1 00:00:00,800 --> 00:00:05,040 Speaker 1: From the heart of where innovation, money and power collide 2 00:00:05,320 --> 00:00:10,600 Speaker 1: in Silicon Valley and beyond. This is Bloomberg Technology with 3 00:00:10,760 --> 00:00:12,240 Speaker 1: Caroline Hyde. 4 00:00:11,840 --> 00:00:13,240 Speaker 2: And Ed Ludlow. 5 00:00:26,200 --> 00:00:28,520 Speaker 3: Live from New York On Caroline Hyde and. 6 00:00:28,400 --> 00:00:31,760 Speaker 4: I'm Mike Shepard in San Francisco. This is Bloomberg Technology 7 00:00:31,920 --> 00:00:32,400 Speaker 4: coming up. 8 00:00:32,720 --> 00:00:35,479 Speaker 5: Deep Seek Day two, we dive into the ripple effects 9 00:00:35,479 --> 00:00:39,080 Speaker 5: in tech markets and the impact on US chip cubs lust. 10 00:00:39,120 --> 00:00:41,800 Speaker 5: We discuss deep Seek's AI model and the future of 11 00:00:41,880 --> 00:00:46,040 Speaker 5: open source with hugging faces. Chief AI scientist and President 12 00:00:46,040 --> 00:00:49,240 Speaker 5: Trump says Microsoft is in talks to acquire the US 13 00:00:49,400 --> 00:00:52,199 Speaker 5: unit of TikTok. Details later this ow but first we 14 00:00:52,280 --> 00:00:56,160 Speaker 5: check in on the markets and we claw our way back. Mike, 15 00:00:56,360 --> 00:00:59,600 Speaker 5: not nearly off setting yesterday's significant sell off. We're up 16 00:00:59,640 --> 00:01:02,600 Speaker 5: just the edge point on the overall NASDAC. Look, Apple 17 00:01:02,640 --> 00:01:04,559 Speaker 5: once again doing heavy lifting to the point side. 18 00:01:04,600 --> 00:01:07,320 Speaker 3: Meta at a new record high, but not enough because. 19 00:01:07,000 --> 00:01:10,440 Speaker 5: In video bounces back but two percent after a seventeen 20 00:01:10,440 --> 00:01:13,679 Speaker 5: percent set off yesterday, but thirty billion being added from 21 00:01:13,680 --> 00:01:15,920 Speaker 5: the market cap that was wiped out by six. 22 00:01:15,760 --> 00:01:17,160 Speaker 3: Hundred billion dollars yesterday. 23 00:01:17,360 --> 00:01:21,440 Speaker 5: This is all surrounding China coming with clearly a very powerful, 24 00:01:21,560 --> 00:01:24,480 Speaker 5: very efficient, very cheap, open source model NIC. 25 00:01:25,560 --> 00:01:28,920 Speaker 4: And yesterday in video called deep Seek's new model quote 26 00:01:29,120 --> 00:01:34,560 Speaker 4: an excellent AI advancement that complies with US technology export controls, 27 00:01:34,800 --> 00:01:37,920 Speaker 4: and that deep Seek's work illustrates how new models can 28 00:01:37,959 --> 00:01:40,880 Speaker 4: be created. Let's bring in Bloomberg, Zee and King now 29 00:01:40,880 --> 00:01:44,120 Speaker 4: to tell us more about this ian. This was a 30 00:01:44,240 --> 00:01:47,840 Speaker 4: rough day yesterday for in Vidia. It bore the brunt 31 00:01:47,960 --> 00:01:50,280 Speaker 4: of the deep Seak shock across the markets. Yet their 32 00:01:50,360 --> 00:01:54,720 Speaker 4: statement really was curious. Tell us what your interpretation was. 33 00:01:54,800 --> 00:01:56,400 Speaker 6: Yeah, I mean, there are a couple of things going 34 00:01:56,440 --> 00:01:59,040 Speaker 6: on here. I mean they could have obviously said nothing. 35 00:01:59,080 --> 00:02:01,040 Speaker 6: They could have just ignored it. They could have said, 36 00:02:01,240 --> 00:02:04,040 Speaker 6: maybe it's kind of dodgy technology that we don't really 37 00:02:04,040 --> 00:02:06,000 Speaker 6: believe in, but they actually endorsed it and said, no, 38 00:02:06,160 --> 00:02:10,280 Speaker 6: this is good. And the sort of underpinning message we 39 00:02:10,320 --> 00:02:13,120 Speaker 6: saw here was that maybe they're telling the US. 40 00:02:13,040 --> 00:02:15,880 Speaker 7: Government, your export controls haven't worked. We should be allowed 41 00:02:15,880 --> 00:02:17,000 Speaker 7: to ship whatever chips we. 42 00:02:16,919 --> 00:02:20,000 Speaker 6: Want to China because guess what, they'll do good stuff 43 00:02:20,040 --> 00:02:21,320 Speaker 6: with whatever they can get hold of. 44 00:02:21,960 --> 00:02:25,639 Speaker 5: Let's just go to that narrative on chip exports potentially 45 00:02:25,680 --> 00:02:26,280 Speaker 5: not working. 46 00:02:26,600 --> 00:02:28,519 Speaker 3: We heard from President Trump. 47 00:02:28,280 --> 00:02:30,080 Speaker 5: Just yesterday about what the impact could be. 48 00:02:30,160 --> 00:02:30,760 Speaker 3: Just take a listening. 49 00:02:31,800 --> 00:02:34,280 Speaker 8: In a very near future, we're going to be placing 50 00:02:34,360 --> 00:02:38,880 Speaker 8: tariffs and farm production of computer chips, semiconductors, and pharmaceuticals 51 00:02:38,880 --> 00:02:42,480 Speaker 8: to return production of these essential goods to the United 52 00:02:42,520 --> 00:02:46,239 Speaker 8: States of America. They left US and they went to Taiwan, 53 00:02:46,400 --> 00:02:49,720 Speaker 8: which is about ninety eight percent of the chip business 54 00:02:49,720 --> 00:02:52,720 Speaker 8: by the way, and we want them to come back. 55 00:02:54,240 --> 00:02:56,240 Speaker 5: Of course, a lot of in video chips are manufactured 56 00:02:56,280 --> 00:02:58,560 Speaker 5: in Taiwan, and so there's going to be an impact there. 57 00:02:58,840 --> 00:03:00,640 Speaker 5: What do you think the regulatory and pat will be 58 00:03:01,520 --> 00:03:03,520 Speaker 5: let alone where the inferencing is really going to be 59 00:03:03,560 --> 00:03:04,920 Speaker 5: the future for GPU demand? 60 00:03:05,880 --> 00:03:08,760 Speaker 6: Yeah, I mean this is the big question here. And 61 00:03:08,800 --> 00:03:12,160 Speaker 6: the expectation I think before the inauguration was that this 62 00:03:12,240 --> 00:03:15,720 Speaker 6: administration would, if anything, double down on this kind of 63 00:03:15,840 --> 00:03:19,520 Speaker 6: crackdown of chips and the flow to China. And here 64 00:03:19,560 --> 00:03:21,760 Speaker 6: we're getting a different aspect of it. So I think 65 00:03:21,800 --> 00:03:24,720 Speaker 6: we'll have to see how much attention the new administration 66 00:03:24,880 --> 00:03:27,799 Speaker 6: is placed is pacing on this issue. But the expectation 67 00:03:27,960 --> 00:03:29,120 Speaker 6: is that they're not going to let. 68 00:03:29,040 --> 00:03:30,800 Speaker 7: Up em back to Nvidia. 69 00:03:30,919 --> 00:03:34,920 Speaker 4: How much is this complicating their efforts to get their 70 00:03:34,960 --> 00:03:38,640 Speaker 4: customers to keep buying all those premium chips like Blackwell 71 00:03:38,800 --> 00:03:41,200 Speaker 4: and the next generations that you've been writing about so much. 72 00:03:41,280 --> 00:03:42,840 Speaker 7: Yeah, well, it creates a massive question. 73 00:03:42,880 --> 00:03:45,760 Speaker 6: It's like, well, if a Chinese startup can fine tune 74 00:03:45,760 --> 00:03:48,720 Speaker 6: their models to this extent using sort of two three 75 00:03:48,800 --> 00:03:51,520 Speaker 6: year old technology on the cheap, why can't you write 76 00:03:51,640 --> 00:03:55,000 Speaker 6: so if you're Microsoft, if you're a WUS and you're 77 00:03:55,200 --> 00:03:57,960 Speaker 6: running all of these systems, and why wouldn't you do 78 00:03:58,040 --> 00:04:01,120 Speaker 6: it cheaper if you can? So obviously has implications for 79 00:04:01,200 --> 00:04:03,160 Speaker 6: what in video is trying to sell. So we'll see 80 00:04:03,160 --> 00:04:04,400 Speaker 6: how they answer those questions. 81 00:04:04,680 --> 00:04:06,560 Speaker 5: Thank you so far the statement is all we get 82 00:04:06,600 --> 00:04:09,640 Speaker 5: in king, We thank you. Now let's just talk about 83 00:04:09,680 --> 00:04:12,680 Speaker 5: how the tech markets, of course roiled by china seemingly cheap, 84 00:04:12,720 --> 00:04:15,840 Speaker 5: powerful AI model. Here's what some Bloomberg Television guests had 85 00:04:15,840 --> 00:04:16,800 Speaker 5: to say about deep Seek. 86 00:04:17,839 --> 00:04:20,680 Speaker 9: I don't think it's about China versus US, right, It's 87 00:04:20,760 --> 00:04:26,080 Speaker 9: really about closed source versus open source and more more, 88 00:04:26,160 --> 00:04:30,200 Speaker 9: American developers should leverage the open source and what deep 89 00:04:30,240 --> 00:04:31,960 Speaker 9: Seak had done and built on top of it. 90 00:04:31,880 --> 00:04:34,240 Speaker 3: And make it greater. Really amazing what they have done. 91 00:04:34,400 --> 00:04:38,279 Speaker 10: The fraction of training cost, fraction of interference costs to 92 00:04:38,279 --> 00:04:41,000 Speaker 10: produce a model like that in deep Seek, but it 93 00:04:41,120 --> 00:04:43,279 Speaker 10: will unravel basically the path of spending. 94 00:04:43,400 --> 00:04:47,279 Speaker 11: Of course, I think a healthy dose of skepticism is 95 00:04:47,480 --> 00:04:52,520 Speaker 11: very very good, it's needed. But even if you multiply 96 00:04:52,960 --> 00:04:57,200 Speaker 11: their spend and their cost by ten times two D time, 97 00:04:57,240 --> 00:05:00,560 Speaker 11: this is still a order and maltitude in terms of 98 00:05:00,640 --> 00:05:01,440 Speaker 11: theotocopy action. 99 00:05:01,680 --> 00:05:04,920 Speaker 12: They're not really a company that's out there trying to 100 00:05:04,960 --> 00:05:08,520 Speaker 12: compete directly with the big US players. So in that sense, 101 00:05:08,560 --> 00:05:11,680 Speaker 12: I think it was an overreaction to the release of 102 00:05:11,720 --> 00:05:12,240 Speaker 12: this model. 103 00:05:13,440 --> 00:05:14,320 Speaker 3: Let's get more reaction. 104 00:05:14,440 --> 00:05:18,040 Speaker 5: Mardin Alton, now chief investment strategist at Empower. The question 105 00:05:18,160 --> 00:05:22,440 Speaker 5: for many is will capital expenditure remain for in video chips, 106 00:05:22,600 --> 00:05:24,320 Speaker 5: for AI hardware writ large. 107 00:05:24,520 --> 00:05:26,040 Speaker 3: Is that what you were tackling yesterday. 