28:16.920 --> 0:28:19.800 the market this week, especially ahead of this FED meeting? 0:28:20.400 --> 0:28:23.480 I think, well, first, I think in terms of the 0:28:23.520 --> 0:28:27.719 FED meeting, not related to AI, I think long our 0:28:27.840 --> 0:28:30.480 view is that there's still inflationary pressures out there in 0:28:30.520 --> 0:28:32.680 the economy when I taught, but I think it's really 0:28:32.720 --> 0:28:35.360 we have this two speed labor market we're seeing. We're 0:28:35.400 --> 0:28:38.880 seeing is for lower end jobs. You know, where's the 0:28:38.960 --> 0:28:41.800 job growth been. It's been restaurants, it's been hotels, It's 0:28:41.800 --> 0:28:43.760 been a lot of service areas. And when I talk 0:28:43.840 --> 0:28:47.120 to companies in those areas, whether you're manufacturing, transportation, still 0:28:47.120 --> 0:28:49.479 a tight labor market. They're having a hard time finding workers. 0:28:49.680 --> 0:28:52.160 If you're talking about more of a college job, professional 0:28:52.160 --> 0:28:55.000 and business services, where are we not getting job growth? 0:28:55.040 --> 0:28:56.640 It obviously has been down. 0:28:56.760 --> 0:29:00.200 We've seen you know, them focusing on cash flow more 0:29:00.280 --> 0:29:02.840 than job growth. So we kind of have this this 0:29:02.920 --> 0:29:04.720 two speed labor market where you still have a tight 0:29:04.800 --> 0:29:08.120 lower end, and the problem for inflation is a lot 0:29:08.160 --> 0:29:10.560 of those industries are very labor intensive. 0:29:10.600 --> 0:29:12.240 You know, childcare for example. 0:29:12.520 --> 0:29:15.440 It's something where labor or restaurants, labor makes up a 0:29:15.520 --> 0:29:18.880 huge percentage of your cost structure, and they're still feeling that pressure. 0:29:18.920 --> 0:29:22.240 So I think we've probably been on the low end 0:29:22.440 --> 0:29:26.640 of the FED saying in terms of cuts for twelve 0:29:26.680 --> 0:29:29.040 months now. Yeah, now I think we're more in line 0:29:29.080 --> 0:29:31.960 with the two cuts roughly this year. But I still 0:29:32.000 --> 0:29:34.400 think that pause holds for a while because I think 0:29:34.440 --> 0:29:40.400 there is that fundamental underlying inflationary pressure that is not 0:29:40.560 --> 0:29:43.600 really when you tease out inflation, you look category by category. 0:29:43.680 --> 0:29:44.800 It's not been durable goods. 0:29:45.160 --> 0:29:48.000 Durable goods have been coming down, but it's also stuff 0:29:48.040 --> 0:29:49.480 that consumers can choose not to buy. 0:29:49.600 --> 0:29:51.960 So I just want to stay with AII for a moment. 0:29:52.000 --> 0:29:55.200 That AI spend the economic you know, activity, it may 0:29:55.240 --> 0:29:57.960 impact on the upper down side. You know, how does it, 0:29:58.040 --> 0:30:00.440 if at all, factor into any of your ex economic 0:30:00.880 --> 0:30:03.320 models when it comes to growth, when it comes to inflation, 0:30:03.400 --> 0:30:04.760 and then ultimately to FED policy. 0:30:04.880 --> 0:30:06.920 You know, AI caused me to get one call really 0:30:06.960 --> 0:30:09.600 wrong last year. It was I thought construction employment was 0:30:09.640 --> 0:30:12.360 going to come down. Yeah, I thought, you know, multifamilies 0:30:12.400 --> 0:30:15.400 coming down. We saw warehouse space is over supplied. Now 0:30:15.440 --> 0:30:18.960 obviously office has its issues, and I underestimated how much 0:30:19.000 --> 0:30:20.120 it would be chipsacked. 0:30:20.560 --> 0:30:21.080 It would be the. 0:30:21.120 --> 0:30:24.280 Chip plants like Columbus for example, or we're headquartered. They 0:30:24.280 --> 0:30:26.520 have that massive intel plant, a couple of the largest 0:30:26.520 --> 0:30:31.280 cranes in the world operating there. And AI was a 0:30:31.360 --> 0:30:33.480 huge part of this. And I still think there's this 0:30:33.640 --> 0:30:36.320 huge first mover advantage. You don't want to be the 0:30:36.400 --> 0:30:38.520 last company. You want to be the last company that 0:30:38.800 --> 0:30:41.520 still gets a power connection versus the first where they 0:30:41.560 --> 0:30:44.880 say no, we don't have enough juice, we can't afford 0:30:45.000 --> 0:30:47.600 or we can't build, you know, connect you for five years. 0:30:48.440 --> 0:30:50.800 And I think that that case still remains. 0:30:51.000 --> 0:30:51.320 Yeah. 0:30:51.360 --> 0:30:53.960 I also think there's another bullish case for power that 0:30:54.000 --> 0:30:56.800 when you look at the forecasts and really baked invert. 0:30:56.520 --> 0:30:57.800 What is that about? Transportation? 0:30:58.440 --> 0:31:01.480 So so, okay, if you look at US energy consumption, 0:31:01.600 --> 0:31:04.640 roughly fifteen percent of US energy consumptions residential, that's what 0:31:04.680 --> 0:31:06.920 we always think of our power. Thirty seven percent of 0:31:07.000 --> 0:31:11.360 US energy consumption is transportation. Right now, over the last 0:31:11.440 --> 0:31:14.200 six months roughly ten percent of US car sales we're 0:31:14.240 --> 0:31:17.680 up to have been plug in vehicles now. Vehicle fleet 0:31:17.720 --> 0:31:19.640 takes a long time to turn over in the US 0:31:19.680 --> 0:31:23.080 average vehicle ages over twelve years now, it takes a while. 