WEBVTT - AI Spending Ripples Across Tech Stack; Nvidia Acquires Hugging Face 0:00:05.670 --> 0:00:09.200 Bloomberg Tech is live from the heart of Silicon Valley 0:00:09.740 --> 0:00:12.020 with Ed Ludlow in San Francisco. 0:00:14.730 --> 0:00:17.349 This is Bloomberg Tech coming up from chips to servers 0:00:17.430 --> 0:00:21.329 to software. AI spending is rippling across the tech stack. 0:00:21.530 --> 0:00:24.810 What earnings from Broadcom, Snowflake and HPE tell us about 0:00:24.829 --> 0:00:27.250 the strength of the AI boom? We'll discuss with the 0:00:27.330 --> 0:00:31.470 CEOs of Snowflake and HPE later this hour. Plus, NVIDIA 0:00:31.630 --> 0:00:34.730 has agreed to acquire Hugging Face in a transaction valued 0:00:34.810 --> 0:00:38.410 about $ 13 billion. We speak with Hugging Face co-founder... Thomas 0:00:38.479 --> 0:00:42.500 Wolfe about the deal and self-driving tech startup Wave reaches 0:00:42.520 --> 0:00:45.919 a deal with Uber to offer robo taxis in London 0:00:46.220 --> 0:00:49.500 with a human in the driver's seat. Wave CEO later 0:00:49.900 --> 0:00:53.779 this hour. This is what technology stocks are doing high 0:00:53.840 --> 0:00:56.840 on the Nasdaq 100 modestly six tenths of one percent. 0:00:57.190 --> 0:01:00.630 It is earnings from software to chips to other parts 0:01:00.670 --> 0:01:04.330 of the infrastructure stack that we're focused on. In Snowflake's case, 0:01:04.590 --> 0:01:06.490 this is a blowout. The stock is on track for 0:01:06.530 --> 0:01:09.929 its best day since May and trading at a 2021 high. 0:01:10.790 --> 0:01:13.810 AI is translating into faster growth. The revenue outlook up 0:01:13.860 --> 0:01:17.900 to $ 6 billion for the year. HPE, strong AI server demand, 0:01:18.120 --> 0:01:20.280 but not as strong as Dell. And there wasn't the 0:01:20.319 --> 0:01:22.339 read-through from that name. We're going to speak to the 0:01:22.380 --> 0:01:24.920 CEOs of Snowflake and HPE later in the hour. Then 0:01:24.959 --> 0:01:29.890 there's Broadcom. Huge AI chip growth ahead. Doubling in fiscal 27. 0:01:29.890 --> 0:01:34.089 Doubling again in fiscal 28. but not huge enough for 0:01:34.130 --> 0:01:36.370 this market. And that's where I want to start today's show. 0:01:36.610 --> 0:01:40.110 Let's get more with Bloomberg's Dina Bass. Broadcom is a 0:01:40.190 --> 0:01:44.240 custom silicon story, right? And they posted some big numbers 0:01:44.280 --> 0:01:46.200 in terms of what's going to come, but it just 0:01:46.240 --> 0:01:48.100 wasn't enough for this market. It was a high bar. 0:01:49.300 --> 0:01:49.520 Sure. 0:01:49.600 --> 0:01:51.840 First of all, their forecast for the current quarter, their 0:01:51.880 --> 0:01:55.800 fiscal fourth, was pretty lackluster in terms of both regular 0:01:55.860 --> 0:01:58.630 sales and AI chip sales. And so it That came 0:01:58.670 --> 0:02:01.310 out initially right after the bell and the stock headed downward. 0:02:01.670 --> 0:02:04.110 Then we got into the call and CEO Hock Tan 0:02:04.170 --> 0:02:08.220 gave the forecast that you just repeated about AI chip 0:02:08.240 --> 0:02:10.280 growth for fiscal 27 and 28 doubling each year, which 0:02:12.919 --> 0:02:16.419 You know, again, seemed to at the time, please, investors more. 0:02:16.460 --> 0:02:19.300 The stock started coming back. But this morning it's back 0:02:19.320 --> 0:02:22.820 down again. And, you know, there are some concerns. One, 0:02:22.840 --> 0:02:26.250 in light of NVIDIA's blowout quarter last week that neither 0:02:26.310 --> 0:02:29.030 the Q4 numbers nor perhaps the forecaster may be good 0:02:29.070 --> 0:02:32.420 enough compared to that. Two, some of this was priced 0:02:32.460 --> 0:02:35.740 into the stock already. There's also concern being raised about 0:02:35.840 --> 0:02:39.090 where that AI chip growth is coming from. You know, 0:02:39.100 --> 0:02:42.750 Tan basically said that most of it is coming from 0:02:42.850 --> 0:02:46.230 Anthropic and OpenAI, who are set to become the company's 0:02:46.270 --> 0:02:50.160 number one and number two customers in 27 and 28, respectively. 