108 00:05:27,080 --> 00:05:29,760 Speaker 2: Well, I think that is one of the critical questions, right, 109 00:05:29,800 --> 00:05:33,000 Speaker 2: because this was a very surgical sell off, a very 110 00:05:33,080 --> 00:05:36,320 Speaker 2: surgical development in the sense that it's attacking really not 111 00:05:36,480 --> 00:05:40,880 Speaker 2: the case for using AI, but really the supply chain 112 00:05:40,960 --> 00:05:44,080 Speaker 2: around AI and all roads on supply chain lean back 113 00:05:44,320 --> 00:05:47,280 Speaker 2: to Nvidia, and so this question of whether companies really 114 00:05:47,320 --> 00:05:49,279 Speaker 2: need to spend this much. I think that is the 115 00:05:49,360 --> 00:05:51,680 Speaker 2: question that markets are grappling with. But it's hard to 116 00:05:51,720 --> 00:05:54,239 Speaker 2: imagine that we're going to see, and of course we'll 117 00:05:54,279 --> 00:05:56,240 Speaker 2: see in earnings this week, but it's hard to imagine 118 00:05:56,240 --> 00:05:59,880 Speaker 2: that we're going to see companies massively change their spending 119 00:06:00,320 --> 00:06:04,640 Speaker 2: patterns simply because of this news. I think there's still 120 00:06:04,680 --> 00:06:07,599 Speaker 2: this arms race going on, So of course that's something 121 00:06:07,600 --> 00:06:10,520 Speaker 2: that we're going to have to watch especially closely. But 122 00:06:10,680 --> 00:06:13,760 Speaker 2: we may need more than just this news to unravel that. 123 00:06:14,279 --> 00:06:18,760 Speaker 5: But the unraveling could start when you have earnings expectations 124 00:06:18,800 --> 00:06:22,040 Speaker 5: as they are and valuations at forty one times future 125 00:06:22,040 --> 00:06:25,560 Speaker 5: earnings for a namelike in video. How much you worried 126 00:06:25,560 --> 00:06:28,520 Speaker 5: that this could have a similarity to what we saw 127 00:06:28,520 --> 00:06:29,360 Speaker 5: in the dot com era. 128 00:06:30,400 --> 00:06:33,520 Speaker 2: I mean, that's really something that we're watching really closely. 129 00:06:33,560 --> 00:06:36,560 Speaker 2: When you're looking at earnings expectations for the broad market overall, 130 00:06:36,800 --> 00:06:39,480 Speaker 2: you're seeing an acceleration of earnings and of course maybe 131 00:06:39,520 --> 00:06:42,480 Speaker 2: not the same for MAGS seven, but certainly a continuation 132 00:06:42,640 --> 00:06:45,120 Speaker 2: of the powerful earnings that we've seen there. And then 133 00:06:45,160 --> 00:06:48,280 Speaker 2: you couple that with valuations that are extreme. When we 134 00:06:48,360 --> 00:06:51,440 Speaker 2: do our analysis of valuations and we break the time 135 00:06:51,520 --> 00:06:54,320 Speaker 2: series for our price to expected earnings down into death siles. 136 00:06:54,600 --> 00:06:56,880 Speaker 2: We're looking at the tech sector that's in the ninetieth 137 00:06:56,880 --> 00:07:01,520 Speaker 2: percentile of valuations. Now associating that with forward one year returns, 138 00:07:01,560 --> 00:07:05,120 Speaker 2: you can still see valuations grind higher and companies perform. 139 00:07:05,240 --> 00:07:08,240 Speaker 2: But ultimately that's where you also start to see some 140 00:07:08,320 --> 00:07:11,480 Speaker 2: of these more meaningful selloffs. And as pedantic as it sounds, 141 00:07:11,560 --> 00:07:14,640 Speaker 2: I think a lot of these developments over the past 142 00:07:14,640 --> 00:07:19,000 Speaker 2: few days relate more to evaluation argument than anything that 143 00:07:19,200 --> 00:07:23,720 Speaker 2: relates to kind of technology in the sense that this 144 00:07:23,800 --> 00:07:26,040 Speaker 2: is big news and technology, but the bigger news is 145 00:07:26,040 --> 00:07:28,800 Speaker 2: that there's no room in the narrative to account for 146 00:07:28,880 --> 00:07:31,360 Speaker 2: anything but world domination when it comes to MAGS seven. 147 00:07:31,600 --> 00:07:33,640 Speaker 2: And this is allowing us to think about the range 148 00:07:33,640 --> 00:07:35,920 Speaker 2: of outcomes Martin. 149 00:07:35,960 --> 00:07:38,520 Speaker 4: To play the movie forward a few years. Where do 150 00:07:38,560 --> 00:07:41,080 Speaker 4: you see this taking us? Is this a parallel to 151 00:07:41,120 --> 00:07:45,040 Speaker 4: the dot com era, which you know, eventually, after some 152 00:07:45,720 --> 00:07:48,920 Speaker 4: troubles right around two thousand, really set us up for 153 00:07:49,120 --> 00:07:50,679 Speaker 4: a long period of growth. 154 00:07:50,880 --> 00:07:52,880 Speaker 7: What about AI? Are we in that moment? 155 00:07:53,880 --> 00:07:56,520 Speaker 2: Well, that's that's the corollary that I reference. I know 156 00:07:56,600 --> 00:07:58,600 Speaker 2: a lot of us are pointing back to the dot 157 00:07:58,640 --> 00:08:01,920 Speaker 2: Com era and looking at evaluation argument. I think that's important, 158 00:08:02,120 --> 00:08:04,520 Speaker 2: But I think the other important element is that this 159 00:08:04,640 --> 00:08:07,600 Speaker 2: is an innovation supercycle. It feels like just the way 160 00:08:07,640 --> 00:08:09,880 Speaker 2: the dot Com era was, and what we saw in 161 00:08:09,880 --> 00:08:14,840 Speaker 2: that period was a massive expansion or growth for the 162 00:08:14,880 --> 00:08:17,560 Speaker 2: broad market, not just for the tech area. So if 163 00:08:17,560 --> 00:08:20,440 Speaker 2: we're looking at the trend growth from the nineteen nineties 164 00:08:20,480 --> 00:08:23,120 Speaker 2: and comparing that to the realized trend growth of the 165 00:08:23,320 --> 00:08:27,440 Speaker 2: subsequent years, we saw massive expansion, whether that's margins or 166 00:08:27,560 --> 00:08:31,640 Speaker 2: enterprise value. That Internet world had a huge impact on 167 00:08:31,680 --> 00:08:33,679 Speaker 2: the broad economy, and I think that's what we can 168 00:08:33,720 --> 00:08:36,280 Speaker 2: expect to see with AI. And in many ways, this 169 00:08:36,440 --> 00:08:39,199 Speaker 2: news around deepseek is part of that in the sense 170 00:08:39,240 --> 00:08:41,120 Speaker 2: that it says we can economize, we can do this 171 00:08:41,200 --> 00:08:43,480 Speaker 2: a lot better, and we can get deployment a lot 172 00:08:43,480 --> 00:08:45,920 Speaker 2: more quickly than maybe we had otherwise thought. 173 00:08:46,559 --> 00:08:50,920 Speaker 4: And on that idea of getting cheaper AI out to 174 00:08:51,040 --> 00:08:54,840 Speaker 4: companies and businesses across the economy maybe a little bit faster. 175 00:08:55,200 --> 00:08:57,599 Speaker 4: What sort of impact do you see and where do 176 00:08:57,640 --> 00:08:59,280 Speaker 4: you see growth potential from that? 177 00:09:00,760 --> 00:09:03,160 Speaker 2: Well, I do think there's this broad market where you 178 00:09:03,240 --> 00:09:10,040 Speaker 2: have to understand spaces that we can't perfectly anticipate. I 179 00:09:10,080 --> 00:09:12,559 Speaker 2: think we're seeing a lot hearing, a lot more conversations 180 00:09:12,600 --> 00:09:19,480 Speaker 2: around retailers and the deployment of kind of acres more 181 00:09:19,600 --> 00:09:22,360 Speaker 2: quickly with fewer steps along the way. I think that's 182 00:09:22,440 --> 00:09:26,520 Speaker 2: one application we're also based square stocks hold up quite 183 00:09:26,520 --> 00:09:30,040 Speaker 2: a bit better. I think that software can easily deploy 184 00:09:30,400 --> 00:09:33,720 Speaker 2: or maybe more easily deploy AI within their world. I 185 00:09:33,720 --> 00:09:36,679 Speaker 2: think that's a possibility. So I think that applications are 186 00:09:36,720 --> 00:09:39,480 Speaker 2: somewhat in less the imagination. It has to be pretty 187 00:09:39,480 --> 00:09:42,200 Speaker 2: broad here, and so there's some argument for kind of 188 00:09:42,200 --> 00:09:45,120 Speaker 2: an equal weight exposure to try to capture that range 189 00:09:45,160 --> 00:09:47,800 Speaker 2: of experiences, but with long term horizon. I don't think 190 00:09:47,840 --> 00:09:50,120 Speaker 2: this is necessarily something that we're going to see in 191 00:09:50,240 --> 00:09:52,160 Speaker 2: Q two Q three of twenty twenty five. 192 00:09:52,559 --> 00:09:54,680 Speaker 5: We're taking a few tech hits to your line, but 193 00:09:54,679 --> 00:09:56,560 Speaker 5: we're going to stick with it matter because I have 194 00:09:56,640 --> 00:09:59,760 Speaker 5: this crucial point for the here and now with retail 195 00:10:00,080 --> 00:10:03,000 Speaker 5: we buying back into in video institutionals not right now? 196 00:10:03,559 --> 00:10:06,520 Speaker 5: What about a so called black swad event or ripple effect. 197 00:10:06,600 --> 00:10:09,160 Speaker 5: We talk of course to nasin Talib yesterday. 198 00:10:09,200 --> 00:10:11,120 Speaker 3: Just have a listen to what he said, right. 199 00:10:12,080 --> 00:10:17,240 Speaker 13: The beginning of an adjustment of people for reality because 200 00:10:17,280 --> 00:10:21,320 Speaker 13: now they realize now it's no longer flawless. You have 201 00:10:21,559 --> 00:10:25,200 Speaker 13: a small little trip on the glass. 202 00:10:25,600 --> 00:10:28,200 Speaker 5: A little chip in the glass of the valuation question, 203 00:10:28,400 --> 00:10:30,160 Speaker 5: a little chip in the glass as to whether this 204 00:10:30,240 --> 00:10:34,200 Speaker 5: is akin to a dot com crisis, but where could 205 00:10:34,400 --> 00:10:38,520 Speaker 5: be sustained? Will Energy survive this? We saw such significant 206 00:10:38,520 --> 00:10:39,600 Speaker 5: sell off to certain of those. 207 00:10:39,520 --> 00:10:44,280 Speaker 2: Names, you know, I think when we're taking a look 208 00:10:44,280 --> 00:10:48,080 Speaker 2: at Nvidia or Energy, I think there's more vulnerability there, 209 00:10:48,400 --> 00:10:51,480 Speaker 2: simply because this is getting at the heart of the 210 00:10:51,480 --> 00:10:54,400 Speaker 2: thesis there that they are the you know, especially in Vidia, 211 00:10:54,480 --> 00:10:58,199 Speaker 2: the dominant provider of this capability that our roads lead 212 00:10:58,240 --> 00:11:00,679 Speaker 2: back to in video when it comes to supply, And 213 00:11:00,720 --> 00:11:04,439 Speaker 2: this story is really tackling that idea that potentially there's 214 00:11:04,480 --> 00:11:06,199 Speaker 2: more ways to skin a cat. You can focus on 215 00:11:06,240 --> 00:11:09,720 Speaker 2: efficiency rather than share mine. Of course, so I think 216 00:11:09,960 --> 00:11:11,840 Speaker 2: the idea of a chip in the glass, I think 217 00:11:11,880 --> 00:11:14,559 Speaker 2: that's important, and it's getting at the heart of Nvidia. 