WEBVTT - Google Cloud Debuts New AI Chips 0:00:02.520 --> 0:00:13.480 Bloomberg Audio Studios, Podcasts, radio news. Bloomberg Tech is alive 0:00:13.560 --> 0:00:17.360 from coast to coast, with Caroline Hide in New York 0:00:17.640 --> 0:00:19.560 and Eva low In sent frances Go. 0:00:21.640 --> 0:00:23.200 This is Bloomberg Tech coming up. 0:00:23.239 --> 0:00:26.400 Alphabet's Google Cloud division on veils the latest generation of 0:00:26.440 --> 0:00:30.400 its tensor processing unit, or TPU, and announces a slew 0:00:30.440 --> 0:00:32.400 of partnership deals or break it all down. 0:00:32.640 --> 0:00:37.360 Plus, a small group of unauthorized users has accessed anthropics 0:00:37.400 --> 0:00:40.680 new AI model Mythos, which the company says can enable 0:00:40.760 --> 0:00:41.839 dangerous cyber attacks. 0:00:41.880 --> 0:00:44.720 And we'll catch up with Rivian CEO RJ. Scarringe to 0:00:44.760 --> 0:00:47.840 discuss a big milestone for the company's new R two 0:00:48.159 --> 0:00:49.120 that's later this hour. 0:00:49.400 --> 0:00:52.159 First we check in on these markets, which managed to 0:00:52.159 --> 0:00:54.880 be shrugging off yet further chao sort occurs in the 0:00:54.920 --> 0:00:57.279 straight orfor moves. The oil market doesn't shrug it off, 0:00:57.280 --> 0:00:59.400 and we're currently seeing Brint up almost three percent. 0:00:59.560 --> 0:01:01.480 We're back above one hundred dollars a barrel. 0:01:01.720 --> 0:01:04.960 What's interesting is equities managed to push through that inflationary 0:01:04.959 --> 0:01:08.800 headwind and that geopolitical anxiety. In fact, we've got new 0:01:08.880 --> 0:01:10.800 record high on then as that one hundred s and 0:01:10.800 --> 0:01:13.240 P five hundred very close to that record, So ed 0:01:13.600 --> 0:01:14.920 there is a lot of optimism. 0:01:15.000 --> 0:01:16.680 A lot of it's around earnings and other news. 0:01:16.880 --> 0:01:18.800 Yeah, news is driving stocks. 0:01:18.920 --> 0:01:22.200 Google had a big jump in the pre market, carried 0:01:22.240 --> 0:01:25.520 over into the main session and is sustained up one 0:01:25.560 --> 0:01:28.160 point seven percent. There is a lot of Google news 0:01:28.160 --> 0:01:31.640 to weigh through. First, let's talk about tensor processing units 0:01:31.720 --> 0:01:35.640 or TPUs. Alphabet's Google Cloud division unveiled the latest generation 0:01:36.000 --> 0:01:40.040 of its TPUs design to make AI computing services faster 0:01:40.600 --> 0:01:43.160 and more efficient. The U lineup comes in two versions, 0:01:43.360 --> 0:01:47.240 the TPU eight T for creating AI software and the 0:01:47.280 --> 0:01:51.200 TPUAI for running AI services that have been created. In 0:01:51.240 --> 0:01:53.960 other words, carro training and inference specific. 0:01:54.320 --> 0:01:56.280 I mean it sounds like Amazon in many ways of 0:01:56.280 --> 0:01:58.360 their chips they've been making, but Google has led the 0:01:58.440 --> 0:02:01.160 charge in terms of who designed their own types and 0:02:01.200 --> 0:02:02.360 then vertically integrates. 0:02:02.360 --> 0:02:03.320 That's really led the point. 0:02:03.160 --> 0:02:04.440 We talked to in the week that there would be 0:02:04.480 --> 0:02:07.400 an inferant specific TPU. We got both a training and 0:02:07.480 --> 0:02:08.760 an inference specific TPU. 0:02:08.800 --> 0:02:09.160 What else? 0:02:09.240 --> 0:02:11.639 I think it's important that Google had been Google had 0:02:11.639 --> 0:02:13.480 been reporting Bloomberg and reporting all of this. 0:02:13.639 --> 0:02:14.840 But also let's talk. 0:02:14.720 --> 0:02:17.640 About more in the world of Google Cloud because the 0:02:17.680 --> 0:02:22.800 company has just expanded several partnerships Oracle to give joint 0:02:22.840 --> 0:02:25.000 customers a simpler way to interact with their Oracle data 0:02:25.040 --> 0:02:27.400 using natural language. But there's also within video to help 0:02:27.440 --> 0:02:30.120 power agenic and physical AI and AI factories. There are 0:02:30.120 --> 0:02:33.600 also agreements with Salesforce, crowd Strike, broad comm startups such 0:02:33.600 --> 0:02:37.080 as Mira Marati's Thinking Machines. There was a whole wealth 0:02:37.120 --> 0:02:39.160 of headlines we had to navigate through. 