WEBVTT - AMD Inks Chip Deal With OpenAI That Triggers Explosive Rally 0:00:02.520 --> 0:00:10.520 Bloomberg Audio Studios, Podcasts, radio News. This is Bloomberg business 0:00:10.520 --> 0:00:14.400 Week Daily reporting from the magazine that helps global leaders 0:00:14.440 --> 0:00:18.360 stay ahead with insight on the people, companies, and trends 0:00:18.360 --> 0:00:23.360 shaping today's complex economy. Plus global business finance and tech 0:00:23.440 --> 0:00:27.320 news as it happens. The Bloomberg Business Week Daily Podcast 0:00:27.640 --> 0:00:32.479 with Carol Masser and Tim Stenebek on Bloomberg Radio and 0:00:32.640 --> 0:00:33.199 of course. 0:00:33.000 --> 0:00:34.240 Co host of Bloomberg Tech. 0:00:34.440 --> 0:00:37.520 Joining us now from San Francisco where he is at 0:00:37.680 --> 0:00:41.519 open AI's developer Day. Hey, Ed, Huh, feels like we 0:00:41.560 --> 0:00:43.959 keep getting these big announcements. We're all getting a bit 0:00:44.040 --> 0:00:46.720 wary of them. What feels like this circular finance that 0:00:46.800 --> 0:00:49.960 we've talked with you about. Among those involved in the 0:00:49.960 --> 0:00:52.479 AI spend and build out, I heard Lisa sussay a 0:00:52.479 --> 0:00:53.400 win for AMD. 0:00:53.240 --> 0:00:55.600 A win for open Ai, feels like a win for everybody. 0:00:55.720 --> 0:00:57.640 Is it really? Do we know for certain? 0:00:58.600 --> 0:00:58.800 Well? 0:00:58.840 --> 0:01:01.560 She also said that she has confidence in open Ai, 0:01:01.880 --> 0:01:04.520 and the reason she said that is I had made 0:01:04.560 --> 0:01:07.720 the point to her that open ai still has to 0:01:07.760 --> 0:01:10.639 come up with the money for this. The difference between 0:01:10.680 --> 0:01:14.199 the open ai and AMD deal compared with the open 0:01:14.240 --> 0:01:17.960 Ai and Video deal as the structure, so it's almost 0:01:18.000 --> 0:01:23.000 the reverse. AMD will sign over stock to open Ai 0:01:23.680 --> 0:01:27.440 only when they've met specific milestones, both operational and financial. 0:01:27.520 --> 0:01:30.200 So in the first instance, open a has to fund, 0:01:30.319 --> 0:01:33.760 finance and build one gig awad of capacity. If they 0:01:33.800 --> 0:01:36.760 do that, AMD will give open Ai some stock, or 0:01:36.880 --> 0:01:39.880 rather technically, open ai has the right to buy a 0:01:39.959 --> 0:01:43.160 trantras shares at a penny apiece, but the burden is 0:01:43.200 --> 0:01:45.960 still on them. And so the point I was making 0:01:45.959 --> 0:01:49.160 with Lisa is have you seen into their books? Where 0:01:49.160 --> 0:01:51.080 are they going to get the money from to fund 0:01:51.160 --> 0:01:52.400 this to bring it online? 0:01:54.160 --> 0:01:54.800 Is it? 0:01:54.840 --> 0:01:57.440 Could this be weird on the circular financing thing that 0:01:57.880 --> 0:02:01.120 you know, in Vidia couldn't have this deal that could 0:02:01.160 --> 0:02:03.840 be worth up to one hundred billion dollars with open 0:02:03.880 --> 0:02:09.160 Ai in videos investment five billion dollars or will invest 0:02:09.200 --> 0:02:13.160 five billion dollars in Intel, And now you have what 0:02:13.280 --> 0:02:15.800 does that mean that in Vidia, if they have open 0:02:15.960 --> 0:02:19.040 Ai equity, then they're kind of an owner or could 0:02:19.040 --> 0:02:22.000 be an owner of AMD. Exactly. 