218 00:11:14,640 --> 00:11:17,960 Speaker 2: But whether this is the beginning of the unraveling of 219 00:11:18,000 --> 00:11:20,959 Speaker 2: all things AI, I think that's a little bit more 220 00:11:21,160 --> 00:11:24,360 Speaker 2: extreme or a bridge too far today, simply because that 221 00:11:24,640 --> 00:11:27,880 Speaker 2: unraveling would depend on the use case, the use case 222 00:11:27,960 --> 00:11:29,960 Speaker 2: of AI being intact, and I don't think that's what 223 00:11:30,000 --> 00:11:32,800 Speaker 2: we're seeing here. It's more of the supply chain question. 224 00:11:33,160 --> 00:11:36,520 Speaker 2: And really at this point, I think there's still more 225 00:11:36,559 --> 00:11:39,000 Speaker 2: that we would need to see to really suggest that 226 00:11:39,040 --> 00:11:41,640 Speaker 2: companies are going to pull back on their capital expenditures 227 00:11:41,880 --> 00:11:43,080 Speaker 2: in any meaningful way. 228 00:11:44,280 --> 00:11:47,840 Speaker 4: Martin Norton Chief Investment Strategies that Empowered Thank You. 229 00:11:48,120 --> 00:11:49,480 Speaker 7: Coming up, we're. 230 00:11:49,320 --> 00:11:52,400 Speaker 4: Going to discuss deep seaks open source approach and what 231 00:11:52,440 --> 00:11:55,600 Speaker 4: it means for competitors. The chief science officer at AI 232 00:11:55,720 --> 00:11:57,880 Speaker 4: platform Hugging Face will join us on that. 233 00:11:58,280 --> 00:11:59,880 Speaker 7: This is Bloomberg. 234 00:12:12,080 --> 00:12:16,200 Speaker 5: China's deepseek AI model up ended markets largely though those 235 00:12:16,320 --> 00:12:17,680 Speaker 5: AI infrastructure plays. 236 00:12:17,920 --> 00:12:19,319 Speaker 3: What about model providers? 237 00:12:19,520 --> 00:12:22,520 Speaker 5: Clearly open source is impacting a HI development. We're joined 238 00:12:22,559 --> 00:12:25,320 Speaker 5: by Thomas Wolf Now here's the chief science officer at 239 00:12:25,360 --> 00:12:28,079 Speaker 5: hugging Face and open source and collaborative platform for AI 240 00:12:28,160 --> 00:12:32,200 Speaker 5: builders is our one the huge breakthrough that the market 241 00:12:32,240 --> 00:12:33,040 Speaker 5: clearly thinks it is. 242 00:12:35,120 --> 00:12:37,280 Speaker 13: I think there is a thin bida for overreaction. 243 00:12:37,440 --> 00:12:40,760 Speaker 10: We've seen a steady increase in open source model performance, 244 00:12:41,280 --> 00:12:44,200 Speaker 10: but we also have to be honest. It's the first model, 245 00:12:44,200 --> 00:12:46,800 Speaker 10: I would say that really reached the performance of closed source. 246 00:12:46,800 --> 00:12:49,080 Speaker 10: So the gap is closed now between the clos source 247 00:12:49,120 --> 00:12:51,040 Speaker 10: models and the open source performance. 248 00:12:52,080 --> 00:12:54,960 Speaker 5: Let's just talk about how the derivatives have expanded. Because 249 00:12:55,400 --> 00:12:58,160 Speaker 5: you are all about open source, you're about community, you're 250 00:12:58,200 --> 00:13:01,920 Speaker 5: about the innovation one model can provide to many others. 251 00:13:02,000 --> 00:13:04,240 Speaker 5: From what we understand from Clem the CEO over with 252 00:13:04,360 --> 00:13:07,760 Speaker 5: hugging face, he's saying that there's been five hundred derivatives 253 00:13:07,960 --> 00:13:11,679 Speaker 5: already created from deep seek. How are you seeing an 254 00:13:11,679 --> 00:13:14,080 Speaker 5: exponential growth of the use of this model. 255 00:13:15,559 --> 00:13:18,319 Speaker 13: Yeah, it's even growing, right, So I was just checking. 256 00:13:18,360 --> 00:13:21,200 Speaker 10: I can give you a fresh new stats just now. 257 00:13:21,520 --> 00:13:25,040 Speaker 10: We're over six six hundred and seventeen models already created 258 00:13:25,040 --> 00:13:28,960 Speaker 10: by the community, more than three point two million downloads 259 00:13:28,960 --> 00:13:32,240 Speaker 10: of all these models, with seven hundred thousand dollars for 260 00:13:32,280 --> 00:13:33,079 Speaker 10: the original model. 261 00:13:33,320 --> 00:13:36,360 Speaker 13: So I think that's really the power of open source. Right. 262 00:13:36,520 --> 00:13:38,079 Speaker 13: We see basically a growth of. 263 00:13:38,640 --> 00:13:41,839 Speaker 10: More than thirty person of downloads day to day, which 264 00:13:41,920 --> 00:13:44,880 Speaker 10: kind of is a testament to all this ecosystem that 265 00:13:45,080 --> 00:13:48,200 Speaker 10: is already starting to build around deep seek with basically 266 00:13:48,240 --> 00:13:51,280 Speaker 10: all these companies, all these teams, all these. 267 00:13:51,160 --> 00:13:53,000 Speaker 13: Organizations that are already you. 268 00:13:52,880 --> 00:13:56,160 Speaker 10: Know, taking this open source model and finding in it, 269 00:13:56,280 --> 00:13:59,199 Speaker 10: adapting it, testing it on the unused cases. So that's 270 00:13:59,640 --> 00:14:02,600 Speaker 10: that's maybe the most beautiful thing about these open source capabilities, 271 00:14:02,600 --> 00:14:06,720 Speaker 10: which is still this ecosystem grow live day by day 272 00:14:06,880 --> 00:14:08,679 Speaker 10: around the original Deepsick model. 273 00:14:09,960 --> 00:14:12,640 Speaker 4: Thomas, we can't separate the story of deep seek from 274 00:14:12,679 --> 00:14:16,440 Speaker 4: the geopolitics here. What concerns do you have that a 275 00:14:16,520 --> 00:14:21,080 Speaker 4: reaction from Washington or other governments might be to restrict 276 00:14:21,160 --> 00:14:24,280 Speaker 4: open source use in some fashion as a way of 277 00:14:24,360 --> 00:14:26,680 Speaker 4: keeping technology out of China's hands. 278 00:14:28,400 --> 00:14:31,000 Speaker 10: Yeah, I think there's a lot of I would say 279 00:14:31,600 --> 00:14:34,880 Speaker 10: a lot of way of reframing these stories as a 280 00:14:35,000 --> 00:14:38,720 Speaker 10: US China. But really the more general story here is 281 00:14:38,760 --> 00:14:41,120 Speaker 10: the open source effect of this model. It could have 282 00:14:41,200 --> 00:14:44,280 Speaker 10: come out of almost any countries, and I would expect 283 00:14:44,320 --> 00:14:47,080 Speaker 10: actually open source model to keep coming out of China. 284 00:14:47,160 --> 00:14:50,080 Speaker 10: But we also have European companies like Mixed Trial starting 285 00:14:50,120 --> 00:14:52,600 Speaker 10: to open source model and very small teams. 286 00:14:52,640 --> 00:14:55,920 Speaker 13: So mordinally, I think hopefully. 287 00:14:55,560 --> 00:14:59,640 Speaker 10: We'll see move from a geopolitical inpretation of this story 288 00:14:59,720 --> 00:15:03,360 Speaker 10: to really interpretation at open source versus closed source, and 289 00:15:03,520 --> 00:15:06,840 Speaker 10: our belief attacking face is open source is really the 290 00:15:06,920 --> 00:15:10,920 Speaker 10: way to foster development, to foster a breakthrough technology, to 291 00:15:11,000 --> 00:15:14,360 Speaker 10: foster growing communities, and a lot more business use cases. 292 00:15:14,480 --> 00:15:19,080 Speaker 4: Basically, Thomas, as a software creator yourself, you must have 293 00:15:19,160 --> 00:15:22,200 Speaker 4: some admiration for what Deep Seek has pulled off, apparently 294 00:15:22,520 --> 00:15:24,720 Speaker 4: at a much lower cost and rivals. 295 00:15:24,360 --> 00:15:25,160 Speaker 7: Here in the US. 296 00:15:25,360 --> 00:15:28,320 Speaker 4: But I bet you have some questions too, what would 297 00:15:28,360 --> 00:15:30,760 Speaker 4: you like to learn more about what the company did 298 00:15:30,800 --> 00:15:31,840 Speaker 4: and how it pulled it off? 299 00:15:33,480 --> 00:15:36,520 Speaker 10: So what is interesting is we already started a tragging 300 00:15:36,560 --> 00:15:40,080 Speaker 10: pace to actually explore, you know, can we reproduce this model? 301 00:15:40,120 --> 00:15:42,960 Speaker 10: Can we reproduce it in poticlar on the open right? 302 00:15:43,040 --> 00:15:45,400 Speaker 10: So we have right the diepsick model. I was trying 303 00:15:45,400 --> 00:15:49,920 Speaker 10: its last year, last day yesterday, and it does really, 304 00:15:50,240 --> 00:15:52,720 Speaker 10: you know, offer what we see in the benchmark, which 305 00:15:52,800 --> 00:15:55,640 Speaker 10: is it's a very powerful model. But what we would 306 00:15:55,720 --> 00:15:58,040 Speaker 10: like to know is exactly can you retrain it? Can 307 00:15:58,040 --> 00:16:00,920 Speaker 10: you actually apply the same recess? Is two more models, 308 00:16:00,960 --> 00:16:03,800 Speaker 10: So I've study this project called open Air one, which 309 00:16:03,840 --> 00:16:07,760 Speaker 10: is basically a reproduction and open reprodiction of the deep 310 00:16:07,800 --> 00:16:10,920 Speaker 10: Sick pipeline. Thankfully, they shared a lot of details on 311 00:16:10,960 --> 00:16:11,440 Speaker 10: how they. 312 00:16:11,320 --> 00:16:14,280 Speaker 13: Train the model, much more than than we see recently. 313 00:16:14,600 --> 00:16:17,440 Speaker 10: So I'm quite confident in the coming month we'll be 314 00:16:17,480 --> 00:16:20,760 Speaker 10: able to basically, you know, understand all the breakthrough that 315 00:16:20,800 --> 00:16:22,040 Speaker 10: we need to make in this model. 316 00:16:22,800 --> 00:16:26,040 Speaker 5: What does it mean for a closed source focused open 317 00:16:26,080 --> 00:16:29,440 Speaker 5: AI or for anthropic in this. 318 00:16:29,520 --> 00:16:33,480 Speaker 13: Moment, Well, I think it's quite positive. And you saw 319 00:16:33,520 --> 00:16:35,760 Speaker 13: probably the reaction of some outman, right. 