0:02:50.180 --> 0:02:53.560 So there's some concern about those two as the customers. 0:02:54.740 --> 0:02:59.020 The stock's down 6% right now, pretty much on track 0:02:59.040 --> 0:03:01.750 for its biggest drop since the first week of June. 0:03:01.750 --> 0:03:05.490 $ 115 billion in fiscal 27 on AI chips, $ 230 billion 0:03:05.630 --> 0:03:09.730 in fiscal 28. Those are big numbers, but like NVIDIA 0:03:09.790 --> 0:03:14.070 is doing now $ 100 billion of revenue overall in a quarter. 0:03:14.210 --> 0:03:16.910 Is there any sense that this custom chip game is 0:03:17.030 --> 0:03:19.640 chipping away at NVIDIA's lead in that market, that they're 0:03:19.660 --> 0:03:21.440 gaining market share on Broadcom's side? 0:03:22.960 --> 0:03:25.340 It does seem to be chipping away at pieces of it. 0:03:25.680 --> 0:03:28.620 We've had a lot of discussion from OpenAI over the 0:03:28.660 --> 0:03:30.740 last couple of weeks about the ways in which they 0:03:30.820 --> 0:03:33.750 are going to use the custom chip that they're building 0:03:33.850 --> 0:03:37.070 with Broadcom. But also at the same time, OpenAI will 0:03:37.090 --> 0:03:41.470 talk about how much NVIDIA they're using. So there is that. 0:03:41.630 --> 0:03:43.690 The other issue for Broadcom is that even in the 0:03:43.730 --> 0:03:48.410 custom chip market, they're facing greater competition. Google TPU had 0:03:48.470 --> 0:03:50.970 been their largest customer, or it still is, but is 0:03:51.010 --> 0:03:53.760 set to be replaced by Anthropic. And Google just signed 0:03:53.820 --> 0:03:57.480 a deal with Broadcom's competitor, Marvell. Also more competition for 0:03:57.520 --> 0:04:00.480 Broadcom for MediaTek. Both of those companies are backed by NVIDIA. 0:04:02.170 --> 0:04:05.740 Bloomberg's Dean Abbas, thank you very much indeed. Stay with Bloomberg. 0:04:05.760 --> 0:04:07.480 We're going to break down the other earnings. We're going 0:04:07.500 --> 0:04:10.060 to speak with the Snowflake CEO, Sridhar Ramaswamy, and the 0:04:10.280 --> 0:04:14.440 HPE CEO, Antonio Neri, later this hour. By the way, 0:04:14.620 --> 0:04:16.680 when we came to air, there were some headlines on 0:04:16.700 --> 0:04:21.270 the Bloomberg terminal about chatbot outages. Curiously, the headlines all 0:04:21.310 --> 0:04:25.309 hit at the same time. DownDetector is saying that users 0:04:25.350 --> 0:04:29.789 are reporting issues with open AI platforms and SpaceX AI 0:04:29.950 --> 0:04:33.919 is saying that Grok is experiencing issues and is investigating 0:04:33.980 --> 0:04:38.000 the outage. And Anthropic is saying all within the same 0:04:38.040 --> 0:04:41.610 time span, it's investigating Claude errors and working on a fix. 0:04:41.990 --> 0:04:44.190 Those all hit the Bloomberg terminal literally at the same 0:04:44.230 --> 0:04:46.550 time before the show started. We're trying to work out 0:04:46.589 --> 0:04:49.880 what's going on. with all of the chatbots around the world. 0:04:50.620 --> 0:04:53.140 Let's get back to the AI story. Our next guest 0:04:53.160 --> 0:04:56.680 says we're seeing strong evidence of AI adoption and broadening 0:04:56.960 --> 0:05:00.750 from hardware to select software at the infrastructure layer. That's 0:05:00.810 --> 0:05:04.950 clear evidence of it as AI becoming increasingly integrated into 0:05:04.990 --> 0:05:09.210 business models. Alliance Bernstein, Somatic Innovation Equities, Chief Investment Officer 0:05:09.250 --> 0:05:11.550 Lei Chu is back with us on Bloomberg Tech. The 0:05:11.589 --> 0:05:14.610 common theme across all of those earnings we outlined, Lei, 0:05:14.650 --> 0:05:17.510 is that AI demand seems very strong. Is that what 0:05:17.570 --> 0:05:17.950 you see? 0:05:18.480 --> 0:05:18.740 Yes. 0:05:18.839 --> 0:05:22.380 In fact, we continue to see AI solidly moving into 0:05:22.440 --> 0:05:26.060 the adoption phase. And at this moment, when you look 0:05:26.240 --> 0:05:30.280 across the board, the general trend actually is demand is 0:05:30.380 --> 0:05:34.919 outstripping supply. So when you see companies actually from the 0:05:34.960 --> 0:05:37.800 commentary versus the