0:31:23.120 --> 0:31:25.000 But as that starts to kick in, if EVE even 0:31:25.120 --> 0:31:28.600 got a small fraction of that thirty seven percent shifting 0:31:29.120 --> 0:31:33.920 from fossil fuels to electric power, it doesn't really matter. 0:31:33.680 --> 0:31:35.760 The source, whether it's gas or solar. 0:31:35.520 --> 0:31:39.840 Or whatever, you have a very different outlook for electric 0:31:39.880 --> 0:31:42.520 demand at the US. And I think that also means 0:31:42.560 --> 0:31:44.600 utilities are going to have to build a ton of infrastructure. 0:31:44.680 --> 0:31:47.440 Does that get derailed by the Trump administration pushing back 0:31:47.480 --> 0:31:50.880 against EB subsidies and moving trying to get people to 0:31:50.880 --> 0:31:51.680 move away from evs. 0:31:51.760 --> 0:31:53.480 I mean, I think it slows it to some degree. 0:31:53.520 --> 0:31:56.040 I think that's definitely. You know, you're seeing a shift 0:31:56.040 --> 0:31:59.440 administration policy. If you look at Germany, for example, when 0:31:59.440 --> 0:32:03.720 they cut their TV subsidies, that dramatically slowed EV's sales. 0:32:04.440 --> 0:32:07.400 But it's still something where there's a lot of the 0:32:07.400 --> 0:32:10.880 global market is shifting in that direction. Autos are shifting 0:32:10.920 --> 0:32:11.320 that way. 0:32:11.640 --> 0:32:13.760 China's all in, China's all in. 0:32:14.360 --> 0:32:17.000 I mean, you talked to a friend of mine runs 0:32:17.000 --> 0:32:20.080 FedEx distribution, owns a couple of Fredecks distribution places, and 0:32:20.120 --> 0:32:21.960 he talks about, you know, the gas prices on his 0:32:22.040 --> 0:32:24.960 delivery trucks, and he's really looking, how do I lower 0:32:25.000 --> 0:32:30.200 my fuel expenses? And that's really a perfect application for 0:32:30.280 --> 0:32:32.560 EV short range so stop start driving? 0:32:32.640 --> 0:32:34.120 So does he do that with Rivians? 0:32:34.640 --> 0:32:38.120 Like yeah, I mean he's looking at Yeah, he's looking 0:32:38.160 --> 0:32:40.920 I mean, Rivian's just starting to sell beyond Amazon now, 0:32:40.920 --> 0:32:44.160 they're just moving beyond Amazon or filling those initial orders, 0:32:44.440 --> 0:32:48.800 whether it's that or Forward and Stillantis both have products 0:32:48.840 --> 0:32:49.840 in that market as well. 0:32:50.080 --> 0:32:51.920 Sorry here, I know we're supposed to be talking about. 0:32:51.720 --> 0:32:54.120 The I kind of love this because it's a fuller 0:32:54.160 --> 0:32:57.680 picture of kind of the markets and really our economy. 0:32:57.960 --> 0:33:00.280 So having said that, I mean, I don't know this 0:33:00.400 --> 0:33:03.440 January meeting, the decision on Wednesday Ian, is it kind 0:33:03.480 --> 0:33:05.760 of a snoozer because we don't really expect anything? Or 0:33:05.800 --> 0:33:09.360 is there a fine noodle, a noodle, a fine needle 0:33:09.440 --> 0:33:11.400 that j Powell has to thread here? 0:33:11.560 --> 0:33:14.000 In some ways I think thematically we've seen you know, 0:33:14.080 --> 0:33:16.160 a year ago when he did the Palll pivot, you 0:33:16.240 --> 0:33:19.280 felt like there was pushback right after he spoke at 0:33:19.280 --> 0:33:21.320 that presser, and it definitely seemed like he was on 0:33:21.320 --> 0:33:23.680 the duvish end of the spectrum. If you look back 0:33:23.680 --> 0:33:26.000 at the September vote on the dot plot, you can 0:33:26.040 --> 0:33:30.320 see almost half of Fed Governor's only wanted three cuts 0:33:30.400 --> 0:33:33.280 this fall. So that would mean essentially there should have 0:33:33.280 --> 0:33:36.640 been one twenty five in each meeting, and that meant 0:33:36.800 --> 0:33:40.000 they were opposed to fifty in September. Now, obviously they 0:33:40.000 --> 0:33:42.040 all get in board, they all push in the same direction, 0:33:42.120 --> 0:33:44.239 but if you look at the minutes, you look at 0:33:44.240 --> 0:33:47.120 the notes, I think he's still again. I think this 0:33:47.200 --> 0:33:49.560 is a repeat of about a year ago. It's he's 0:33:49.600 --> 0:33:51.760 on the dubvish end of the spectrum. He has to 0:33:51.800 --> 0:33:53.760 get the board to come along with him. And I 0:33:53.800 --> 0:33:56.920 don't think I think. I think the inflationary pressure is 0:33:56.920 --> 0:34:00.480 still there. I still think there's some underlying and when 0:34:00.520 --> 0:34:03.280 I talk to companies in the economy, depending on the 0:34:03.320 --> 0:34:04.960 part of the labor market they're in. 0:34:05.640 --> 0:34:06.600 It's still there. 0:34:06.760 --> 0:34:10.279 It's it's still significant, and I think that helps some 0:34:10.320 --> 0:34:12.600 of the more hawkish voices on the FED, where the 0:34:12.880 --> 0:34:15.200 very narrowly balanced FED I think on this. 0:34:15.360 --> 0:34:19.680 So it's really labor costs that you're most concerned about. 0:34:19.680 --> 0:34:21.840 It's not the price of eggs that everybody keeps talking about. 0:34:22.000 --> 0:34:23.280 It's really labor costs. 