WEBVTT - Meta to Deploy Homegrown Chips, Uber to Offer Zoox Rides 0:00:02.520 --> 0:00:13.520 Bloomberg Audio Studios, podcasts, radio news. Bloomberg Tech is alive 0:00:13.560 --> 0:00:17.360 from coast to coast, with Caroline Hyde in New York 0:00:17.640 --> 0:00:19.640 and Eva Low in San Francisco. 0:00:22.800 --> 0:00:24.320 This is Bloomberg Tech coming up. 0:00:24.360 --> 0:00:27.880 Oracle reports strong sales and issues an outlook that suggests 0:00:27.880 --> 0:00:30.000 little lead up in demand for AI computing. 0:00:30.040 --> 0:00:31.480 We break down the results. 0:00:31.280 --> 0:00:34.720 Plus Meta plans to deploy four new generations of its 0:00:34.720 --> 0:00:37.440 in house AI chips by the end of twenty twenty 0:00:37.440 --> 0:00:40.040 seven to help power its rapidly expanding AI workloads. 0:00:40.640 --> 0:00:44.000 And Uber will let customers hail robo taxis from Amazon 0:00:44.120 --> 0:00:45.519 zekes starting. 0:00:45.120 --> 0:00:46.920 In Las Vegas this summer. 0:00:46.960 --> 0:00:50.160 We'll hear from both companies in an exclusive interview later 0:00:50.240 --> 0:00:54.360 this hour. There's one big mover in technology, and it's Oracle, 0:00:54.400 --> 0:00:56.920 on track for its best days since September, up around 0:00:56.920 --> 0:00:58.080 eleven twelve percent. 0:00:58.560 --> 0:00:59.760 The numbers are very clear. 0:01:00.400 --> 0:01:03.280 Sales of ninety billion dollars for the fiscal year starting 0:01:03.280 --> 0:01:07.360 in June, capital expenditures holding at fifty billion dollars. The 0:01:07.440 --> 0:01:09.920 backlog is rising, and so the market's looking at this 0:01:10.000 --> 0:01:13.960 and saying, okay, the AI demand storing story is intact. 0:01:14.280 --> 0:01:17.160 Now it's about execution risk and credit risk, and how 0:01:17.240 --> 0:01:19.920 is Oracle going to manage cash to fund all of this. 0:01:20.200 --> 0:01:21.559 There's a lot more behind the numbers. 0:01:21.560 --> 0:01:23.200 Carry there is, and we can dig into them. 0:01:23.319 --> 0:01:25.920 And the results of Bloomberg's Brodie Ford that they did 0:01:26.040 --> 0:01:29.120 enough to stay off at any anxiety. They're sticking to 0:01:29.200 --> 0:01:32.480 just that fifty billion in capital expenditure, and the numbers 0:01:32.480 --> 0:01:36.200 thus far, particularly in the offering of cloud and the 0:01:36.240 --> 0:01:38.200 infrastructure side, really strong. 0:01:39.840 --> 0:01:42.400 Exactly right now. It's all about, as you both said, 0:01:42.400 --> 0:01:45.080 that execution risk. I mean, we have to remember that 0:01:45.200 --> 0:01:47.840 Oracle was kind of the poster child for Wall Street 0:01:47.880 --> 0:01:50.520 getting scared about all these big data centers not going 0:01:50.640 --> 0:01:54.080 quite right or the margins being upside down. But the 0:01:54.120 --> 0:01:56.720 results last night gave a lot of comfort, right They 0:01:56.760 --> 0:02:01.080 showed that things are progressing, They're recognizing revenue, the margin 0:02:01.200 --> 0:02:03.960 is reasonable, and so these are kind of all the 0:02:04.080 --> 0:02:06.920 messages that investors were hoping to hear from Oracle. 0:02:08.200 --> 0:02:09.880 Brady, take us inside the earning school. 0:02:10.040 --> 0:02:12.480 What was it like, you know, going into it, we 0:02:12.520 --> 0:02:16.920 had broken this story about Abilene not expanding from what 0:02:17.080 --> 0:02:20.640 was a set one point two gigawatts to potentially two gigawatts, 0:02:20.639 --> 0:02:23.040 and the markets reacted as far as I can tell 0:02:23.040 --> 0:02:25.119 on the transcripts, it didn't come up, but there were 0:02:25.160 --> 0:02:26.760 lots of questions about like how. 0:02:26.680 --> 0:02:28.040 Oracle was going to pull this off. 0:02:29.400 --> 0:02:32.360 Yeah, for sure. I think the big question underlying a 0:02:32.360 --> 0:02:35.560 lot of it is how are the big builds going right? 