WEBVTT - ASML Boosts Outlook as AI Chip Demand Soars

0:00:02.520 --> 0:00:04.210
<v Speaker 1>Bloomberg Audio Studios.

0:00:04.690 --> 0:00:05.370
<v Speaker 2>Podcasts.

0:00:05.690 --> 0:00:06.210
<v Speaker 3>Radio.

0:00:06.650 --> 0:00:07.070
<v Speaker 1>News.

0:00:11.050 --> 0:00:16.980
<v Speaker 4>Bloomberg Tech is live from the heart of Silicon Valley

0:00:17.500 --> 0:00:19.779
<v Speaker 4>with Ed Ludlow in San Francisco.

0:00:22.390 --> 0:00:25.700
<v Speaker 1>This is Bloomberg Tech. Coming up, ASML raises its sales

0:00:25.730 --> 0:00:29.280
<v Speaker 1>forecast for 2026 and aims to increase production of its

0:00:29.320 --> 0:00:33.000
<v Speaker 1>chip-making machines to keep up with soaring AI spending. Plus,

0:00:33.020 --> 0:00:35.620
<v Speaker 1>we hear from NVIDIA CEO Jensen Wang, fresh on the

0:00:35.659 --> 0:00:39.620
<v Speaker 1>heels of announcing plans to invest $ 2 billion in European deals.

0:00:39.950 --> 0:00:42.850
<v Speaker 1>And we take you inside of the making of Grand

0:00:42.909 --> 0:00:46.110
<v Speaker 1>Theft Auto 6, expected to be the biggest video game

0:00:46.150 --> 0:00:50.530
<v Speaker 1>blockbuster of all time. Okay, welcome to Wednesday. I'm taking

0:00:50.550 --> 0:00:54.440
<v Speaker 1>a look at shares of ASML. Up 2.6% at one

0:00:54.480 --> 0:00:57.800
<v Speaker 1>point in the session, up 3.6% in Europe and Amsterdam,

0:00:57.840 --> 0:01:01.440
<v Speaker 1>and one point fell 1%. The producer of lithography machines,

0:01:01.520 --> 0:01:05.170
<v Speaker 1>chip making tools to manufacture cutting edge chips, raised its

0:01:05.209 --> 0:01:08.630
<v Speaker 1>sales outlook, but said it's struggling to meet demand. Joining

0:01:08.650 --> 0:01:12.089
<v Speaker 1>us for the Deets, Bloomberg's Sarah Jacob, who is in Amsterdam.

0:01:12.110 --> 0:01:14.970
<v Speaker 1>For ASML, this is all about capacity, right? They had

0:01:15.030 --> 0:01:17.550
<v Speaker 1>plans to boost capacity. They're saying that they're still in

0:01:17.569 --> 0:01:20.450
<v Speaker 1>this place where demand is running ahead of their ability

0:01:20.500 --> 0:01:21.899
<v Speaker 1>to supply. Give us the basics.

0:01:22.319 --> 0:01:22.619
<v Speaker 5>Hi.

0:01:22.720 --> 0:01:23.640
<v Speaker 1>Yes, that's right.

0:01:23.980 --> 0:01:26.940
<v Speaker 6>So ASML, again, basically just, you know, like you said,

0:01:27.000 --> 0:01:30.440
<v Speaker 6>raising the low end of its annual sales forecast for

0:01:30.440 --> 0:01:35.240
<v Speaker 6>2026 because it seems soaring artificial intelligence-linked demand. The CEO

0:01:35.260 --> 0:01:38.440
<v Speaker 6>today saying that he's getting, you know, customers who are

0:01:38.550 --> 0:01:41.890
<v Speaker 6>asking for more tools every day, and it's become the

0:01:41.910 --> 0:01:45.170
<v Speaker 6>key focus of life of ASML to get that capacity going.

0:01:45.630 --> 0:01:49.400
<v Speaker 6>Just as a reminder, in July, ASML said that it

0:01:49.440 --> 0:01:52.820
<v Speaker 6>expected about 65 units in terms of capacity for 2026,

0:01:52.800 --> 0:01:53.720
<v Speaker 6>a 30% increase on that for 2027, and it was

0:01:53.740 --> 0:02:01.850
<v Speaker 6>investigating a further 30% increase for 2028. Today, it says

0:02:01.890 --> 0:02:05.890
<v Speaker 6>that those plans and those current supply patterns would not

0:02:05.950 --> 0:02:09.540
<v Speaker 6>be enough to fulfill demand from what it's hearing from customers.

0:02:09.580 --> 0:02:15.360
<v Speaker 6>So now they're investigating going beyond 30% in 2028. So

0:02:15.419 --> 0:02:18.450
<v Speaker 6>that's going to be interesting because it's hearing more from

0:02:18.510 --> 0:02:22.550
<v Speaker 6>its customers in terms of the short-term and the medium-term requirements.

0:02:23.880 --> 0:02:27.169
<v Speaker 1>Away from the financials, China is the story, right? There

0:02:27.230 --> 0:02:31.310
<v Speaker 1>is a piece of proposed legislation to deepen restrictions, export

0:02:31.330 --> 0:02:35.730
<v Speaker 1>restrictions on DUV, emerging machines to China. This came up

0:02:35.790 --> 0:02:38.940
<v Speaker 1>in the call. What did ASML have to say about China, Sarah?

0:02:41.000 --> 0:02:41.280
<v Speaker 7>Yes.

0:02:41.440 --> 0:02:44.680
<v Speaker 6>So China, again, the third largest market for ASML in

0:02:44.700 --> 0:02:48.639
<v Speaker 6>the third quarter after South Korea and Taiwan, still a

0:02:48.680 --> 0:02:52.690
<v Speaker 6>significant contributor. And then we spoke to Roger Dessen, who

0:02:52.770 --> 0:02:55.990
<v Speaker 6>is the CFO of ASML, in a call with journalists today.

0:02:56.010 --> 0:02:58.850
<v Speaker 6>And he basically said, according to him, they believe that

0:02:58.870 --> 0:03:02.290
<v Speaker 6>the Match Act is unbalanced. And he said that, you know,

0:03:02.510 --> 0:03:07.980
<v Speaker 6>by introducing such restrictions, essentially, you could slow it going

0:03:08.060 --> 0:03:10.280
<v Speaker 6>to China, but it would sort of increase, but not

0:03:10.380 --> 0:03:13.519
<v Speaker 6>stop what China can do in developing technology and might actually,

0:03:13.940 --> 0:03:17.560
<v Speaker 6>you know, sort of help homegrown technology in that regard.

0:03:18.180 --> 0:03:21.800
<v Speaker 6>And by preventing, you know, ASML's ability to service its

0:03:21.880 --> 0:03:24.620
<v Speaker 6>own tools in China, there could potentially be, you know,

0:03:24.700 --> 0:03:26.780
<v Speaker 6>intellectual property leakage or theft.

0:03:28.320 --> 0:03:30.770
<v Speaker 1>Bloomberg's Sarah Jacobs out in Amsterdam. A very long day

0:03:30.810 --> 0:03:34.630
<v Speaker 1>for you. ASML, Europe's most valuable company, reporting its earnings.

0:03:34.910 --> 0:03:37.870
<v Speaker 1>Let's hear from ASML CEO Christophe Fouguet, who said in

0:03:37.930 --> 0:03:40.680
<v Speaker 1>a video released with the company's results that customers are

0:03:40.780 --> 0:03:43.980
<v Speaker 1>asking every day for more tools to meet their own needs.

0:03:45.380 --> 0:03:48.840
<v Speaker 8>Today, they expect that their supply will not meet the

0:03:48.880 --> 0:03:52.820
<v Speaker 8>demand at least until 2028. So this is a very,

0:03:52.860 --> 0:03:55.940
<v Speaker 8>very strong demand on their side. So a lot of

0:03:56.020 --> 0:03:59.040
<v Speaker 8>our discussion with our customer now are shifting, in fact,

0:03:59.320 --> 0:04:01.520
<v Speaker 8>beyond 2027. A lot of order intake for 2028, but

0:04:01.540 --> 0:04:03.100
<v Speaker 8>also some strong interest to already place order for 2029

0:04:03.100 --> 0:04:05.740
<v Speaker 8>and sometime even 2030.

0:04:11.580 --> 0:04:14.740
<v Speaker 1>So ASML already talking to its customers about 2029. Let's

0:04:14.780 --> 0:04:17.640
<v Speaker 1>discuss what that means for ASML, the AI boom with

0:04:17.700 --> 0:04:22.990
<v Speaker 1>Antonella Frangelo and Listerne and Taser Sanpaolo. Again, it's about capacity, right?

0:04:23.010 --> 0:04:25.250
<v Speaker 1>And I think what the market's really struggling with here

0:04:25.330 --> 0:04:28.270
<v Speaker 1>is like your customers are asking for more. You're trying

0:04:28.290 --> 0:04:32.430
<v Speaker 1>to address the capacity issue. Where can they find more capacity?

0:04:35.770 --> 0:04:39.950
<v Speaker 9>Hi, thank you for this opportunity. This is a key

0:04:39.990 --> 0:04:46.990
<v Speaker 9>point for them because it's an opportunity to increase their

0:04:47.029 --> 0:04:49.729
<v Speaker 9>capacity to match the strong orders demand.

0:04:49.790 --> 0:04:50.610
<v Speaker 1>What can they do?

0:04:51.270 --> 0:04:55.310
<v Speaker 9>They can optimize their.

0:04:56.630 --> 0:04:57.290
<v Speaker 1>Production.

0:04:57.610 --> 0:05:03.909
<v Speaker 9>They can release product upgrades. They can they are already

0:05:04.450 --> 0:05:11.400
<v Speaker 9>receiving strong demand for productivity.

0:05:10.820 --> 0:05:15.500
<v Speaker 1>For product upgrades, which.

0:05:15.420 --> 0:05:22.990
<v Speaker 9>Is the reason why in the third quarter the installed

0:05:23.029 --> 0:05:30.969
<v Speaker 9>base management where I then expected. they are really pushing

0:05:31.190 --> 0:05:35.500
<v Speaker 9>their effort to increase the capacity which remains for them

0:05:35.720 --> 0:05:39.920
<v Speaker 9>and for the whole value chain, the key constraint to

0:05:39.960 --> 0:05:42.180
<v Speaker 9>match demand, which remains very strong.