0:02:39.320 --> 0:02:42.040 But maybe it's the fact that they're making their own chips. 0:02:42.200 --> 0:02:44.760 It's becoming more efficient and that makes more people want 0:02:44.800 --> 0:02:46.400 to come in on the cloud reglationship. 0:02:46.480 --> 0:02:49.200 Let's be honest, this is a massive news from Google Cloud, 0:02:49.360 --> 0:02:53.840 and there are signs that internally more ARI code is 0:02:53.840 --> 0:02:56.480 being written with AI. But they're doing deals and Cloud 0:02:56.600 --> 0:02:59.320 could be taking market share. Let's stay on alphabet man 0:02:59.360 --> 0:03:02.359 deep seeing the Bloomberg intelligence is with us. I don't 0:03:02.400 --> 0:03:04.520 know if you were going to publish one react or 0:03:04.560 --> 0:03:07.680 it doesn't react because there are so many pieces of 0:03:07.760 --> 0:03:10.920 new information. I suppose that start on the TPU angle. 0:03:11.320 --> 0:03:14.400 That's a big focus for Google Cloud. Why why do 0:03:14.480 --> 0:03:16.280 we care so much about their custom silicon? 0:03:17.040 --> 0:03:18.240 I mean, look right now. 0:03:18.280 --> 0:03:22.200 It is about having the capacity to deploy AI workloads. 0:03:22.320 --> 0:03:25.320 That's what everyone is scrambling for. You have seen you know, 0:03:25.400 --> 0:03:29.560 Anthropic really kind of change their models because they don't 0:03:29.600 --> 0:03:33.360 have the capacity to serve their best model. And look, 0:03:33.760 --> 0:03:37.080 Google has that advantage when it comes to owning that 0:03:37.280 --> 0:03:41.160 TPU supply and also deploying it at scale with their 0:03:41.160 --> 0:03:43.920 family or apps. And I think the best data point 0:03:44.000 --> 0:03:46.280 we are going to get in this earning season is 0:03:46.320 --> 0:03:50.400 the cloud growth for Google segment already was close to 0:03:50.640 --> 0:03:54.960 fifty percent last quarter when Azure grew forty percent, So 0:03:55.040 --> 0:03:56.840 that gap most likely. 0:03:56.680 --> 0:03:57.680 Is going to increase. 0:03:58.120 --> 0:04:01.040 And not only are they growing fams than the other 0:04:01.080 --> 0:04:05.000 cloud players, their margins are expanding. And that's where people 0:04:05.040 --> 0:04:08.880 will realize they are the lowest cost token provider when 0:04:08.960 --> 0:04:11.040 you compare them to everyone else out there. 0:04:11.680 --> 0:04:14.360 We don't have to wait long to get that data. 0:04:14.440 --> 0:04:17.479 April twenty ninth is when the earnings come. What do 0:04:17.480 --> 0:04:19.839 you make of all the expanded deals? Then? Just this 0:04:20.080 --> 0:04:22.480 just hint. The Mormon people are coming to them for 0:04:22.520 --> 0:04:24.560 the cloud. Does it matter whether they're startups or whether 0:04:24.600 --> 0:04:26.160 they're legacy businesses. 0:04:26.400 --> 0:04:28.880 Yeah, Look, I think a lot of the deals involve 0:04:28.960 --> 0:04:32.919 vertical integration where they are providing the TPUs, they're doing 0:04:32.960 --> 0:04:37.360 it on their cloud and it's using Gemini. So you know, 0:04:37.400 --> 0:04:41.240 if you are using Google Cloud, you're getting that complete package, 0:04:41.279 --> 0:04:44.880 whereas with everyone else you're using either in video chips 0:04:45.000 --> 0:04:48.719 or a BLUs cloud. And to me, the best illustration is, 0:04:48.760 --> 0:04:52.239 you know the deal that Amazon has had with open 0:04:52.279 --> 0:04:56.960 Ai and Entropic, where they are giving Entropic five billion 0:04:57.000 --> 0:05:00.320 dollars money to spend on their cloud, whereas the deal 0:05:00.360 --> 0:05:04.839 that Entropic had with Google three point five gigawards. Google 0:05:04.920 --> 0:05:05.360 is not. 0:05:05.240 --> 0:05:06.440 Giving Entthropic anymore. 