0:02:35.600 --> 0:02:38.240 And the indicator that some of these are going well 0:02:38.400 --> 0:02:41.600 is that the revenue is getting recognized. That means capacity 0:02:41.720 --> 0:02:44.440 is getting handed over to big customers like Open AI. 0:02:45.400 --> 0:02:48.040 The line in the transcript I thought was interesting was 0:02:48.080 --> 0:02:51.600 that their new co CEO, Clay mcglurk, said that ninety 0:02:51.639 --> 0:02:54.240 percent of what we delivered this quarter was on time 0:02:54.360 --> 0:02:57.360 or ahead of schedule. I think that was probably a 0:02:57.840 --> 0:03:00.200 kind of response to a lot of anxiety out there 0:03:00.280 --> 0:03:04.000 about potential delays or things kind of going sideways that 0:03:04.639 --> 0:03:08.120 despite the noise, we're handing over capacity just like we 0:03:08.160 --> 0:03:08.800 certainly would. 0:03:09.840 --> 0:03:12.079 Brody's got the Tech and Depth newsletter out about how 0:03:12.120 --> 0:03:15.840 Oracle's pulling every lever in its AI cash crunch. Really 0:03:15.840 --> 0:03:18.959 worth reading, Brody, Thank you very much. That's the story 0:03:19.040 --> 0:03:21.440 right now, right in markets. What is happening on the 0:03:21.480 --> 0:03:24.040 demand side, what is happening in the supply side, and 0:03:24.080 --> 0:03:26.840 what is happening in the cash supporting all of it. 0:03:27.080 --> 0:03:30.679 Matthew Weird Goldman Sachs Managing director joins us. Now, it's 0:03:30.720 --> 0:03:33.240 really interesting because for a very long time we were 0:03:33.320 --> 0:03:35.600 zeroed in on capital expenditures and that was the be 0:03:35.720 --> 0:03:37.800 all and the end all. There are stories on the 0:03:37.840 --> 0:03:42.120 Bloomberg terminal today about the credit profile and CDs for 0:03:42.160 --> 0:03:46.880 example on Oracle. Our investors starting to look deeper and 0:03:46.960 --> 0:03:49.520 at different data sets to assess the health of what's 0:03:49.560 --> 0:03:49.920 going on. 0:03:50.160 --> 0:03:50.360 Yeah. 0:03:50.400 --> 0:03:54.640 Absolutely, we look though at the structure of the national 0:03:54.680 --> 0:03:57.520 economy and we look at CAPPAX as a share of GDP. 0:03:57.720 --> 0:04:01.240 What's interesting to note is that capital formation as a 0:04:01.280 --> 0:04:04.040 sharf GDP is the same level as a percent as 0:04:04.040 --> 0:04:06.600 it was in twenty eighteen. So we've seen a tremendous 0:04:06.640 --> 0:04:09.880 amount of investment, especially in the tech inn AI space. 0:04:10.000 --> 0:04:13.200 From a national standpoint, it doesn't feel imbalance. Now rightly, 0:04:13.240 --> 0:04:16.680 investors are starting to wonder what is the return going 0:04:16.720 --> 0:04:19.400 to be on all of the investment going into AI. 0:04:19.920 --> 0:04:22.120 We think that there has been a lot of excitement, 0:04:22.279 --> 0:04:26.000 perhaps over exuberance in places, but clearly, as you mentioned 0:04:26.120 --> 0:04:28.360 ed that there is a lot of demand still and 0:04:28.360 --> 0:04:30.120 that's something that we think can continue to go. 0:04:30.640 --> 0:04:32.799 You have a very clear line which is the primary 0:04:32.839 --> 0:04:37.920 determinant in equity markets and equity prices is earnings percent. 0:04:38.400 --> 0:04:41.200 We just had an astonishingly detailed earning sprint from Oracle. 0:04:41.600 --> 0:04:43.240 What did you learn about the world from it? 0:04:43.520 --> 0:04:45.920 Yeah, well, I think going back to what you just said, 0:04:45.920 --> 0:04:49.000 that demand for AI capex is still very very high. 0:04:49.279 --> 0:04:52.280 When we think about what determines equity returns, whether you're 0:04:52.320 --> 0:04:55.039 looking at history since World War Two or even the 0:04:55.120 --> 0:04:56.960 last year. For the s and P five hundred and 0:04:56.960 --> 0:04:59.960 eighty to ninety percent of the return comes from earnings, 0:05:00.080 --> 0:05:03.080 and so the outlook for prices is really predicated on 0:05:03.560 --> 0:05:05.719 do you expect earnings to grow and we do think so. 0:05:05.839 --> 0:05:08.400 We think earnings growth for the SMP five hundred will 0:05:08.440 --> 0:05:12.400 be about ten percent this year. That is largely predicated 0:05:12.400 --> 0:05:14.920 on an economy that we think will be growing above trend. 