0:05:43.529 --> 0:05:49.430
<v Speaker 1>Antonella, how unusual is it for ASML or anyone in

0:05:49.529 --> 0:05:52.520
<v Speaker 1>the chip making equipment space to already be talking to

0:05:52.580 --> 0:05:53.599
<v Speaker 1>customers about.

0:05:53.400 --> 0:05:53.920
<v Speaker 2>2029, 2030?

0:05:53.920 --> 0:05:57.400
<v Speaker 1>They haven't taken orders, right? They're pretty clear on that.

0:05:57.640 --> 0:06:01.520
<v Speaker 1>But they are talking to customers about 2029 and 2030. Yeah, this.

0:06:03.339 --> 0:06:07.659
<v Speaker 9>Is, you know, something which is becoming quite, you know,

0:06:08.260 --> 0:06:16.770
<v Speaker 9>common to hear from testing equipment players. We heard from

0:06:16.800 --> 0:06:23.540
<v Speaker 9>many of them speaking about full capacity already reached in

0:06:23.560 --> 0:06:27.740
<v Speaker 9>terms of orders for 2027 and still very stretched for 2028.

0:06:33.720 --> 0:06:38.700
<v Speaker 9>You have to consider that we are across an unprecedented

0:06:38.740 --> 0:06:46.989
<v Speaker 9>tech revolution with AI deployment. It's a multi-year upcycle, so

0:06:47.910 --> 0:06:53.180
<v Speaker 9>it's something that we have to look at a multi-year

0:06:53.279 --> 0:06:58.279
<v Speaker 9>time horizon and for it's correct to look at 2029, 2030,

0:06:58.279 --> 0:07:07.470
<v Speaker 9>and even beyond in that case, considering the potential technological

0:07:07.510 --> 0:07:12.530
<v Speaker 9>upgrades they are implementing and the new initiatives they launched

0:07:12.730 --> 0:07:15.780
<v Speaker 9>with TSMC and Samsung.

0:07:17.420 --> 0:07:21.140
<v Speaker 1>And so let's talk about ASML's customers, right? We know

0:07:21.180 --> 0:07:28.550
<v Speaker 1>who they are. High concentration, TSMC, Samsung, and Intel. Of

0:07:28.610 --> 0:07:31.710
<v Speaker 1>those three, do we have a sense of who needs

0:07:32.210 --> 0:07:36.870
<v Speaker 1>these lithography machines most? Who's pushing ASML hardest?

0:07:39.040 --> 0:07:43.080
<v Speaker 9>They don't provide this information, but we can deduct that

0:07:43.140 --> 0:07:48.350
<v Speaker 9>the memory makers are now running ahead of all the others.

0:07:48.530 --> 0:07:53.790
<v Speaker 9>They confirmed that during today's call. But also advanced logic

0:07:54.210 --> 0:08:00.960
<v Speaker 9>are needed for the advanced chips. So at the end,

0:08:01.220 --> 0:08:07.810
<v Speaker 9>it's something that is related to all the main advanced

0:08:07.910 --> 0:08:09.170
<v Speaker 9>chip manufacturers.

0:08:10.500 --> 0:08:14.660
<v Speaker 1>ASML is Europe's most valuable company, publicly traded company. It's

0:08:14.720 --> 0:08:18.420
<v Speaker 1>a stock that's up 67% year to date. But what

0:08:18.440 --> 0:08:21.300
<v Speaker 1>I found interesting in kind of the bigger picture story

0:08:22.040 --> 0:08:26.550
<v Speaker 1>of what's happening in the chip equipment makers, the semiconductor

0:08:26.570 --> 0:08:30.190
<v Speaker 1>industry overall is ASML hit a record in June. And

0:08:30.290 --> 0:08:33.330
<v Speaker 1>it is essentially traded sideways since it's made attempts to

0:08:33.350 --> 0:08:36.510
<v Speaker 1>get to those levels, but not hit fresh records since

0:08:36.550 --> 0:08:39.630
<v Speaker 1>that June 30th record. Where's your head at with the

0:08:39.650 --> 0:08:43.750
<v Speaker 1>valuation of this company right now and the forward trajectory

0:08:43.790 --> 0:08:44.530
<v Speaker 1>of its growth?

0:08:45.890 --> 0:08:50.670
<v Speaker 9>Yeah, the decline in the share price from the May

0:08:50.730 --> 0:08:58.840
<v Speaker 9>peak is something that also regarded the other most of

0:08:59.059 --> 0:09:06.670
<v Speaker 9>semiconductor players. But considering the strong earnings upgrades we have

0:09:06.730 --> 0:09:11.650
<v Speaker 9>implemented since then, now valuation looks very appealing in my view.

0:09:12.420 --> 0:09:17.040
<v Speaker 9>The stock is trading at multiples, which are in line

0:09:17.240 --> 0:09:22.900
<v Speaker 9>with the last five-year average, which in our view doesn't

0:09:23.179 --> 0:09:28.860
<v Speaker 9>take into consideration properly the fact that we are now

0:09:29.020 --> 0:09:32.940
<v Speaker 9>in an up cycle. So, the stock should trade above

0:09:33.280 --> 0:09:39.450
<v Speaker 9>the historical average. Moreover, the stock has almost closed the

0:09:39.500 --> 0:09:45.189
<v Speaker 9>historical premium to the US peers, which we also consider

0:09:45.370 --> 0:09:48.230
<v Speaker 9>as not, you.

0:09:48.150 --> 0:09:49.210
<v Speaker 1>Know, adequate.

0:09:49.250 --> 0:09:54.579
<v Speaker 9>So, for all these reasons, we believe that stock correction,

0:09:54.720 --> 0:09:59.420
<v Speaker 9>and on the other hand, the strong earnings upgrades. and

0:09:59.640 --> 0:10:04.380
<v Speaker 9>the potential further upside ahead could prompt a re-rating of

0:10:04.400 --> 0:10:05.400
<v Speaker 9>the share price.

0:10:06.200 --> 0:10:09.600
<v Speaker 1>And so let's talk about China. You heard Sarah's reporting

0:10:09.800 --> 0:10:13.720
<v Speaker 1>on ASML's position on the Match Act and the burden

0:10:13.760 --> 0:10:17.540
<v Speaker 1>that they feel it unduly places on them. China was 23%

0:10:17.540 --> 0:10:19.620
<v Speaker 1>of 3Q system sales, and by the end of the year,

0:10:19.640 --> 0:10:24.260
<v Speaker 1>it will be 20%. What outlook do you have for

0:10:24.370 --> 0:10:25.800
<v Speaker 1>ASML's business in China?

0:10:28.050 --> 0:10:32.450
<v Speaker 9>The company was very detailed today, and they said that

0:10:32.730 --> 0:10:37.150
<v Speaker 9>China should account for around 20% of sales this year.

0:10:37.210 --> 0:10:40.929
<v Speaker 9>They had been more cautious in previous quarters, but this

0:10:41.000 --> 0:10:46.160
<v Speaker 9>quarter China performed a bit better than, for example, the

0:10:46.200 --> 0:10:50.459
<v Speaker 9>second quarter. The key point on China is not short-term,

0:10:50.580 --> 0:10:55.400
<v Speaker 9>but is on the, as you said, the match act.

0:10:56.220 --> 0:10:59.120
<v Speaker 9>And in my view, this is the main reason why

0:10:59.600 --> 0:11:06.370
<v Speaker 9>the stock is currently not expressing its full potential, because

0:11:06.809 --> 0:11:14.620
<v Speaker 9>investors are fearing that this implementation of this new regulatory

0:11:16.480 --> 0:11:22.780
<v Speaker 9>constraint could have an impact, a double-digit impact on ASML.

0:11:22.840 --> 0:11:29.209
<v Speaker 9>And Maybe we'll have more clarity on that after the

0:11:29.270 --> 0:11:34.790
<v Speaker 9>midterm elections. So this could be one of the reasons

0:11:34.850 --> 0:11:40.400
<v Speaker 9>why the stock is still a bit kept for.

0:11:40.300 --> 0:11:40.940
<v Speaker 6>The time being.

0:11:41.740 --> 0:11:44.060
<v Speaker 1>Antonella Frangila from Intesa San Paolo. It's great to have

0:11:44.080 --> 0:11:46.260
<v Speaker 1>you with us here on Bloomberg Tech. Thank you very much. Now,

0:11:46.300 --> 0:11:48.050
<v Speaker 1>coming up, we're going to hear from NVIDIA CEO Jensen

0:11:48.070 --> 0:11:51.390
<v Speaker 1>Wang as he announces plans to invest an additional $ 2

0:11:51.090 --> 0:11:54.589
<v Speaker 1>billion in European deals during a tour of the region.

0:11:54.610 --> 0:11:57.220
<v Speaker 1>A little bit of gen sanity going on in the continent.

0:11:57.500 --> 0:12:11.240
<v Speaker 1>That's next. This is Bloomberg Tech. Let's take a look

0:12:11.260 --> 0:12:14.960
<v Speaker 1>at today's relatively big number, $ 21 billion. That's how much

0:12:15.000 --> 0:12:18.520
<v Speaker 1>Japan could put behind two new global foundries chip plants

0:12:18.900 --> 0:12:22.160
<v Speaker 1>in upstate New York. The agreement taps Japan's $ 550 billion

0:12:22.179 --> 0:12:26.040
<v Speaker 1>investment pledge to the US, boosting American production of chips

0:12:26.140 --> 0:12:30.170
<v Speaker 1>used in everything from cars to communications gear. And Washington

0:12:30.450 --> 0:12:32.270
<v Speaker 1>is getting a piece of the action. The US government

0:12:32.290 --> 0:12:36.209
<v Speaker 1>will take a roughly 4% stake in global foundries. deepening

0:12:36.250 --> 0:12:39.230
<v Speaker 1>its direct investment into American chip making. Shares of Global

0:12:39.270 --> 0:12:41.670
<v Speaker 1>Foundries up 4%. There was a much bigger spike in

0:12:41.690 --> 0:12:44.069
<v Speaker 1>the pre-market. They kind of fell away pretty quickly. There's

0:12:44.090 --> 0:12:47.360
<v Speaker 1>a lot to understand here. Bloomberg's Ian King, thankfully, is

0:12:47.400 --> 0:12:51.939
<v Speaker 1>with us. In answering, explain Global Foundries' place in the

0:12:52.140 --> 0:12:55.920
<v Speaker 1>sort of chip making world. But, you know, this is

0:12:56.059 --> 0:12:59.620
<v Speaker 1>them establishing more of a domestic footprint, right? But with

0:12:59.640 --> 0:13:00.199
<v Speaker 1>Japanese money.