0:34:23.320 --> 0:34:25.279 Well, yeah, I mean, I'm the grocery shopper in the family, 0:34:25.320 --> 0:34:27.400 so I definitely noticed the price of eggs. And I 0:34:27.480 --> 0:34:29.719 know used to work in Georgia, so I know, you know, 0:34:29.800 --> 0:34:33.080 top egg economy, top egg producing state in the country 0:34:33.239 --> 0:34:36.279 is basically offline right now because of bird flu. So 0:34:36.320 --> 0:34:38.719 there's some big things going on there. But now I'm 0:34:38.800 --> 0:34:41.160 much more concerned about rent growth. I mean, that's a 0:34:41.200 --> 0:34:43.360 good sign we're seeing on the inflationary front. If you 0:34:43.360 --> 0:34:47.680 look at the market based rent indexes, you can see 0:34:47.760 --> 0:34:49.799 you're at one one and a half percent. The problem is, 0:34:50.000 --> 0:34:52.040 you know, two thirds of Americans on their home A 0:34:52.040 --> 0:34:54.640 lot of that single family we're still seeing and the 0:34:54.880 --> 0:34:59.160 implicit rent deflator from that comes from rents on single family, 0:34:59.200 --> 0:35:01.680 not multi family, and that's still feeling pressure. 0:35:01.680 --> 0:35:03.279 What did you want to say, I know I jumped in. 0:35:03.360 --> 0:35:06.120 Oh, I just I wanted to just say in videos 0:35:06.120 --> 0:35:10.239 down eighteen percent, everybody, that's the that's the day. Looking 0:35:10.280 --> 0:35:13.960 at Broadcoms down eighteen percent, Oracles down seventeen. 0:35:13.480 --> 0:35:15.480 Percent, but no inflation there. 0:35:16.080 --> 0:35:18.120 Then that's a backdrop of some would say. 0:35:18.000 --> 0:35:20.320 It's on sale or maybe not, depending on the outlook. 0:35:20.880 --> 0:35:23.279 This was really fun. Come back soon and why are 0:35:23.719 --> 0:35:26.239 senior vice president and director of economics loved where the 0:35:26.320 --> 0:35:29.680 economic conversation went over at Huntington Commercial Bank. 0:35:29.760 --> 0:35:31.600 Joining us here Tim in studio. 0:35:31.920 --> 0:35:35.600 You're listening to the Bloomberg Business Week podcast. Catch us 0:35:35.680 --> 0:35:38.680 live weekday afternoons from two to five y's during Listen 0:35:38.719 --> 0:35:42.280 on Applecarplay and Android Auto with the Bloomberg Business app, 0:35:42.440 --> 0:35:44.759 or watch us live on YouTube. 0:35:45.560 --> 0:35:47.799 Susy dan Is Cudy of America's over at Wi Bro. 0:35:47.880 --> 0:35:50.719 It's a global consulting firm that provides services related to 0:35:50.760 --> 0:35:53.840 IT cloud and yes AI. And that's really what we 0:35:53.880 --> 0:35:55.560 want to talk to her about because that is why 0:35:55.560 --> 0:35:58.600 we're seeing the selloff concern that a cheaper AI model 0:35:58.640 --> 0:36:01.719 from the Chinese startup Seek could make valuations of these 0:36:01.760 --> 0:36:05.200 tech that has powered the bull market pretty tough to justify. 0:36:05.239 --> 0:36:07.360 It's what we spoke to Gena Martin Adams about just 0:36:07.360 --> 0:36:11.440 a minute ago. Suzanne joins us from East Brunswick, New Jersey. Susanne, 0:36:11.440 --> 0:36:13.680 you are just back from Davos, where we know AI 0:36:13.840 --> 0:36:16.480 was front and center. Given comments that we areed from 0:36:16.600 --> 0:36:19.640 Alexander Wong, who is the founder at CEO over at 0:36:19.680 --> 0:36:23.320 Scale AI, how would you though characterize the AI conversation 0:36:23.400 --> 0:36:25.120 that was happening over at Davos. 0:36:26.200 --> 0:36:29.640 Well, thank you both for having me. Everything was AI 0:36:29.960 --> 0:36:32.239 last week. You could walk down the street and every 0:36:32.280 --> 0:36:36.200 storefront had something about AI or was talking about AI. 0:36:36.560 --> 0:36:37.960 So it was front and center. 0:36:38.560 --> 0:36:41.239 This is clearly a defining year for AI and for 0:36:41.360 --> 0:36:45.880 jen AI, and businesses are looking to figure out how 0:36:45.920 --> 0:36:50.200 do they drive value from this. Most CEOs believe that 0:36:50.400 --> 0:36:53.600 it's going to significantly change the way they do business 0:36:53.960 --> 0:36:57.719 in terms of how they create and deliver value and 0:36:57.800 --> 0:37:01.040 obtain value from the work that they do. And there 0:37:01.080 --> 0:37:04.719 was a clear momentum around, you know, defining the strategy 0:37:04.719 --> 0:37:08.080 in which they're going to invest in AI to create 0:37:08.120 --> 0:37:09.320 that value. 0:37:08.719 --> 0:37:11.640 Why do you think it's that they were saying or 0:37:11.680 --> 0:37:14.400 why is it that this is going to be a 0:37:14.440 --> 0:37:17.200 defining year, especially on a day when we kind of 0:37:17.239 --> 0:37:19.560 just got you know, a company that might say, hey, 0:37:19.600 --> 0:37:22.200 you don't need to do the big span or you know, 0:37:22.280 --> 0:37:25.120 we can do it. It's a different maybe financial model 0:37:25.280 --> 0:37:27.120 in terms of the buildout, I feel like we're still 0:37:27.200 --> 0:37:29.520 kind of finding our way. So why why is it 0:37:30.200 --> 0:37:32.840 that twenty twenty five is the defining way. I'm not 0:37:32.920 --> 0:37:33.839 quite sure how to read that. 0:37:33.880 --> 0:37:34.080 Well. 0:37:34.160 --> 0:37:36.440 I think last year everybody was talking about AI and 0:37:36.480 --> 0:37:39.560 just piloting with genai. It was, you know, it became 0:37:39.760 --> 0:37:42.560 it's democratized. Everybody has access to it, we can use it. 0:37:43.120 --> 0:37:45.800 I think the GENAI races and during a new phase. 