0:05:14.960 --> 0:05:16.919 We're expecting two point four percent real. 0:05:16.760 --> 0:05:19.400 GDP growth versus trend growth for the US of about 0:05:19.440 --> 0:05:22.720 two percent that we think will generate that ten percent 0:05:23.240 --> 0:05:25.520 earnings growth. We think, though, in our base case, that 0:05:25.600 --> 0:05:28.560 the total return though for a SMP five hundred, will 0:05:28.600 --> 0:05:31.400 be about seven percent, which means not all that earnings 0:05:31.440 --> 0:05:34.599 growth makes its way into the total return, primarily because 0:05:34.640 --> 0:05:36.960 we think multiples will come down a little bit from 0:05:36.960 --> 0:05:38.239 their currently elevated levels. 0:05:38.240 --> 0:05:38.640 Interesting. 0:05:38.640 --> 0:05:40.640 Interesting because you've sort of been talking about the tech 0:05:40.680 --> 0:05:42.599 sect to not being in a bubble math. You talking 0:05:42.640 --> 0:05:46.200 about how valuations are relatively reasonable and Sunny Oracles come 0:05:46.279 --> 0:05:47.919 down some in terms of its own. 0:05:48.160 --> 0:05:50.200 Pe ratios, but taught. 0:05:50.080 --> 0:05:51.920 Us a little bit about the anxiety and the software 0:05:51.960 --> 0:05:55.800 part of the business. Oracle spoke and pushed back against that. 0:05:56.160 --> 0:05:58.000 Have you pushed back against that more broadly? 0:05:58.880 --> 0:06:01.279 Yeah, we do think that there should be some concern 0:06:01.320 --> 0:06:04.200 about the software sector. This, though, is reminding us of 0:06:04.240 --> 0:06:06.680 a few episodes historically. Whether it was the global financial 0:06:06.720 --> 0:06:10.480 crisis and there were concerns about the US financial sector. 0:06:10.200 --> 0:06:11.800 And existential concerns. 0:06:11.960 --> 0:06:13.760 If we go back to COVID, there are concerns about 0:06:13.760 --> 0:06:16.160 the office real estate sector and whether it was dead. 0:06:16.600 --> 0:06:18.120 We don't think this time is different. 0:06:18.200 --> 0:06:22.600 Investors unfortunately, tend to move in mass, especially when pessimism 0:06:22.800 --> 0:06:23.440 takes over. 0:06:23.960 --> 0:06:25.520 We think this time is no different. 0:06:25.680 --> 0:06:28.719 Reality is probably somewhere in between what we've seen in 0:06:28.760 --> 0:06:32.000 terms of the status quo and the existential concerns that 0:06:32.040 --> 0:06:35.200 investors have. We think of this more as an evolution 0:06:35.880 --> 0:06:38.840 of the sector. There's certainly a competitive risk, but we 0:06:38.920 --> 0:06:41.599 certainly don't think that this is an existential one where 0:06:41.600 --> 0:06:44.440 there's so much focus on the terminal value. We think 0:06:44.440 --> 0:06:49.039 these are absolutely very competitive companies that will continue to 0:06:49.200 --> 0:06:50.320 evolve over time. 0:06:50.920 --> 0:06:53.080 There was a lot of anxiety in the market initially 0:06:53.080 --> 0:06:55.760 about software, and then look, there was the geopolitical risk 0:06:55.800 --> 0:06:57.919 that we're now combating day in, day out for twelve 0:06:57.920 --> 0:07:01.200 state drips straight day over with the Middle East and 0:07:01.320 --> 0:07:04.360 conflict with Iran. Matthew, does that change in any way 0:07:04.400 --> 0:07:08.000 your perspective on investing into this moment in where you 0:07:08.080 --> 0:07:09.359 are in the market. 