0:02:22.680 --> 0:02:26.120 Yeah, it's much more fair to say that the Nvidia 0:02:26.200 --> 0:02:30.760 open Ai circumstance is circular financing because in Vidia commits 0:02:30.760 --> 0:02:34.280 to making a one hundred billion dollar investment into open Ai. 0:02:34.800 --> 0:02:36.640 We actually don't know if that's in the form of 0:02:36.720 --> 0:02:39.200 cash or if it would be like chips in lieu 0:02:39.240 --> 0:02:43.080 of cash, but they do take open Ai equity. In 0:02:43.120 --> 0:02:46.760 this case, AMD passes its own shares to open Ai, 0:02:47.280 --> 0:02:50.160 but not capital. And so if you're going to build 0:02:50.200 --> 0:02:53.640 six gigawats of capacity over many years, you know that 0:02:53.840 --> 0:02:56.280 is like the peak energy demand of like a major 0:02:56.400 --> 0:02:59.680 US city. We're talking lots of money up front in 0:02:59.720 --> 0:03:02.799 cap to expenditure, and open ay has got to do that. Now, 0:03:02.800 --> 0:03:05.960 we pose many questions to both Lisa sou and Open 0:03:05.960 --> 0:03:08.119 Eye present Greg Brockman, who was sat next to her, 0:03:08.400 --> 0:03:11.040 Will it be debt, will you issue more equity? Will 0:03:11.080 --> 0:03:13.480 you do something else? And he basically Greg Brockman said, 0:03:13.600 --> 0:03:15.760 we'll look at all those mechanisms and so there are 0:03:15.760 --> 0:03:18.200 many more questions still that we don't have answers to. 0:03:18.520 --> 0:03:20.800 Well, that's I guess. My thing is like, why don't 0:03:20.800 --> 0:03:22.720 these companies just do deals. 0:03:22.720 --> 0:03:24.840 I'll buy you chips, I'll do my bild, I'll do 0:03:24.919 --> 0:03:27.919 this thing the way we do it normally, rather than 0:03:28.360 --> 0:03:32.280 all of these big announcements with a lot of questions 0:03:32.320 --> 0:03:35.440 that feel like We don't really know that there's anything 0:03:35.520 --> 0:03:36.040 there there. 0:03:39.280 --> 0:03:44.000 The market is simply focused, irrespective of the mechanism for financing, 0:03:44.400 --> 0:03:46.880 on the revenue growth that comes out the other side, 0:03:47.000 --> 0:03:51.000 particularly like software names. And so that was the justification 0:03:51.080 --> 0:03:53.600 that Greg Brockman made. He said, the revenue will come. 0:03:53.920 --> 0:03:55.520 As you know, we've been writing a lot about the 0:03:55.560 --> 0:03:59.120 debt markets underpinning this, and debt investors feel the same way. 0:03:59.600 --> 0:04:01.880 All I say is there is also a technology story. 0:04:02.040 --> 0:04:03.600 You know, if you're going to build a data center 0:04:03.640 --> 0:04:05.400 and one of these providing the chips and the others 0:04:05.480 --> 0:04:08.400 is going to use the capacity for software, it does 0:04:08.480 --> 0:04:10.720 make sense to partner and work together to make the 0:04:10.760 --> 0:04:12.040 engineering make sense. 0:04:12.640 --> 0:04:15.840 If you're watching on Bloomberg TV or on youtubeer Bloomberg 0:04:15.880 --> 0:04:18.359 Originals got this great graphic up that kind of helps 0:04:18.400 --> 0:04:20.480 us keep up. This is by no means exhaustive, but 0:04:20.560 --> 0:04:24.159 shows where some of the money has gone in recent months. 0:04:24.160 --> 0:04:27.320 When it comes to the space, again, by no means exhaustive. 0:04:27.320 --> 0:04:29.800 There it's kind of like Carol, we're connecting. We're like 0:04:29.800 --> 0:04:33.159 out a whiteboard and we're kind of connecting pieces because 0:04:33.600 --> 0:04:35.560 to keep these things straight is pretty difficult. 0:04:35.640 --> 0:04:38.120 Well yeah, I mean that's the thing, ed. You know, 0:04:38.640 --> 0:04:40.720 we need to all be smart. We are smart, and 0:04:40.800 --> 0:04:42.640 you know we keep seeing these deals coming at us. 0:04:42.640 --> 0:04:45.679 We hear people saying, well when like a Meta Reports 0:04:45.720 --> 0:04:48.360 and some other you know, the big hyperscalers. We do 0:04:48.440 --> 0:04:52.200 see the impact of the AI spend, but I do 0:04:52.279 --> 0:04:55.600 wonder if it's going to carry through our economy like 0:04:55.640 --> 0:04:57.360 everyone expects. 