0:05:07.000 --> 0:05:11.120 So that is a sign that illustrates that Google probably 0:05:11.240 --> 0:05:13.560 is in a better position when it comes to negotiating 0:05:13.600 --> 0:05:15.599 with these frontier and. 0:05:15.600 --> 0:05:18.440 For our client base of a Bloomberg terminal, for example, 0:05:18.520 --> 0:05:22.320 you think about just the finance industry. Citadel's over there 0:05:22.320 --> 0:05:24.599 talking about how they're using Google. But there's a story 0:05:24.640 --> 0:05:26.760 that's really well read on the terminal talking about Vista 0:05:26.800 --> 0:05:28.960 striking a deal to speed up Google AI in. 0:05:28.960 --> 0:05:30.080 Its software portfolio. 0:05:30.080 --> 0:05:32.560 This is very much financially focused in terms of the 0:05:32.560 --> 0:05:33.479 industry that it's changing. 0:05:33.560 --> 0:05:36.760 So let's just call Vista a private equity firm or 0:05:36.760 --> 0:05:39.120 an investor that has a portfolio of companies. This is 0:05:39.120 --> 0:05:41.839 a kind of tried and tested format where Vista makes 0:05:41.880 --> 0:05:44.240 deal with Google Cloud, but the net result is the 0:05:44.279 --> 0:05:48.800 portfolio companies get access to either the compute capacity or 0:05:48.839 --> 0:05:51.960 something on customs, silicon, whatever it might be. But Vista 0:05:52.040 --> 0:05:54.920 has the cloud to do that on the portfolio company's behalf. 0:05:55.160 --> 0:05:58.559 That's a playbook we've seen before. But yeah, yeah, spend 0:05:58.560 --> 0:06:01.680 a bit of money on AI and spending. 0:06:01.279 --> 0:06:02.400 They are mandy. 0:06:02.760 --> 0:06:04.880 Just briefly, how much do you think that this adds 0:06:04.920 --> 0:06:06.560 to the revenue run rate? How much does all of 0:06:06.600 --> 0:06:09.680 this really matter to Alphabet's huge business model, which isn't 0:06:09.680 --> 0:06:10.520 oil about. 0:06:10.200 --> 0:06:11.880 It so real quick, like I'm not going to read 0:06:11.920 --> 0:06:15.720 you every single headline insert company name here. Expanded pack 0:06:15.800 --> 0:06:16.720 with Google Cloud. 0:06:16.800 --> 0:06:17.600 Some reports in. 0:06:17.560 --> 0:06:20.039 The press, like with Thinking Machines, it's a multi billion 0:06:20.080 --> 0:06:22.920 dollar deal that will show up, right, surely it will 0:06:22.920 --> 0:06:23.320 show up. 0:06:23.360 --> 0:06:27.400 I mean Google's share of infrastructure cloud where you know, 0:06:27.480 --> 0:06:31.520 think of the triopoly there was Amazon, Microsoft and Google. 0:06:32.800 --> 0:06:34.800 Way, well they will move up. 0:06:35.480 --> 0:06:39.040 I think with AI workload share, they will certainly have 0:06:39.080 --> 0:06:41.760 a higher share compared to what they had in their 0:06:41.800 --> 0:06:44.960 traditional cloud and this will probably be the biggest driver 0:06:45.120 --> 0:06:48.440 of that owning the TPUs, and then obviously their chips 0:06:48.440 --> 0:06:52.080 doing far better than other ASA providers, including Amazon. 0:06:52.400 --> 0:06:54.000 I just want to go back to the basics of 0:06:54.040 --> 0:06:54.680 the TPU. 0:06:55.320 --> 0:06:57.359 You and I have discussed that body of research that 0:06:57.400 --> 0:07:01.479 you did comparing the existing generation TPU performance against other 0:07:01.560 --> 0:07:05.600 A six and Nvidia. What's the basic story of the TPU. 0:07:05.760 --> 0:07:08.960 What's the selling point of Google saying you should use 0:07:09.040 --> 0:07:12.320 this instead of using the latest in video gear. 0:07:12.560 --> 0:07:15.800 I mean, look, these are not confirmed reports, but Mythos 0:07:16.160 --> 0:07:20.560 may have been trained entirely