numbers, the common theme what we are 0:05:37.820 --> 0:05:41.570 seeing is that Demand is so strong that supply simply 0:05:41.610 --> 0:05:44.000 cannot keep up. And that's not just for this year, 0:05:44.420 --> 0:05:47.360 but it's for multiple years to come, whether it's at 0:05:47.380 --> 0:05:50.280 the chip level or other parts of the infrastructure as well. 0:05:51.900 --> 0:05:53.760 I want to go back to NVIDIA last week just 0:05:53.820 --> 0:05:57.270 as an illustrative example, because they said that they would 0:05:57.310 --> 0:06:00.630 have top line growth of 70 percent in the next 0:06:00.670 --> 0:06:04.969 fiscal year. But that was a forecast modeled on supply, 0:06:05.070 --> 0:06:09.060 their ability to supply, not growth modeled on demand. I 0:06:09.140 --> 0:06:12.480 found that math very interesting. How did you interpret it? 0:06:13.110 --> 0:06:15.090 Yes, and then so this is actually we're at a 0:06:15.130 --> 0:06:17.790 very interesting moment and then that moment where you're seeing 0:06:17.830 --> 0:06:20.530 some volatility in the market as well. The market by 0:06:20.550 --> 0:06:24.110 and large, we often, generally speaking, we reward bid and 0:06:24.130 --> 0:06:28.080 raises and we want to see estimate revision, particularly in 0:06:28.140 --> 0:06:32.500 technology sector and fast growing sectors. But currently, we're in 0:06:32.580 --> 0:06:35.599 this really interesting dynamic where the demand is so strong, 0:06:36.060 --> 0:06:39.170 but we simply do not have enough supply to support that. 0:06:39.550 --> 0:06:43.029 So we're in this interesting transition phase where for the 0:06:43.070 --> 0:06:46.290 near term, you don't see as much estimate revision, even 0:06:46.330 --> 0:06:50.410 though the demand outlook is extremely strong. But that just 0:06:50.450 --> 0:06:54.370 means we just have to look at everything in the 0:06:54.410 --> 0:06:57.770 context of valuation as well, and then in the context 0:06:57.810 --> 0:07:00.870 of horizon. You know, when we think about long term 0:07:00.930 --> 0:07:04.900 durability of growth and profitability and combine that with a 0:07:04.920 --> 0:07:08.380 reasonable valuation. And that's when you find the best performing stocks. 0:07:08.500 --> 0:07:10.400 And that's what we are really focused on. 0:07:11.720 --> 0:07:14.680 We started the conversation with your writing about how this 0:07:14.760 --> 0:07:15.520 is broadening out. 0:07:15.700 --> 0:07:15.940 Right. 0:07:16.280 --> 0:07:19.500 Basically from chips to other layers of the stack. How 0:07:19.560 --> 0:07:22.580 do changes in portfolios reflect that right now? 0:07:23.440 --> 0:07:26.710 I think we are in the adoption phase. So when 0:07:26.750 --> 0:07:28.850 you're in the adoption phase and then you look back, 0:07:28.910 --> 0:07:32.620 we started simply at the physical infrastructure layer. But as 0:07:32.720 --> 0:07:35.340 we move to adoption phase, and you've already seen it, 0:07:35.380 --> 0:07:38.980 we went from the broadening out, even at the infrastructure layer, 0:07:39.060 --> 0:07:42.640 we have gone from simply GPU to TPU to CPU 0:07:42.700 --> 0:07:45.390 to the networking layer. But now we're in the phase 0:07:45.430 --> 0:07:48.590 where you're also seeing as enterprise and consumers starting to 0:07:48.730 --> 0:07:52.450 use more AI, you will see opportunities at the infrastructure layer, 0:07:52.590 --> 0:07:55.810 at the software layer. at the cybersecurity layer and many 0:07:55.890 --> 0:07:59.210 other areas as well. And actually, I also think we're 0:07:59.230 --> 0:08:02.550 starting to see evidence of AI being adopted in some 0:08:02.570 --> 0:08:06.030 of the industries. So if you look so far, what 0:08:06.050 --> 0:08:10.250 we have seen is Some industries, such as infotech, financials, 0:08:10.570 --> 0:08:14.290 or healthcare, these are historically some of the fastest adopters 0:08:14.390 --> 0:08:18.070 of new technology. And we do believe that AI is 0:08:18.210 --> 0:08:23.680 once-in-a-lifetime transformative technology. Whoever uses it to their advantage will 0:08:23.720 --> 0:08:27.940 be able to increase their profit pool, accelerate their growth. 