0:13:00.440 --> 0:13:00.640
<v Speaker 2>Yeah.

0:13:00.920 --> 0:13:05.309
<v Speaker 10>Global Foundries is basically a foundry. It's a TSMC type company.

0:13:06.110 --> 0:13:10.990
<v Speaker 10>although it concentrates on less advanced chip manufacturing, but like TSMT,

0:13:11.030 --> 0:13:14.469
<v Speaker 10>tries to supply as many people as it can.

0:13:14.470 --> 0:13:17.630
<v Speaker 1>$ 21 billion, actually, it's not that big a number when

0:13:17.650 --> 0:13:19.670
<v Speaker 1>you think about what it takes to stand up a

0:13:19.710 --> 0:13:23.010
<v Speaker 1>fab and get it online and ramp. There was a

0:13:23.070 --> 0:13:27.429
<v Speaker 1>separate kind of parallel headline about money that Global Foundries

0:13:27.470 --> 0:13:30.390
<v Speaker 1>had received through the CHIPS Act, CHIPS Act from the

0:13:30.429 --> 0:13:33.510
<v Speaker 1>Biden administration, that essentially, as I understand it now, freer

0:13:33.570 --> 0:13:35.590
<v Speaker 1>to spend it on different things.

0:13:36.170 --> 0:13:37.410
<v Speaker 2>That's right. I mean, there's a couple of things.

0:13:37.410 --> 0:13:41.939
<v Speaker 10>$ 21 billion, if you're TSMC or Samsung or Intel, isn't

0:13:41.980 --> 0:13:44.220
<v Speaker 10>enough to even buy you one plant. But in the

0:13:44.340 --> 0:13:48.130
<v Speaker 10>kind of chip that Global Foundry specialize in, it is.

0:13:48.429 --> 0:13:50.430
<v Speaker 10>It's a good chunk of money and it's an order

0:13:50.470 --> 0:13:52.890
<v Speaker 10>of magnitude change for them in terms.

0:13:52.690 --> 0:13:55.490
<v Speaker 2>Of what they can deploy. And yes, because They've got

0:13:55.530 --> 0:13:55.910
<v Speaker 2>this money.

0:13:55.950 --> 0:13:58.580
<v Speaker 10>They can now use the Chips Act money for other

0:13:58.630 --> 0:14:00.540
<v Speaker 10>specialist projects that they're working on.

0:14:00.940 --> 0:14:04.459
<v Speaker 1>This is two projects in upstate New York since this

0:14:04.500 --> 0:14:08.880
<v Speaker 1>kind of focus on re-onshoring American chip making capacity. Are

0:14:08.920 --> 0:14:11.650
<v Speaker 1>these just announcements from Global Foundries or in the kind

0:14:11.690 --> 0:14:14.210
<v Speaker 1>of last few years, have they ramped up what they're

0:14:14.230 --> 0:14:15.790
<v Speaker 1>doing here domestically?

0:14:16.090 --> 0:14:18.570
<v Speaker 2>I mean, they're kind of an unusual company.

0:14:18.650 --> 0:14:22.400
<v Speaker 10>Remember, they started off life as the plants of AMD,

0:14:22.980 --> 0:14:25.479
<v Speaker 10>bought by the government of Abu Dhabi. Then they bought

0:14:25.500 --> 0:14:29.080
<v Speaker 10>Charter Semiconductor in Singapore. So then they built this plant

0:14:29.280 --> 0:14:31.320
<v Speaker 10>in upstate New York, and now they're adding to that.

0:14:31.360 --> 0:14:34.200
<v Speaker 2>So, you know, the name's above the door.

0:14:34.240 --> 0:14:37.260
<v Speaker 10>It is Global Foundries, but they are very much riding

0:14:37.280 --> 0:14:40.130
<v Speaker 10>this wave of public money being used to bring back

0:14:40.210 --> 0:14:41.130
<v Speaker 10>this to the U.S.

0:14:41.410 --> 0:14:44.050
<v Speaker 1>The final element of this story is bringing back to

0:14:44.070 --> 0:14:47.150
<v Speaker 1>the U.S. with the U.S. taking a stake in global foundries.

0:14:47.670 --> 0:14:49.610
<v Speaker 1>Explain that landscape a bit, because it's not just global

0:14:49.630 --> 0:14:52.270
<v Speaker 1>foundries where the U.S. government has made an investment.

0:14:52.490 --> 0:14:54.930
<v Speaker 10>Yeah, I mean, the obvious other one is Intel, where

0:14:55.090 --> 0:14:59.670
<v Speaker 10>basically the Trump administration helped Intel out, gave Intel more

0:14:59.720 --> 0:15:01.420
<v Speaker 10>freedom and said, in return, the.

0:15:01.400 --> 0:15:03.080
<v Speaker 2>Taxpayer gets a slice of the action.

0:15:03.100 --> 0:15:06.040
<v Speaker 10>And as we've seen, the president has talked up the

0:15:06.080 --> 0:15:09.400
<v Speaker 10>returns that they've already sort of realized from that investment.

0:15:09.540 --> 0:15:11.980
<v Speaker 10>And we're seeing an attempt to repeat that, albeit at

0:15:12.020 --> 0:15:12.920
<v Speaker 10>a smaller scale here.

0:15:13.200 --> 0:15:15.240
<v Speaker 1>Okay, Bloomberg's Ian King, thank you very much for being

0:15:15.280 --> 0:15:18.290
<v Speaker 1>with us on set. NVIDIA CEO Jensen Wang says the

0:15:18.330 --> 0:15:22.450
<v Speaker 1>company's planning to announce $ 2 billion in European investments over

0:15:22.470 --> 0:15:25.230
<v Speaker 1>the next week. At a press gaggle in Paris yesterday,

0:15:25.250 --> 0:15:28.050
<v Speaker 1>Wang reiterated that this will add to the $ 8 billion

0:15:28.070 --> 0:15:30.950
<v Speaker 1>in investments the chip makers already put into the continent

0:15:30.970 --> 0:15:34.620
<v Speaker 1>in the last year. This is the beginning, the beginning

0:15:34.760 --> 0:15:36.060
<v Speaker 1>of AI Europe.

0:15:36.100 --> 0:15:40.300
<v Speaker 11>In the last year alone, we have invested $ 8 billion

0:15:40.990 --> 0:15:44.490
<v Speaker 11>in European startups. And in the next week, I'm going

0:15:44.510 --> 0:15:48.170
<v Speaker 11>to announce another two more billion dollars. And so we've

0:15:48.190 --> 0:15:52.010
<v Speaker 11>invested tremendously in just about every single layer of the

0:15:52.110 --> 0:15:54.250
<v Speaker 11>AI infrastructure here in Europe.

0:15:54.430 --> 0:15:55.660
<v Speaker 1>And so the next layer.

0:15:55.510 --> 0:15:58.680
<v Speaker 11>Above that is investments. These are going to be the

0:15:58.780 --> 0:16:00.760
<v Speaker 11>superstar companies of the next generation.

0:16:00.920 --> 0:16:03.320
<v Speaker 2>What about the speculative bubble on AI?

0:16:03.820 --> 0:16:04.800
<v Speaker 1>Are you frightened about it?

0:16:05.020 --> 0:16:09.650
<v Speaker 11>Today, demand is far greater than supply. And the reason

0:16:09.670 --> 0:16:11.170
<v Speaker 11>for that is because we're in the beginning of this

0:16:11.230 --> 0:16:13.810
<v Speaker 11>new industrial build-out. And so we need to build AI

0:16:13.870 --> 0:16:17.550
<v Speaker 11>infrastructure around the world. And so demand is much, much

0:16:17.580 --> 0:16:20.540
<v Speaker 11>greater than supply. Someday, of course, the two will come

0:16:20.560 --> 0:16:24.760
<v Speaker 11>into balance. Now, whether it's five years from now or seven,

0:16:24.820 --> 0:16:27.340
<v Speaker 11>ten years from now, which is kind of the duration

0:16:27.380 --> 0:16:29.900
<v Speaker 11>of a new technology build-out, except this one requires a

0:16:29.960 --> 0:16:34.000
<v Speaker 11>lot of factories and requires energy and construction. So there

0:16:34.040 --> 0:16:38.990
<v Speaker 11>are many natural limiters. that keeps the industry from building

0:16:39.030 --> 0:16:43.600
<v Speaker 11>too fast. But eventually, it surely will happen that demand

0:16:43.640 --> 0:16:46.360
<v Speaker 11>and supply will come into balance. But you're not going

0:16:46.380 --> 0:16:47.570
<v Speaker 11>to see that in the next several years.

0:16:49.460 --> 0:16:52.990
<v Speaker 1>That was NVIDIA CEO Jensen Wang. Applied Materials gave a

0:16:53.000 --> 0:16:56.790
<v Speaker 1>bullish outlook on spending of wafer fab equipment, but said

0:16:56.830 --> 0:16:59.850
<v Speaker 1>a number of factors make it harder to provide a

0:16:59.910 --> 0:17:03.800
<v Speaker 1>detailed outlook. Overall, yesterday's investor briefing was kind of light

0:17:03.860 --> 0:17:07.119
<v Speaker 1>on specifics. Applied Materials CEO Gary Dickinson spoke with us

0:17:07.160 --> 0:17:09.139
<v Speaker 1>yesterday after the event and gave a little bit more.

0:17:10.220 --> 0:17:13.920
<v Speaker 7>We didn't give a specific forecast because it's changing so quickly.

0:17:14.100 --> 0:17:17.560
<v Speaker 7>The workloads are changing quickly. All of these technologies, models

0:17:17.619 --> 0:17:23.870
<v Speaker 7>are changing quickly. So picking a precise point is really challenging. However,

0:17:23.940 --> 0:17:28.469
<v Speaker 7>for us, we are investing to prepare to double our

0:17:28.570 --> 0:17:30.070
<v Speaker 7>output over the next two years.