on TPUs or majority of 0:07:20.600 --> 0:07:24.960 the cluster for that Mythos training was Google GPUs. So 0:07:25.080 --> 0:07:28.760 from that perspective, if the best model out there, whether 0:07:28.800 --> 0:07:33.880 it's their own, Google, Gemini or Anthropics models, is trained 0:07:33.920 --> 0:07:38.400 on TPU, that is a pretty good validation of how 0:07:38.440 --> 0:07:41.240 good their chips are both for training and then they 0:07:41.280 --> 0:07:44.680 deployed at scale across their family of apps, search being 0:07:44.720 --> 0:07:48.480 that biggest kind of surface where they deploy their own 0:07:48.480 --> 0:07:50.760 models at scale, and that's all inferencing. 0:07:51.080 --> 0:07:53.120 Okay, we want to go back to something that you 0:07:53.160 --> 0:07:55.760 said at the very top of the conversation. You hinted 0:07:55.840 --> 0:08:00.840 that mythos maybe being unfolded as it is a compute 0:08:01.080 --> 0:08:03.080 rather than perhaps it's all powerful nature. 0:08:03.400 --> 0:08:04.640 What gives you that feeling? 0:08:04.720 --> 0:08:08.120 And how will the token economics change that for some 0:08:08.200 --> 0:08:09.600 of these lms in the future. 0:08:09.760 --> 0:08:12.120 Yeah, so right now, if you look at the last 0:08:12.200 --> 0:08:15.480 data point that Google has shared, they're around fifteen hundred 0:08:15.520 --> 0:08:19.160 trillion tokens per month. That's about three to four times 0:08:19.200 --> 0:08:22.080 that of open ai and Tropic based on our work. 0:08:22.160 --> 0:08:26.040 So they're already operating at such a high scale. Why 0:08:26.440 --> 0:08:29.880 Because they are the lowest cost token provider. They don't 0:08:29.960 --> 0:08:32.600 have to get those chips from Nvidia to you know, 0:08:32.720 --> 0:08:36.000 deploy it for inferencing, and that is the difference. I 0:08:36.040 --> 0:08:38.800 think that will become more and more obvious as we 0:08:38.840 --> 0:08:44.079 go towards AGENTYKEAI, which is basically doing inferencing much longer 0:08:44.160 --> 0:08:46.680 than you know, using it in a copilot or a 0:08:46.760 --> 0:08:49.080 chat pot type of inference format. 0:08:49.280 --> 0:08:51.439 All night, all day, all night, man deep saying and 0:08:51.480 --> 0:08:54.600 bloemberg intelligence. Great to have you break down what is 0:08:54.640 --> 0:08:57.360 a whole swofth of news around alphabet coming up? 0:08:57.440 --> 0:08:59.760 Rivian's long promised DOR two starts. 0:08:59.520 --> 0:09:02.000 Rolling off the line's sweet with the CEO about the 0:09:02.000 --> 0:09:03.760 next chapter for the electric vehicle maker. 0:09:04.280 --> 0:09:05.240 It's a bloomberg tech. 0:09:14.800 --> 0:09:17.520 Rivian's new R two is rolling off the line in 0:09:17.600 --> 0:09:18.319 normal Illinois. 0:09:18.360 --> 0:09:19.920 Big milestone for the company. 0:09:20.040 --> 0:09:23.720 It's a smaller, more affordable suv and for many this 0:09:23.800 --> 0:09:27.200 is the vehicle that will define Rivian's path to profitability. 0:09:27.559 --> 0:09:31.200 The company continues to burn through cash. Joining us now 0:09:31.360 --> 0:09:35.280 from the factory is CEO RJ. Scaringe, who has driven 0:09:35.360 --> 0:09:37.560 quite a few R twos off the line already today. 0:09:38.320 --> 0:09:40.760 It's a big moment, but ourj we should point out 0:09:40.800 --> 0:09:44.800 that on Friday the facility was hit by a tornado, 0:09:45.559 --> 0:09:49.040 by a big storm. There was damage to certain parts 0:09:49.040 --> 0:09:52.280 of the factory. What was the net result of that? 0:09:52.480 --> 0:09:56.360 Is everyone okay? And has there been any material impact 0:09:56.440 --> 0:09:57.960 to the startup production plan? 0:10:00.240 --> 0:10:02.640 I mean, of course, of course, you can't plan for 0:10:02.679 --> 0:10:05.920 these kinds of things, and we didn't expect to tornado 0:10:05.960 --> 0:10:08.440 to hit Friday