0:08:28.220 --> 0:08:30.800 And we're actually seeing early evidence of that. So you 0:08:30.880 --> 0:08:35.280 see the portfolio naturally transition following along this type of 0:08:35.360 --> 0:08:38.189 transformation and along the curve of adoption. 0:08:39.800 --> 0:08:43.830 Alliance Bernstein, Sphematic Innovation Equity, CAI, Laichu, back on the show. 0:08:43.870 --> 0:08:47.490 Thank you very much indeed. Coming up, NVIDIA is going 0:08:47.510 --> 0:08:50.410 to acquire Hugging Face in a transaction valued at about 0:08:50.410 --> 0:08:54.040 $ 13 billion. We're going to speak with Hugging Face's co-founder, 0:08:54.080 --> 0:09:02.890 Thomas Wolff, about the deal. This is Bloomberg Tech. NVIDIA 0:09:02.980 --> 0:09:05.920 has agreed to acquire Hugging Face in a transaction valued 0:09:05.960 --> 0:09:10.040 at about $ 13 billion, confirming earlier Bloomberg reports. Here to 0:09:10.080 --> 0:09:14.640 discuss is Hugging Face co-founder Thomas Wolff, back on Bloomberg Tech. 0:09:15.260 --> 0:09:17.640 There's a piece of this that's really interesting, that Hugging 0:09:17.679 --> 0:09:21.540 Face continues to be Hugging Face, right Thomas? Owned by NVIDIA, 0:09:22.210 --> 0:09:25.089 but is an open platform. Could we start with the 0:09:25.130 --> 0:09:27.970 mechanics of that? How Hugging Face is going to run 0:09:28.030 --> 0:09:31.700 under the proposal that you and Clem and the others agreed? 0:09:32.140 --> 0:09:35.640 Do you and Clem stay with Hugging Face and the 0:09:35.700 --> 0:09:37.580 specifics of how you're going to structure this? 0:09:39.910 --> 0:09:40.370 Yeah, of course. 0:09:40.410 --> 0:09:43.530 So, I mean, the first thing to say is we 0:09:43.790 --> 0:09:47.400 are staying indeed, like all the three co-founders are staying. 0:09:47.420 --> 0:09:51.819 You know, we're doubling down with six years at least 0:09:51.900 --> 0:09:54.640 retention at Hugging Face and the idea is all the 0:09:54.660 --> 0:09:57.250 team is going to join. And the goal is really, 0:09:57.530 --> 0:10:02.110 as Jensen said it in his tweet and interview this morning, 0:10:02.250 --> 0:10:03.890 and as you can read it on the blog post 0:10:03.950 --> 0:10:09.100 as well, is to keep an open, independent, compute agnostic 0:10:09.340 --> 0:10:13.800 platform and infrastructure for Hugging Face. So practically for our user, 0:10:13.840 --> 0:10:17.630 this is going to change. really very little. The main 0:10:17.670 --> 0:10:24.140 difference is we'll have the backing, this power of NVIDIA 0:10:24.540 --> 0:10:27.680 to be able to double down and to accelerate our work. 0:10:30.720 --> 0:10:33.470 What is it that Hugging Face is going to be 0:10:33.530 --> 0:10:37.310 able to do now, Thomas, like either scaling or in 0:10:37.350 --> 0:10:40.069 the speed of movement that you just wouldn't have otherwise 0:10:40.110 --> 0:10:41.850 been able to do independently? 0:10:44.510 --> 0:10:49.179 I think we've been extremely efficient along our journey. We've 0:10:49.200 --> 0:10:52.400 been able to, we've raised a very minimal amount if 0:10:52.420 --> 0:10:56.160 you compare to any AI company nowadays. We've raised less 0:10:56.200 --> 0:10:58.830 than half a billion dollars. I think we've been able 0:10:58.850 --> 0:11:02.670 to achieve amazing results with that. We scaled to 3 0:11:02.670 --> 0:11:05.630 million models on the hub, like 18 million users. We 0:11:05.650 --> 0:11:08.989 have a We have 200,000 companies using us. And I 0:11:09.030 --> 0:11:12.910 think the idea is now what we realized this year 0:11:12.950 --> 0:11:16.330 was that, on the one hand, open source was having 0:11:16.350 --> 0:11:20.650 a huge moment, like open source was really on the rise. 0:11:21.000 --> 0:11:24.880 And also, when we were attacked, you remember earlier this summer, 0:11:24.920 --> 0:11:28.060 we realized that open source was actually critical for safety 0:11:28.160 --> 0:11:31.319 and critical for many aspects. Happy to dive into that 0:11:31.440 --> 0:11:35.300 from safety to company being able to do research to technology. 