0:37:45.840 --> 0:37:49.279 We saw it this morning and I think we're still 0:37:49.280 --> 0:37:52.720 in the beginning, you know. To me, it really calls 0:37:52.760 --> 0:37:55.719 to everyone that this emphasizes the pace we're going to 0:37:55.719 --> 0:37:59.399 see innovation and really figuring out how we can create 0:37:59.440 --> 0:38:02.480 agile and environments to deliver on AI. And at the 0:38:02.600 --> 0:38:04.840 end of the day, it's going to be about the 0:38:04.880 --> 0:38:07.279 quality of the data that these AI are trained on 0:38:07.480 --> 0:38:10.640 and deliver for our clients to really deliver value to 0:38:10.680 --> 0:38:13.799 the organizations, regardless of how much they spend. And I 0:38:13.840 --> 0:38:16.400 think you know where we see it is our clients 0:38:16.480 --> 0:38:19.760 really are talking about the data they need to fuel AI, 0:38:19.880 --> 0:38:21.919 whether it's a large language model or a small one. 0:38:22.680 --> 0:38:24.960 I'm just just for context, because we do speak to 0:38:25.000 --> 0:38:27.160 a lot of folks who run consulting firms or run 0:38:27.600 --> 0:38:29.760 parts of consulting firms. Just give us an idea of 0:38:29.800 --> 0:38:32.480 your clients, the type of clients you have, and you 0:38:32.520 --> 0:38:34.239 know who's out there working with you. 0:38:34.320 --> 0:38:35.920 Yeah, so I work, so my responsibility. 0:38:35.920 --> 0:38:38.920 I have clients across financial services, oil and gas, energy 0:38:38.920 --> 0:38:42.759 and utilities, high tech, and manufacturing and automotive. So I 0:38:42.760 --> 0:38:47.560 have clients across many different industries and many different challenges 0:38:47.600 --> 0:38:49.160 that can be solved with them without AI. 0:38:49.280 --> 0:38:51.239 So who would you say is leveraging AI the best 0:38:51.280 --> 0:38:51.680 right now? 0:38:52.440 --> 0:38:56.000 Well, you know, AI is a proven and it's been 0:38:56.040 --> 0:38:58.759 around for ten years. We've been using in AI, and 0:38:58.760 --> 0:39:01.000 many of our financial service as clients have been using 0:39:01.040 --> 0:39:04.359 AI and machine learning models for years. It's only now 0:39:04.360 --> 0:39:08.720 with GENAI we're seeing this acceleration and that's what's truly 0:39:08.760 --> 0:39:11.640 exciting is how we can leverage the genai applications. You know, 0:39:11.680 --> 0:39:14.920 we've been also leveraging it internally so that we can 0:39:15.000 --> 0:39:18.319 learn and then deliver those learnings to our clients. We're 0:39:18.360 --> 0:39:22.280 leveraging in HR, we're leveraging it illegal across the board 0:39:22.400 --> 0:39:25.160 to make sure that we understand how AI can be 0:39:25.320 --> 0:39:30.440 used and understand the foundation and the governance and responsible 0:39:30.600 --> 0:39:32.439 nature in which we need to deploy it and train 0:39:32.480 --> 0:39:34.600 our people to make sure that it's successful. 0:39:35.160 --> 0:39:37.800 So, Susan, what I'm wondering is like on a daylight today, 0:39:37.840 --> 0:39:41.200 where this story a company that nobody had really are 0:39:41.280 --> 0:39:44.080 I should say nobody because our Bloomberg intelligence teams like, yeah, 0:39:44.160 --> 0:39:47.000 our team overseas has been writing about deep Seek for 0:39:47.120 --> 0:39:48.880 you know, the last couple of years, so it's not 0:39:49.200 --> 0:39:51.640 necessarily new to them, but certainly new to I feel 0:39:51.680 --> 0:39:55.040 like our conversations when it comes to AI in general. 0:39:55.120 --> 0:39:55.760 So I am. 0:39:55.640 --> 0:39:59.040 Curious, have you guys your team, you know your phone's 0:39:59.080 --> 0:40:02.279 been ringing today's say, wait, what is this deep Seek? 0:40:02.800 --> 0:40:05.840 And you know I've been spending X on my AI 0:40:06.040 --> 0:40:09.480 infrastructure build or you know, making plans? Do I need 0:40:09.520 --> 0:40:11.920 to rethink it? Do I not need to spend as much, 0:40:12.080 --> 0:40:15.399 you know, on site versus not? You know what tell 0:40:15.480 --> 0:40:18.759 us about today and the conversations perhaps that you guys 0:40:18.760 --> 0:40:19.520 have all been having. 0:40:20.400 --> 0:40:23.640 So today it's all over the news. It's absolutely correct, 0:40:23.640 --> 0:40:26.480 everyone's talking about it. I think the reality is over 0:40:26.520 --> 0:40:29.239 time there's going to be winners and losers. You know, 0:40:29.320 --> 0:40:31.640 the role that we play is helping our clients make 0:40:31.640 --> 0:40:34.480 sure they have the foundations in place. One of the biggest, 0:40:34.480 --> 0:40:36.200 and I'll go back to it, is the data. 0:40:36.040 --> 0:40:36.640 And the people. 0:40:37.239 --> 0:40:40.080 You know, training your teams to think about AI first 0:40:40.080 --> 0:40:43.680 and how to bring AI into a business to add value, 0:40:43.760 --> 0:40:46.799 so it could make an AGENC that could do make 0:40:46.840 --> 0:40:51.319 autonomous decisions and be managed appropriately. The data foundations, so 0:40:51.440 --> 0:40:53.799 the models don't hallucinate. All that has to be done 0:40:53.920 --> 0:40:57.879 regardless of the technology that you choose. So that's where 0:40:57.880 --> 0:41:00.360 I think the real area that we can add value 0:41:00.360 --> 0:41:03.960 in helping our clients navigate both the frameworks, the governance, 0:41:04.080 --> 0:41:06.480 the training of the people, all the change management, and 0:41:06.520 --> 0:41:07.160 the data. 0:41:07.320 --> 0:41:10.480 All that still has to happen regardless. 