0:17:32.160 --> 0:17:34.610
<v Speaker 1>So initially investors were less than impressed. The stock had

0:17:34.670 --> 0:17:37.609
<v Speaker 1>their worst decline in about a month after that original

0:17:37.650 --> 0:17:40.830
<v Speaker 1>kind of less detailed forecast. And the shares continued to slide,

0:17:40.930 --> 0:17:43.830
<v Speaker 1>actually down 2.4% in the session right now, but on

0:17:43.850 --> 0:17:46.649
<v Speaker 1>track for their seventh straight day of declines, which for

0:17:46.670 --> 0:17:49.310
<v Speaker 1>Applied Materials is the longest losing streak going back to

0:17:49.369 --> 0:17:52.770
<v Speaker 1>January of 2024. If ASML does the patterns on the wafer,

0:17:53.600 --> 0:17:57.600
<v Speaker 1>then Applied Materials fills and builds the material, the actual

0:17:57.619 --> 0:18:01.300
<v Speaker 1>physical material onto that chip. Let's get out to New York.

0:18:01.320 --> 0:18:03.139
<v Speaker 1>There's so much more going on around the world. Bloomberg's

0:18:03.160 --> 0:18:05.060
<v Speaker 1>Yahira Anand has that news.

0:18:05.080 --> 0:18:05.600
<v Speaker 2>Hi, Yahira.

0:18:05.980 --> 0:18:06.420
<v Speaker 1>Hi, Ed.

0:18:06.480 --> 0:18:08.879
<v Speaker 12>It's time now for Talking Tech. First up, Bank of

0:18:08.940 --> 0:18:11.820
<v Speaker 12>America is no longer part of the potential roster of

0:18:11.900 --> 0:18:16.460
<v Speaker 12>banks working on Moonshot AI's IPO. That's according to people familiar.

0:18:16.540 --> 0:18:19.840
<v Speaker 12>The move comes as the China AI darling faces increased

0:18:19.900 --> 0:18:23.100
<v Speaker 12>scrutiny here in the U.S. Goldman Sachs is now the

0:18:23.400 --> 0:18:27.850
<v Speaker 12>only U.S. bank expected to have a role in the deal. Plus,

0:18:27.900 --> 0:18:31.430
<v Speaker 12>Mark Zuckerberg says Meta had to commit to spending hundreds

0:18:31.550 --> 0:18:34.970
<v Speaker 12>of billions of dollars last year on AI infrastructure before

0:18:35.010 --> 0:18:38.710
<v Speaker 12>the company had any idea as to whether they would succeed.

0:18:38.730 --> 0:18:41.710
<v Speaker 12>It's a revealing disclosure for a company that has struggled

0:18:41.770 --> 0:18:46.360
<v Speaker 12>to persuade investors to accept Meta. its ballooning expenses for

0:18:46.440 --> 0:18:50.940
<v Speaker 12>its AI ambitions. And the Brooklyn District Attorney's Office has

0:18:51.000 --> 0:18:56.440
<v Speaker 12>disrupted a multinational cryptocurrency scam network centered in Eastern Europe.

0:18:56.510 --> 0:19:00.150
<v Speaker 12>This caps a years-long effort to stop a fraud operation

0:19:00.190 --> 0:19:04.010
<v Speaker 12>that has stolen tens of millions of dollars from Americans.

0:19:04.330 --> 0:19:04.429
<v Speaker 2>Ed.

0:19:05.450 --> 0:19:10.550
<v Speaker 1>Thank you very much, Yoharan. A security breach from inside OpenAI.

0:19:10.630 --> 0:19:14.030
<v Speaker 1>One of the company's own employees has been arrested. accused

0:19:14.070 --> 0:19:17.659
<v Speaker 1>of planting a recording device inside an executive conference room

0:19:17.700 --> 0:19:21.240
<v Speaker 1>at its San Francisco offices. OpenAI says there's no sign

0:19:21.280 --> 0:19:24.460
<v Speaker 1>that sensitive data was compromised. Bloomberg's Rachel Metz has been

0:19:24.500 --> 0:19:26.580
<v Speaker 1>reporting and has the details. And let's just start with

0:19:26.600 --> 0:19:30.340
<v Speaker 1>the details, Rachel. Everything we know to this point about

0:19:30.359 --> 0:19:30.850
<v Speaker 1>the incident.

0:19:31.290 --> 0:19:31.590
<v Speaker 2>Sure.

0:19:31.720 --> 0:19:35.050
<v Speaker 13>So an OpenAI staffer, as you say, was arrested for

0:19:35.710 --> 0:19:39.830
<v Speaker 13>allegedly planting a hidden recording device. It was found in

0:19:39.869 --> 0:19:42.910
<v Speaker 13>the back of a TV in an executive conference room

0:19:42.970 --> 0:19:45.210
<v Speaker 13>in one of OpenAI's San Francisco offices.

0:19:46.190 --> 0:19:49.710
<v Speaker 1>The company reported it to the FBI. The FBI responded.

0:19:50.030 --> 0:19:54.619
<v Speaker 13>Has taken over in a criminal complaint that was unsealed yesterday.

0:19:55.080 --> 0:19:56.620
<v Speaker 1>They made clear why.

0:19:56.440 --> 0:20:01.800
<v Speaker 13>They were arresting this now former employee, including that there

0:20:01.820 --> 0:20:03.540
<v Speaker 13>was a whole timeline in there of all the different

0:20:03.560 --> 0:20:07.020
<v Speaker 13>times that He had allegedly been near this conference room.

0:20:07.040 --> 0:20:08.620
<v Speaker 13>There was a picture that you're showing on the screen

0:20:08.640 --> 0:20:11.520
<v Speaker 13>right now of him peering behind the TV. This device

0:20:11.600 --> 0:20:15.600
<v Speaker 13>is really inexpensive. It was a $ 14 device and it

0:20:15.619 --> 0:20:19.440
<v Speaker 13>was disguised in toilet paper. I'm not sure why it

0:20:19.480 --> 0:20:22.689
<v Speaker 13>was disguised in toilet paper, but it was. So it's

0:20:22.710 --> 0:20:24.909
<v Speaker 13>going to be really interesting to see what happens further.

0:20:25.070 --> 0:20:28.290
<v Speaker 13>And there is a hearing in the case in federal

0:20:28.350 --> 0:20:30.330
<v Speaker 13>court in San Francisco on Monday. So we'll get some

0:20:30.350 --> 0:20:30.889
<v Speaker 13>more info then.

0:20:32.270 --> 0:20:35.760
<v Speaker 1>What OpenAI was saying is that they use their own

0:20:35.880 --> 0:20:40.900
<v Speaker 1>AI models to make an assessment themselves of this device.

0:20:40.920 --> 0:20:44.540
<v Speaker 1>And we were showing images as provided by the authorities

0:20:44.580 --> 0:20:46.659
<v Speaker 1>of that device. Could you try and explain that a

0:20:46.680 --> 0:20:47.060
<v Speaker 1>little bit?

0:20:47.880 --> 0:20:50.930
<v Speaker 13>I don't know too much more about that, but I

0:20:50.950 --> 0:20:54.520
<v Speaker 13>am going to assume that they used one of their

0:20:54.540 --> 0:20:57.399
<v Speaker 13>AI models, took a picture of it, and asked for

0:20:57.440 --> 0:20:59.460
<v Speaker 13>more information about the device. I will say that the

0:20:59.480 --> 0:21:03.379
<v Speaker 13>device says Seed with three E's, Seed Streams. on it,

0:21:03.390 --> 0:21:09.290
<v Speaker 13>which is a company that makes a variety or sells

0:21:09.330 --> 0:21:14.300
<v Speaker 13>a variety of devices, including this device. So I saw

0:21:14.340 --> 0:21:17.100
<v Speaker 13>the number of the device in the filing, and once

0:21:17.140 --> 0:21:19.560
<v Speaker 13>you take that number, it's pretty easy to look it up.

0:21:19.619 --> 0:21:22.560
<v Speaker 13>It looked to be available for sale, and this is

0:21:22.660 --> 0:21:27.170
<v Speaker 13>a small computing device that builds a microphone and has

0:21:27.210 --> 0:21:28.850
<v Speaker 13>a camera and stuff on it, so you could use

0:21:28.890 --> 0:21:29.650
<v Speaker 13>it for surveillance.

0:21:30.530 --> 0:21:32.550
<v Speaker 1>This is a story that I'm sure we will keep

0:21:32.590 --> 0:21:35.590
<v Speaker 1>tracking as we get more developments. Bluebergs, Rachel Metz, thank

0:21:35.630 --> 0:21:38.330
<v Speaker 1>you very much indeed. Another OpenAI battle is playing out

0:21:38.350 --> 0:21:41.650
<v Speaker 1>in court today. This one with Apple. The iPhone maker

0:21:41.950 --> 0:21:45.630
<v Speaker 1>has accused OpenAI of using former Apple employees to get

0:21:45.670 --> 0:21:48.760
<v Speaker 1>its hand on trade secrets as it pushes into consumer hardware.

0:21:49.160 --> 0:21:52.280
<v Speaker 1>Apple wants a judge to block OpenAI from using any

0:21:52.320 --> 0:21:55.840
<v Speaker 1>allegedly stolen information. OpenAI denies the claims and and wants

0:21:55.859 --> 0:21:59.160
<v Speaker 1>the case thrown out. More on OpenAI coming up after

0:21:59.200 --> 0:22:02.520
<v Speaker 1>the break. We'll also have an interview with CFO Sarah

0:22:02.540 --> 0:22:05.639
<v Speaker 1>Fryer from the International AI Summit over in Dublin. Up

0:22:05.760 --> 0:22:09.810
<v Speaker 1>next on Bloomberg Tech, Modal CEO Eric Bernhardsen joins us

0:22:09.830 --> 0:22:13.030
<v Speaker 1>to discuss the AI cloud company's latest fundraise. There's a

0:22:13.070 --> 0:22:14.930
<v Speaker 1>lot going on. It's halftime here in the city of

0:22:15.050 --> 0:22:31.560
<v Speaker 1>San Francisco. This is Bloomberg Tech. Bloomberg Tech Okay, welcome

0:22:31.580 --> 0:22:33.240
<v Speaker 1>back to Bloomberg Tech. I'm going to try and squeeze

0:22:33.260 --> 0:22:37.800
<v Speaker 1>this story in. Yep, bending spoons up 41%. Italian, let's

0:22:37.820 --> 0:22:41.900
<v Speaker 1>call it a tech conglomerate. Berkshire Hathaway investing more money,

0:22:41.900 --> 0:22:47.960
<v Speaker 1>$ 333 million, building what represents a 4% stake in bending spoons.

0:22:48.840 --> 0:22:51.000
<v Speaker 1>And clearly that's a jolt in the arm of the stock.