0:11:36.030 --> 0:11:38.740 basically a company be able to own their data. And 0:11:38.770 --> 0:11:42.420 we thought, okay, we have a few choice here. Either 0:11:42.480 --> 0:11:45.320 we raise more money and we double down or we 0:11:45.420 --> 0:11:48.699 find someone who share our value, share our mission very 0:11:48.740 --> 0:11:51.980 deeply and we partner with them. We were very lucky 0:11:52.020 --> 0:11:56.350 that Jensen was active receptive to this discussion. 0:11:57.250 --> 0:11:59.390 So it sounds like, you know, Hugging Face looked at 0:11:59.630 --> 0:12:03.209 a strategy for the future and approached NVIDIA or Jensen 0:12:03.250 --> 0:12:05.050 and NVIDIA approached Hugging Face? 0:12:06.050 --> 0:12:09.450 Yeah, we approached him. Very open discussion. I think everything 0:12:09.470 --> 0:12:10.210 was on the table. 0:12:11.610 --> 0:12:15.189 We often have, you know, this offer to invest, to 0:12:15.490 --> 0:12:19.390 basically raise more money. And the discussion was, OK, where 0:12:19.429 --> 0:12:22.069 should we go? And at some point we were, you know, 0:12:22.290 --> 0:12:24.610 we're so aligned. There is so many things we can 0:12:24.650 --> 0:12:27.860 do here that let's go all the way. 0:12:29.740 --> 0:12:34.239 What was interesting, as outlined by Jensen in his various communications, like, 0:12:34.260 --> 0:12:37.200 for example, a blog post on LinkedIn, is that the 0:12:37.240 --> 0:12:40.880 Hugging Face community will still have the same choices, right? 0:12:41.380 --> 0:12:45.400 Including the cloud platform backing, the inference platform. In other words, 0:12:45.780 --> 0:12:51.890 there is no mandatory requirement to use NVIDIA Compute. Was 0:12:51.929 --> 0:12:55.350 that something that you guys on your side insisted on, 0:12:55.770 --> 0:12:57.730 or what's the backstory to that? 0:12:59.590 --> 0:13:01.220 Yeah, I mean, we'll talk about that. 0:13:01.250 --> 0:13:05.380 But the idea really here is Jensen believed like us 0:13:05.740 --> 0:13:09.000 in the power of community, the power of collaboration. You know, 0:13:09.040 --> 0:13:12.640 he wrote this great letter about open source model. He 0:13:12.860 --> 0:13:17.809 thinks really widely. We have this same, you know, wide 0:13:17.850 --> 0:13:22.570 scope of thinking. We think there is a huge collaboration 0:13:22.650 --> 0:13:25.090 that we think actually open source models have to be 0:13:25.130 --> 0:13:30.980 built through collaboration that transcends normal competition. So here also 0:13:31.500 --> 0:13:34.680 there is a very big alignment in how we think 0:13:34.740 --> 0:13:40.179 about these problems. Now the exact specifics will be defined 0:13:40.320 --> 0:13:40.800 very soon. 0:13:42.920 --> 0:13:45.830 This is a slight tangent or an aside, but last 0:13:45.870 --> 0:13:47.250 week you were on the show with me and we 0:13:47.260 --> 0:13:50.990 were talking about your robotics business. How does the robotics 0:13:51.050 --> 0:13:56.050 business transfer with NVIDIA? Because you had massive demand for 0:13:56.090 --> 0:14:01.030 Microdot quite quickly. NVIDIA's got massive scale. They are a 0:14:01.050 --> 0:14:04.520 hardware company as well. Explain the opportunity there. 0:14:06.550 --> 0:14:09.050 Yeah, I think you're also the same, you know, you've 0:14:09.090 --> 0:14:12.650 seen Jensen on stage with some ducks, you know, there 0:14:12.670 --> 0:14:15.130 was this Disney, not really ducks, but what they were 0:14:15.190 --> 0:14:19.970 called BDX robot last year, GT. We know Jensen also 0:14:19.990 --> 0:14:23.750 love robotics. We love robotics as well. We see here 0:14:23.840 --> 0:14:26.420 also it's very, you know, it's something that we can 0:14:26.500 --> 0:14:30.120 keep developing very similarly to how we've been building it 0:14:31.040 --> 0:14:33.950 at Hugging Face. I think the surprising news for us 0:14:34.030 --> 0:14:38.690 was that this robotic release almost eclipsed or even, you know, 0:14:39.110 --> 0:14:43.140 was even more successful than some rumors about. 0:14:42.860 --> 0:14:44.700 The join, the acquisition. 