0:04:59.160 --> 0:04:59.320 Well. 0:04:59.480 --> 0:05:02.919 Metas for example, they have a core legacy business that 0:05:03.040 --> 0:05:06.120 is advertising, and in most of the most recent quarters 0:05:06.160 --> 0:05:10.080 they've demonstrated significant top line growth and to a certain extent, 0:05:10.120 --> 0:05:13.719 bottom line growth, which is them saying we invested in AI. 0:05:14.160 --> 0:05:17.000 AI made our ads better and more valuable, we're making 0:05:17.040 --> 0:05:19.640 more money from it. With chat GPT, it's a lot 0:05:19.680 --> 0:05:21.640 harder to see that. But that's why here at develop 0:05:21.680 --> 0:05:24.200 a Day, it's really important. So the piece of news 0:05:24.240 --> 0:05:26.680 that came out from the keynote was they now have 0:05:26.800 --> 0:05:30.360 eight hundred million weekly users active users of chat GPT. 0:05:31.080 --> 0:05:33.960 The data point that market participants keep hitting me over 0:05:34.000 --> 0:05:37.080 the head with is how many of those free users 0:05:37.400 --> 0:05:41.000 are converted to being paid subscribers. Because unless you have 0:05:41.120 --> 0:05:44.039 that data, we can't know what the top line growth 0:05:44.080 --> 0:05:46.400 for open AI looks like. But look in the markets 0:05:46.480 --> 0:05:47.840 right now. I mean, you guys will have to tell me. 0:05:47.880 --> 0:05:49.479 I don't have a Bloomberg in front of me, but 0:05:49.560 --> 0:05:52.040 I saw a pool of about a dozen stocks that 0:05:52.160 --> 0:05:54.279 jumped in the last hour just because they got name 0:05:54.360 --> 0:05:58.840 checked by open ai here today. And it's by association. 0:05:59.040 --> 0:06:02.039 The idea that if using chat GPT you can access 0:06:02.080 --> 0:06:06.320 third party apps or third party company offerings, everyone's going 0:06:06.360 --> 0:06:06.880 to make money. 0:06:07.000 --> 0:06:07.159 You know. 0:06:07.200 --> 0:06:09.679 The market has a pretty simple logic in that respect. 0:06:09.839 --> 0:06:11.680 Yeah, for those who do have Bloombergs in front of them, 0:06:11.760 --> 0:06:15.239 just type in NI movers into the Bloomberg terminal. Cisco 0:06:15.320 --> 0:06:18.080 system shares up to session high during Open Air event. 0:06:18.200 --> 0:06:22.120 Figma shares Extend gained to sixteen percent after being mentioned 0:06:22.120 --> 0:06:25.520 at Open Ai event. Salesforce shares also jump up two 0:06:25.560 --> 0:06:28.760 point six percent during the Open Ai event. HubSpot, Zillo, 0:06:28.960 --> 0:06:32.320 Trip Advisor, Carol all moving higher as a result of 0:06:32.720 --> 0:06:34.480 what was talked about at the open ai event. 0:06:34.560 --> 0:06:36.320 And this goes back to ed. You know, you were 0:06:36.360 --> 0:06:41.920 trying to press we Sue, and you know in your 0:06:41.960 --> 0:06:45.760 conversation and I just right, and I just think about, like, 0:06:45.839 --> 0:06:48.360 what are the big questions that still remain for you? You 0:06:48.279 --> 0:06:50.360 want your you know, you're reporting on these ins and outs. 0:06:50.400 --> 0:06:53.080 You're there at this event. We think about open ai 0:06:53.200 --> 0:06:55.560 a lot. We talk about chat, GPT a lot. Who's 0:06:55.600 --> 0:06:58.120 to say that there isn't some other s, you know, 0:06:59.600 --> 0:07:02.000 chat that comes in and takes So I don't know 0:07:02.040 --> 0:07:04.279 if there's just a lot going on. What are the 0:07:04.360 --> 0:07:08.240 big questions that remain in a day when we get 0:07:08.240 --> 0:07:09.240 this big announcement. 0:07:09.360 --> 0:07:11.360 What wasn't answered for you? Still? 0:07:13.800 --> 0:07:17.120 Yeah, I think the financing, you know, I think in 0:07:17.240 --> 0:07:21.320 any corner of the Bloomberg newsroom, you've got to understand 0:07:21.320 --> 0:07:23.600 the debt markets because look at the role that debt 0:07:23.640 --> 0:07:28.200 played in previous financial crisis, the dot com bubble, what 0:07:28.280 --> 0:07:32.720 happened with Silicon Valley Bank, what happened in Asia earlier 0:07:32.720 --> 0:07:36.239 in the two thousands. I wasn't around then, but debt 0:07:36.280 --> 0:07:38.760 is a big underpinning here. And I think the signal 0:07:38.800 --> 0:07:41.440 that Greg Brockman gave us was they'll look at all mechanisms, 0:07:41.440 --> 0:07:44.080 including debt. So that's one part of it. On the 0:07:44.120 --> 0:07:47.800 other hand, this is I'm a technology journalist and technology matters. 