320 00:16:35,840 --> 00:16:38,160 Speaker 10: I think for a good competition in the field is 321 00:16:38,200 --> 00:16:41,320 Speaker 10: something that is just a net positive for developing a technology. 322 00:16:41,760 --> 00:16:43,680 Speaker 13: And in particular here because a lot of. 323 00:16:43,600 --> 00:16:45,480 Speaker 10: This model and as I was saying, a lot of 324 00:16:45,560 --> 00:16:48,280 Speaker 10: recipes are open, it means I would I would be 325 00:16:48,400 --> 00:16:51,440 Speaker 10: very surprised they don't, you know, directly take the recipes 326 00:16:51,480 --> 00:16:54,760 Speaker 10: that can be useful for improving both open AI, anthropic 327 00:16:55,000 --> 00:16:59,160 Speaker 10: Google or the coming LAMMA models, and that this model basically. 328 00:16:58,680 --> 00:17:01,800 Speaker 13: You know, quickly catch So I think it's a good thing. 329 00:17:01,920 --> 00:17:04,399 Speaker 10: That's also the good thing about open sources because you 330 00:17:04,760 --> 00:17:07,679 Speaker 10: share a lot of information about your models. You know, 331 00:17:07,800 --> 00:17:11,959 Speaker 10: you actually lift the whole fields up with you by 332 00:17:12,040 --> 00:17:14,160 Speaker 10: you know, explaining how how you make your breakthrough. 333 00:17:15,040 --> 00:17:18,320 Speaker 5: But some people don't want to see everyone rise up. 334 00:17:18,560 --> 00:17:22,280 Speaker 5: They don't want to see China rise up. Ultimately, is 335 00:17:22,320 --> 00:17:25,480 Speaker 5: that a false narrative a strong man to be, even 336 00:17:25,560 --> 00:17:27,800 Speaker 5: saying that China is behind the US when there's open 337 00:17:27,840 --> 00:17:29,160 Speaker 5: source communities such as yours. 338 00:17:30,880 --> 00:17:34,400 Speaker 10: Yeah, the Open Source Committee generally don't really know any border, right, 339 00:17:34,880 --> 00:17:37,440 Speaker 10: as I was saying today, it's a Chinese team, but generally, 340 00:17:37,520 --> 00:17:39,960 Speaker 10: you know, it's just it's just a very smart, edge 341 00:17:40,320 --> 00:17:40,920 Speaker 10: edge round team. 342 00:17:41,000 --> 00:17:41,200 Speaker 13: Right. 343 00:17:41,400 --> 00:17:43,800 Speaker 10: There are teams like that in many countries, and that's 344 00:17:43,800 --> 00:17:46,439 Speaker 10: why we keep saying it would keep seeing new actors 345 00:17:46,480 --> 00:17:50,919 Speaker 10: in AI, right, So I think ultimately, I believe in 346 00:17:51,000 --> 00:17:54,800 Speaker 10: fair market. I think having competition, sharing more actively, you know, 347 00:17:54,920 --> 00:17:58,960 Speaker 10: is a good way to make progress together. So you know, 348 00:17:59,040 --> 00:18:01,200 Speaker 10: in your future, we will like to see much more 349 00:18:01,240 --> 00:18:03,840 Speaker 10: actors active in AI. We would like to see basically 350 00:18:04,000 --> 00:18:07,280 Speaker 10: a larger company a bit everywhere in the world. And 351 00:18:07,320 --> 00:18:09,320 Speaker 10: I think that's maybe the first step, you know, in 352 00:18:09,359 --> 00:18:09,919 Speaker 10: this direction. 353 00:18:11,280 --> 00:18:15,280 Speaker 4: Are we expecting to see similar innovations coming from Europe? 354 00:18:15,359 --> 00:18:18,760 Speaker 4: There have been questions and concerns raised in the European 355 00:18:18,880 --> 00:18:21,680 Speaker 4: Union about the level of regulation and then it might 356 00:18:21,720 --> 00:18:23,920 Speaker 4: be restricting development in AI. 357 00:18:26,280 --> 00:18:28,440 Speaker 13: Yeah, it's obviously a big discussion, right. 358 00:18:28,720 --> 00:18:30,919 Speaker 10: I was a divorce a bit last week, and if 359 00:18:30,960 --> 00:18:32,960 Speaker 10: you see a lot of the discussion was, Okay, what 360 00:18:33,480 --> 00:18:34,680 Speaker 10: is about your regulation? 361 00:18:35,080 --> 00:18:36,760 Speaker 13: I just think the teams are really good. 362 00:18:36,840 --> 00:18:39,600 Speaker 10: I would be surprised, for instance, in the UK, in Germany, 363 00:18:39,720 --> 00:18:41,760 Speaker 10: you know in France that we that we don't see 364 00:18:41,760 --> 00:18:44,879 Speaker 10: all these teams that basically helped train you know, the 365 00:18:44,960 --> 00:18:48,879 Speaker 10: models of Mistual, at Meta, at Opena, at Entropic. 366 00:18:48,920 --> 00:18:51,040 Speaker 13: We've seen a lot of people also living op Ai 367 00:18:51,200 --> 00:18:52,440 Speaker 13: to start the und startup. 368 00:18:52,520 --> 00:18:55,840 Speaker 10: So I think one takeaway from Deep Sick is that 369 00:18:55,920 --> 00:19:00,280 Speaker 10: basically the recipe to build a very good quality large 370 00:19:00,320 --> 00:19:04,400 Speaker 10: language model nowadays is something that's almost accessible to everyone, right, 371 00:19:04,480 --> 00:19:07,520 Speaker 10: So I think all these people leaving the big tech 372 00:19:07,520 --> 00:19:10,720 Speaker 10: companies will start the out team, and because you kind 373 00:19:10,720 --> 00:19:12,920 Speaker 10: of just need a few millions, I would be surprised 374 00:19:12,920 --> 00:19:14,639 Speaker 10: we don't see a lot more teams coming out of 375 00:19:14,640 --> 00:19:17,400 Speaker 10: Europe and Botska, but also as a region this year. 376 00:19:18,760 --> 00:19:21,840 Speaker 4: Thomas Wolf, chief science officer at Hugging Face, thanks for 377 00:19:21,920 --> 00:19:33,160 Speaker 4: joining us. Microsoft is in talks that acquire the US 378 00:19:33,280 --> 00:19:36,080 Speaker 4: arm of TikTok. At least that's according to President Trump 379 00:19:36,160 --> 00:19:38,440 Speaker 4: last night, who did not elaborate further. 380 00:19:38,800 --> 00:19:40,240 Speaker 7: Microsoft declient to comment. 381 00:19:40,520 --> 00:19:43,479 Speaker 4: Let's bring in Bloomberg's Balance and Power co host Kaylee 382 00:19:43,520 --> 00:19:47,040 Speaker 4: lines for more. Kaylee, is this another case of Trump 383 00:19:47,320 --> 00:19:50,560 Speaker 4: trying to make a deal happen? He casts himself as 384 00:19:50,600 --> 00:19:53,960 Speaker 4: the consummate deal maker. We are so short on details, though, 385 00:19:54,240 --> 00:19:56,920 Speaker 4: walk us through what more we know, if anything beyond 386 00:19:57,000 --> 00:19:57,480 Speaker 4: last night. 387 00:19:58,880 --> 00:20:00,960 Speaker 3: Well, the details are pretty limited, Mike. 388 00:20:01,000 --> 00:20:03,720 Speaker 14: He was asked directly by a reporter if Microsoft was 389 00:20:03,760 --> 00:20:06,359 Speaker 14: in talks to acquire TikTok, and the President said, quote, 390 00:20:06,359 --> 00:20:08,520 Speaker 14: I would say yes, And that's really all we got 391 00:20:08,520 --> 00:20:11,120 Speaker 14: on that specific matter. As he said, Microsoft isn't commenting 392 00:20:11,440 --> 00:20:13,440 Speaker 14: on this either. Keep in mind, though, that during the 393 00:20:13,480 --> 00:20:16,040 Speaker 14: first Trump administration, back in the summer of twenty twenty, 394 00:20:16,040 --> 00:20:19,600 Speaker 14: Bloomberg did report Microsoft was looking at acquiring TikTok's US 395 00:20:19,640 --> 00:20:22,760 Speaker 14: operations back then, when the first kind of pressure around 396 00:20:22,760 --> 00:20:26,399 Speaker 14: divesting or banning it over national security concerns was a rising. 397 00:20:26,440 --> 00:20:28,600 Speaker 14: Oracle was reportedly looking at it too, And we know 398 00:20:28,720 --> 00:20:31,560 Speaker 14: just last week when the Stargate project was being announced 399 00:20:31,560 --> 00:20:33,439 Speaker 14: at the White House and Larry Ellison of Oracle was 400 00:20:33,440 --> 00:20:35,399 Speaker 14: in the room, Donald Trump suggested he'd be open to 401 00:20:35,440 --> 00:20:38,720 Speaker 14: Ellison buying TikTok or Elon Musk when he was asked 402 00:20:38,760 --> 00:20:41,640 Speaker 14: about that. We know there are other players involved as well, 403 00:20:41,680 --> 00:20:44,439 Speaker 14: including Frank McCourt, who you speak with frequently on this program, 404 00:20:44,440 --> 00:20:46,159 Speaker 14: who have made bus and it does seem that is 405 00:20:46,160 --> 00:20:49,040 Speaker 14: what Donald Trump's preference really is here, a bidding war. 406 00:20:49,119 --> 00:20:51,280 Speaker 14: He said as much last night that he thinks bidding 407 00:20:51,280 --> 00:20:54,080 Speaker 14: war's result in the best deals. As for what that 408 00:20:54,119 --> 00:20:56,919 Speaker 14: deal ultimately looks like, that remains unclear. What he was 409 00:20:56,960 --> 00:20:59,720 Speaker 14: clear on yesterday when speaking to the House Republican Conference 410 00:20:59,760 --> 00:21:02,479 Speaker 14: in his to Raw Club in Florida is that he 411 00:21:02,560 --> 00:21:05,640 Speaker 14: does not want China involved in whatever happens here. Of course, 412 00:21:05,720 --> 00:21:08,040 Speaker 14: China is going to have some say in that ultimately, 413 00:21:08,280 --> 00:21:11,040 Speaker 14: would the US government have involvement still, Katie. 414 00:21:11,760 --> 00:21:12,639 Speaker 3: Well, that's the question. 