night, but you know, the teams responded 0:10:08.480 --> 0:10:12.839 really well. Fortunately everyone was safe. We followed our protocols 0:10:12.840 --> 0:10:14.560 as soon as we knew there were tornadoes in the area. 0:10:15.679 --> 0:10:18.200 But yeah, tornado went through the south end of the 0:10:18.200 --> 0:10:21.800 plant and ripped the roof off the building and knocked 0:10:21.840 --> 0:10:23.640 down some of the some of the plants as well, 0:10:24.320 --> 0:10:26.599 and so the last you know, seventy two hours have 0:10:26.679 --> 0:10:30.199 been around the clock. You're getting all the water out 0:10:30.200 --> 0:10:33.920 of the facility and getting ready to rebuild the south end. 0:10:34.200 --> 0:10:37.640 But you know, it's like the ultimate example of resilience 0:10:37.640 --> 0:10:39.920 of the business where the teams have come together and 0:10:40.160 --> 0:10:42.200 we still had vehicles come off the line this morning. 0:10:42.720 --> 0:10:44.680 Is the plan still the same You can get through 0:10:44.720 --> 0:10:47.480 the same kind of ramp strategy that you had the 0:10:47.520 --> 0:10:50.360 same volumes in that initial output. 0:10:51.040 --> 0:10:53.400 Yeah, our ramp up this week and into next week. 0:10:53.600 --> 0:10:57.160 We're not making changes to the plan. We're obviously having 0:10:57.200 --> 0:10:59.720 to make some accommodations for the fact that a big 0:10:59.760 --> 0:11:02.400 it's the logistics area of the plant and so the 0:11:02.800 --> 0:11:05.080 way that materira was flowing into the plant through the 0:11:05.120 --> 0:11:07.760 dock doors does not have to come in a different way. 0:11:08.440 --> 0:11:11.520 But you know, we're working through it. We're working through it. 0:11:12.160 --> 0:11:17.800 What's interesting is orders demand configuration, ah J. What do 0:11:17.880 --> 0:11:20.880 people want and how willingly are they putting in the orders? 0:11:21.880 --> 0:11:25.200 Yeah, I mean, I'm as you said, I drove a 0:11:25.200 --> 0:11:27.360 bunch of our twos off the line today. I've been 0:11:27.440 --> 0:11:28.920 driving in our two for a while, and the car 0:11:29.000 --> 0:11:32.439 is just incredible. It's the combination of the package, the 0:11:32.920 --> 0:11:38.320 driving dynamics, the efficiency of course, the design, and just 0:11:38.360 --> 0:11:41.520 the aesthetic feel. So we're really excited to get folks 0:11:41.559 --> 0:11:44.520 into it. There's a tremendous amount of enthusiasm. We've had 0:11:44.520 --> 0:11:49.400 a few events where we've hosted potential customers to experience it. 0:11:50.040 --> 0:11:52.439 And you know, we were in Denver a couple of 0:11:52.480 --> 0:11:54.800 days ago and there's like a super long line just 0:11:54.840 --> 0:11:55.520 to sit in the car. 0:11:55.600 --> 0:11:57.199 So we'd love to see the enthusiasm. 0:11:58.000 --> 0:12:01.120 And I can't wait to continue to ramp up production 0:12:01.200 --> 0:12:02.160 start seeing these on the road. 0:12:02.360 --> 0:12:04.480 I mean, I loved Ed's enthusiasm when he came to 0:12:04.520 --> 0:12:06.200 have a look at your in house AI chip, and 0:12:06.240 --> 0:12:09.880 I'm interested as to where therefore that vertical integration comes in. 0:12:09.960 --> 0:12:12.760 You're also including light O. Look, there's there's a roadmap 0:12:13.040 --> 0:12:15.360 for these variants. When does that ender production? 