0:07:09.840 --> 0:07:12.360 Yeah, We're very focused on your on and it's terrible 0:07:12.400 --> 0:07:16.440 from a humanity standpoint. I think though, we would focus 0:07:16.520 --> 0:07:20.720 though on from an investment standpoint, that this resembles what 0:07:20.840 --> 0:07:24.080 typically happens when there's a geopolitical shock over the near term, 0:07:24.520 --> 0:07:27.720 there is a reaction by financial markets. What we've seen 0:07:27.880 --> 0:07:31.239 historically and specifically we've looked at the last forty years, 0:07:31.600 --> 0:07:34.240 there are twenty one air strikes in terms of the 0:07:34.400 --> 0:07:37.160 US in the Middle East, and we do see some 0:07:37.280 --> 0:07:39.880 patterns emerge. The first is that the initial reaction in 0:07:39.920 --> 0:07:43.320 markets is at risky assets like equities initially sell off, 0:07:43.720 --> 0:07:47.480 but then we see the dollar appreciate, we see oil rise, 0:07:47.600 --> 0:07:52.520 we see bond prices rise. But after about eight weeks, fundamentals, 0:07:52.560 --> 0:07:57.040 whether economic or corporate fundamentals, reassert themselves and the prior 0:07:57.200 --> 0:08:01.559 trend in financial markets resumes such that after eight weeks 0:08:01.640 --> 0:08:04.560 ninety five percent of the time, US equities are actually 0:08:04.640 --> 0:08:08.840 above their pre strike levels by an average of about 0:08:08.880 --> 0:08:12.880 four percent. So as a consequence, we haven't changed our views. 0:08:13.000 --> 0:08:16.080 We have a key investment tended at Golden sachs that 0:08:16.320 --> 0:08:20.480 history is a useful guide. The words this time is different, 0:08:20.520 --> 0:08:24.160 we think are very dangerous. We think that history will 0:08:24.160 --> 0:08:27.040 repeat itself here and eventually markets will resume their work. 0:08:27.080 --> 0:08:30.480 Trend every day trying to understand the long term impact 0:08:30.480 --> 0:08:34.200 of AI yes on inflation, YEP and labor markets. And 0:08:34.240 --> 0:08:36.400 I know the Goldman House cool is that AI may 0:08:36.440 --> 0:08:39.560 display a million jobs a year, but there is also 0:08:39.679 --> 0:08:43.199 job creation happening in parallel. Just what you said about 0:08:43.440 --> 0:08:46.439 how one might approach that as an investor from your desk, 0:08:46.520 --> 0:08:49.839 in your perspective, what happens in the short, medium and 0:08:49.920 --> 0:08:50.679 long term. 0:08:50.440 --> 0:08:53.400 So we think technological innovation is a feature and not 0:08:53.480 --> 0:08:56.160 a bug of the US equity market. In the US economy, 0:08:56.200 --> 0:08:58.840 the US economy is by far the most innovative globally. 0:08:58.880 --> 0:09:01.880 It's a key reason that we recommend our clients be 0:09:01.920 --> 0:09:06.320 strategically overweight US assets. Now, certainly AI is going to 0:09:06.360 --> 0:09:09.959 boost productivity, and unfortunately it is going to displace workers. 0:09:09.960 --> 0:09:12.880 But as you mentioned ed, there's a tremendous amount of 0:09:13.040 --> 0:09:16.400 new job creation each year. Bear with me some numbers here. 0:09:16.440 --> 0:09:18.240 There are one hundred and seventy million jobs in the 0:09:18.360 --> 0:09:21.480 US each year, they are about twenty five to thirty 0:09:21.520 --> 0:09:24.920 million jobs either lost or destroyed. But on the other 0:09:24.960 --> 0:09:28.040 side of that, and this is creative destruction, new jobs 0:09:28.080 --> 0:09:30.760 are created, and there are more new jobs created than 0:09:30.840 --> 0:09:31.760 jobs destroyed. 0:09:31.760 --> 0:09:32.800 So as we put that one. 0:09:32.720 --> 0:09:37.120 Million number into context, we think that a lot, if 0:09:37.160 --> 0:09:39.920 not all, of those jobs lost to AI will be 0:09:40.040 --> 0:09:42.960 replaced by new jobs. Think about the investment we've seen 0:09:43.040 --> 0:09:46.240 so far, whether it's the construction of data centers, think 0:09:46.280 --> 0:09:49.920 about rising financial market prices. That's increased net worth, and 0:09:49.960 --> 0:09:54.520 it's increased consumer consumption. And so as we think about 0:09:54.559 --> 0:09:57.040 the net impact longer term, we're bullish and we think 0:09:57.080 --> 0:09:59.560 that this is something that will be additive, even though 0:09:59.600 --> 0:10:01.840 over the show there will be some tough adjustments. 