415 00:21:12,680 --> 00:21:15,560 Speaker 14: He's talked about this fifty percent ownership structure, whoever buys 416 00:21:15,600 --> 00:21:18,440 Speaker 14: it sharing equity with the United States essentially, so it 417 00:21:18,440 --> 00:21:20,600 Speaker 14: would be some kind of half and half deal, But 418 00:21:20,680 --> 00:21:22,560 Speaker 14: it remains unclear whether or not that can be done 419 00:21:22,560 --> 00:21:25,120 Speaker 14: by the now April fourth deadline, which has been extended 420 00:21:25,160 --> 00:21:28,040 Speaker 14: by Trump's executive order. There's also a massive question around 421 00:21:28,080 --> 00:21:30,399 Speaker 14: whether anti trust concerns would be raised by some of 422 00:21:30,400 --> 00:21:35,040 Speaker 14: these individuals or companies like Microsoft acquiring operations of this size, 423 00:21:35,040 --> 00:21:37,080 Speaker 14: which could be tens of billions of dollars and obviously, 424 00:21:37,080 --> 00:21:38,920 Speaker 14: as you guys will know, one hundred and seventy million 425 00:21:39,000 --> 00:21:42,280 Speaker 14: users in the US. While Donald Trump and the people 426 00:21:42,280 --> 00:21:44,720 Speaker 14: he's instilled at places like the FTC or at the 427 00:21:44,760 --> 00:21:47,720 Speaker 14: Department of Justice may have some alignment on this, it 428 00:21:47,800 --> 00:21:51,600 Speaker 14: still could face scrutiny, especially from big tech skeptical lawmakers 429 00:21:51,760 --> 00:21:53,200 Speaker 14: on Capitol Hill and. 430 00:21:53,320 --> 00:21:56,679 Speaker 5: Whether bite don'ts whatever said it, Katie Hines, Yeah, thanks 431 00:21:56,680 --> 00:21:57,080 Speaker 5: so much. 432 00:22:05,480 --> 00:22:07,840 Speaker 3: Welcome back to med Technology. I'm Caroline Hyde in New 433 00:22:07,920 --> 00:22:09,200 Speaker 3: York and I'm. 434 00:22:09,040 --> 00:22:10,520 Speaker 7: Mike Shepard in San Francisco. 435 00:22:10,960 --> 00:22:13,600 Speaker 4: Let's discuss the impact that Deep seak is having on 436 00:22:13,720 --> 00:22:17,640 Speaker 4: US tech companies now with Bloomberg Sharen Gafari. Sharen, thanks 437 00:22:17,680 --> 00:22:20,840 Speaker 4: so much and thanks for all your reporting today. There's 438 00:22:20,920 --> 00:22:23,600 Speaker 4: a lot of soul searching going on in Silicon Valley 439 00:22:23,760 --> 00:22:26,200 Speaker 4: over the past forty eight hours. Tell us a little 440 00:22:26,240 --> 00:22:28,680 Speaker 4: bit more about what's happening. They set up war rooms. 441 00:22:28,920 --> 00:22:30,399 Speaker 4: What is the conversation like? 442 00:22:31,640 --> 00:22:34,439 Speaker 15: So the top a I labs right now in the 443 00:22:34,600 --> 00:22:37,920 Speaker 15: US are trying to figure out how the Chinese startup 444 00:22:38,000 --> 00:22:41,160 Speaker 15: Deep Seek was able to catch up so quickly. Right, 445 00:22:41,200 --> 00:22:43,720 Speaker 15: you have their latest R one model. Really exploding on 446 00:22:43,760 --> 00:22:46,600 Speaker 15: the scene in the past week, and everyone is astounded 447 00:22:46,680 --> 00:22:49,960 Speaker 15: by how competitive it is. It is actually leading by 448 00:22:49,960 --> 00:22:53,720 Speaker 15: some metrics on these AI models that have taken Western companies, 449 00:22:53,960 --> 00:22:55,679 Speaker 15: you know, years end, a lot more money. 450 00:22:56,280 --> 00:22:57,520 Speaker 3: The Chinese company. 451 00:22:57,240 --> 00:23:00,320 Speaker 5: Deep Sea claims to build well, we can old bait 452 00:23:00,359 --> 00:23:04,080 Speaker 5: as to how much it was copying building off Western technology. 453 00:23:04,119 --> 00:23:06,919 Speaker 5: But the key question is will capex remain the same 454 00:23:07,160 --> 00:23:09,800 Speaker 5: and ultimately what it means for open AI and Anthropic 455 00:23:09,880 --> 00:23:11,719 Speaker 5: in terms of reducing the cost of their models. 456 00:23:11,800 --> 00:23:13,680 Speaker 3: What do you think the ripple effects Ascherene. 457 00:23:15,119 --> 00:23:17,359 Speaker 16: Yeah, I think we're starting to see people really question 458 00:23:18,440 --> 00:23:23,080 Speaker 16: whether we need these astronomical budgets for building the most 459 00:23:23,080 --> 00:23:26,119 Speaker 16: advanced AI models and whether, you know, because. 460 00:23:25,760 --> 00:23:27,960 Speaker 3: Deep Seek was able to make some gains in. 461 00:23:27,960 --> 00:23:32,480 Speaker 15: How efficiently they use the computing power, if those you know, 462 00:23:32,520 --> 00:23:34,680 Speaker 15: gains can be applied now to US companies and sort 463 00:23:34,680 --> 00:23:37,040 Speaker 15: of questioning why US companies didn't come up with that first. 464 00:23:38,600 --> 00:23:39,920 Speaker 3: It's a fascinating big take. 465 00:23:40,000 --> 00:23:43,000 Speaker 5: Go and read it from Sharene Gafari and the meta 466 00:23:43,040 --> 00:23:44,679 Speaker 5: war rooms that are currently being set up. 467 00:23:44,720 --> 00:23:45,479 Speaker 3: We appreciate it. 468 00:23:45,520 --> 00:23:48,480 Speaker 5: Now we've got to assess what President Trump had to 469 00:23:48,480 --> 00:23:50,280 Speaker 5: say about the competition from deep Seek. 470 00:23:51,560 --> 00:23:54,080 Speaker 8: I think if it's if it's fact and if it's true, 471 00:23:54,720 --> 00:23:57,080 Speaker 8: and nobody really knows what it is, but I view 472 00:23:57,119 --> 00:23:59,000 Speaker 8: that as a positive because you'll be doing that too, 473 00:23:59,080 --> 00:24:00,960 Speaker 8: so you won't be spending much and you'll get the 474 00:24:00,960 --> 00:24:06,720 Speaker 8: same result. Hopefully, the release of Deepseek AI from a 475 00:24:06,880 --> 00:24:09,600 Speaker 8: Chinese company should be a wake up call for our 476 00:24:09,680 --> 00:24:13,280 Speaker 8: industries that we need to be laser focused on competing 477 00:24:13,320 --> 00:24:13,600 Speaker 8: to win. 478 00:24:14,080 --> 00:24:15,200 Speaker 3: Let's talk about that competition. 479 00:24:15,280 --> 00:24:17,639 Speaker 5: Jacqueline Rice Nelson is with us CO Founracy of Tribe 480 00:24:17,640 --> 00:24:20,800 Speaker 5: Ai provides AI services for leading enterprises, and I have 481 00:24:20,840 --> 00:24:22,720 Speaker 5: a feeling a lot of them are rushing to understand 482 00:24:22,720 --> 00:24:25,640 Speaker 5: how they can use this cheap and innovative AI model. 483 00:24:25,960 --> 00:24:30,119 Speaker 17: Absolutely first, thrilled to be here. I think this twenty 484 00:24:30,160 --> 00:24:33,840 Speaker 17: twenty five is the year for enterprises to get ROI 485 00:24:34,040 --> 00:24:38,159 Speaker 17: from their AI efforts, and the ROI calculation just changed dramatically, 486 00:24:38,560 --> 00:24:41,679 Speaker 17: So I think that kind of is the headline for businesses, 487 00:24:41,800 --> 00:24:44,440 Speaker 17: which is use cases that weren't possible just a few 488 00:24:44,440 --> 00:24:48,840 Speaker 17: weeks ago or even last week are now possible. And 489 00:24:49,200 --> 00:24:51,440 Speaker 17: I think we're about to see an explosion of AI 490 00:24:51,560 --> 00:24:55,920 Speaker 17: experimentation and also value delivery to organizations and enterprises away. 491 00:24:56,040 --> 00:24:58,199 Speaker 5: Have any of those organizations been reticent to use it 492 00:24:58,200 --> 00:24:59,640 Speaker 5: simply because of where it was born? 493 00:24:59,720 --> 00:25:01,479 Speaker 3: And it's a great question. 494 00:25:02,160 --> 00:25:05,760 Speaker 17: I actually had the same question and dove into this myself, 495 00:25:06,359 --> 00:25:08,560 Speaker 17: and I think it all comes down to how you 496 00:25:08,680 --> 00:25:12,240 Speaker 17: use these models. So there are lots of ways to 497 00:25:12,280 --> 00:25:16,240 Speaker 17: be using LAMA or open source models today in ways 498 00:25:16,240 --> 00:25:18,320 Speaker 17: that have lots of guardrails that are set up to 499 00:25:18,359 --> 00:25:20,879 Speaker 17: be more secure and can be run locally on your 500 00:25:20,920 --> 00:25:23,639 Speaker 17: own environment. In many ways, actually it's almost easier with 501 00:25:23,680 --> 00:25:26,520 Speaker 17: open source than it is with closed source models, And 502 00:25:26,600 --> 00:25:31,000 Speaker 17: so everything comes down to the implementations of these models. 503 00:25:31,040 --> 00:25:34,399 Speaker 17: I think today a lot of businesses have these same questions, 504 00:25:35,119 --> 00:25:37,600 Speaker 17: but that's in many ways why they're sort of coming 505 00:25:37,680 --> 00:25:40,880 Speaker 17: to Tribe and trying to help navigate the landscape across 506 00:25:40,920 --> 00:25:43,280 Speaker 17: all of these different models. What are best to use 507 00:25:43,400 --> 00:25:46,280 Speaker 17: and for what types of use cases and how to 508 00:25:46,280 --> 00:25:49,480 Speaker 17: set it up in the optimal ways Jacqueline. 509 00:25:49,480 --> 00:25:51,840 Speaker 4: One of the key questions that's come up in relation 510 00:25:51,960 --> 00:25:54,760 Speaker 4: to deep seek is the one on spending. How is 511 00:25:54,800 --> 00:25:57,960 Speaker 4: that factoring into the way you are guiding people through 512 00:25:58,000 --> 00:25:59,119 Speaker 4: the AI landscape? 513 00:25:59,119 --> 00:25:59,320 Speaker 2: Now? 514 00:25:59,640 --> 00:26:01,919 Speaker 4: Is it a different message that you're having to deliver 515 00:26:02,480 --> 00:26:07,600 Speaker 4: about efficiency and about bringing a product to market for 516 00:26:07,640 --> 00:26:08,320 Speaker 4: a lot less. 517 00:26:09,240 --> 00:26:12,880 Speaker 17: So the short answer is yes, I think that we are. 518 00:26:13,440 --> 00:26:16,199 Speaker 17: I compare it to a highway. We have just added 519 00:26:16,280 --> 00:26:18,600 Speaker 17: a lane or multiple lanes to the highway. When you 520 00:26:18,640 --> 00:26:21,719 Speaker 17: do that, the traffic does not go down. Actually it 521 00:26:21,760 --> 00:26:25,159 Speaker 17: only increases, and that's because driving and going on road 522 00:26:25,200 --> 00:26:28,160 Speaker 17: trips gets more attractive. The same is true right now 523 00:26:28,280 --> 00:26:31,320 Speaker 17: for AI. So I think we're about to see really 524 00:26:31,359 --> 00:26:36,000 Speaker 17: an explosion of activity and that sort of across companies, 525 00:26:36,040 --> 00:26:39,760 Speaker 17: across enterprises within the large and to your point on 526 00:26:39,840 --> 00:26:44,320 Speaker 17: efficiency and for product innovation, I think we're going to 527 00:26:44,359 --> 00:26:48,680 Speaker 17: see proliferation across these use cases. And then that doesn't 528 00:26:48,680 --> 00:26:51,679 Speaker 17: even include the large hyper scalers who are still on 529 00:26:51,760 --> 00:26:55,440 Speaker 17: this quest for AGI and nothing changes there. I don't 530 00:26:55,440 --> 00:26:58,280 Speaker 17: think anyone is taking their foot off the gas. And 531 00:26:58,320 --> 00:27:00,760 Speaker 17: if anything, competition is just in dramatically. 