0:07:48.600 --> 0:07:51.560 Open ai is going to use in AMD chips for inference. 0:07:51.680 --> 0:07:54.080 Inference is where you run the model not training, you 0:07:54.120 --> 0:07:58.280 are actually running people's queries and workloads. And if you 0:07:58.320 --> 0:08:00.840 look at all the movers that Timless did today, what 0:08:00.880 --> 0:08:03.960 open ai was basically saying is chat GPT is becoming 0:08:04.000 --> 0:08:07.040 an operating system. You use it for a number of functions, 0:08:07.040 --> 0:08:08.760 but you can also integrate it with all the other 0:08:08.800 --> 0:08:12.200 software you're using. And actually that's a pretty clear answer 0:08:12.280 --> 0:08:16.240 to some of the questions about will revenue be generated 0:08:16.280 --> 0:08:19.000 from this. People use all of these platforms for hours 0:08:19.040 --> 0:08:22.360 each day, and you know, the market seems to buy 0:08:22.440 --> 0:08:24.800 that if they can build the capacity of AMD in 0:08:24.840 --> 0:08:28.000 the timeline that they've said they will, they can actually 0:08:28.080 --> 0:08:29.280 start doing things of use. 0:08:29.680 --> 0:08:30.440 Isn't it wild? 0:08:30.440 --> 0:08:30.600 Though? 0:08:30.640 --> 0:08:33.320 Ryan Blestelica has this great story and the headline on 0:08:33.440 --> 0:08:35.640 it is open ai is fast becoming a whale in 0:08:35.679 --> 0:08:37.800 the stock market that it is shunned. I mean, open 0:08:37.840 --> 0:08:41.880 ai is not a publicly traded company, and yet Tim 0:08:42.000 --> 0:08:43.920 put down a bunch of names that are moving as 0:08:43.960 --> 0:08:48.080 a result of this this deal. Last week we saw 0:08:48.160 --> 0:08:51.319 open ai sending shares of e commerce companies like Shopify 0:08:51.440 --> 0:08:54.600 Etsy soaring after unveiling an instant by. 0:08:54.440 --> 0:08:58.199 Option in chat GPT. I mean, it is remarkable. 0:08:59.520 --> 0:09:04.080 Then the market moves that are are evolving around open ai, 0:09:04.320 --> 0:09:07.439 which again makes me say, what are we One of 0:09:07.520 --> 0:09:08.160 the things. 0:09:07.880 --> 0:09:08.720 That I do each day. 0:09:09.920 --> 0:09:11.839 Yeah, one of the things I do each day, and 0:09:11.880 --> 0:09:13.240 I know you guys do it as well. When I 0:09:13.240 --> 0:09:15.960 first long onto my Bloomberg, I do NI Globe rap 0:09:16.000 --> 0:09:18.360 and I look at the market's wrap because they always 0:09:18.360 --> 0:09:21.120 have a pretty good handle on the sentiment of the morning. 0:09:21.480 --> 0:09:25.320 And last Wednesday or Thursday, it was equity markets push 0:09:25.400 --> 0:09:29.400 higher on open ai enthusiasm. It was the confirmation of 0:09:29.559 --> 0:09:32.720 open AI's five hundred billion dollar valuation in the private 0:09:32.760 --> 0:09:36.439 markets that was driving sentiment in the public equity markets, 0:09:36.840 --> 0:09:39.560 and that was well sourced. It's based on lots of 0:09:39.559 --> 0:09:43.480 conversations with market participants. That is again a reflection of 0:09:43.480 --> 0:09:46.640 the names that Tim read out today. Open Ai, a 0:09:46.679 --> 0:09:50.440 private company, is driving all kinds of markets at the 0:09:50.440 --> 0:09:53.720 moment with what it is saying and to an equal extent, 0:09:53.800 --> 0:09:56.880 the activity it is doing in both hardware and software. 0:09:57.360 --> 0:09:59.480 Well, there's a lot of activity happening behind you. Ed 0:09:59.520 --> 0:10:02.160 at the op ai Developer Day conference. In just about 0:10:02.160 --> 0:10:04.040 an hour, we're going to hear from you again. You're 0:10:04.040 --> 0:10:06.960 going to be speaking with open Ai COO Brad Lightcap. 