0:12:15.520 --> 0:12:18.679 A J sour our in house silicon, which is a 0:12:18.960 --> 0:12:20.959 eight hundred tops per chip. We have two of those 0:12:21.000 --> 0:12:24.080 chips in the vehicle and we couple that with cameras, 0:12:24.160 --> 0:12:26.280 radar and a lighter that sits to the top of 0:12:26.280 --> 0:12:29.560 the windshield that comes in at the end of this year, 0:12:29.800 --> 0:12:31.800 and so our launch edition still has a very highly 0:12:31.840 --> 0:12:35.880 capable self driving platform, which is essentially a further development 0:12:35.920 --> 0:12:39.520 of what's in our our one vehicle now. But across 0:12:39.520 --> 0:12:42.640 the vehicle, you know, R two just represents so many 0:12:42.760 --> 0:12:45.160 learnings for us as a business, whether it's the drive line, 0:12:45.160 --> 0:12:47.720 the power electronics, just the way it's being built in 0:12:47.800 --> 0:12:51.200 terms of cost and cost efficiency, and so there's there's 0:12:51.240 --> 0:12:53.400 like not a there's not a detail in the car 0:12:53.440 --> 0:12:57.400 that hasn't been thoughtfully evolved or developed relative to what 0:12:57.559 --> 0:12:58.800 we've done in our one. 0:13:00.040 --> 0:13:01.480 Oh Ja, I want to go a bit more specific 0:13:01.559 --> 0:13:04.000 on what car is talking about. The story with R 0:13:04.040 --> 0:13:06.520 two has changed a little bit. I think you're probably 0:13:06.960 --> 0:13:10.120 thinking about an autonomous future a bit more than you were. 0:13:10.760 --> 0:13:13.199 But the idea was to get close to that forty 0:13:13.240 --> 0:13:16.640 five thousand dollars mark a vehicle for the masses. You 0:13:16.679 --> 0:13:20.480 start with a premium and more expensive variant, But what 0:13:20.520 --> 0:13:23.400 are you seeing in the demand pockets for getting to 0:13:23.480 --> 0:13:25.920 a lower price point that is where you want to 0:13:25.920 --> 0:13:28.240 take on the established lowems. 0:13:27.800 --> 0:13:30.600 Right, Yeah, Well, the average price of a new car 0:13:30.960 --> 0:13:34.120 in the United States is about fifty thousand dollars and 0:13:34.200 --> 0:13:37.440 so really important to us as we're developing R two 0:13:37.559 --> 0:13:42.080 is recognizing we wanted to be right in that range, 0:13:42.120 --> 0:13:44.880 and so having a variant that starts at forty five 0:13:45.679 --> 0:13:48.600 and then going up to our top speck for the 0:13:48.640 --> 0:13:52.800 performance version with all the kit for the interior is 0:13:52.840 --> 0:13:55.880 at fifty seven, it allows us to really nicely straddle that. 0:13:55.920 --> 0:13:58.120 And of course, you know, launching any product, you have 0:13:58.160 --> 0:13:59.920 the decision of what do you launch first? You launch 0:14:00.040 --> 0:14:03.880 the base back, midspec top spec right, and it's impossible 0:14:03.920 --> 0:14:06.160 to sort of make everyone happy with when you have 0:14:06.160 --> 0:14:08.080 a launch edition. So we decided to launch with the 0:14:08.080 --> 0:14:13.240 premium spec, but we'll be introducing like a middle like 0:14:13.280 --> 0:14:17.640 a midspec vehicle, and then the forty five thousand version 0:14:17.640 --> 0:14:20.640 will come thereafter. But that all rolls out over roughly 0:14:20.680 --> 0:14:24.320 the next year, and you know, there's so much enthusiasm 0:14:24.400 --> 0:14:27.560 for the product that we're very like, we feel really 0:14:27.600 --> 0:14:31.240 confident with the way we've configured the specifications and even 0:14:31.280 --> 0:14:31.720 the pricing. 0:14:31.760 --> 0:14:34.080 We feel quite strong that we've we've. 0:14:33.840 --> 0:14:37.560 Priced it appropriately, especially when you look at the capabilities 0:14:37.640 --> 0:14:40.280 and then the cross shop of what other vehicles are 0:14:40.320 --> 0:14:40.800 in the space. 