0:14:45.160 --> 0:14:49.760 And so it's, I think, a good indication of, you know, 0:14:49.820 --> 0:14:52.210 how we'll be able to manage these different things going 0:14:52.270 --> 0:14:53.310 on forward in the future. 0:14:55.000 --> 0:14:59.000 Thomas, this is a deal, right? $ 13 billion, $ 1 billion 0:14:59.040 --> 0:15:03.540 of which is tied to employee retention through stock. How 0:15:04.020 --> 0:15:07.760 did this discussion go around regulation, antitrust? How do you 0:15:07.800 --> 0:15:12.050 think that your competitors, NVIDIA's competitors, will view this deal? 0:15:12.490 --> 0:15:15.770 And what level of confidence or concern is there on 0:15:15.790 --> 0:15:18.990 this passing through the regulatory process with ease? 0:15:22.420 --> 0:15:23.900 Yeah, all of that is a bit early to tell 0:15:23.940 --> 0:15:28.540 your answer, right? We announced this just today, Ed. I 0:15:28.680 --> 0:15:33.670 think it's a process we'll see. We're quite confident. We think, 0:15:33.710 --> 0:15:37.510 you know, there is a really nice path where basically 0:15:37.550 --> 0:15:42.850 Hugging Face continues to operate as this independent, open platform 0:15:42.890 --> 0:15:45.970 that's been in the past. We think this is a 0:15:46.000 --> 0:15:47.520 very good direction. 0:15:47.540 --> 0:15:48.620 It makes a lot of sense. 0:15:49.660 --> 0:15:52.970 Now, you know, I'm not yet to be to be 0:15:53.710 --> 0:15:56.370 answering really like deep regulation questions right now. 0:15:57.670 --> 0:16:00.460 Okay, so I appreciate that. Very quickly, what do Hugging 0:16:00.540 --> 0:16:03.540 Face staff feel about this? What's the response internally then? 0:16:04.720 --> 0:16:07.890 I mean, ecstatic, you know, you maybe saw the picture 0:16:07.950 --> 0:16:12.130 we shared on Twitter, you know, Jensen joined our old hands. 0:16:13.010 --> 0:16:15.320 For a lot of the team members, it was kind 0:16:15.360 --> 0:16:20.460 of a once-in-a-lifetime opportunity, basically, to talk with Jensen. I 0:16:21.100 --> 0:16:24.850 think the whole team is really very excited about this 0:16:25.330 --> 0:16:26.530 new direction. 0:16:27.030 --> 0:16:28.490 Everyone is ready to. 0:16:28.450 --> 0:16:32.330 Go 10 years again, like building to the next step 0:16:32.450 --> 0:16:36.170 of open-source models. We think this is a great chance 0:16:36.220 --> 0:16:39.220 for the whole open-source model ecosystem, to be honest. 0:16:39.260 --> 0:16:40.720 We think it's kind of incredible. 0:16:42.430 --> 0:16:45.950 Thomas Wolfe of Hugging Face, back on the show a 0:16:45.990 --> 0:16:48.910 week later after the confirmation that NVIDIA has acquired Hugging 0:16:48.950 --> 0:16:52.180 Face for about $ 13 million. Thank you so much. Coming up, 0:16:52.200 --> 0:16:54.780 we're going to go to London. London riders are about 0:16:54.800 --> 0:16:57.480 to get robo-taxi access through the UK capital. 0:16:57.500 --> 0:16:59.780 I'm going to talk to the CEO of Wave. 0:16:59.480 --> 0:17:02.020 About a deal with Uber that's making that possible. This 0:17:02.120 --> 0:17:13.180 is Bloomberg Tech. Self-driving technology startup Wave has reached a 0:17:13.200 --> 0:17:16.480 deal with Uber to offer robotaxi rides in London with 0:17:16.520 --> 0:17:18.720 a safety driver. It's the latest move in the global 0:17:18.760 --> 0:17:23.070 race to commercialize autonomous ride hailing. Wave CEO Alex Kendall 0:17:23.109 --> 0:17:25.730 is back with us on Bloomberg Tech. You know, Alex, 0:17:26.590 --> 0:17:29.090 in covering this industry, I know that this is a 0:17:29.130 --> 0:17:32.230 source of frustration. But the first question is always, what's 0:17:32.270 --> 0:17:35.130 the path for the safety driver coming out of the vehicle? 