0:41:09.840 --> 0:41:11.640 Of which model that you choose to use and the 0:41:11.680 --> 0:41:12.799 foundations you build on. 0:41:13.760 --> 0:41:16.960 So the spend is going to still happen. Nobody's going 0:41:17.000 --> 0:41:17.440 to put back. 0:41:17.880 --> 0:41:20.800 Yeah, I think it'll depend on what the use cases 0:41:20.840 --> 0:41:23.160 and what the spend will be. So I think that 0:41:23.239 --> 0:41:25.279 is yet to be determined, depending on the models that 0:41:25.320 --> 0:41:27.040 our clients use. And many of our clients are using 0:41:27.120 --> 0:41:29.319 large models, and many of our clients are using their 0:41:29.400 --> 0:41:32.920 data and developing small models. But you know, I do 0:41:33.000 --> 0:41:35.640 think this is a race that is continuing and we're 0:41:35.640 --> 0:41:37.879 going to see this evolution and many more to come. 0:41:38.440 --> 0:41:43.200 Where do you see the actual money savings happening with 0:41:43.280 --> 0:41:45.319 these companies that you work with? Is it from a 0:41:45.320 --> 0:41:48.680 customer service perspective? The whole rise of the AI agent? 0:41:50.239 --> 0:41:53.240 Where do companies where would our margins be helped? 0:41:54.040 --> 0:41:55.400 Yeah? So I think there's two things. 0:41:55.440 --> 0:41:59.560 You know, there's a whole site of productivity and efficiency. 0:42:00.160 --> 0:42:03.560 Seeing that in development, you know, making our developers more 0:42:03.560 --> 0:42:06.080 productive and more efficient, and we're seeing it in testing 0:42:06.120 --> 0:42:10.719 of technology and applications. Absolutely, we're seeing that today. There's 0:42:10.760 --> 0:42:14.920 also things like contact center where we're helping clients in 0:42:15.040 --> 0:42:18.359 servicing and using AI to make the agent smarter and 0:42:18.400 --> 0:42:21.919 delivering higher quality engagement. So there's lots of area where 0:42:21.960 --> 0:42:25.440 we're seeing productivity and efficiency and the question is does 0:42:25.440 --> 0:42:28.000 that translate into our clients being able to add more 0:42:28.080 --> 0:42:31.160 higher value work to those or driving the costs down 0:42:31.160 --> 0:42:31.880 in their business. 0:42:31.880 --> 0:42:35.759 And we're seeing both. You know, we launched just as I. 0:42:36.040 --> 0:42:39.319 You know, one example is whippro now, which is a 0:42:39.719 --> 0:42:43.160 HR intelligent agent and we've put all of our HR 0:42:43.239 --> 0:42:47.080 policies and procedures into this agent and it's serviced over 0:42:47.120 --> 0:42:50.239 eight million queries to date at greater than ninety percent accuracy. 0:42:50.560 --> 0:42:54.120 We've been able to reduce the number of of HR 0:42:55.480 --> 0:42:59.160 employees to service those queries and to service those and 0:42:59.239 --> 0:43:01.520 be able to reap purpose. So you can say that's 0:43:01.560 --> 0:43:04.600 a great efficiency play, but it's also a great employee experience. 0:43:05.040 --> 0:43:07.680 Our employees are able to get real time information, real 0:43:07.760 --> 0:43:11.600 time insights into HR and be able to have a 0:43:11.680 --> 0:43:15.560 better experience than maybe looking for documents in lots of 0:43:15.600 --> 0:43:16.160 flou volders. 0:43:16.200 --> 0:43:17.719 Well, that's what I wanted to ask you. What was 0:43:17.760 --> 0:43:20.719 the kind of satisfaction rate as I play around or 0:43:20.800 --> 0:43:25.879 increasingly am you know, working with chats chatbots to deal 0:43:25.880 --> 0:43:28.399 with service customer service issues. I mean, sometimes it's really 0:43:28.440 --> 0:43:30.719 great and sometimes it's not. So I'm curious from your 0:43:30.719 --> 0:43:33.040 internal team how did it go or what you guys 0:43:33.080 --> 0:43:35.520 have learned in the process, because it's never perfect right 0:43:35.560 --> 0:43:36.080 out of the gate. 0:43:36.880 --> 0:43:38.160 No, it isn't, and it takes time. 0:43:38.200 --> 0:43:42.960 In the early days, you know, the the the quality 0:43:43.239 --> 0:43:45.640 wasn't as good as it is today, and the satisfaction 0:43:45.760 --> 0:43:47.920 is really high. I don't have the specific numbers. I 0:43:47.920 --> 0:43:50.319 know I've used it and it has been great. The 0:43:50.360 --> 0:43:52.840 reality is it's not the chat bots, which is based 0:43:52.840 --> 0:43:56.400 on rules. It's the it's an intelligence agent, so it 0:43:56.480 --> 0:43:58.319 learns over time and we all know. 0:43:58.320 --> 0:44:00.320 That's the power of AI. AM that's what it a 0:44:00.360 --> 0:44:01.080 great experience. 