0:10:02.240 --> 0:10:03.880 Matthew, I'm Goldman Sachs. 0:10:04.240 --> 0:10:06.319 Great to have your expertise today. Thanks for coming in. 0:10:07.040 --> 0:10:11.040 While our next Meta plans to deploy homegrown chips handle 0:10:11.120 --> 0:10:13.480 AI workloads. I'll have the details next. 0:10:13.679 --> 0:10:31.640 This is Bloomberg Tech. This is Meta's chip lab in Fremont, California. 0:10:32.080 --> 0:10:35.479 Inside the company is developing the next generations of MTIA, 0:10:35.840 --> 0:10:39.160 short for Meta Training and Inference Accelerator. It's in house 0:10:39.200 --> 0:10:41.480 AI chip program. It's a long term effort to build 0:10:41.520 --> 0:10:45.080 the most efficient architecture for Meta's internal workloads, with four 0:10:45.160 --> 0:10:47.800 new generations of chips planned over the next two years, 0:10:47.840 --> 0:10:51.960 from ranking and recommendations to large scale gen AI inferants. 0:10:52.000 --> 0:10:53.840 When chips come in from the fab, this is where 0:10:53.840 --> 0:10:57.520 they're validated, tested at the chip rack and workload level 0:10:57.559 --> 0:11:01.480 before deployment into metas data centers. MTIA three hundred is 0:11:01.480 --> 0:11:05.920 already in production, supporting ranking and recommendations, training helping decide 0:11:06.040 --> 0:11:08.640 what shows up in your social feed Those go into 0:11:08.679 --> 0:11:12.640 liquid called servers like these. MTIA four hundred is moving 0:11:12.679 --> 0:11:17.560 towards deployment, expanding into broader AI workloads, including Genai Future 0:11:17.600 --> 0:11:21.920 versions four fifty and five hundred push further into Genai inference, 0:11:21.960 --> 0:11:25.120 with deployments planned in twenty twenty seven. The effort hasn't 0:11:25.120 --> 0:11:28.440 always moved as quickly as Mark Zuckerberger Meta had hoped. 0:11:28.640 --> 0:11:31.320 Meta has made some acquisitions, it has tried to make 0:11:31.360 --> 0:11:34.760 some others in an effort to strengthen its in house 0:11:34.840 --> 0:11:38.680 chip talent and accelerate progress. AI models are evolving faster 0:11:38.760 --> 0:11:41.720 than traditional chip cycles, so Meta is speeding up the 0:11:41.760 --> 0:11:46.080 design process, aiming to improve performance, cost and power efficiency 0:11:46.120 --> 0:11:49.040 at scale. At the same time, the company is striking 0:11:49.080 --> 0:11:52.400 major supply deals with leading chip makers, securing gigawats of 0:11:52.440 --> 0:11:56.000 AI computing capacity. The strategy is by compute at scale 0:11:56.000 --> 0:11:59.480 from Nvidia and AMD, but also use custom silicon, where 0:11:59.480 --> 0:12:02.880 metas works are uniquely its own, because in the AI race, 0:12:03.000 --> 0:12:05.600 it isn't just about the models, It's about the compute 0:12:05.640 --> 0:12:08.280 behind them. Okay, so we broke the story on this 0:12:08.320 --> 0:12:11.000 one this morning. The reaction in Meta shares kind of muted. 0:12:11.080 --> 0:12:16.040 Remember Meta still a big buyer of Nvidia and AMD GPUs. 0:12:16.400 --> 0:12:19.080 Let's get more on Meta's own chip plans the in 0:12:19.160 --> 0:12:22.720 house and custom silicon with Bluemos Riley Griffin who came 0:12:22.760 --> 0:12:25.200 with me to Fremont, and it was an interesting experience, right, 0:12:25.360 --> 0:12:29.760 Like this is both they want to do custom silicon 0:12:29.800 --> 0:12:32.120 because they have a lot of internal workloads, lots of 0:12:32.160 --> 0:12:34.600 AI inference, but they're going to have to keep buying 0:12:34.600 --> 0:12:35.800 from the big names as well. 0:12:36.200 --> 0:12:38.960 Yeah, it's a strategy of more is more is more, 0:12:39.080 --> 0:12:42.199 and that is for the chips, that is for gigawatts. 