0:17:35.510 --> 0:17:38.440 Take that and explain the London case study to us. 0:17:40.130 --> 0:17:43.020 Hi, Ed. Well, firstly, as you know, London is an 0:17:43.220 --> 0:17:47.340 incredibly difficult place to drive. It's a 2,000-year-old city. It's 0:17:47.380 --> 0:17:51.750 full of narrow roads, frequent pedestrians and cyclists. And so 0:17:52.270 --> 0:17:54.730 I'm delighted that we've been able to bring our technology here. 0:17:54.750 --> 0:17:57.560 We've been learning to drive here for a decade, and it's, 0:17:57.890 --> 0:17:59.660 I think, one of the hardest places in the world 0:17:59.720 --> 0:18:02.790 to learn to drive. But we're starting with a supervised 0:18:02.830 --> 0:18:06.270 service this week with safety operators. And the path to 0:18:06.350 --> 0:18:09.250 launch a driverless service is really three things. The first 0:18:09.390 --> 0:18:12.260 one is to move to a scalable vehicle platform. Our 0:18:12.280 --> 0:18:15.520 business model works with automakers so that they build the 0:18:15.580 --> 0:18:20.480 robot taxis, no retrofits or built-in mass production hardware. Second, 0:18:20.540 --> 0:18:24.160 of course, validating the safety metrics we need for driverless operation. 0:18:24.180 --> 0:18:27.980 And third, We've got a permit today for supervised operations, 0:18:28.040 --> 0:18:30.840 and we're working with regulators towards a driverless permit. So 0:18:30.920 --> 0:18:33.880 those three things, which we're working on very quickly over 0:18:33.920 --> 0:18:34.359 here in London. 0:18:35.859 --> 0:18:38.119 At the same time, you know, you have multiple projects 0:18:38.160 --> 0:18:40.940 in parallel, right? There is the driverless platform of Nissan, 0:18:41.230 --> 0:18:45.010 for example, where there's a future in London, but other 0:18:45.050 --> 0:18:47.670 cities come into play in the R & D of that. 0:18:48.190 --> 0:18:49.290 Explain that relationship. 0:18:50.810 --> 0:18:53.949 Yeah, that's right. We've... We've looked to build a business 0:18:54.010 --> 0:18:56.990 where we want to provide intelligence across anyone who wants 0:18:57.010 --> 0:19:00.370 to build intelligent machines. And uniquely, we've got an AI 0:19:00.410 --> 0:19:03.409 that can scale from consumer vehicles to robo-taxis. So we 0:19:03.470 --> 0:19:05.550 are deploying our AI in consumer vehicles with a number 0:19:05.570 --> 0:19:09.250 of leading automakers like Nissan and Stellantis. And of course, 0:19:09.869 --> 0:19:13.760 as of today, starting robo-taxi services today in London. And 0:19:13.780 --> 0:19:15.280 by the end of the year, we'll be scaling this 0:19:15.359 --> 0:19:18.240 to Tokyo. And we've agreed with Uber to take it 0:19:18.260 --> 0:19:19.620 to 10 cities beyond that as well. 0:19:20.390 --> 0:19:22.810 It's amazing. Alex, it's been almost 10 years since I 0:19:22.850 --> 0:19:26.670 left London and moved to SF. What's your relationship like 0:19:26.710 --> 0:19:29.340 with that city, with the mayor? But also, you know, 0:19:29.380 --> 0:19:32.580 under a new government with Andy Burnham, is there a 0:19:32.640 --> 0:19:37.500 pathway for sort of comprehensive framework regulation for your industry 0:19:37.660 --> 0:19:39.300 in the UK and other UK cities? 0:19:39.320 --> 0:19:43.290 One thing people might not realize is that earlier this year, 0:19:43.350 --> 0:19:46.929 the UN published a technical regulation that allows for a 0:19:46.970 --> 0:19:50.770 global framework for high levels of autonomy and consumer vehicles 0:19:50.970 --> 0:19:54.840 and robot taxis. And many countries around the world, UK, Europe, Japan, 0:19:56.119 --> 0:19:58.359 all sign up to these regulations. And they also provide 0:19:58.380 --> 0:20:01.950 a basis for in the US as well. So this 0:20:02.010 --> 0:20:05.560 is fantastic. And then the UK itself has been adopting these. 0:20:05.840 --> 0:20:08.080 There was an act passed in 2024, and we've been 0:20:08.119 --> 0:20:11.500 working with Transport for London and government here to bring 0:20:11.540 --> 0:20:14.000 this into fruition. So we're pretty happy with how this 0:20:14.040 --> 0:20:16.720 is growing. And London is one of Uber's five largest 0:20:16.780 --> 0:20:20.639 markets by revenue. So it's a really important market to 0:20:20.680 --> 0:20:22.800 bring autonomy into. I think Londoners are going to love 0:20:22.820 --> 0:20:27.260 this service. And what we've got is a safe and 0:20:27.280 --> 0:20:30.480 responsible plan to bring this technology, introduce it to Londoners 0:20:31.140 --> 0:20:33.220 and people through the UK with regulators here. 0:20:34.119 --> 0:20:37.199 You know, the Uber relationships, key to underscore, right, it's 0:20:37.240 --> 0:20:40.190 through the Uber app. You said a moment ago that 0:20:40.250 --> 0:20:43.510 there's a plan for 10 cities with Uber going forward. 