532 00:27:02,040 --> 00:27:04,800 Speaker 4: And the open source model, is this something that is 533 00:27:05,320 --> 00:27:10,119 Speaker 4: feasible to be scaled at large enterprises, Big corporations that 534 00:27:10,240 --> 00:27:14,680 Speaker 4: also may have security considerations and other factors to bring 535 00:27:14,760 --> 00:27:17,440 Speaker 4: in them might make them question whether open source is 536 00:27:17,480 --> 00:27:18,600 Speaker 4: the right choice for them. 537 00:27:19,160 --> 00:27:21,480 Speaker 17: Yeah, I think there are a lot of considerations that 538 00:27:21,560 --> 00:27:24,800 Speaker 17: go into what is the right model selection, and so 539 00:27:25,240 --> 00:27:27,760 Speaker 17: I think it's really clear that we have entered an era, 540 00:27:27,840 --> 00:27:29,920 Speaker 17: and actually I've always thought we were in this era, 541 00:27:30,280 --> 00:27:36,159 Speaker 17: which is that we are in a multimodel world. Businesses 542 00:27:36,240 --> 00:27:39,159 Speaker 17: need to be able to build, to swap between models 543 00:27:39,160 --> 00:27:45,640 Speaker 17: seamlessly to optimize for performance, for latency, for accuracy, and 544 00:27:45,680 --> 00:27:48,800 Speaker 17: for cost. And what that means is they also need 545 00:27:49,080 --> 00:27:52,600 Speaker 17: the guardrails, the security, the infrastructure to set those things 546 00:27:52,680 --> 00:27:55,919 Speaker 17: up in proper ways. They also need the evals, the 547 00:27:55,960 --> 00:27:59,120 Speaker 17: evaluation framework to actually be able to compare and contrast 548 00:27:59,119 --> 00:28:02,919 Speaker 17: across models so they know what models to use. And 549 00:28:02,960 --> 00:28:06,520 Speaker 17: so I think there, while there might be some hesitance, 550 00:28:07,560 --> 00:28:10,080 Speaker 17: we have done lots of development on closed source models 551 00:28:10,080 --> 00:28:12,760 Speaker 17: and lots of development on open source models, and the 552 00:28:12,840 --> 00:28:16,480 Speaker 17: considerations still come down to what is that company trying 553 00:28:16,480 --> 00:28:18,760 Speaker 17: to achieve and what are the things we're optimizing for. 554 00:28:19,040 --> 00:28:22,159 Speaker 17: That's what determines the model. So I, oh, please go ahead. 555 00:28:22,520 --> 00:28:24,159 Speaker 5: Well I'm interested in that because I want to go 556 00:28:24,200 --> 00:28:27,560 Speaker 5: back to sort of the Jevons paradox piece that you know, 557 00:28:27,760 --> 00:28:30,280 Speaker 5: the more AI innovation there is, ultimately the more people 558 00:28:30,280 --> 00:28:33,080 Speaker 5: on those highways the more innovation, but also the same 559 00:28:33,119 --> 00:28:34,919 Speaker 5: amount of compute. Can you speak a bit to like 560 00:28:34,960 --> 00:28:37,240 Speaker 5: the inferencing part of this, because that seems to be 561 00:28:37,280 --> 00:28:39,840 Speaker 5: the silver lining that indeeded videos trying to see from 562 00:28:39,880 --> 00:28:41,800 Speaker 5: this look, you're still going to need our GPUs, just 563 00:28:41,840 --> 00:28:42,479 Speaker 5: not for training. 564 00:28:42,480 --> 00:28:43,200 Speaker 3: It's the inference. 565 00:28:43,880 --> 00:28:46,120 Speaker 17: I think that's fair. Look, I think we have not 566 00:28:46,240 --> 00:28:49,160 Speaker 17: cracked the code on the hardware that's needed. Like it's 567 00:28:49,200 --> 00:28:51,920 Speaker 17: clear that compute is still needed at really high levels. 568 00:28:52,320 --> 00:28:55,120 Speaker 17: There is no way that we have even with these 569 00:28:55,160 --> 00:28:58,239 Speaker 17: new deep seek models, we haven't found the answers yet. Right, 570 00:28:58,280 --> 00:29:00,800 Speaker 17: We're not stopping. And I think that that's the goal 571 00:29:00,920 --> 00:29:04,760 Speaker 17: case for how and why you're going to see continued 572 00:29:04,840 --> 00:29:08,440 Speaker 17: investment here is that we are in a continuous innovation 573 00:29:08,640 --> 00:29:13,520 Speaker 17: cycle and ultimately, I think that drives utilization and I 574 00:29:13,560 --> 00:29:16,760 Speaker 17: think the question is for what. And then the other 575 00:29:16,800 --> 00:29:20,360 Speaker 17: piece is that ultimately consumers are the biggest winners here, 576 00:29:21,160 --> 00:29:23,920 Speaker 17: and I think that is the most exciting story. There's 577 00:29:23,920 --> 00:29:26,280 Speaker 17: a lot of fear, there's a lot of you know, 578 00:29:26,360 --> 00:29:30,200 Speaker 17: China versus US dynamics. I think the reality is that 579 00:29:30,600 --> 00:29:34,240 Speaker 17: competition is good. And what I'm hearing from the US 580 00:29:34,360 --> 00:29:38,240 Speaker 17: companies is you know, nothing has kind of the US 581 00:29:38,400 --> 00:29:41,000 Speaker 17: large AI companies, nothing has kind of blown their socks 582 00:29:41,080 --> 00:29:44,400 Speaker 17: deef right. This is innovation they at least, you know, 583 00:29:44,520 --> 00:29:48,040 Speaker 17: feel that they already have. And the difference is that 584 00:29:48,080 --> 00:29:50,280 Speaker 17: there has been an optimization for open source and an 585 00:29:50,320 --> 00:29:53,280 Speaker 17: optimization for cost in ways that they have not done. 586 00:29:53,600 --> 00:29:55,360 Speaker 17: And so I think we're about to see a lot 587 00:29:55,400 --> 00:29:59,120 Speaker 17: of fast follow and Sam Alman signaled as much last night. 588 00:29:59,240 --> 00:30:01,880 Speaker 3: Yeah, he said, we're going to drop models faster or is. 589 00:30:01,880 --> 00:30:03,920 Speaker 5: It going to drop the cost because that it must 590 00:30:03,960 --> 00:30:05,080 Speaker 5: be something that puts off your. 591 00:30:04,960 --> 00:30:07,680 Speaker 17: Client's definitely going to drop the cost. And I actually 592 00:30:07,680 --> 00:30:12,880 Speaker 17: think we are likely to see a cost curve decrease 593 00:30:13,000 --> 00:30:16,280 Speaker 17: like that's dramatic from them, but also a latency decrease. 594 00:30:16,600 --> 00:30:18,200 Speaker 17: And I think that that's how they're going to play 595 00:30:18,240 --> 00:30:20,680 Speaker 17: the game is to try to now one up completely 596 00:30:21,040 --> 00:30:22,160 Speaker 17: on both dimensions. 597 00:30:23,560 --> 00:30:28,480 Speaker 4: Jacqueline Rice Nelson, CEO of Tribe AI, Thank you, Caraw's just. 598 00:30:28,560 --> 00:30:31,080 Speaker 5: Check on these markets because these are all the questions 599 00:30:31,280 --> 00:30:33,480 Speaker 5: that investors are trained on at the moment. What does 600 00:30:33,520 --> 00:30:35,560 Speaker 5: it mean for compute going forwards, what does it mean 601 00:30:35,600 --> 00:30:38,520 Speaker 5: for the competitors in the generative AI model space? And 602 00:30:38,560 --> 00:30:40,680 Speaker 5: what does it ultimately mean for software too? Now we're 603 00:30:40,680 --> 00:30:42,600 Speaker 5: clawing back some of our losses of yesterday, but not 604 00:30:42,640 --> 00:30:44,440 Speaker 5: March eight tens percent higher on than as that one 605 00:30:44,520 --> 00:30:45,560 Speaker 5: hundred Meta. 606 00:30:45,320 --> 00:30:46,120 Speaker 3: In a new record high. 607 00:30:46,160 --> 00:30:49,200 Speaker 5: We'll talk more about Lama there, but Bitcoin up nine 608 00:30:49,240 --> 00:30:50,840 Speaker 5: tens percent. We're going to delve into the world of 609 00:30:50,880 --> 00:30:52,400 Speaker 5: crypto and the ripple effects in a moment where one 610 00:30:52,480 --> 00:30:53,600 Speaker 5: hundred and two thousand move on. 611 00:30:53,800 --> 00:30:55,080 Speaker 3: The individual movers. 612 00:30:54,800 --> 00:30:56,800 Speaker 5: That you've got to keep your eyes trained on have been, 613 00:30:56,840 --> 00:30:59,640 Speaker 5: of course some of the points contributors we've seen Meta 614 00:30:59,800 --> 00:31:02,320 Speaker 5: new record. As I mentioned, Apple has been doing significantly 615 00:31:02,360 --> 00:31:04,600 Speaker 5: well in video bounces back, but hardly at all. Up 616 00:31:04,640 --> 00:31:07,360 Speaker 5: two point eight percent after seventeen percent sell off, We're 617 00:31:07,400 --> 00:31:09,600 Speaker 5: only up about thirty billion. It will six hundred billion 618 00:31:09,640 --> 00:31:12,320 Speaker 5: yesterday Tesla of by two percent. Remember, earnings are almost 619 00:31:12,400 --> 00:31:14,920 Speaker 5: upon us Tesla tomorrow out of the gate. We're gonna 620 00:31:14,920 --> 00:31:16,760 Speaker 5: have the likes of Microsoft to chew on and Meta 621 00:31:16,840 --> 00:31:19,080 Speaker 5: as well. But coming up more on musks World and 622 00:31:19,160 --> 00:31:23,479 Speaker 5: questions around his approach to efficiency, because apparently it's rather 623 00:31:23,520 --> 00:31:50,560 Speaker 5: good at it, don't It would seem this is gluebog technology. 624 00:31:52,440 --> 00:31:55,360 Speaker 5: Elil musks X has announced a new partnership with Visa. 625 00:31:55,440 --> 00:31:57,600 Speaker 5: The social media platform is tapping the payment company to 626 00:31:57,720 --> 00:32:00,680 Speaker 5: enable digital wallets on its app and website, in what 627 00:32:00,720 --> 00:32:02,920 Speaker 5: marks the first step for the company to create the 628 00:32:02,920 --> 00:32:04,040 Speaker 5: so called Everything app. 629 00:32:04,320 --> 00:32:04,480 Speaker 3: Now. 630 00:32:04,480 --> 00:32:07,080 Speaker 5: The move comes as bankers are now looking to offload 631 00:32:07,120 --> 00:32:10,360 Speaker 5: some free billion dollars of X's debt. According to sources, 632 00:32:10,360 --> 00:32:12,800 Speaker 5: Pimco Apollo are now said to be among the asset 633 00:32:12,800 --> 00:32:15,200 Speaker 5: managers looking to purchase a portion of the debt being 634 00:32:15,240 --> 00:32:17,600 Speaker 5: sold by group of banks at by Morgan Stanley. Some 635 00:32:17,680 --> 00:32:21,840 Speaker 5: interesting sweetness in that offering too, and staying with Elil Musk, 636 00:32:22,080 --> 00:32:25,600 Speaker 5: Many are now scrutinizing the tech billionaires approach to efficiency 637 00:32:26,040 --> 00:32:27,160 Speaker 5: with his own companies. 