0:12:42.240 --> 0:12:47.319 That is a total diversification approach to securing that compute 0:12:47.640 --> 0:12:49.280 given the insatiable demand. 0:12:49.520 --> 0:12:52.640 Ed right now for that, talk to us a little 0:12:52.640 --> 0:12:54.480 bit about the m and A that was involved Riley 0:12:54.520 --> 0:12:56.440 to beef up the team because Ed hinted at it 0:12:56.480 --> 0:12:59.400 in his piece there. But they were pushed back by 0:12:59.720 --> 0:13:02.160 one Japanese company. But then we're able to buy a 0:13:02.240 --> 0:13:04.400 local chip making focus. 0:13:05.040 --> 0:13:05.360 Yeah. 0:13:05.400 --> 0:13:07.920 So early last year, Meta had made an approach for 0:13:08.160 --> 0:13:12.360 furios Ai, a Korean chip maker that rebuked it. It 0:13:12.400 --> 0:13:14.960 turned down that eight hundred million dollar offer. We then 0:13:15.040 --> 0:13:17.760 reported in March. Later towards the end of the year, 0:13:18.160 --> 0:13:21.440 we also reported that Meta was successfully able to acquire 0:13:21.559 --> 0:13:24.360 Revos and with that it's bench of talent and that 0:13:24.480 --> 0:13:26.240 was really the key here. They were able to bring 0:13:26.240 --> 0:13:28.720 in more than four hundred employees who are now peppered 0:13:28.760 --> 0:13:34.120 across MTIA. This team, some of whom we met in Fremont, 0:13:34.520 --> 0:13:38.079 that are making these four different chips, and they're able 0:13:38.120 --> 0:13:40.560 to do that in parallel with that additional headcount. 0:13:41.280 --> 0:13:44.120 I think the specific is really interesting, Like MTIA three 0:13:44.200 --> 0:13:46.640 hundred is real. You know, Meta tells us it's in 0:13:46.679 --> 0:13:48.720 production and they still regard that as. 0:13:48.640 --> 0:13:50.880 AI right training the AI. 0:13:51.000 --> 0:13:55.600 That results in what shows in your timeline targeted ads, 0:13:55.840 --> 0:13:59.840 but actually the next step is LAMA running the model. 0:13:59.880 --> 0:14:02.280 So when you put a request into Meta AI, what 0:14:02.400 --> 0:14:04.520 else did you see there? I mean, the facility was 0:14:04.600 --> 0:14:07.920 large and from a dollar perspective, they seem very committed here. 0:14:08.400 --> 0:14:10.760 It seems like a massive investment. They wouldn't give us 0:14:10.800 --> 0:14:13.120 a specific figure to how much they're spending on their 0:14:13.160 --> 0:14:16.199 own in house chip work, but just seeing the products 0:14:16.240 --> 0:14:19.040 on the table, you got to believe there are billions 0:14:19.040 --> 0:14:21.520 at play here. One thing I would have loved to 0:14:21.560 --> 0:14:24.520 see was MTIA for fifty to five hundred. These are 0:14:24.680 --> 0:14:27.440 the next next generation, and. 0:14:27.600 --> 0:14:31.080 We asked they showed us the cads or the EDAs. 0:14:31.120 --> 0:14:34.240 But yeah, no, no finished product yet and that we're 0:14:34.240 --> 0:14:36.640 looking to twenty twenty seven for that. So two more 0:14:37.080 --> 0:14:41.640 being released next year and in six month increments, which 0:14:41.720 --> 0:14:42.840 is really notable. 0:14:43.080 --> 0:14:46.720 Yeah, it was a great deep dive. I'm looking forward 0:14:46.720 --> 0:14:50.000 to seeing the next chips. Roddy Griffin, thank you very much. Indeed, 0:14:50.040 --> 0:14:53.960 Now in other news, in Nvidia will invest two billion 0:14:54.000 --> 0:14:56.840 dollars into Nevius now as part of a strategic partnership 0:14:56.960 --> 0:14:59.640 to develop and build AI data centers. You guessed it 0:15:00.000 --> 0:15:02.840 will help Nevius deploy more than get this five gigawatts 0:15:03.160 --> 0:15:05.800 in video systems by the end of twenty thirty. Again, 0:15:06.480 --> 0:15:10.960 this age old circular financing question comes once more, and 0:15:11.080 --> 0:15:14.480 in video putting it in to the end customer for 0:15:14.560 --> 0:15:17.520 its chips. Nebe also has a fascinating history. Of course, 0:15:17.560 --> 0:15:20.360 it's the Dutch company. It was the holding company of 0:15:20.440 --> 0:15:22.720 what was once Yandex. In fact, it's still led by 0:15:22.760 --> 0:15:24.840 Yandex's co founder of the Russian business. 