0:20:43.530 --> 0:20:45.990 Are any of those cities in the United States, Alex? 0:20:47.210 --> 0:20:49.230 We will be bringing it globally, including to the US. 0:20:49.270 --> 0:20:52.510 We haven't announced any other specific cities yet, but stay tuned. 0:20:52.550 --> 0:20:53.889 I'll give you a call when we do, Ed. 0:20:55.619 --> 0:20:58.439 Real quick, how are you guys doing on the funding side? 0:20:58.920 --> 0:21:01.660 There was reports that some Wave staff sold off stock 0:21:02.100 --> 0:21:04.780 on the Pisces exchange, for example. Is that an employee 0:21:04.800 --> 0:21:08.109 liquidity thing? Are you out there doing secondaries? What's going on? 0:21:09.590 --> 0:21:14.070 Yeah, we're well funded. We've raised almost $ 3 billion of cash. 0:21:14.130 --> 0:21:17.449 We just did our second employee liquidity round. We've got 0:21:17.470 --> 0:21:20.390 a great global team. And actually, one of the things 0:21:20.430 --> 0:21:22.530 I think we're well set up to do is to 0:21:22.570 --> 0:21:26.560 really continue with our frontier embodied AI science. We've been 0:21:26.580 --> 0:21:29.300 the first team to build vision language action models and 0:21:29.320 --> 0:21:33.209 world models for robotics, and now bring this new contrarian 0:21:33.230 --> 0:21:36.210 approach to the technology to market, as well, of course, 0:21:36.290 --> 0:21:40.100 continuing to grow in production applications like in consumer vehicles 0:21:40.140 --> 0:21:43.679 all around the world and for robo-taxi services. So it's 0:21:43.720 --> 0:21:46.109 full steam ahead. It's been a phenomenal year for our 0:21:46.130 --> 0:21:48.310 team so far, and this is a pretty proud moment 0:21:48.350 --> 0:21:48.729 for us. 0:21:49.650 --> 0:21:52.510 Alex, just very quickly, what you just said, world models, 0:21:52.609 --> 0:21:55.350 remind the audience what Wave's core competence is. 0:21:57.090 --> 0:21:59.609 What we've built is an end-to-end AI driver. So it's 0:21:59.650 --> 0:22:02.520 a single neural network that's capable of reasoning and making 0:22:02.560 --> 0:22:05.960 decisions in real time. It can generalize across any vehicle 0:22:05.980 --> 0:22:08.540 we work with, small, large vehicles all around the world. 0:22:09.040 --> 0:22:11.060 We're the first company to have driven this technology in 0:22:11.119 --> 0:22:15.149 over 500 cities, zero shot. So no high-definite maps. It's 0:22:15.190 --> 0:22:17.470 an AI that's capable of driving any vehicle all the 0:22:17.490 --> 0:22:18.250 way around the world. 0:22:19.410 --> 0:22:21.700 Alex Kendall, Wave CEO, back on the show. It's great 0:22:21.720 --> 0:22:24.639 to have you now deployed modestly in London. Coming up, 0:22:24.700 --> 0:22:28.400 snowflakes earnings, beating analyst estimates, a big move in the stock. 0:22:28.420 --> 0:22:31.389 We're going to speak to the CEO, Sridhar Ramaswamy. That's next. 0:22:31.910 --> 0:22:34.389 Halftime here in the city of San Francisco. And this 0:22:34.410 --> 0:22:42.220 is what markets look like. This is Bloomberg Tech. Welcome 0:22:42.240 --> 0:22:45.440 back to Bloomberg Tech. The NASDAQ 100, really tech-heavy index, 0:22:45.520 --> 0:22:48.700 pushing up to session highs, a gain of more than 1% now. 0:22:49.080 --> 0:22:50.840 A lot of what's happening in the market is Fed 0:22:50.880 --> 0:22:53.530 speak and what the Fed will or won't do in 0:22:53.550 --> 0:22:56.590 terms of the outlook for rates. But earnings is also 0:22:56.710 --> 0:23:00.609 a very big part of this story. Elsewhere, what is 0:23:00.650 --> 0:23:03.840 happening in the release of different generations of AI model, 0:23:04.180 --> 0:23:07.659 also massively important. Meta's up 4%. This one was a 0:23:07.680 --> 0:23:10.659 big news item this morning. It's launched its latest and 0:23:10.700 --> 0:23:13.420