0:22:51.060 --> 0:22:53.160
<v Speaker 1>It's one way, like hit this morning, redhead, the stock

0:22:53.180 --> 0:22:55.780
<v Speaker 1>goes bananas, and I thought we should get it in.

0:22:56.060 --> 0:22:58.940
<v Speaker 1>But there is a lot happening in private markets as well.

0:22:59.420 --> 0:23:03.169
<v Speaker 1>AI cloud startup Modal has closed a $ 750 million Series

0:23:03.210 --> 0:23:07.850
<v Speaker 1>D fundraise, bringing its post-money valuation to $ 15. 75 billion. So

0:23:07.890 --> 0:23:10.930
<v Speaker 1>Modal says it's seen tremendous growth over the past year,

0:23:11.390 --> 0:23:15.010
<v Speaker 1>pointing to examples like the more than 3 billion sandboxes

0:23:15.410 --> 0:23:19.060
<v Speaker 1>it has run for customers. Delighted to say that Modal's CEO,

0:23:19.500 --> 0:23:23.020
<v Speaker 1>Eric Van Hudson, is with us. Let's talk the round.

0:23:23.040 --> 0:23:26.570
<v Speaker 1>It had been in the press a little bit earlier

0:23:26.590 --> 0:23:30.600
<v Speaker 1>and clearly a lot of momentum. The basics of why

0:23:30.630 --> 0:23:35.430
<v Speaker 1>you needed to raise capital and what you'll do with it. Yeah,

0:23:35.520 --> 0:23:36.380
<v Speaker 1>and thanks for having me.

0:23:36.700 --> 0:23:39.020
<v Speaker 14>We see so much demand coming from so many different

0:23:39.080 --> 0:23:43.240
<v Speaker 14>types of companies. There's really some incredible AI applications getting built,

0:23:43.359 --> 0:23:46.400
<v Speaker 14>and I really think we've entered a new era of software.

0:23:46.420 --> 0:23:49.580
<v Speaker 14>There's so many new capabilities that this new software enables

0:23:49.640 --> 0:23:51.740
<v Speaker 14>us to do, and we provide a lot of those

0:23:51.800 --> 0:23:53.760
<v Speaker 14>building blocks, and you really kind of need a new

0:23:53.800 --> 0:23:56.400
<v Speaker 14>set of building blocks for all these phases, whether that's

0:23:56.630 --> 0:23:59.590
<v Speaker 14>in this phase, we're talking about inference and training and

0:23:59.650 --> 0:24:02.129
<v Speaker 14>sandboxes and a lot of these building blocks are to

0:24:02.170 --> 0:24:04.710
<v Speaker 14>power this new set of AI applications. And that's where

0:24:05.050 --> 0:24:07.070
<v Speaker 14>there's just so much demand. And we're raising money right

0:24:07.090 --> 0:24:10.880
<v Speaker 14>now to invest in really growing faster, launching new products,

0:24:10.980 --> 0:24:12.320
<v Speaker 14>and getting a lot of capacity.

0:24:13.840 --> 0:24:18.160
<v Speaker 1>The three billion sandboxes. Could you just explain the basics

0:24:18.200 --> 0:24:19.470
<v Speaker 1>of what we're talking about there?

0:24:20.840 --> 0:24:24.300
<v Speaker 14>Yeah, so sandboxes is something we started building actually three

0:24:24.340 --> 0:24:26.679
<v Speaker 14>years ago. We started seeing a lot of companies explore

0:24:26.760 --> 0:24:28.680
<v Speaker 14>using LLMs for generating code.

0:24:29.609 --> 0:24:32.330
<v Speaker 1>And we figured they need a place to encapsulate.

0:24:31.830 --> 0:24:35.410
<v Speaker 14>And run those things in isolation to control these agents

0:24:35.490 --> 0:24:37.010
<v Speaker 14>in a way where they can execute code but not

0:24:37.070 --> 0:24:39.419
<v Speaker 14>escape these sandboxes.

0:24:39.440 --> 0:24:43.660
<v Speaker 1>So we launched it in 2023. I didn't see a lot.

0:24:43.560 --> 0:24:45.660
<v Speaker 14>Of growth until about two years later with, I think,

0:24:45.700 --> 0:24:48.580
<v Speaker 14>Sonnet 4.5 or one of those models that really could

0:24:48.619 --> 0:24:52.570
<v Speaker 14>output full apps. That created an explosion in demand for

0:24:53.090 --> 0:24:56.129
<v Speaker 14>code execution primitives that you can couple with these AI agents.

0:24:56.190 --> 0:24:57.490
<v Speaker 1>We've seen incredible demand.

0:24:57.550 --> 0:25:00.350
<v Speaker 14>In fact, it's about 300 million run rate revenue just

0:25:00.390 --> 0:25:03.050
<v Speaker 14>from sandboxes, which is one of our building blocks that

0:25:03.070 --> 0:25:03.609
<v Speaker 14>we've been launching.

0:25:04.560 --> 0:25:07.080
<v Speaker 1>But I find that really interesting. Obviously, you know, for

0:25:07.119 --> 0:25:09.950
<v Speaker 1>our audience, they might have become more familiar with sandbox

0:25:09.970 --> 0:25:14.070
<v Speaker 1>environments because of the open AI hugging face incident. But

0:25:14.150 --> 0:25:16.150
<v Speaker 1>at the root of that story was the idea, like

0:25:16.210 --> 0:25:19.230
<v Speaker 1>many in industry thought it was a good thing. Like

0:25:19.410 --> 0:25:22.909
<v Speaker 1>open AI is testing the capabilities of these two frontier

0:25:22.950 --> 0:25:27.369
<v Speaker 1>models where their guardrails were reduced deliberately, right? So that

0:25:27.609 --> 0:25:32.480
<v Speaker 1>the full cyber capability is. could be deployed. I'm assuming

0:25:32.520 --> 0:25:35.460
<v Speaker 1>from Modal's standpoint, Eric, you want that kind of testing

0:25:35.540 --> 0:25:38.440
<v Speaker 1>activity to be going on. That's the driver of that

0:25:38.500 --> 0:25:39.160
<v Speaker 1>business unit.

0:25:40.560 --> 0:25:42.860
<v Speaker 14>Yeah, and I think it's exciting to see how good

0:25:42.920 --> 0:25:44.379
<v Speaker 14>agents are at cybersecurity.

0:25:44.540 --> 0:25:45.920
<v Speaker 1>But it's really much broader than that.

0:25:45.960 --> 0:25:47.640
<v Speaker 14>I mean, you know, we see so much demand also

0:25:47.680 --> 0:25:50.730
<v Speaker 14>for inference and reinforcement learning. And I think if you

0:25:50.770 --> 0:25:53.510
<v Speaker 14>couple those things together, we have so many companies building

0:25:53.530 --> 0:25:58.730
<v Speaker 14>incredible coding models. Companies like Cognition is training some incredible

0:25:58.780 --> 0:26:01.639
<v Speaker 14>coding models on modal. Ramp is using it to do

0:26:01.660 --> 0:26:03.880
<v Speaker 14>a lot of internal developments. I think Sandbox is really

0:26:03.920 --> 0:26:06.820
<v Speaker 14>becoming more than just, you know, a way to test

0:26:06.840 --> 0:26:10.450
<v Speaker 14>these agents hacking, it's also becoming a pretty foundational primitive

0:26:10.490 --> 0:26:13.530
<v Speaker 14>in coding models and really pushing forward a lot of

0:26:13.570 --> 0:26:14.649
<v Speaker 14>these AI applications, a.

0:26:14.650 --> 0:26:18.250
<v Speaker 1>Lot of which are running on Modal. Can I just

0:26:18.310 --> 0:26:22.020
<v Speaker 1>ask you, Eric, to explain sort of the Modal's perspective

0:26:22.060 --> 0:26:25.399
<v Speaker 1>on tokenomics? You know, your business is to provide the

0:26:25.440 --> 0:26:28.620
<v Speaker 1>infrastructure that helps to build and run AI applications and

0:26:28.680 --> 0:26:31.439
<v Speaker 1>run AI applications being the key bit. You know, when

0:26:31.460 --> 0:26:34.719
<v Speaker 1>the token economics change, how does that impact you, if

0:26:34.790 --> 0:26:35.129
<v Speaker 1>at all?

0:26:36.820 --> 0:26:40.340
<v Speaker 14>Yeah, I mean, I think inference is an incredible business.

0:26:40.420 --> 0:26:43.080
<v Speaker 14>It's been growing so fast, especially with the open source models.

0:26:43.100 --> 0:26:46.990
<v Speaker 14>We've seen so much demand for inference. There's obviously a

0:26:47.010 --> 0:26:49.210
<v Speaker 14>lot of providers in this space. And we think inference,

0:26:49.630 --> 0:26:51.250
<v Speaker 14>for a lot of reasons, is just one part of

0:26:51.270 --> 0:26:53.190
<v Speaker 14>the package. We think a lot about sort of the

0:26:53.210 --> 0:26:56.630
<v Speaker 14>reinforcement learning and the sandboxes as like one bigger platform

0:26:56.650 --> 0:26:59.450
<v Speaker 14>that we provide. What we've seen continuously over the last

0:26:59.500 --> 0:27:02.220
<v Speaker 14>few years is like as token prices drop, it just

0:27:02.260 --> 0:27:04.100
<v Speaker 14>creates so much more demand for these things.

0:27:04.480 --> 0:27:06.140
<v Speaker 1>We've seen some incredible customers.

0:27:06.260 --> 0:27:09.930
<v Speaker 14>TypeSafe, most recently with Jav, basically exploding overnight, launching a

0:27:09.970 --> 0:27:14.930
<v Speaker 14>whole new type of inference, which is classification. Launched on modal,

0:27:15.210 --> 0:27:18.060
<v Speaker 14>got to millions of customers overnight, which is very incredible.

0:27:19.940 --> 0:27:22.300
<v Speaker 1>On October 1st, I saw the headline that you guys

0:27:22.359 --> 0:27:25.400
<v Speaker 1>have launched, the VM Sandbox is right. The basic idea

0:27:25.420 --> 0:27:28.510
<v Speaker 1>is that it gives AI agents access to virtual computers.