0:15:25.120 --> 0:15:27.840 And it's basically you know, it's one of the European names. 0:15:27.920 --> 0:15:30.120 The stock reaction, I guess not that surprising, right. In 0:15:30.200 --> 0:15:34.280 Video has made several multi billion dollar investments in one 0:15:34.360 --> 0:15:36.960 or another layer of what's happening in AI infrastructure, but 0:15:37.000 --> 0:15:39.040 at one point of eighteen percent on track for his 0:15:39.040 --> 0:15:41.840 biggest jobs since September. I guess a vote of confidence 0:15:41.880 --> 0:15:45.400 from the main market leader for GPUs that go into 0:15:45.400 --> 0:15:46.160 the data centers. 0:15:46.360 --> 0:15:46.920 Yeah, certainly. 0:15:47.000 --> 0:15:49.240 What we've just had news this week of in scale 0:15:49.320 --> 0:15:51.640 and the fact that they've got new funding got more 0:15:51.680 --> 0:15:54.720 people on the board. Nick klegg Sheryl Samberg. So there's 0:15:54.720 --> 0:15:57.320 a real heating up of those neo clouds over in 0:15:57.320 --> 0:15:59.440 Europe as well as cool we've hover in the US 0:16:00.640 --> 0:16:03.200 another day. It is another big bit of AI agent news. 0:16:03.320 --> 0:16:06.000 Data Bricks is launching Geni Code. It's a built in 0:16:06.200 --> 0:16:10.040 autonomous AI assistant for really technical talent. Joining us now 0:16:10.080 --> 0:16:12.720 in New York CEO Ali Goertzi, who's here for a 0:16:12.720 --> 0:16:14.560 big sales event that you host here in the city 0:16:14.600 --> 0:16:18.920 and Alik, what is this particular bit of code equipment 0:16:19.000 --> 0:16:19.880 going to do for your talent? 0:16:20.120 --> 0:16:20.360 Yeah? 0:16:20.400 --> 0:16:21.800 So what it is is, you know, we've heard a 0:16:21.840 --> 0:16:24.360 lot about cloud code and coding agents and so on, 0:16:24.600 --> 0:16:26.160 but what they do, they're kind of like interns that 0:16:26.200 --> 0:16:28.680 can write code for you. That's great, but how do 0:16:28.680 --> 0:16:31.560 you get that code into production? And for that you 0:16:31.640 --> 0:16:34.040 need to be sure that you know nothing's going to 0:16:34.040 --> 0:16:35.320 go wrong, that you can measure it. 0:16:35.280 --> 0:16:35.840 And so on. 0:16:36.200 --> 0:16:38.040 So that's gen code is really about, how do you 0:16:38.080 --> 0:16:41.160 actually get your code into production, how do you measure 0:16:41.200 --> 0:16:43.760 the data pipelines that power all of the dashboards that 0:16:43.800 --> 0:16:47.760 you see, and how do you even start building AI models. 0:16:47.920 --> 0:16:49.800 So what gen code can do is it can build 0:16:49.800 --> 0:16:52.000 a machine learning model for you that can predict prices 0:16:52.560 --> 0:16:54.320 or estimate, you know, how much you're going to sell 0:16:54.320 --> 0:16:57.120 of this, and it can do this you know itself. 0:16:57.200 --> 0:16:59.680 It just goes build machine learning models, iterates on them, 0:16:59.720 --> 0:17:01.960 does a lot of the things that data scientists previously 0:17:01.960 --> 0:17:03.000 had to do themselves. 0:17:03.240 --> 0:17:05.280 It just automates all of that for you, and. 0:17:05.240 --> 0:17:09.400 It's not getting things incorrect. As the work continues, there's 0:17:09.440 --> 0:17:11.760 still been a lot of concern that these things are 0:17:11.760 --> 0:17:13.600 hallucinating the longer term that they work. 0:17:13.760 --> 0:17:16.040 Yeah, this is a big problem in the industry. So 0:17:16.680 --> 0:17:18.320 with this, this is like one piece of the puzzle, 0:17:18.359 --> 0:17:19.639 is what I talked about. The other piece of the 0:17:19.680 --> 0:17:23.080 puzzle is we also acquired a company called Quotient, and 0:17:23.119 --> 0:17:27.080 these are the folks behind GitHub Copilot's quality measurement. 0:17:27.600 --> 0:17:29.960 So there are folks focused. 