0:14:41.520 --> 0:14:44.720 I've experienced the R two that does have the lde 0:14:44.840 --> 0:14:48.240 R and the in house silicon, and it's so interesting 0:14:48.280 --> 0:14:50.240 there's going to be a generation of R twos that 0:14:50.320 --> 0:14:52.800 don't have it. I know that you said they're very 0:14:52.840 --> 0:14:57.280 capable with the Nvidia silicon and without LIDR, but that's 0:14:57.320 --> 0:14:59.920 a big call to make, right. Is there any flexible 0:15:00.400 --> 0:15:03.920 to accelerate the timing of where you can start production 0:15:04.040 --> 0:15:06.400 on the R two line in Illinois and bring in 0:15:06.520 --> 0:15:11.120 that gear, because what you're pitching for the autonomous capabilities, 0:15:11.280 --> 0:15:15.120 even the advanced driver assistance capabilities of R two, they 0:15:15.160 --> 0:15:18.320 are predicated on the on that next gen gear. 0:15:19.880 --> 0:15:22.520 Well, the launch edition still has a very high ceiling, 0:15:22.640 --> 0:15:25.800 and so this will get to and our Gen two 0:15:25.880 --> 0:15:26.840 R one will also do this. 0:15:27.560 --> 0:15:29.280 Later this year, will start to. 0:15:29.320 --> 0:15:32.480 Roll out point to point level two, so I meaning 0:15:32.880 --> 0:15:35.360 hands off wheel, eyes on road, but you type the 0:15:35.360 --> 0:15:37.640 address and the car completely drives you there. 0:15:38.320 --> 0:15:39.360 And then into next. 0:15:39.280 --> 0:15:42.600 Year we'll start to roll out level three, so hands off, 0:15:42.640 --> 0:15:46.320 eyes off for specific domains in this case, you know, 0:15:46.560 --> 0:15:50.080 starting with highways, and that will be true for Gen 0:15:50.120 --> 0:15:51.960 two of R one, It'll be true for the launch 0:15:52.080 --> 0:15:53.880 edition of R two, and will of course be true 0:15:53.880 --> 0:15:56.520 for the varying of the vehicle that has the higher 0:15:56.560 --> 0:16:01.280 comput stack and more capable perception. That more capable perception 0:16:01.360 --> 0:16:03.160 is going to allow for is a higher ceiling of 0:16:03.160 --> 0:16:05.680 what the vehicle can ultimately achieve, and so it'll achieve 0:16:06.360 --> 0:16:10.560 higher levels of capability. And importantly, it serves as a 0:16:10.640 --> 0:16:13.120 really viable part of our data flywheel. 0:16:13.280 --> 0:16:15.000 And so think of v R two fleet. 0:16:15.040 --> 0:16:18.840 It's a high volume product as being part of training 0:16:19.000 --> 0:16:21.920 are what we call our large driving model and the 0:16:22.000 --> 0:16:24.880 existence of a light our plus the enhanced inference just 0:16:24.920 --> 0:16:28.040 allows us to do a better job of capturing lots 0:16:28.080 --> 0:16:30.040 and lots of driving miles to train our model. 0:16:30.920 --> 0:16:35.560 Look, other automakers are probably looking at that flywheel, ah Jay, 0:16:35.680 --> 0:16:37.880 I'm really interested in some of the hints we're getting, 0:16:37.960 --> 0:16:40.720 some of the coverage in the markets about discussions to 0:16:40.840 --> 0:16:44.920 license your software, your autonomy stack, your underlying technology to 0:16:44.960 --> 0:16:45.680 other automakers. 0:16:45.760 --> 0:16:46.520 Is that in the works? 0:16:48.880 --> 0:16:50.520 Well, I think in the fullness of time. 0:16:51.160 --> 0:16:54.000 When you look at what we're building as a business, 0:16:53.400 --> 0:16:57.840 there's really two revenue opportunities. One is we continue to 0:16:57.840 --> 0:17:01.520 grow by selling vehicles. So of course R two is 0:17:01.520 --> 0:17:04.760 the embodiment of that. That's what we're driving towards But 0:17:04.840 --> 0:17:08.600 there's another element of our business which is leveraging the 0:17:08.640 --> 0:17:11.080 technology we're developing. And the first example that was what 0:17:11.119 --> 0:17:13.640 we did with Volkswagen. We did a five point eight 0:17:13.640 --> 0:17:17.919 billion dollar software licensing deal and that allows us to 0:17:17.920 --> 0:17:21.399 deploy this stout, our self driving but our compute stack 0:17:21.640 --> 0:17:26.320 and our software platform, our oas across a wide range 0:17:26.600 --> 0:17:30.240 of electric vehicles within the Volkswagen portfolio. And so we're 0:17:30.240 --> 0:17:32.560 excited about the potential to do more deals like that 0:17:32.640 --> 0:17:36.400 with our operating system in the vehicle, with our vehicle 0:17:36.640 --> 0:17:39.880 and certainly as we think about autonomy, the potential that's 0:17:39.920 --> 0:17:43.400 there for us also license that in the long term. 0:17:44.000 --> 0:17:45.520 Ravian CEO RJ scarrange. 