0:10:07.080 --> 0:10:08.840 You're going to have a conversation with him that I'm 0:10:08.840 --> 0:10:10.559 sure is going to touch on a lot of these themes. 0:10:10.840 --> 0:10:14.320 Before then, though, the reason all of these companies are moving, 0:10:14.559 --> 0:10:16.800 can you talk a little bit about this idea of 0:10:16.880 --> 0:10:20.160 open Ai, or rather chat GPT becoming an operating system 0:10:20.160 --> 0:10:23.880 and how these publicly traded third party companies are starting 0:10:23.880 --> 0:10:27.520 to use the technology to make their software to make 0:10:27.600 --> 0:10:29.720 whatever they're offering the services better. 0:10:32.320 --> 0:10:34.959 Yeah, for those listening on radio and watching on TV 0:10:35.040 --> 0:10:38.320 that are familiar with software, what OpenAI announced was very 0:10:38.400 --> 0:10:43.079 simple API access for several third parties to chat GPT, 0:10:43.559 --> 0:10:45.640 and they demonstrated it on stage. So you know, I 0:10:45.720 --> 0:10:48.520 use chat GPT for hours each day. But the idea 0:10:48.559 --> 0:10:51.800 is you link it to say, your Figma platform or 0:10:51.800 --> 0:10:55.199 your Spotify account even and you can use some of 0:10:55.240 --> 0:10:58.400 the features of those platforms within chat GPT. That was 0:10:58.440 --> 0:11:01.600 a sort of direct catalyst to moving. But what they 0:11:01.640 --> 0:11:04.120 also spoke about on stage, and Sam Altman did pretty 0:11:04.120 --> 0:11:07.640 good job of kind of explaining it, is that chat 0:11:07.720 --> 0:11:11.760 GPT is going to basically act much like the App 0:11:11.760 --> 0:11:15.000 Store does on OS. Right with Apple, you have a 0:11:15.000 --> 0:11:17.480 community of developers, those are all the people that are 0:11:17.520 --> 0:11:21.040 here today, the thousands of them that are building applications, 0:11:21.480 --> 0:11:23.920 and where will they publish those applications where they'll be 0:11:23.960 --> 0:11:27.440 reviewed by open ai and they will be integrated either 0:11:27.480 --> 0:11:31.280 as APIs or standalone applications within chat gpt. That is 0:11:31.320 --> 0:11:34.320 how the economy around the Android play Store, the Apple 0:11:34.360 --> 0:11:38.120 iOS store works. You have people build software, and then 0:11:38.160 --> 0:11:40.920 you have a place a library catalog where all of 0:11:40.920 --> 0:11:43.800 those things list. In some cases you charge an ad 0:11:43.800 --> 0:11:47.599 hoc subscription to use those applications. And that's kind of 0:11:47.640 --> 0:11:51.000 what open ai is saying here that chat GPT can 0:11:51.080 --> 0:11:55.480 be very interesting, although it's also contingent on the developer 0:11:55.520 --> 0:11:57.040 community stepping up and doing it. 0:11:57.360 --> 0:11:59.719 Who needs to be worried? Alphabet? Like, who needs to 0:11:59.720 --> 0:12:00.319 be work worried? 0:12:00.320 --> 0:12:03.560 As we watch this unfold and everything seemed to kind 0:12:03.559 --> 0:12:08.120 of revolve around chatchypt, open ai and as you said, like, 0:12:08.160 --> 0:12:11.160 if this is one of the new big search engines 0:12:11.200 --> 0:12:14.480 going forward, I mean Alphabet with its own search engines 0:12:14.520 --> 0:12:17.240 and making its own pivot, do they need to be worried? 0:12:17.280 --> 0:12:17.439 Though? 0:12:18.160 --> 0:12:18.280 Right? 0:12:20.040 --> 0:12:20.240 Yeah? 0:12:20.240 --> 0:12:21.880 I mean you guys, I actually should have checked, and 0:12:21.920 --> 0:12:24.600 I just didn't, But I don't know what Apple, Microsoft 0:12:24.960 --> 0:12:28.120 or Alphabet did during that keynote. I mean, did they 0:12:28.120 --> 0:12:28.920 get down or up? 0:12:29.040 --> 0:12:31.720 Alphabets right now up about two percent, so it's near 0:12:32.000 --> 0:12:33.520 its highs of the session, and. 0:12:33.520 --> 0:12:35.720 For your Microsoft's up two point four percent. Yeah, so 0:12:35.960 --> 0:12:37.400 I had a high of the session. 