0:27:28.970 --> 0:27:30.889
<v Speaker 1>I've experienced this a little bit recently where I've been

0:27:30.930 --> 0:27:33.650
<v Speaker 1>trying to give my agents, it could be Grogbot or Muse,

0:27:34.130 --> 0:27:36.429
<v Speaker 1>access to things that are actually helpful to me. And

0:27:37.380 --> 0:27:39.219
<v Speaker 1>there's a limit to that sometimes, but could you just

0:27:39.260 --> 0:27:42.520
<v Speaker 1>explain that technology, why that was kind of a meaningful

0:27:42.560 --> 0:27:43.280
<v Speaker 1>launch for you guys?

0:27:44.720 --> 0:27:46.790
<v Speaker 14>Yeah, I mean, agents are getting more and more capable

0:27:46.830 --> 0:27:50.570
<v Speaker 14>at controlling computers. So we're launching VM sandboxes, virtual machines,

0:27:50.690 --> 0:27:54.689
<v Speaker 14>and that basically provides a full computer environment, Linux. But

0:27:54.750 --> 0:27:57.250
<v Speaker 14>over time, we hope to launch also support for Windows

0:27:57.310 --> 0:28:00.370
<v Speaker 14>and Mac. And the idea is like basic agents are

0:28:00.609 --> 0:28:02.930
<v Speaker 14>they're getting very good at performing so many different tasks.

0:28:02.950 --> 0:28:04.830
<v Speaker 14>And we think this is a big part in, you know,

0:28:04.850 --> 0:28:07.210
<v Speaker 14>getting them to do more long running tasks, maybe personal

0:28:07.250 --> 0:28:10.950
<v Speaker 14>assistant tasks or booking tickets or but also a lot

0:28:10.970 --> 0:28:13.340
<v Speaker 14>of coding models, of course, is a big driver right now.

0:28:14.660 --> 0:28:16.600
<v Speaker 1>Eric Ben Hudson of the Model C. It's been great

0:28:16.619 --> 0:28:18.480
<v Speaker 1>to have you with us. I really enjoyed the conversation.

0:28:18.520 --> 0:28:22.619
<v Speaker 1>Thank you very much. OpenAI CFO Sarah Fryer says European

0:28:22.680 --> 0:28:25.530
<v Speaker 1>businesses are trailing when it comes to AI adoption. She

0:28:25.550 --> 0:28:28.750
<v Speaker 1>was speaking to Bloomberg's Tom McKenzie at the International AI

0:28:28.770 --> 0:28:32.369
<v Speaker 1>Summit in Dublin. Fryer also discussed recent concerns around security

0:28:32.410 --> 0:28:36.090
<v Speaker 1>and AI, saying the firm's focusing more on safety and alignment.

0:28:37.150 --> 0:28:39.350
<v Speaker 4>First of all, on what we're doing to make sure

0:28:39.450 --> 0:28:43.230
<v Speaker 4>inside of OpenAI, we're doing things that are more safe,

0:28:43.310 --> 0:28:46.500
<v Speaker 4>more aligned, but also to have transparency when things happen.

0:28:46.580 --> 0:28:49.280
<v Speaker 4>So I think in Hugging Face's case, we really wanted

0:28:49.300 --> 0:28:52.420
<v Speaker 4>to get the information out into the world. But internally,

0:28:52.560 --> 0:28:54.800
<v Speaker 4>we are doubling down on how we think about safety

0:28:54.820 --> 0:28:58.990
<v Speaker 4>and alignment, putting more compute to bear. Our newest model, Astra,

0:28:59.030 --> 0:29:01.510
<v Speaker 4>is our most aligned model. But there's even times we

0:29:01.530 --> 0:29:04.390
<v Speaker 4>pace the frontier. So we might hold back. and next

0:29:04.430 --> 0:29:07.070
<v Speaker 4>generation model, which we're actually doing at the moment, because

0:29:07.110 --> 0:29:09.080
<v Speaker 4>we want to do a few more turns on safety

0:29:09.120 --> 0:29:13.620
<v Speaker 4>and alignment. As we think about your question about liability

0:29:13.660 --> 0:29:15.480
<v Speaker 4>and so on, I often think about it with a

0:29:15.480 --> 0:29:17.420
<v Speaker 4>bit more of a car analogy, right? There's clearly the

0:29:17.520 --> 0:29:19.780
<v Speaker 4>parts that go into the car, and if a screw

0:29:19.820 --> 0:29:23.380
<v Speaker 4>fails or a part fails, there's liability with the supply chain.

0:29:23.750 --> 0:29:26.710
<v Speaker 4>The car itself, car companies have put seatbelts in action

0:29:26.730 --> 0:29:29.150
<v Speaker 4>and so on to keep us safe inside cars, but

0:29:29.190 --> 0:29:31.150
<v Speaker 4>if something goes wrong because of what they did, there's

0:29:31.170 --> 0:29:35.240
<v Speaker 4>liability there. And then there's obviously user liability too, right?

0:29:35.260 --> 0:29:38.850
<v Speaker 4>If someone drives while perhaps having drunk too many Guinnesses,

0:29:39.540 --> 0:29:41.760
<v Speaker 4>that also sits with the user. So I think the

0:29:41.800 --> 0:29:44.740
<v Speaker 4>world is going to break into pieces to think about

0:29:45.200 --> 0:29:49.960
<v Speaker 4>where the ultimate liability originated and then with that, who

0:29:50.000 --> 0:29:52.200
<v Speaker 4>should be making sure it doesn't happen again.

0:29:52.260 --> 0:29:54.120
<v Speaker 11>And to be clear on the compute side, are you

0:29:54.160 --> 0:29:57.800
<v Speaker 11>shifting more spend and more compute towards alignment away from

0:29:57.960 --> 0:29:58.280
<v Speaker 11>R &amp; D?

0:29:58.380 --> 0:29:59.500
<v Speaker 2>Because that could be one solution.

0:29:59.980 --> 0:30:03.530
<v Speaker 4>Think about when you're training a model in pre-training or post-training.

0:30:03.870 --> 0:30:06.030
<v Speaker 4>Safety and alignment is an inherent part of it. So

0:30:06.070 --> 0:30:09.210
<v Speaker 4>we've actually always done a lot of investment in that area.

0:30:09.250 --> 0:30:11.710
<v Speaker 4>So I wouldn't say it's net new vast amounts of

0:30:11.750 --> 0:30:15.610
<v Speaker 4>more compute. There's probably more just when you look at

0:30:15.990 --> 0:30:19.090
<v Speaker 4>before and after. But it's not a major shift. But

0:30:19.250 --> 0:30:22.010
<v Speaker 4>generally speaking, we're spending a lot more on compute now.

0:30:22.250 --> 0:30:25.340
<v Speaker 4>to keep driving frontier research. And that could be across

0:30:25.500 --> 0:30:28.640
<v Speaker 4>every type of model, not just in LLM, but voice models,

0:30:28.760 --> 0:30:32.760
<v Speaker 4>image models, even specific models to areas like life science.

0:30:33.220 --> 0:30:34.830
<v Speaker 4>And then, of course, a lot more on compute to

0:30:34.870 --> 0:30:38.490
<v Speaker 4>serve our customers because our demand is growing so quickly.

0:30:40.030 --> 0:30:43.660
<v Speaker 1>That was OpenAI CFO Sarah Fryer speaking with Bloomberg's Tom McKenzie.

0:30:45.400 --> 0:30:48.380
<v Speaker 1>Cisco CEO Chuck Robbins says it's paramount for companies to

0:30:48.420 --> 0:30:50.810
<v Speaker 1>work closely with government. Speaking at a panel on tech

0:30:50.850 --> 0:30:54.950
<v Speaker 1>sovereignty with Bloomberg's Hazlin Amin at the New Economic Forum,

0:30:54.990 --> 0:30:57.750
<v Speaker 1>he spoke about the company's work with foreign governments to

0:30:57.790 --> 0:31:00.690
<v Speaker 1>work collaboratively on sovereign critical infrastructure.

0:31:01.370 --> 0:31:05.910
<v Speaker 5>When you talk about sovereignty, in many cases we talk

0:31:05.950 --> 0:31:09.670
<v Speaker 5>about software assets, cloud-delivered software solutions, and that's a good

0:31:09.710 --> 0:31:14.660
<v Speaker 5>example to address this question. When we began initial conversations,

0:31:14.680 --> 0:31:17.460
<v Speaker 5>this started primarily in Europe, I'd say two to three

0:31:17.480 --> 0:31:21.820
<v Speaker 5>years ago. It really began in earnest. And the early

0:31:21.860 --> 0:31:25.740
<v Speaker 5>conversations were, we want everything that you deliver today, we

0:31:25.900 --> 0:31:28.920
<v Speaker 5>want it on-prem, we want total control, we want it

0:31:29.050 --> 0:31:33.229
<v Speaker 5>operated by citizens of this country. And so then you

0:31:33.270 --> 0:31:36.630
<v Speaker 5>start whittling down a little bit from there because the

0:31:36.670 --> 0:31:40.010
<v Speaker 5>reality is that to deliver that technology on a country

0:31:40.050 --> 0:31:43.070
<v Speaker 5>by country by country by country basis at that level

0:31:43.110 --> 0:31:48.380
<v Speaker 5>of capability is going to be probably a non-starter from

0:31:48.410 --> 0:31:52.170
<v Speaker 5>a cost perspective by the time you're finished. And so

0:31:52.830 --> 0:31:55.130
<v Speaker 5>what we have done over the years is actually sat,

0:31:55.290 --> 0:31:58.510
<v Speaker 5>we've sat with the governments and said, what is it you're,

0:31:58.670 --> 0:32:01.490
<v Speaker 5>what are you trying to accomplish? Let's figure this out together.

0:32:01.530 --> 0:32:05.030
<v Speaker 5>Instead of you giving us a feature list, let's talk

0:32:05.050 --> 0:32:06.490
<v Speaker 5>about what are the things you're worried about? What are

0:32:06.510 --> 0:32:09.960
<v Speaker 5>the things that you feel adamant about? And, you know,

0:32:10.400 --> 0:32:12.120
<v Speaker 5>the two big things that always come up, and then

0:32:12.160 --> 0:32:13.620
<v Speaker 5>there's some other things that we have to deal with.

0:32:13.640 --> 0:32:16.000
<v Speaker 5>The two big things are always data. We want to

0:32:16.040 --> 0:32:18.280
<v Speaker 5>maintain our data and data sovereignty. We don't want it

0:32:18.320 --> 0:32:20.400
<v Speaker 5>exiting the country. We don't want it out of our control.