0:17:45.560 --> 0:17:47.240 We'll let you get back to those R twos that 0:17:47.280 --> 0:17:48.480 are rolling off the shop floor. 0:17:48.760 --> 0:17:51.160 We so appreciate your time today from MENINOI. 0:17:51.280 --> 0:17:55.480 Meanwhile, electric vehicle adoption it has been accelerating, particularly in 0:17:55.560 --> 0:17:58.879 China and increasingly spilling into global markets, driven by a 0:17:58.920 --> 0:18:01.720 wave of lower cost models and the country's leading cheaper 0:18:01.720 --> 0:18:02.399 battery production. 0:18:02.680 --> 0:18:04.800 It's the topic of tonight's episode of Primer. 0:18:05.119 --> 0:18:07.879 Take a listen, you're driving a car, you're driving a 0:18:07.920 --> 0:18:12.200 battery and China right now is the leader in batteries. 0:18:12.880 --> 0:18:17.159 China's evs are already cheaper on average than similar gas cards, 0:18:17.880 --> 0:18:20.520 and it's no surprise that four of the top five 0:18:20.600 --> 0:18:24.000 battery makers in the world are located here. Also, one 0:18:24.040 --> 0:18:26.960 of them, BYD, has been jocking with Tesla for years 0:18:27.000 --> 0:18:30.360 for the title of the world's biggest EV maker. All 0:18:30.400 --> 0:18:34.440 that success means lots of competition at home, propelling China 0:18:34.520 --> 0:18:36.359 to look outward for expansion. 0:18:37.040 --> 0:18:38.960 And while some big markets like the US and. 0:18:38.920 --> 0:18:43.480 The EU have resisted China's cheaper evs, many emerging markets 0:18:43.560 --> 0:18:44.320 have brought them in. 0:18:46.000 --> 0:18:48.520 You can catch the full episode of Primer tonight on 0:18:48.560 --> 0:18:52.680 Bloomberg six pm Eastern and on Bloomberg Originals at eight pm. 0:18:52.880 --> 0:18:57.000 Terrific work from Caro and the team. Now coming up dogs. 0:18:57.040 --> 0:19:00.000 The AI moth model the Anthropic warned was two big 0:19:00.160 --> 0:19:05.200 cybersecuritists to release widely has been accessed by unauthorized users. 0:19:05.280 --> 0:19:06.680 That's according to a Bloomberg sauce. 0:19:06.720 --> 0:19:07.639 We have that report. 0:19:07.760 --> 0:19:09.320 Next, this is Bomberg Tech. 0:19:17.280 --> 0:19:20.840 A small group of unauthorized users have accessed to anthropics 0:19:20.920 --> 0:19:23.680 new AI model mythos. It's according to a person familiar 0:19:23.720 --> 0:19:26.639 with the matter, and documentation viewed by Bloomberg News. 0:19:26.840 --> 0:19:28.520 Now, as you probably remember. 0:19:28.320 --> 0:19:31.680 And Throberic has said mythos can enable dangerous cyber attacks, 0:19:31.840 --> 0:19:34.320 and because of that, the company only gave access to 0:19:34.359 --> 0:19:37.520 select users to help them probe their own systems for vulnerabilities. 0:19:37.760 --> 0:19:40.800 Let's get the latest. Bloomberg's AI reporter Rachel Mets. This 0:19:40.920 --> 0:19:44.360 caused quite the stir when you reported it last night. 0:19:44.680 --> 0:19:47.920 And what's so extraordinary is who's done it and how. 0:19:47.720 --> 0:19:50.439 They did it, Rachel. These aren't people wanting to wreak havoc. 0:19:50.560 --> 0:19:53.560 They want to play with a thing exactly. Yeah. 0:19:53.640 --> 0:19:57.440 These are people that enjoy playing with new EI models. 0:19:57.560 --> 0:20:01.600 And they had been try trying to figure out the 0:20:01.720 --> 0:20:06.440 names of various AI models for Anthropic, in particular, through 0:20:06.480 --> 0:20:10.560 a variety of techniques such as having some familiarity with 0:20:10.920 --> 0:20:15.840 what the online location of Anthropics old models looked like, 0:20:16.160 --> 0:20:18.440 you know what that URL would look like, and using 0:20:18.480 --> 0:20:20.800 that to guess as to what it might be in 0:20:20.840 --> 0:20:21.320