0:17:29.520 --> 0:17:32.880 On quality measurement making sure that they can do monitoring 0:17:33.000 --> 0:17:34.960 of how things are going. So this is the other 0:17:35.000 --> 0:17:36.919 piece of the puzzle, so that we can actually do that. 0:17:37.000 --> 0:17:40.080 So when you launch these things and they're running, how 0:17:40.119 --> 0:17:41.800 can we oversee to make sure that nothing is going 0:17:41.840 --> 0:17:43.639 wrong or it's sort of if it's going off. 0:17:43.520 --> 0:17:45.639 Track, we can stop it, restart it. 0:17:45.680 --> 0:17:48.240 So that's equally important, maybe even more important than the 0:17:48.280 --> 0:17:48.840 first part. 0:17:49.640 --> 0:17:49.880 Alian. 0:17:49.920 --> 0:17:51.760 It's great to have you back on on Bloomberg Tech, 0:17:52.119 --> 0:17:56.280 like taking in aggregate, launch of gene code, acquisition of quoting, 0:17:56.359 --> 0:17:59.080 and how much at this stage for data breaks. Is 0:17:59.119 --> 0:18:02.119 it about product diversity, you know, kind of broadening the 0:18:02.200 --> 0:18:05.840 offering or suite or platform for your customers. 0:18:06.359 --> 0:18:08.680 Yeah, I mean that's super important, right because the space 0:18:08.760 --> 0:18:11.439 is moving so fast, and you know it just you know, 0:18:11.520 --> 0:18:13.480 just I think six months ago we were talking about 0:18:13.680 --> 0:18:16.280 things that were out of completing code. To now everybody 0:18:16.280 --> 0:18:18.359 has agents that can write the code. So the question 0:18:18.440 --> 0:18:20.720 is how do we actually make these the code that 0:18:20.760 --> 0:18:23.439 has been written into production, make sure that they're powering 0:18:23.480 --> 0:18:25.720 up everything inside of the enterprise and making sure that 0:18:25.760 --> 0:18:28.439 we're monitoring it, making sure that we can actually start 0:18:28.600 --> 0:18:30.640 doing more interesting things with it. Can we build AI 0:18:30.760 --> 0:18:34.080 models automatically for the business that can predict the business? 0:18:34.880 --> 0:18:37.719 So you know, to complete that puzzle and have all 0:18:37.760 --> 0:18:40.600 the different puzzle pieces, you know, you have to expand. 0:18:40.720 --> 0:18:42.920 So it is about that and you know that's why 0:18:43.000 --> 0:18:45.800 Quotient and also gene code that have been added to this. 0:18:46.560 --> 0:18:49.640 How how defensive was that move? You know, I think 0:18:49.640 --> 0:18:54.240 about the interesting cursor right now, what Claude is offering 0:18:54.600 --> 0:18:58.440 on or anthropic through the coding side, your awareness of 0:18:59.359 --> 0:19:00.760 the battle field in AI. 0:19:01.200 --> 0:19:03.359 Yeah, I mean those are great partners and customers that 0:19:03.359 --> 0:19:05.280 we work with and we use actually cursor and we 0:19:05.400 --> 0:19:06.760 use cloud code internally at theater but. 0:19:06.880 --> 0:19:07.760 As well, Yeah we do. 0:19:08.560 --> 0:19:10.399 But those things they focus on how can we write 0:19:10.400 --> 0:19:13.760 how can we help software engineers? What gene code really 0:19:13.760 --> 0:19:15.720 can do is it brings it to the knowledge worker, 0:19:16.040 --> 0:19:18.400 the people that create your dashboards inside of an organization 0:19:18.440 --> 0:19:20.560 and they make sure that your revenue numbers are correct 0:19:20.880 --> 0:19:22.560 or the people that make sure that the data that's 0:19:22.560 --> 0:19:25.640 coming in every day into the organization is correct. Nothing breaks, 0:19:26.000 --> 0:19:27.440 you know, there's not an outage. You don't have like 0:19:27.480 --> 0:19:29.800 a blue screen. The dashboard is out and you can't 0:19:29.800 --> 0:19:32.040 see it, or the people that are building machine learning 0:19:32.080 --> 0:19:35.200 models that can predict your prices or your costs or 0:19:35.240 --> 0:19:39.040 you know, doing assessment with machine learning models. It automates 0:19:39.080 --> 0:19:41.480