0:44:01.080 --> 0:44:03.719 So when something needs to be adjusted, it can be 0:44:04.080 --> 0:44:07.680 and when there's feedback, it constantly is iterating, so it's 0:44:07.719 --> 0:44:10.400 only getting better over time and becoming that that, you know, 0:44:10.440 --> 0:44:14.400 the most intelligent HR agent it can be. So I 0:44:14.400 --> 0:44:17.800 think overall it's the right direction in terms of creating 0:44:17.800 --> 0:44:21.880 a great employee experience and enabling HR to do what 0:44:21.920 --> 0:44:23.960 they should be doing and taking care of people and 0:44:24.000 --> 0:44:27.160 not just answering simple questions. And I think that makes 0:44:27.200 --> 0:44:29.600 it for a more interesting HR experience for that employee 0:44:29.640 --> 0:44:30.840 as well. 0:44:31.200 --> 0:44:34.040 What about the administration and the role of the Trump 0:44:34.080 --> 0:44:37.759 administration in this. We started the announcement of Stargate last week, 0:44:37.800 --> 0:44:39.880 one hundred billion up to five hundred billion dollars of 0:44:39.920 --> 0:44:43.600 investment from AI companies. What was the conversation happening around 0:44:43.600 --> 0:44:45.680 this administration with regard to AI. 0:44:46.680 --> 0:44:47.279 I'll tell you. 0:44:47.320 --> 0:44:50.640 At Davos, there was optimism around the global economy and 0:44:50.640 --> 0:44:54.400 the resilience as well as the US economy, and overall, 0:44:54.520 --> 0:44:57.759 there was interest around what we're doing around AI as 0:44:57.800 --> 0:45:01.160 well as regulatory policies, and I think overall there was 0:45:01.200 --> 0:45:06.040 really just great optimism in Davos around the US economy. 0:45:08.280 --> 0:45:12.120 So in terms of kind of where we go from here, 0:45:12.360 --> 0:45:14.160 you know you're in it. You guys are doing it 0:45:14.200 --> 0:45:16.759 at your own company, you're working with companies consulting when 0:45:16.760 --> 0:45:19.160 it comes to AI. You know this is your number 0:45:19.160 --> 0:45:22.040 three right that this continues certainly to be an investment theme, 0:45:22.120 --> 0:45:25.399 although taking a big hit in today's session, no doubt 0:45:25.440 --> 0:45:28.719 about it. So top of mind that you think investors 0:45:28.760 --> 0:45:30.640 should be watching out for. And just got about thirty 0:45:30.680 --> 0:45:31.480 forty seconds here. 0:45:32.680 --> 0:45:37.120 Yeah, I see the investment in AI to continue and 0:45:37.160 --> 0:45:39.680 I and you know, I think the speculations is going 0:45:39.760 --> 0:45:41.160 to double in the next couple of years. 0:45:41.200 --> 0:45:42.960 I don't believe that test changed. 0:45:44.520 --> 0:45:47.200 Our clients are investing A and AI, and I think 0:45:47.239 --> 0:45:52.799 today's announcements and what we saw today only emphasizes the 0:45:52.840 --> 0:45:53.520 importance of. 0:45:53.480 --> 0:45:54.960 The innovation moving at pace. 0:45:55.840 --> 0:45:59.000 And I really do believe it shows that those who 0:45:59.040 --> 0:46:03.320 invest now in bringing AI into and learning and experimenting 0:46:03.360 --> 0:46:06.200 and scaling will be the market leaders. It's going to 0:46:06.239 --> 0:46:10.320 disrupt every industry and that's what We're encouraging our clients 0:46:10.320 --> 0:46:13.120 to do to lean in and you know, power their 0:46:13.200 --> 0:46:16.160 organizations with AI. And that's why we've added you know, 0:46:16.400 --> 0:46:20.000 whip Pro powered AI and we really believe us being 0:46:20.040 --> 0:46:22.719 powered by AI will only help our clients be successful 0:46:22.760 --> 0:46:24.360 and implementing in their own business. 0:46:24.719 --> 0:46:28.719 Susan Dad, CEO of America's over at the global consulting 0:46:28.840 --> 0:46:32.280 firm whip Ro, joining us from New Jersey this afternoon. 0:46:35.239 --> 0:46:37.239 Macout you let me drive. 0:46:37.480 --> 0:46:43.080 Oh no, no, no, no, this is not a toy, Judge, Honey, please. 0:46:42.880 --> 0:46:42.960 I. 0:46:45.600 --> 0:46:46.160 Want to drive. 0:46:48.400 --> 0:46:49.320 It's a good question. 0:46:53.160 --> 0:46:56.360 This is the drive to the clothes that plungs for MEFIC. 0:46:58.280 --> 0:47:01.200 On Bloomberg Radio, all right, everybody got about eighteen minutes 0:47:01.239 --> 0:47:03.560 left to end the Monday trade. You are listening and 0:47:03.600 --> 0:47:06.080 watching Bloomberg Business with Carol Master along with Tim Stenovek 0:47:06.160 --> 0:47:08.719 live here in our Bloomberg Interactive Broker studio. You heard 0:47:08.760 --> 0:47:12.080 Charlie and Bill Maloney breaking it down. S ANDP down 0:47:12.080 --> 0:47:14.920 about one point seven percent, the Nawstak one hundred down 0:47:15.320 --> 0:47:18.080 the most when it comes to the index trade looking 0:47:18.160 --> 0:47:21.320 about three point two percent decline here. I am curious 0:47:21.400 --> 0:47:23.160 to see what Aaron Kennon has to say about all 0:47:23.160 --> 0:47:26.560 of this. Co founder and CEO at Clear Harbor Asset Management, 0:47:26.600 --> 0:47:29.640 they've got about one and a half billion in assets 0:47:29.680 --> 0:47:32.800 under management. Joining us once again from Stanford, Connecticut. Aaron, 0:47:33.239 --> 0:47:37.120 the story that we're all talking about deep seek, I 0:47:37.160 --> 0:47:37.800 am curious. 0:47:37.960 --> 0:47:42.279 I know, you know, we're trying to put it into context, trying. 