0:12:37.120 --> 0:12:42.120 All in right, Yeah, yeah, yeah, you know, I don't 0:12:42.160 --> 0:12:44.400 know the answer to that about whether they'd be worried. 0:12:44.480 --> 0:12:47.280 I mean, we've done a lot of reporting on open 0:12:47.320 --> 0:12:51.559 ai going its own way and having more independence from Microsoft. 0:12:51.559 --> 0:12:54.640 The original idea was that Microsoft would use open AI's 0:12:54.720 --> 0:12:58.760 underlying work in models to improve their own software capabilities. 0:12:58.920 --> 0:13:01.559 Of course, one of the recent elements is Microsoft is 0:13:01.679 --> 0:13:05.000 using Anthropic for its own in house software applications and 0:13:05.040 --> 0:13:07.640 has looked at one of open AI's rivals. But you 0:13:07.760 --> 0:13:10.640 are right, it's about open ai wanting to do more 0:13:10.920 --> 0:13:11.840 across software. 0:13:12.760 --> 0:13:12.960 You know. 0:13:13.040 --> 0:13:16.720 One of the other beneficiaries in the keynote was I 0:13:16.760 --> 0:13:19.840 think Salesforce. You can correct me to him if I'm wrong. 0:13:19.960 --> 0:13:22.400 I saw Salesforce as one of the movers service now 0:13:22.440 --> 0:13:26.400 as well, you know, it's just another sales channel in 0:13:26.440 --> 0:13:30.600 some sense. You can access another person software through chat, GPT, 0:13:31.240 --> 0:13:33.719 and you know, the market again its logic is very 0:13:33.760 --> 0:13:37.000 simple that it drives use and that's good. You know, 0:13:37.040 --> 0:13:38.640 when you have a partnership with open. 0:13:38.480 --> 0:13:41.520 Ai, salesforce shares up three percent right now, up as 0:13:41.600 --> 0:13:45.000 much as four point two percent during that keynote, Ed Ludlow, 0:13:45.120 --> 0:13:47.520 he's at open AI's Developer Day. 0:13:48.920 --> 0:13:52.040 Stay with us. More from Bloomberg Business Weekdaily coming up 0:13:52.280 --> 0:13:52.800 after this. 0:13:57.240 --> 0:14:01.120 You're listening to the Bloomberg Business Week Daily podcast. Catch 0:14:01.160 --> 0:14:04.360 us live weekday afternoons from two to five eastering. Listen 0:14:04.400 --> 0:14:08.000 on Applecarplay and Android Auto with the Bloomberg Business app, 0:14:08.160 --> 0:14:10.400 or watch us live on YouTube. 0:14:11.120 --> 0:14:13.679 Let's bring in Mandy Singh, global head of Tech research 0:14:13.720 --> 0:14:17.000 at Bloomberg Intelligency, joins us here in the Bloomberg Interactive 0:14:17.080 --> 0:14:20.600 Brokers studio. So I just want to start with the 0:14:20.920 --> 0:14:21.960 validation or. 0:14:21.880 --> 0:14:23.560 Not of AMD's technology here. 0:14:23.680 --> 0:14:26.520 Ed reminded us that this is an inference deal, not 0:14:26.600 --> 0:14:29.360 a training deal. But how should we look at this 0:14:29.440 --> 0:14:34.200 in the context of Nvidia being the standout leader when 0:14:34.240 --> 0:14:34.920 it comes to this. 0:14:34.960 --> 0:14:39.240 Tech Look, I mean for AMD, there weren't many options 0:14:39.680 --> 0:14:43.040 when you look at you know, Nvidia and Broadcom partnering 0:14:43.160 --> 0:14:47.240 with all the hyperscalers. AMD clearly has just two percent 0:14:47.520 --> 0:14:50.200 market share right now. And what do you do when 0:14:50.320 --> 0:14:53.760 Nvidia announces a big investment in open ai to the 0:14:53.800 --> 0:14:58.280 two zero hundred billion or multi years. And in this case, 0:14:58.320 --> 0:15:01.440 i think AMD had to do some thing. So I'm 0:15:01.480 --> 0:15:04.200 sure they gave a very sweet deal to open Ai 0:15:04.400 --> 0:15:07.120 in terms of, you know, the margins they're going to 0:15:07.240 --> 0:15:11.000 charge open Ai, the level of custom silicon work that 0:15:11.040 --> 0:15:14.360 they are willing to do for open Ai. And open 0:15:14.400 --> 0:15:16.880 ai has a great advantage that it can, you know, 0:15:16.960 --> 0:15:21.040 bargain and really get these sweet deals right now with everyone. 