638 00:32:27,200 --> 00:32:29,120 Speaker 3: This as he gears up to cut spending and waste 639 00:32:29,120 --> 00:32:29,360 Speaker 3: in the. 640 00:32:29,400 --> 00:32:33,400 Speaker 5: US government with of course, the Dose initiative for mores Cretrudel, 641 00:32:33,520 --> 00:32:36,000 Speaker 5: it's a perfect person to discuss as to how good 642 00:32:36,080 --> 00:32:38,280 Speaker 5: is mascut efficiency in the private sector. 643 00:32:39,800 --> 00:32:42,200 Speaker 18: Well, I think it's it's worth kind of, you know, 644 00:32:42,240 --> 00:32:45,560 Speaker 18: taking a look at his track record and maybe sort 645 00:32:45,600 --> 00:32:47,920 Speaker 18: of you know, being open to the idea that there's 646 00:32:48,160 --> 00:32:52,240 Speaker 18: maybe some more similarities in how he has run his 647 00:32:52,320 --> 00:32:56,800 Speaker 18: companies and you know, how his criticism of the government 648 00:32:57,480 --> 00:33:02,200 Speaker 18: is you know, characterized, then then he is possibly let on. 649 00:33:02,320 --> 00:33:05,680 Speaker 18: You know, this is it's been the case that Tesla, 650 00:33:05,760 --> 00:33:09,160 Speaker 18: for example, you know, took a good decade before it 651 00:33:09,200 --> 00:33:13,040 Speaker 18: was making its investors any money. Uh And and you 652 00:33:13,080 --> 00:33:16,080 Speaker 18: know there was plenty of waste, you know, even according 653 00:33:16,120 --> 00:33:19,960 Speaker 18: to him, or inefficiency according to him within the company 654 00:33:20,600 --> 00:33:23,120 Speaker 18: that was sort of allowed to. 655 00:33:22,640 --> 00:33:24,880 Speaker 7: Fester for some time before he turned things around. 656 00:33:24,920 --> 00:33:27,560 Speaker 18: And so you know, no one's disputing that Tesla and 657 00:33:27,600 --> 00:33:31,680 Speaker 18: SpaceX aren't you know, quite efficient and now quite you know, 658 00:33:31,800 --> 00:33:35,800 Speaker 18: lucrative companies for their shareholders. But you can find, you know, 659 00:33:35,880 --> 00:33:40,360 Speaker 18: your fair share of inefficiency or waste within his own companies. 660 00:33:40,400 --> 00:33:43,200 Speaker 18: And you know, even on his own earnings calls, you know, 661 00:33:43,280 --> 00:33:47,200 Speaker 18: him sort of talking about ways in which his companies 662 00:33:47,320 --> 00:33:50,960 Speaker 18: haven't necessarily been as efficient as you would expect from 663 00:33:51,000 --> 00:33:54,280 Speaker 18: somebody who's about to run run Doge. 664 00:33:55,080 --> 00:34:00,360 Speaker 4: Craig, your reporting pointed to some pretty visible example of 665 00:34:00,480 --> 00:34:04,240 Speaker 4: maybe a pro fligod approach flying tires from the Czech 666 00:34:04,280 --> 00:34:07,400 Speaker 4: Republic to the US to ensure their delivery that had 667 00:34:07,400 --> 00:34:10,480 Speaker 4: to have been expensive. How does that kind of approach 668 00:34:10,640 --> 00:34:15,080 Speaker 4: translate to to government where the consequences and stakeholders are 669 00:34:15,320 --> 00:34:16,000 Speaker 4: very different. 670 00:34:17,520 --> 00:34:21,440 Speaker 18: I think what what Denni's Danna Hole in San Francisco, uh, 671 00:34:21,600 --> 00:34:24,319 Speaker 18: you know, pointed out in this story is essentially that 672 00:34:24,400 --> 00:34:28,920 Speaker 18: you know, there's there's a willingness to sort of allow 673 00:34:29,040 --> 00:34:33,400 Speaker 18: for you know, some inefficiency if it means sort of uh, 674 00:34:33,480 --> 00:34:36,920 Speaker 18: you know, other deliverables and uh, you know within the government. 675 00:34:37,000 --> 00:34:39,919 Speaker 18: I think you know, the answer that we're clearly going 676 00:34:39,960 --> 00:34:42,799 Speaker 18: to get is you know, something different in terms of 677 00:34:43,160 --> 00:34:46,400 Speaker 18: you know, not so much forgiveness of Okay, yes we 678 00:34:46,440 --> 00:34:49,640 Speaker 18: were inefficient, but we delivered X or y uh. 679 00:34:49,840 --> 00:34:51,760 Speaker 7: You know what we're what we're seeing from. 680 00:34:51,640 --> 00:34:54,640 Speaker 18: Doze early on is is this you know, relatively new 681 00:34:54,800 --> 00:34:57,360 Speaker 18: X account that is just you know, kind of highlighting 682 00:34:57,400 --> 00:34:59,560 Speaker 18: all the ways that the government is you know, so 683 00:34:59,760 --> 00:35:03,000 Speaker 18: as of blowing taxpayer dollars and all the ways in 684 00:35:03,040 --> 00:35:05,480 Speaker 18: which you know, Musk and his team are coming in 685 00:35:05,920 --> 00:35:09,479 Speaker 18: and sort of you know, writing those wrongs in his view. 686 00:35:09,560 --> 00:35:14,000 Speaker 18: And so I think, you know, he's not showing a 687 00:35:14,000 --> 00:35:17,239 Speaker 18: whole lot of willingness to you know, cut slack in 688 00:35:17,320 --> 00:35:20,359 Speaker 18: terms of trying different things or you know, trying things 689 00:35:20,360 --> 00:35:22,839 Speaker 18: and then not going well as he maybe has been 690 00:35:22,920 --> 00:35:26,640 Speaker 18: in his sort of you know times as CEO of 691 00:35:26,680 --> 00:35:27,560 Speaker 18: many of these companies. 692 00:35:28,800 --> 00:35:30,520 Speaker 7: Bloomberg's Craig Trudell, thank you. 693 00:35:30,920 --> 00:35:33,600 Speaker 4: Turning to crypto, deep seek sell off pose a threat 694 00:35:33,640 --> 00:35:38,000 Speaker 4: across all markets, including crypto, with bitcoin seeing its biggest 695 00:35:38,040 --> 00:35:40,520 Speaker 4: intra day drop in more than a month. All this 696 00:35:40,640 --> 00:35:44,040 Speaker 4: comes just as President Trump signed an executive order last 697 00:35:44,080 --> 00:35:46,799 Speaker 4: week calling for the creation of a White House Advisory 698 00:35:46,840 --> 00:35:50,640 Speaker 4: Group on digital currencies. Joining us now to discuss all 699 00:35:50,680 --> 00:35:54,360 Speaker 4: of this is Melton Demr's Crucible Capital Group General partner 700 00:35:54,440 --> 00:35:56,440 Speaker 4: and founder, Melton we. 701 00:35:56,440 --> 00:35:57,920 Speaker 7: Really have to get right to it. 702 00:35:58,000 --> 00:36:01,480 Speaker 4: We were talking about Elon Musk Department of Government efficiency. 703 00:36:01,640 --> 00:36:05,120 Speaker 4: One of the biggest questions facing your industry is regulation. 704 00:36:05,680 --> 00:36:08,640 Speaker 4: David Sachs, the new AI and cryptos are what are 705 00:36:08,680 --> 00:36:11,880 Speaker 4: you looking for him and the Trump team to clear away? 706 00:36:13,200 --> 00:36:15,600 Speaker 19: Well, I think I'm going to say something controversial. 707 00:36:15,880 --> 00:36:18,759 Speaker 20: The last four years by an administration, the lack of 708 00:36:18,840 --> 00:36:22,800 Speaker 20: clarity on regulatory policy in many ways I think was 709 00:36:22,840 --> 00:36:25,719 Speaker 20: actually constructive for the crypto industry in the sense that 710 00:36:26,320 --> 00:36:30,080 Speaker 20: there was not a lot of external pressure. Clarity I 711 00:36:30,080 --> 00:36:32,600 Speaker 20: think can be really challenging for markets. 712 00:36:32,800 --> 00:36:34,560 Speaker 19: Yes, there are obviously bright spots. 713 00:36:34,600 --> 00:36:37,960 Speaker 20: The rollback of SAB one twenty one will now allow 714 00:36:38,280 --> 00:36:40,680 Speaker 20: banks to hold crypto on their balance sheet, make it 715 00:36:40,760 --> 00:36:45,719 Speaker 20: easier for institutions to hold crypto. Sure, that's great, but clarity, 716 00:36:45,760 --> 00:36:50,239 Speaker 20: I think also creates more perspective on what will work 717 00:36:50,280 --> 00:36:52,120 Speaker 20: and what won't work in crypto. 718 00:36:52,239 --> 00:36:54,719 Speaker 19: So far, the view has been number go up. 719 00:36:54,800 --> 00:36:54,960 Speaker 13: Right. 720 00:36:55,000 --> 00:36:57,759 Speaker 20: We sometimes joke that coin as number go up technology, 721 00:36:58,160 --> 00:37:00,160 Speaker 20: and there's a meme in the industry. 722 00:37:00,160 --> 00:37:01,279 Speaker 19: Me we're all going to make it. 723 00:37:01,760 --> 00:37:03,959 Speaker 20: I think what's happening in week one of the Trump 724 00:37:04,000 --> 00:37:08,640 Speaker 20: administration is this clarity is helping people and particularly markets 725 00:37:08,719 --> 00:37:11,600 Speaker 20: understand that we are not all going to make it. 726 00:37:11,680 --> 00:37:14,000 Speaker 20: They're going to be winners and they're going to be losers. 727 00:37:14,280 --> 00:37:17,359 Speaker 20: And this administration has made it very clear that they're 728 00:37:17,400 --> 00:37:19,840 Speaker 20: perfectly content picking winners and losers. 729 00:37:20,760 --> 00:37:23,040 Speaker 3: AI coin is going to be winners in the future. 730 00:37:23,080 --> 00:37:26,560 Speaker 5: There's been this Steve tailing of the AI trade within crypto, 731 00:37:26,840 --> 00:37:29,000 Speaker 5: and they've taken a brutal hit of course, in the 732 00:37:29,080 --> 00:37:29,720 Speaker 5: CAR narrative. 733 00:37:30,600 --> 00:37:33,120 Speaker 20: Now, look, I think the AI crypto narrative is a 734 00:37:33,160 --> 00:37:36,480 Speaker 20: confusing one. Two things I look at One, how does 735 00:37:36,560 --> 00:37:40,040 Speaker 20: crypto make AI better or safer? And the jury is 736 00:37:40,080 --> 00:37:43,000 Speaker 20: out there and then the reverse is how does. 737 00:37:42,920 --> 00:37:45,880 Speaker 19: AI make crypto better or safer? 738 00:37:46,400 --> 00:37:50,440 Speaker 20: There I think we've used AI to create new casinos 739 00:37:50,520 --> 00:37:54,279 Speaker 20: in crypto in the form of agents or online accounts 740 00:37:54,600 --> 00:37:57,920 Speaker 20: tied to lms that are creating their own coins and 741 00:37:58,000 --> 00:37:59,200 Speaker 20: launching their own coins. 