0:32:21.260 --> 0:32:24.830
<v Speaker 5>And we want operational control of these systems, right? We

0:32:24.850 --> 0:32:27.410
<v Speaker 5>want to be able to, we want to know how

0:32:27.450 --> 0:32:31.070
<v Speaker 5>they're operating. That doesn't mean you have to have every,

0:32:31.110 --> 0:32:34.410
<v Speaker 5>like if it's an in-country cloud service, it doesn't necessarily

0:32:34.450 --> 0:32:37.320
<v Speaker 5>mean it has to be run by local citizens. Maybe

0:32:37.670 --> 0:32:40.840
<v Speaker 5>it is, maybe it isn't. But you need to understand

0:32:40.900 --> 0:32:43.850
<v Speaker 5>enough to know that You have operational control and in

0:32:43.930 --> 0:32:46.990
<v Speaker 5>certain situations you actually want to, you may want to

0:32:47.070 --> 0:32:51.110
<v Speaker 5>tighten that system down where you limit the external communication

0:32:51.310 --> 0:32:54.729
<v Speaker 5>outside the country under times of stress. And so you

0:32:54.750 --> 0:32:58.230
<v Speaker 5>get into these discussions about what it is you're trying

0:32:58.250 --> 0:33:01.160
<v Speaker 5>to truly accomplish and I think we can deliver it.

0:33:01.250 --> 0:33:05.490
<v Speaker 5>So we've created a portfolio of what we call sovereign

0:33:06.570 --> 0:33:10.090
<v Speaker 5>critical infrastructure that we have launched in several countries in

0:33:10.190 --> 0:33:14.240
<v Speaker 5>Europe over the last year. We just launched it in Canada,

0:33:14.320 --> 0:33:16.590
<v Speaker 5>and we haven't really announced it publicly yet, but over

0:33:16.610 --> 0:33:18.230
<v Speaker 5>the next few months, we'll be bringing that to India

0:33:18.290 --> 0:33:18.670
<v Speaker 5>as well.

0:33:20.590 --> 0:33:23.270
<v Speaker 1>That was Cisco CEO Chuck Robbins at the Bloomberg New

0:33:23.310 --> 0:33:25.930
<v Speaker 1>Economy Forum. There's been a lot of robo-taxi news this week.

0:33:26.090 --> 0:33:29.130
<v Speaker 1>Tesla's facing a new U.S. safety probe over its robo-taxis,

0:33:29.170 --> 0:33:32.620
<v Speaker 1>while in Europe, it's dropped the full self-driving name as

0:33:32.670 --> 0:33:36.120
<v Speaker 1>it seeks wider approval for that consumer product. Bloomberg Intelligence...

0:33:35.880 --> 0:33:38.860
<v Speaker 1>has research out that says China could see more robotaxi

0:33:38.880 --> 0:33:41.640
<v Speaker 1>permits from the second half of next year. And our

0:33:41.680 --> 0:33:45.040
<v Speaker 1>next guest argues that Europe is now emerging as the

0:33:45.120 --> 0:33:49.690
<v Speaker 1>next battleground for U.S. and Chinese robotaxi companies. Frank McCleary

0:33:49.730 --> 0:33:51.990
<v Speaker 1>is a partner at Arthur D. Little, and I found

0:33:52.030 --> 0:33:54.890
<v Speaker 1>your research to be really interesting. Thank you. That's what's

0:33:54.910 --> 0:33:58.390
<v Speaker 1>happening in Europe, right? If the Chinese robotaxi operators can't

0:33:58.410 --> 0:34:01.650
<v Speaker 1>be in the U.S., but they're expanding outside of China domestically,

0:34:02.120 --> 0:34:04.060
<v Speaker 1>then Europe is where they're going to try and take

0:34:04.080 --> 0:34:05.220
<v Speaker 1>a hold. Yeah.

0:34:05.440 --> 0:34:08.860
<v Speaker 3>And this is where the fragmentation that we're seeing globally

0:34:08.920 --> 0:34:11.419
<v Speaker 3>in terms of different industries, whether it's automotive or other

0:34:11.440 --> 0:34:16.500
<v Speaker 3>industrial sectors, Europe is still where China is, Chinese OEMs,

0:34:16.540 --> 0:34:20.109
<v Speaker 3>Chinese manufacturers are pushing in from a infrastructure perspective, from

0:34:20.250 --> 0:34:23.170
<v Speaker 3>a FDI perspective. And you've got a lot of announcements

0:34:23.370 --> 0:34:28.070
<v Speaker 3>of Uber trying to lock up partnerships with different technology

0:34:28.090 --> 0:34:31.530
<v Speaker 3>providers to ensure that they have the right to play

0:34:31.790 --> 0:34:36.060
<v Speaker 3>in that space. Asset light model, but continuing to have

0:34:36.520 --> 0:34:38.890
<v Speaker 3>that ability to shift demand from one to the other.

0:34:40.130 --> 0:34:42.790
<v Speaker 1>What kind of a regulatory environment does... I think we're

0:34:42.810 --> 0:34:45.310
<v Speaker 1>talking about the European Union. We can talk about the

0:34:45.390 --> 0:34:49.359
<v Speaker 1>UK maybe separately. What does that present to a Chinese

0:34:49.430 --> 0:34:53.840
<v Speaker 1>AV operator that wants to deploy commercially? How is it working?

0:34:54.200 --> 0:34:57.000
<v Speaker 3>So what's interesting with Europe is that each country is

0:34:57.020 --> 0:35:00.500
<v Speaker 3>sort of developing differently, but they're developing quickly. So it

0:35:00.520 --> 0:35:04.239
<v Speaker 3>started with Germany laying out a level four infrastructure or

0:35:04.760 --> 0:35:10.080
<v Speaker 3>regulatory framework. Croatia recently with Zagreb pushing their pilots. You

0:35:10.100 --> 0:35:13.850
<v Speaker 3>have the UK with their national infrastructure framework. Spain, I think,

0:35:13.910 --> 0:35:18.910
<v Speaker 3>most recently. And then you've got smaller markets like Norway, Austria, Switzerland.

0:35:19.390 --> 0:35:22.890
<v Speaker 3>And so you don't have an EU-wide regulatory framework. And

0:35:23.310 --> 0:35:25.490
<v Speaker 3>do I see that happening the way the EU works?

0:35:25.910 --> 0:35:26.410
<v Speaker 1>Typically, no.

0:35:26.590 --> 0:35:30.939
<v Speaker 3>You're going to have these market-by-market regulations. regulatory frameworks. But

0:35:31.200 --> 0:35:35.530
<v Speaker 3>what is consistent is the desire to get from pilot

0:35:35.590 --> 0:35:38.130
<v Speaker 3>to operation. So Croatia is probably the best example where

0:35:38.650 --> 0:35:42.170
<v Speaker 3>they laid out a framework that allowed Pony AI and

0:35:42.210 --> 0:35:48.430
<v Speaker 3>Vern to quickly get from pilot to commercial operations relatively fast.

0:35:48.840 --> 0:35:51.600
<v Speaker 1>Spain is more of a stepwise. And so the key

0:35:51.640 --> 0:35:52.520
<v Speaker 1>with regulatory.

0:35:52.200 --> 0:35:54.660
<v Speaker 3>Frameworks is providing the transparency of how to do it.

0:35:55.040 --> 0:35:57.049
<v Speaker 3>I think the difference in the U.S., it's a 50

0:35:57.050 --> 0:35:59.569
<v Speaker 3>state by state. You know, we haven't seen NHTSA come

0:35:59.610 --> 0:36:01.470
<v Speaker 3>in and say, okay, here's how we're going to do,

0:36:01.489 --> 0:36:05.750
<v Speaker 3>you know, sort of manage safety and regulation around.

0:36:06.790 --> 0:36:11.820
<v Speaker 1>Well, in this country, outside of codified regulation. You know, essentially,

0:36:11.900 --> 0:36:14.600
<v Speaker 1>is the door shut to a Chinese AV operator?

0:36:15.420 --> 0:36:20.500
<v Speaker 3>So with the connected vehicle and ADAS software regulations that

0:36:20.540 --> 0:36:23.779
<v Speaker 3>came in during the end of the Biden administration, it

0:36:23.820 --> 0:36:27.290
<v Speaker 3>sort of shut down Chinese OEMs bringing in from China.

0:36:27.330 --> 0:36:32.330
<v Speaker 3>Does that limit a Chinese player from coming in and manufacturing,

0:36:32.410 --> 0:36:34.169
<v Speaker 3>developing and so on in the U.S.

0:36:34.969 --> 0:36:37.130
<v Speaker 1>maybe not they did it but they have to do.

0:36:37.010 --> 0:36:41.080
<v Speaker 3>It completely in the u.s full software development stack or

0:36:41.180 --> 0:36:44.759
<v Speaker 3>being able to to leverage from other technology providers so

0:36:44.800 --> 0:36:49.069
<v Speaker 3>that's where i don't see that happening soon uh or

0:36:49.390 --> 0:36:51.290
<v Speaker 3>in the foreseeable future and i think that not to

0:36:51.310 --> 0:36:55.240
<v Speaker 3>get into politics but it The initial regulations on connected

0:36:55.260 --> 0:36:58.759
<v Speaker 3>vehicle came in during the first Trump administration. It was

0:36:58.780 --> 0:37:01.359
<v Speaker 3>finally pushed out through the Biden administration. So it's playing

0:37:01.400 --> 0:37:03.450
<v Speaker 3>on both sides of the aisle. And so I don't

0:37:03.489 --> 0:37:06.270
<v Speaker 3>see it changing or we don't see it changing in

0:37:06.310 --> 0:37:07.610
<v Speaker 3>the short midterm.

0:37:07.770 --> 0:37:10.730
<v Speaker 1>I think part of our conversation, right, is about commercial deployment.

0:37:10.750 --> 0:37:13.489
<v Speaker 1>You know, a real service where a consumer through an

0:37:13.590 --> 0:37:16.890
<v Speaker 1>app hails a robot taxi, gets in and pays a

0:37:16.930 --> 0:37:19.529
<v Speaker 1>fare for it. But from an R &amp; D standpoint,

0:37:19.660 --> 0:37:21.219
<v Speaker 1>you know, for quite a long time, we talked about

0:37:21.260 --> 0:37:24.160
<v Speaker 1>how China was just like, Go after it. You know,

0:37:24.180 --> 0:37:27.400
<v Speaker 1>in cities all across that nation, there was a lot

0:37:27.420 --> 0:37:30.060
<v Speaker 1>of training that was done. There were lots of miles covered.