0:47:42.000 --> 0:47:44.319 Because it's not like the whole market is selling off 0:47:44.360 --> 0:47:47.520 and everybody running for the exit door. So how do 0:47:47.600 --> 0:47:49.200 you think about something like this? 0:47:50.560 --> 0:47:52.719 Yeah, thanks for having me back, Carol. I think at 0:47:52.719 --> 0:47:54.640 the end of the day, we're seeing the evolution of 0:47:54.760 --> 0:47:58.840 artificial intelligence in large language models. We're seeing what I 0:47:58.880 --> 0:48:02.920 would call sort of the gradual commoditization of artificial intelligence, 0:48:03.600 --> 0:48:10.200 and you know, less compute power and you know, less 0:48:10.239 --> 0:48:15.279 GPU power, less electricity demand needed for a product that 0:48:15.560 --> 0:48:17.920 is ninety eight percent as good as the best product 0:48:17.960 --> 0:48:21.000 we have here in the United States is the story 0:48:21.040 --> 0:48:23.400 of the day. And obviously the hyperscalers are under a 0:48:23.440 --> 0:48:25.600 lot of pressure, as you mentioned at the break there 0:48:25.640 --> 0:48:29.800 in Videos down seventeen eighteen percent. Broadcom is down seventeen 0:48:29.880 --> 0:48:34.560 eighteen percent because they were the sort of hyped up 0:48:35.440 --> 0:48:38.279 companies so to speak, that we're going to be the 0:48:38.280 --> 0:48:41.600 beneficiaries of this AI demand. You have data center stocks 0:48:41.640 --> 0:48:44.680 which are very much part of that ecosystem down big today. 0:48:44.760 --> 0:48:48.319 You have sort of the cooling stocks that cool the 0:48:48.400 --> 0:48:50.839 data centers down big today. But then you have other 0:48:50.920 --> 0:48:54.279 parts of the technology landscape today that are interestingly doing 0:48:54.360 --> 0:48:56.880 quite well, and they seem to be almost like the 0:48:56.920 --> 0:48:59.920 beneficiaries of this commoditization of AI as I'm calling it, 0:49:00.880 --> 0:49:03.200 which is that which is to say, if you remember 0:49:03.280 --> 0:49:05.239 just a month or two or even a week ago, 0:49:05.400 --> 0:49:07.719 it was how much is you know, Facebook, how much 0:49:07.719 --> 0:49:10.320 are are Meta? How much are these other companies spending 0:49:10.320 --> 0:49:15.240 on cap x And essentially that's to purchase the hyperscalers, 0:49:15.280 --> 0:49:20.160 the Nvidia chips, the Broadcom chips. And now that narrative 0:49:20.200 --> 0:49:23.440 is changing. And so you have software companies like you know, 0:49:23.800 --> 0:49:28.200 Salesforce and work Day and Service now that are huge 0:49:28.200 --> 0:49:32.520 beneficiaries of the implementation of AI into their software packages. 0:49:32.760 --> 0:49:34.760 But now they're thinking this morning, well, I guess maybe 0:49:34.760 --> 0:49:36.719 this this will cost us less, and those stocks are 0:49:36.719 --> 0:49:38.080 in the green today. 0:49:38.160 --> 0:49:41.080 Yeah, that's that's interesting to hear. I'm wondering if you're 0:49:41.200 --> 0:49:44.719 using this as an opportunity to buy or sell. Are 0:49:44.719 --> 0:49:48.279 you changing asset allocation at all within your portfolios? 0:49:49.760 --> 0:49:54.160 Well, not doing a lot today, but happily it's challenged. 0:49:54.160 --> 0:49:58.239 If you're an active manager in Vidius fifteen percent of 0:49:58.239 --> 0:50:02.839 the Russell one thousand technology, you know if it's six 0:50:02.920 --> 0:50:05.799 percent of the S and P five hundred. So even 0:50:05.840 --> 0:50:07.839 if you're an active manager, and even if you own 0:50:07.880 --> 0:50:10.200 these names that aren't doing well today, you have to 0:50:10.200 --> 0:50:12.799 own more than within if your benchmarks the SMP, you 0:50:12.800 --> 0:50:14.759 have to own more than six percent of Nvidia and 0:50:14.800 --> 0:50:17.080 the SMP to say well, that's a real active bet, right, 0:50:17.160 --> 0:50:22.560 So unless you're underweighted, and so no, I we we 0:50:22.640 --> 0:50:26.840 feel reasonably good today. We have good broad diversification. I 0:50:26.840 --> 0:50:29.480 think there's been a lot of concentration in this market. 0:50:30.280 --> 0:50:33.120 We saw that in twenty twenty four, tim where if 0:50:33.120 --> 0:50:34.800 you look at the S and P five hundred, almost 0:50:34.840 --> 0:50:37.319 sixty five percent of the returns last year came from 0:50:37.400 --> 0:50:39.600 ten names, of which many are the names we're talking 0:50:39.600 --> 0:50:43.279 about right now. And you know to the extent that 0:50:43.280 --> 0:50:46.640 there's a broadening out of the market advancement in twenty 0:50:46.719 --> 0:50:49.200 twenty five, if we are up on the year, we 0:50:49.400 --> 0:50:51.480 welcome that. You look at the equal weight S and 0:50:51.520 --> 0:50:54.200 P five hundred today, it's only down one quarter of 0:50:54.280 --> 0:50:57.680 one percent today, quite telling, So. 0:50:58.400 --> 0:51:01.760 Forgive me, did you so? Would you are you buying? 0:51:01.800 --> 0:51:03.799 Would you be buying an end video? Would you buy 0:51:03.800 --> 0:51:06.400 a broadcom? Because they've all of a sudden been whacked 0:51:06.840 --> 0:51:08.759 in a big way. If you still believe in them 0:51:08.760 --> 0:51:11.919 fundamentally and I'm not sure if you do or I'm 0:51:11.960 --> 0:51:12.560 just curious. 0:51:13.680 --> 0:51:18.800 Yeah, we believe in growth opportunities within the technology segment. 0:51:19.080 --> 0:51:23.280 We would also argue that there has been a significant 0:51:23.600 --> 0:51:28.359 amount of passive dollars flooding into these names, driving them. 0:51:28.400 --> 0:51:31.600 If you look at the top seven names of the 0:51:31.640 --> 0:51:33.719 S and P five hundred, they're trading over thirty times 0:51:33.800 --> 0:51:36.080 forward multiples. You look at the bottom four under ninety 0:51:36.120 --> 0:51:39.560 three names trading at seventeen times. So do we think 0:51:39.560 --> 0:51:40.720 they're quality companies? 