0:15:20.720 --> 0:15:23.920 A sweet deal meaning here's one hundred and sixty million 0:15:23.960 --> 0:15:26.720 shares of our company at a penny apiece. 0:15:26.640 --> 0:15:29.680 Maybe, And I look, I mean, I don't know. I 0:15:29.680 --> 0:15:32.640 mean this is nobody can look that far out. But 0:15:32.720 --> 0:15:36.080 the fact that Nvidia is willing to spend ten billion 0:15:36.200 --> 0:15:41.240 dollars you know in open Ai, and AMD gave them 0:15:41.280 --> 0:15:43.520 this option that they could buy the shares, you know, 0:15:43.640 --> 0:15:46.200 if it hits a certain level, that just goes to 0:15:46.320 --> 0:15:50.000 show that open ai wants to show that it has 0:15:50.040 --> 0:15:53.560 the power, not the chip makers. And so over time, 0:15:53.600 --> 0:15:56.960 if you look at you know, software cycles, the application 0:15:57.120 --> 0:16:00.200 is the sticky part. It's never the chips or the architecture, 0:16:00.360 --> 0:16:02.680 and that's what open the Eye is saying, look, guys, 0:16:03.000 --> 0:16:05.440 we have the best models, right, we have chat cept 0:16:05.600 --> 0:16:09.440 the application, so that's the end product, and they want 0:16:09.680 --> 0:16:12.280 diversification in terms of the inference capacity. 0:16:12.520 --> 0:16:14.680 Listen, we have this chart and for those who are 0:16:14.680 --> 0:16:17.440 on radio, it's just showing all of the deals that 0:16:17.520 --> 0:16:21.120 open ai is doing or some different companies are some 0:16:21.160 --> 0:16:24.120 of the deal them, whether it's Microsoft, although they've been 0:16:24.120 --> 0:16:27.120 pulling back and like lining up other deals. You know, 0:16:28.080 --> 0:16:32.120 I want We're trying to so hard be smart about 0:16:32.160 --> 0:16:34.320 this because it's easy to get caught up in it, 0:16:34.400 --> 0:16:35.160 and it does feel like. 0:16:35.120 --> 0:16:37.040 This has become increasingly a momentum trait. 0:16:37.120 --> 0:16:38.880 I'm not saying people aren't making money out of it, 0:16:38.960 --> 0:16:41.880 but how do we make sense because is there a 0:16:42.080 --> 0:16:47.640 chance that open ai and chatchept is not ultimately the 0:16:47.680 --> 0:16:52.200 big model or the big next generation search engine for everybody? 0:16:51.800 --> 0:16:53.360 Like, how do we be. 0:16:53.400 --> 0:16:57.280 Smart about this spend because some of it potentially a 0:16:57.320 --> 0:16:59.400 fluffy fluffy and not going to go anywhere. 0:16:59.480 --> 0:17:02.480 I know there are our references to circular financing and 0:17:02.520 --> 0:17:04.560 all that, but the best way to think about this 0:17:04.920 --> 0:17:09.240 is the concentration that you see among the players, whether 0:17:09.560 --> 0:17:13.280 it's you know, Nvidia being exposed to the hyperscalers, and 0:17:13.320 --> 0:17:16.560 now the number of hyperscalers are growing because open ai 0:17:17.119 --> 0:17:20.000 was dependent on Microsoft, like you said, now it's trying 0:17:20.040 --> 0:17:23.160 to become an independent hyperscale vendor. In fact, they said 0:17:23.200 --> 0:17:26.080 they may open their own data centers. They have a 0:17:26.080 --> 0:17:29.200 three hundred billion dollar deal with Oracle now, so they 0:17:29.240 --> 0:17:32.440 want to be a hyperscaler who does end to end 0:17:32.480 --> 0:17:35.280 work like Google does. It had its own chip, right 0:17:35.400 --> 0:17:38.560 it processes nine hundred and eighty trillion tokens per month, 0:17:39.000 --> 0:17:41.800 and that's what you know. Open ai wants to get 0:17:41.840 --> 0:17:44.840 to that they have their own chip. They use chips 0:17:44.840 --> 0:17:48.200 from AMD and Nvidia and not just rely on one provider. 