742 00:37:59,560 --> 00:38:01,520 Speaker 19: To me, that's not what's interesting. 743 00:38:01,800 --> 00:38:04,759 Speaker 20: What I'm looking at is how can crypto and some 744 00:38:04,920 --> 00:38:10,600 Speaker 20: of the opportunities around aggregation, optimization, and financialization help crypto 745 00:38:10,800 --> 00:38:13,799 Speaker 20: make AI better. And then I'm really looking at what 746 00:38:13,880 --> 00:38:17,680 Speaker 20: Apple is doing smaller models run locally on devices I 747 00:38:17,680 --> 00:38:21,600 Speaker 20: think is very interesting opportunity for crypto. Obviously, the big 748 00:38:21,719 --> 00:38:24,319 Speaker 20: story around deep seek is what's going to happen to 749 00:38:24,360 --> 00:38:27,399 Speaker 20: all of this energy and compute capex and I think 750 00:38:27,480 --> 00:38:29,680 Speaker 20: there are some of the early efforts we're seeing in 751 00:38:29,719 --> 00:38:34,080 Speaker 20: crypto to aggregate the long tail of compute into these 752 00:38:34,160 --> 00:38:37,320 Speaker 20: marketplaces where you can pay in stable coins or dollars. 753 00:38:37,520 --> 00:38:38,520 Speaker 19: Is really interesting. 754 00:38:38,719 --> 00:38:41,320 Speaker 3: Oh, guys, all the way back to file coin of old. 755 00:38:41,440 --> 00:38:42,879 Speaker 3: But I'm interested. 756 00:38:42,560 --> 00:38:47,920 Speaker 5: In Melton in what the regulatory spirits have meant underlying 757 00:38:48,120 --> 00:38:51,680 Speaker 5: all of what is sent meant and felt like to 758 00:38:51,800 --> 00:38:54,600 Speaker 5: go up on bitcoin, What has the trading told you. 759 00:38:55,719 --> 00:38:57,520 Speaker 19: Yeah, if we look at markets right. 760 00:38:57,840 --> 00:39:00,439 Speaker 20: What I think is always so interesting is you see 761 00:39:00,440 --> 00:39:03,680 Speaker 20: a lot of sentiment online and it's easy to say something, 762 00:39:03,760 --> 00:39:05,760 Speaker 20: but if you want to know what people are truly thinking, 763 00:39:05,800 --> 00:39:07,640 Speaker 20: you have to look at markets, and you have to. 764 00:39:07,560 --> 00:39:08,320 Speaker 19: Look at flows. 765 00:39:08,680 --> 00:39:11,440 Speaker 20: The biggest week of inflows we had in the last 766 00:39:11,520 --> 00:39:14,280 Speaker 20: year into crypto ETFs, right, which were a great proxy 767 00:39:14,719 --> 00:39:17,800 Speaker 20: four million dollars the week that Trump won the election. Okay, 768 00:39:18,120 --> 00:39:20,920 Speaker 20: last week with all of the regulatory quote unquote clarity 769 00:39:21,040 --> 00:39:24,200 Speaker 20: or at least setting of directions with Trump administration and 770 00:39:24,200 --> 00:39:26,160 Speaker 20: the executive orders being signed. 771 00:39:26,239 --> 00:39:27,440 Speaker 19: Two billion in inflows. 772 00:39:27,640 --> 00:39:30,840 Speaker 20: So the expectation is always better than the reality. 773 00:39:31,040 --> 00:39:33,560 Speaker 19: Right. Markets not reacting positively. 774 00:39:33,960 --> 00:39:38,120 Speaker 20: CM Bitcoin futures contract had the biggest drop in open 775 00:39:38,160 --> 00:39:41,839 Speaker 20: interest in its entire trading history yesterday, So it is 776 00:39:42,000 --> 00:39:46,640 Speaker 20: very clear that I'm crypto traders. Markets are feeling overextended, 777 00:39:46,880 --> 00:39:49,000 Speaker 20: just like they did on the big AI names. 778 00:39:48,680 --> 00:39:49,840 Speaker 19: And there's been a pullback. 779 00:39:50,080 --> 00:39:52,200 Speaker 20: The question is how much of that is going to 780 00:39:52,239 --> 00:39:54,719 Speaker 20: come back with clarity, and how much of that is 781 00:39:54,760 --> 00:39:58,000 Speaker 20: going to come back when we actually start to see reality. 782 00:39:57,600 --> 00:40:01,640 Speaker 5: Catch up with hype trade the room of sell the news, 783 00:40:01,640 --> 00:40:04,080 Speaker 5: Malton damares, it's so good to have you on Creaseable 784 00:40:04,080 --> 00:40:05,480 Speaker 5: Capital Group general partner. 785 00:40:06,000 --> 00:40:06,600 Speaker 3: We thank you. 786 00:40:14,920 --> 00:40:17,680 Speaker 4: Meta is set to report fourth quarter earnings after the 787 00:40:17,719 --> 00:40:20,359 Speaker 4: bell tomorrow. For more of what we can expect from 788 00:40:20,360 --> 00:40:23,880 Speaker 4: the social media giant, Bloomberg's Kurt Wagner joins us Now, 789 00:40:24,239 --> 00:40:28,320 Speaker 4: Kurt Metta somehow escaped the big sell off from deep 790 00:40:28,360 --> 00:40:31,439 Speaker 4: Seek yesterday. Why is that and what does that tell 791 00:40:31,520 --> 00:40:33,160 Speaker 4: us going into the results tomorrow? 792 00:40:34,560 --> 00:40:37,239 Speaker 21: Yeah, it might be two things. I think one is, 793 00:40:37,920 --> 00:40:39,759 Speaker 21: you may recall, Mike, at the end of last week, 794 00:40:39,800 --> 00:40:42,160 Speaker 21: they announced their big plan for the year, all the 795 00:40:42,200 --> 00:40:45,360 Speaker 21: spending they were going to be doing on AI. You know, 796 00:40:45,640 --> 00:40:48,800 Speaker 21: they announced that Threads was going to start running ads. 797 00:40:48,840 --> 00:40:50,920 Speaker 21: Like they kind of bront run to their own earnings 798 00:40:50,920 --> 00:40:53,799 Speaker 21: a little bit with some of the biggest news that 799 00:40:53,880 --> 00:40:56,480 Speaker 21: they have planned for twenty twenty five. And so I 800 00:40:56,520 --> 00:40:58,439 Speaker 21: think maybe some people just saw a little bit less 801 00:40:58,480 --> 00:41:00,560 Speaker 21: uncertainty with Meta because some of that was out there. 802 00:41:00,640 --> 00:41:03,800 Speaker 21: I think the second is that this deep Seek stuff 803 00:41:04,040 --> 00:41:07,000 Speaker 21: in a way sort of validates the AI strategy that 804 00:41:07,040 --> 00:41:09,000 Speaker 21: Meta has been pushing towards this whole time. You know, 805 00:41:09,080 --> 00:41:12,439 Speaker 21: Mark Zuckerberg has been arguing for more than a year 806 00:41:12,880 --> 00:41:15,440 Speaker 21: that where this was going was open source, and that 807 00:41:15,880 --> 00:41:17,719 Speaker 21: is where Meta was going to go as well, was 808 00:41:17,760 --> 00:41:20,799 Speaker 21: to create an open source model that other people could 809 00:41:20,800 --> 00:41:23,040 Speaker 21: build on. Now we saw that is what deep Seek 810 00:41:23,520 --> 00:41:25,960 Speaker 21: is building as well. So there's obviously more competition for 811 00:41:26,040 --> 00:41:28,680 Speaker 21: Meta here, but I think the strategy that they were 812 00:41:28,719 --> 00:41:31,920 Speaker 21: employing might be seen as the best way forward here. 813 00:41:31,920 --> 00:41:34,759 Speaker 21: So perhaps some investors were less spooked because they thought 814 00:41:34,760 --> 00:41:36,759 Speaker 21: the road ahead for Meta made a little bit more 815 00:41:36,800 --> 00:41:37,800 Speaker 21: sense than some of the others. 816 00:41:37,880 --> 00:41:41,320 Speaker 5: Gene Munster deep Water Asset Management reflected that exact issue, 817 00:41:41,360 --> 00:41:43,800 Speaker 5: so too did City, and they're also thinking that maybe 818 00:41:43,920 --> 00:41:47,080 Speaker 5: the AI models becoming cheaper and cheaper are also going 819 00:41:47,120 --> 00:41:50,799 Speaker 5: to just really push on the overll profitability the advertising improvements. 820 00:41:50,840 --> 00:41:52,040 Speaker 5: Is that what we're going to have to hear the 821 00:41:52,160 --> 00:41:55,600 Speaker 5: our return on AI investment from Mark as well as 822 00:41:55,680 --> 00:41:57,840 Speaker 5: his investment going forward in Capex. 823 00:41:59,120 --> 00:42:01,239 Speaker 21: Yeah, this has been one of the biggest questions, not 824 00:42:01,320 --> 00:42:03,680 Speaker 21: just for Meta but for all these tech companies is 825 00:42:04,000 --> 00:42:05,319 Speaker 21: when is all this investment going. 826 00:42:05,239 --> 00:42:05,799 Speaker 13: To pay off? 827 00:42:06,120 --> 00:42:08,600 Speaker 21: I think in twenty twenty four, actually Medaid probably a 828 00:42:08,600 --> 00:42:10,479 Speaker 21: better job than a lot of its peers that sort 829 00:42:10,480 --> 00:42:13,799 Speaker 21: of conveying where AI was impacting the product right, not 830 00:42:13,840 --> 00:42:18,440 Speaker 21: only in making the ads more efficient, but in dispersing 831 00:42:18,480 --> 00:42:21,040 Speaker 21: their AI assistant across all of their different apps. Having 832 00:42:21,040 --> 00:42:23,440 Speaker 21: the ray Band smart glasses right, they had sort of 833 00:42:23,480 --> 00:42:26,960 Speaker 21: tangible products that they could roll out and show people. Obviously, 834 00:42:26,960 --> 00:42:28,640 Speaker 21: with this massive investment they're going to be making in 835 00:42:28,680 --> 00:42:30,520 Speaker 21: twenty twenty five, it will be even more important for 836 00:42:30,560 --> 00:42:31,920 Speaker 21: them to continue to show that. 837 00:42:32,320 --> 00:42:33,080 Speaker 13: But given what they did. 838 00:42:33,040 --> 00:42:35,120 Speaker 21: In twenty twenty four, at least it seems like the 839 00:42:35,160 --> 00:42:37,160 Speaker 21: street sort of knows what to expect from them a 840 00:42:37,200 --> 00:42:39,120 Speaker 21: little bit. So we'll see if they're able to continue 841 00:42:39,120 --> 00:42:40,040 Speaker 21: that tomorrow. 842 00:42:40,120 --> 00:42:43,680 Speaker 5: For earnings, shares at a record high, currently trading six 843 00:42:43,800 --> 00:42:46,000 Speaker 5: hundred and seventy seven kut Wagner, We thank you. That 844 00:42:46,080 --> 00:42:48,440 Speaker 5: does it for this edition of Bloomberg Technology. You do 845 00:42:48,520 --> 00:42:50,840 Speaker 5: not want to forget a podcast. You can find it 846 00:42:50,880 --> 00:42:52,960 Speaker 5: on the terminal as well as online on Apple, Spotify, 847 00:42:52,960 --> 00:42:53,560 Speaker 5: and iHeart. 848 00:42:54,080 --> 00:42:55,360 Speaker 3: This is Bloomberg Technology.