0:51:40.800 --> 0:51:40.960 Yes? 0:51:41.000 --> 0:51:43.040 Do we think that some of them are not trading 0:51:43.080 --> 0:51:47.080 at necessarily that the most attractive price point? Yes, and 0:51:47.120 --> 0:51:49.360 we think we can be in the affirmative on both. 0:51:49.680 --> 0:51:53.359 But we do think too that that doesn't necessarily mean 0:51:53.360 --> 0:51:55.600 over the long run that dollars are going to be 0:51:55.640 --> 0:51:59.000 accruing significantly away from these hyper scalers. But this broad 0:51:59.040 --> 0:52:03.480 opportunity across technology. Even the utility landscape today is quite interesting. 0:52:03.520 --> 0:52:06.560 Where this whole narrative around electricity demand, we've been hearing 0:52:06.560 --> 0:52:10.160 it for weeks and months. Look at Constellation Energy today, 0:52:10.160 --> 0:52:10.920 this is the deal. 0:52:10.760 --> 0:52:12.840 That Microsoft did right three Mile Island. 0:52:12.840 --> 0:52:17.520 That's down twenty percent today, yea. 0:52:17.080 --> 0:52:18.320 Down thirty percent today. 0:52:18.440 --> 0:52:20.520 Look at the water utilities today that no one wanted 0:52:20.560 --> 0:52:23.200 to talk about last year. They're up five six seven 0:52:23.239 --> 0:52:26.719 percent today. So it's really interesting. This rotation interesting? 0:52:26.920 --> 0:52:31.200 So interesting rotation? Is it an investible interesting rotation? 0:52:32.440 --> 0:52:34.839 Well, I mean, just look at multiples. I mean, are 0:52:34.840 --> 0:52:38.600 we gonna Are we going We haven't contracted permitted a 0:52:38.800 --> 0:52:43.160 nuclear build out since the nineteen seventies. The markets acting 0:52:43.200 --> 0:52:46.600 as though we're going to have a dozen nuclear buildouts 0:52:48.160 --> 0:52:52.560 contracted in twenty twenty five. It seems highly unlikely. I mean, 0:52:52.600 --> 0:52:55.160 we do think that we're gonna have less regulation. But 0:52:55.840 --> 0:52:57.840 even the announcement at the White House last week was 0:52:57.880 --> 0:53:02.120 quite quite interesting. Where you know, if you look at 0:53:02.120 --> 0:53:04.399 thee hundred billion dollars spend a year for four years, 0:53:04.400 --> 0:53:06.480 actually more than that, it's five hundred billion over four 0:53:06.560 --> 0:53:10.560 years that was announced to build out AI infrastructure. That 0:53:10.680 --> 0:53:12.839 timing looks a little off right now, but to build 0:53:12.840 --> 0:53:16.120 out AI infrastructure, the amount of electricity demand for those 0:53:16.200 --> 0:53:19.400 data centers would be about one point seven gig a 0:53:19.400 --> 0:53:22.239 lots per year based on the dollar spin that they're 0:53:22.280 --> 0:53:25.759 talking about. That's about that's that's over. That's about it. 0:53:25.800 --> 0:53:28.080 Actually it's two point seven excuse me, that's almost three 0:53:28.160 --> 0:53:31.440 nuclear power plants a year just for that project. So 0:53:31.520 --> 0:53:34.319 that we were getting I think over our skis in 0:53:34.360 --> 0:53:36.080 the marketplace to some extent a. 0:53:37.600 --> 0:53:38.080 Short time. 0:53:38.160 --> 0:53:39.959 Okay, if we're getting over ski if we were getting 0:53:39.960 --> 0:53:42.600 over skis now to continue the metaphor, where are we 0:53:42.640 --> 0:53:43.760 on the on the slope? 0:53:45.560 --> 0:53:48.400 Well, I think there are plenty of opportunities in the market, 0:53:48.440 --> 0:53:50.840 as I mentioned, if you look at forward multiples outside 0:53:50.920 --> 0:53:56.239 of this sort of you know, hyper focused area of technology. 0:53:57.160 --> 0:53:59.239 We've even in utilities as I mentioned, you know, the 0:53:59.239 --> 0:54:03.080 water utilities, they have nice roe growth, they're training at 0:54:03.120 --> 0:54:07.240 reasonable multiples. You look at the industrial segment of the market, 0:54:07.480 --> 0:54:11.120 we're seeing a lot of opportunity across the board there. 0:54:11.239 --> 0:54:16.560 If you look at you know, certain segments of the financials, 0:54:16.560 --> 0:54:19.239 we think MNA activity is going to accelerate this year. 0:54:19.480 --> 0:54:24.640 Private equity funds have in many cases reached the end 0:54:24.640 --> 0:54:27.280 of their life in terms of the limited partnership duration 0:54:27.360 --> 0:54:30.239 of ten years or more, and so we think that 0:54:30.640 --> 0:54:33.240 these companies are going to need to be either brought 0:54:33.280 --> 0:54:36.319 into the public market or sold in back into the 0:54:36.320 --> 0:54:40.040 private market this year. Huge opportunities for banks, advisory firms, 0:54:40.480 --> 0:54:44.680 companies that work on valuations of private equity. So you 0:54:44.760 --> 0:54:46.480 have plenty of opportunity there too. 0:54:46.480 --> 0:54:48.080 All right, Aaron, thanks so much. 0:54:48.160 --> 0:54:50.560 Aeron Kennon He is co founder and CEO Clear Harvard 0:54:50.600 --> 0:54:53.320 Asset Management, joining us from Stanford, Connecticut. 0:54:53.680 --> 0:54:58.480 This is the Bloomberg Business Week podcast, available on Apple, Spotify, 0:54:58.600 --> 0:55:02.320 and anywhere else you get your podcasts. Listen live weekday 0:55:02.360 --> 0:55:06.400 afternoons from two to five pm Eastern on Bloomberg dot com, 0:55:06.440 --> 0:55:10.320 the iHeartRadio app, tune In, and the Bloomberg Business App. 0:55:10.560 --> 0:55:13.320 You can also watch us live every weekday on YouTube 0:55:13.560 --> 0:55:15.720 and always on the Bloomberg terminal.