0:17:48.359 --> 0:17:51.160 But they are still a private company. What are things 0:17:51.160 --> 0:17:55.879 that we might not know in their financials right now. 0:17:56.000 --> 0:17:58.960 That they have negative gross martins and so one of 0:17:59.000 --> 0:18:01.280 the way is that they are trying to fix that 0:18:01.600 --> 0:18:05.359 is by arbitraging against the semiconductor chip makers who are 0:18:05.400 --> 0:18:08.800 giving them the compute capacity. And we know that this stuff, 0:18:08.960 --> 0:18:11.960 like one of the things with Generative AI, it's very 0:18:11.960 --> 0:18:15.199 compute intensive, right, and it's not changing even though we 0:18:15.280 --> 0:18:18.760 have GPT five, Like, there's nothing that has changed in 0:18:18.840 --> 0:18:23.280 terms of compute efficiency. It still requires ten times more 0:18:23.280 --> 0:18:26.000 compute than your CPUs that were used for you know, 0:18:26.080 --> 0:18:29.280 traditional software. So it is an expensive technology. 0:18:29.840 --> 0:18:34.200 Okay, hold on, though, you said, you know, I don't Yeah, 0:18:34.200 --> 0:18:36.239 I don't want to push aside the circular financing thing 0:18:36.280 --> 0:18:38.960 because this is kind of blowing my mind here. Yeah, 0:18:39.160 --> 0:18:43.520 if in Vidia has a deal with or with open Ai, 0:18:43.560 --> 0:18:46.320 that's up to one hundred billion dollars. So let's say 0:18:46.359 --> 0:18:48.720 in Vidia gets equity in open Ai as a result 0:18:48.720 --> 0:18:52.560 of that. Yeah, then open Ai then gets equity in 0:18:52.640 --> 0:18:56.280 AMD as a result of this deal. Would that essentially 0:18:56.320 --> 0:18:59.920 mean that in Vidia is an owner of AMD? 0:19:00.320 --> 0:19:02.680 No, I think that's too far fair. I mean, think 0:19:02.760 --> 0:19:07.240 of companies having minority stakes. So there are instances where 0:19:07.320 --> 0:19:10.800 Google has a minority stake in Lift. You know, there's 0:19:10.840 --> 0:19:12.640 so many instances because. 0:19:12.320 --> 0:19:15.560 Google Ventures was an investor in it early on. In 0:19:15.600 --> 0:19:18.480 that's Google's venture capital army. You can see that within 0:19:19.080 --> 0:19:22.119 and in the past, within Google's maps application. It used 0:19:22.160 --> 0:19:24.800 to suggest, if it still does, it would suggest Lift rides. 0:19:24.840 --> 0:19:28.720 It made sense. Yeah, and so look in these competitors though. 0:19:28.800 --> 0:19:32.359 Because the numbers are getting so big, I mean you 0:19:32.600 --> 0:19:36.800 need financing from somewhere right now. It was the Hyperscalers, 0:19:36.840 --> 0:19:41.119 the four Hyperscalers were the biggest buyers of AI data 0:19:41.119 --> 0:19:45.160 center infrastructure capacity. They were the ones who were building it. 0:19:45.560 --> 0:19:49.240 And now it's like Opening Eyes saying we want to 0:19:49.280 --> 0:19:52.840 be in the business of building AI data centers because 0:19:52.880 --> 0:19:56.240 they look at Google and say they're running their data 0:19:56.240 --> 0:19:59.440 centers a lot more efficiently than we are, even though 0:19:59.440 --> 0:20:02.600 we haven't an signs with Microsoft. So Opening Eyes big 0:20:02.640 --> 0:20:05.679 bet is if I have more compute capacity, one, I 0:20:05.680 --> 0:20:09.399 can undercut everyone else in terms of offering my model, 0:20:09.760 --> 0:20:12.600 and I'll be more aggressive with pricing, and once I 0:20:12.680 --> 0:20:16.399 have the customer on board, I can sell them more services. 0:20:16.440 --> 0:20:19.760 But just being a model provider is not a good business. 0:20:20.080 --> 0:20:23.399 That's what open ai is suggesting that having the best 0:20:23.440 --> 0:20:27.360 model is not a good mode. Ultimately, everyone will catch 0:20:27.440 --> 0:20:30.120 up in terms of their model capacity, and then they 0:20:30.119 --> 0:20:31.840 will go across different models. 0:20:31.880 --> 0:20:33.280