0:37:30.920 --> 0:37:33.080
<v Speaker 1>Does that give the Chinese AV operators any kind of

0:37:33.140 --> 0:37:37.620
<v Speaker 1>advantage to deploy in other jurisdictions and markets, to your mind?

0:37:37.840 --> 0:37:39.700
<v Speaker 3>No, because you still have to train in the new environment.

0:37:39.980 --> 0:37:41.890
<v Speaker 3>So you have to. It's the same thing that we

0:37:41.930 --> 0:37:44.629
<v Speaker 3>see with Waymo in the U.S. or, you know, there's

0:37:44.650 --> 0:37:46.870
<v Speaker 3>still a training period. Now, what Waymo is getting better

0:37:46.969 --> 0:37:50.330
<v Speaker 3>at is training and accelerating, getting to commercial operations quicker.

0:37:50.350 --> 0:37:50.450
<v Speaker 1>Right.

0:37:50.850 --> 0:37:54.750
<v Speaker 3>So the Chinese having millions of miles overseas doesn't necessarily

0:37:54.770 --> 0:37:56.910
<v Speaker 3>mean that it's going to work in Croatia or in

0:37:56.950 --> 0:38:01.130
<v Speaker 3>the UK or in a different urban environment, weather conditions

0:38:01.190 --> 0:38:01.930
<v Speaker 3>and so on.

0:38:01.969 --> 0:38:04.290
<v Speaker 1>Just very quick on FSD. There was the news that

0:38:04.370 --> 0:38:08.150
<v Speaker 1>in Europe, in some jurisdictions, Tesla has dropped the name FSD.

0:38:08.250 --> 0:38:12.610
<v Speaker 1>I do FSD 60 miles every day round trip. Is

0:38:12.630 --> 0:38:14.969
<v Speaker 1>there real progress that you're seeing from Tesla on that?

0:38:15.010 --> 0:38:16.450
<v Speaker 1>What is a consumer product right now?

0:38:17.510 --> 0:38:21.000
<v Speaker 3>Well, they're using similar, I haven't looked deeply into the stack,

0:38:21.020 --> 0:38:23.740
<v Speaker 3>but they're using similar technology for the cyber cab today

0:38:24.100 --> 0:38:26.480
<v Speaker 3>in Austin and deploying. So I think they're... They.

0:38:26.540 --> 0:38:28.060
<v Speaker 1>Say it's a part of the development arc for a

0:38:28.080 --> 0:38:29.240
<v Speaker 1>future robo-taxi for sure.

0:38:29.580 --> 0:38:31.609
<v Speaker 3>And it was a story from when the Model 3

0:38:31.610 --> 0:38:33.730
<v Speaker 3>came out that we're going to redeploy Model 3s that

0:38:33.750 --> 0:38:35.510
<v Speaker 3>come back off lease and we're going to use them

0:38:36.190 --> 0:38:40.600
<v Speaker 3>in a robo-taxi environment. So they're going to continue pushing it.

0:38:40.670 --> 0:38:45.569
<v Speaker 3>I think this is the EU regulator's Putting their foot

0:38:45.610 --> 0:38:49.850
<v Speaker 3>down on what product liability and also, I don't want

0:38:49.989 --> 0:38:51.450
<v Speaker 3>to say false advertising, but.

0:38:53.630 --> 0:38:56.580
<v Speaker 1>Calling it what it is. Yeah, and if the European

0:38:56.620 --> 0:38:58.719
<v Speaker 1>regulator doesn't like it, as you said, they put their

0:38:58.739 --> 0:39:01.200
<v Speaker 1>foot down. Frank McCleary, Arthur Dealer, it's been great to

0:39:01.219 --> 0:39:01.620
<v Speaker 1>have you with us.

0:39:01.620 --> 0:39:01.810
<v Speaker 2>Thank you.

0:39:01.820 --> 0:39:03.560
<v Speaker 1>Thank you very much indeed. Coming up on the show,

0:39:03.590 --> 0:39:07.350
<v Speaker 1>the long-awaited Grand Theft Auto 6 releases in November, and

0:39:07.410 --> 0:39:10.489
<v Speaker 1>it's expected to generate over $ 3 billion in just its

0:39:10.530 --> 0:39:13.589
<v Speaker 1>first four months. More details on that upcoming release when

0:39:13.610 --> 0:39:16.070
<v Speaker 1>we return. And if you take a head of break,

0:39:16.110 --> 0:39:17.610
<v Speaker 1>take a look at shares of Netflix. We'll report in

0:39:17.650 --> 0:39:21.540
<v Speaker 1>the journal about... a pretty significant layoff. I think it

0:39:21.560 --> 0:39:26.009
<v Speaker 1>was 6% of global staff, down 1.3%. I don't know,

0:39:26.300 --> 0:39:27.830
<v Speaker 1>it took a while for the market to kind of

0:39:27.850 --> 0:39:29.830
<v Speaker 1>react to that one a little bit, but it's a

0:39:29.850 --> 0:39:40.170
<v Speaker 1>pretty straightforward story. This is Bloomberg Tech. GTA 6 is

0:39:40.210 --> 0:39:42.810
<v Speaker 1>set to release on November 19th, expected to be one

0:39:42.830 --> 0:39:46.529
<v Speaker 1>of the biggest entertainment releases in history. The game, which

0:39:46.590 --> 0:39:49.890
<v Speaker 1>is created by Rockstar Games, a development studio owned by Take-Two,

0:39:50.330 --> 0:39:53.569
<v Speaker 1>is expected to generate more than $ 3 billion in its

0:39:53.610 --> 0:39:57.040
<v Speaker 1>first four months to take us inside the making of

0:39:57.100 --> 0:40:02.020
<v Speaker 1>Grand Theft Auto 6. Bloomberg's Jason Schreier, who published the

0:40:02.040 --> 0:40:05.690
<v Speaker 1>pretty much definitive backstory this morning. My first question is,

0:40:05.780 --> 0:40:09.069
<v Speaker 1>why did this take eight years Man, you.

0:40:09.010 --> 0:40:13.030
<v Speaker 15>Know, I think $ 3 billion actually might be a conservative estimate,

0:40:13.070 --> 0:40:15.330
<v Speaker 15>just to put things in perspective. Yeah, it took eight

0:40:15.350 --> 0:40:18.469
<v Speaker 15>years for a bunch of reasons. COVID was a big one.

0:40:18.730 --> 0:40:21.549
<v Speaker 15>We don't think about it too much these days, or

0:40:21.590 --> 0:40:23.520
<v Speaker 15>we try not to think about it too much these days,

0:40:23.560 --> 0:40:26.180
<v Speaker 15>but having a pandemic early in production of a game

0:40:26.200 --> 0:40:28.529
<v Speaker 15>adds probably a year or two to the schedule. They

0:40:28.590 --> 0:40:30.900
<v Speaker 15>ran into issues with their vision, trying to figure out

0:40:30.940 --> 0:40:33.100
<v Speaker 15>what the game was going to look like. They lost

0:40:33.120 --> 0:40:36.140
<v Speaker 15>some key people along the way. They actually doubled in

0:40:36.219 --> 0:40:40.259
<v Speaker 15>size from around 2,400 people working at Rockstar Games on

0:40:40.300 --> 0:40:46.530
<v Speaker 15>Grand Theft Auto to 4,700, which is a monumental thing.

0:40:46.590 --> 0:40:48.950
<v Speaker 15>And that actually helps you make a lot more stuff,

0:40:48.989 --> 0:40:50.730
<v Speaker 15>but it also brings with it a whole lot of

0:40:50.790 --> 0:40:55.530
<v Speaker 15>communication issues and coordination issues that can make things take

0:40:55.650 --> 0:40:56.490
<v Speaker 15>a little bit longer.

0:40:56.890 --> 0:40:58.190
<v Speaker 1>And really games are just.

0:40:58.450 --> 0:41:01.530
<v Speaker 15>Complicated and detailed, especially a game like this with probably

0:41:01.570 --> 0:41:04.620
<v Speaker 15>the biggest virtual world that we have ever seen in

0:41:04.640 --> 0:41:05.460
<v Speaker 15>a video game.

0:41:06.100 --> 0:41:09.920
<v Speaker 1>So it's just like an amazing experience. story of what

0:41:09.940 --> 0:41:11.680
<v Speaker 1>it was like to be on the Rockstar team in

0:41:11.719 --> 0:41:15.730
<v Speaker 1>that time. You talked about the headcount almost doubling. Management

0:41:15.890 --> 0:41:19.450
<v Speaker 1>tried to avoid the brutal crunch, right? You talk about that.

0:41:19.770 --> 0:41:22.430
<v Speaker 1>There were some disputes, like things weren't easy. Five day

0:41:22.670 --> 0:41:25.890
<v Speaker 1>a week office requirements. Speak a little about that. Yeah.

0:41:25.950 --> 0:41:28.240
<v Speaker 15>So on Red Dead Redemption 2, which was the last

0:41:28.280 --> 0:41:31.819
<v Speaker 15>game from Rockstar Games, the maker of GTA, there was

0:41:31.860 --> 0:41:35.040
<v Speaker 15>a big controversy surrounding the overtime that they had to

0:41:35.080 --> 0:41:38.609
<v Speaker 15>put in for that game. not just overtime here and there,

0:41:38.670 --> 0:41:41.910
<v Speaker 15>but nights and weekends for weeks, sometimes months on end.

0:41:42.290 --> 0:41:45.509
<v Speaker 15>And Rockstar as a company tried to mitigate that. And

0:41:45.590 --> 0:41:47.899
<v Speaker 15>some of the things they did, including hiring many more

0:41:47.940 --> 0:41:51.320
<v Speaker 15>people to try to fix that workload, they added a

0:41:51.360 --> 0:41:54.660
<v Speaker 15>new policy offering paid time off to people in exchange

0:41:54.700 --> 0:41:57.930
<v Speaker 15>for extra hours worked. So they've been trying to fix that.

0:41:57.989 --> 0:42:00.250
<v Speaker 15>But one problem they ran into is was a big

0:42:00.590 --> 0:42:05.050
<v Speaker 15>leak back in 2022. Dozens of videos of unfinished gameplay

0:42:05.090 --> 0:42:07.530
<v Speaker 15>wound up on the internet, and that led the company

0:42:07.570 --> 0:42:10.620
<v Speaker 15>to crack down on remote work, which has been very

0:42:10.700 --> 0:42:14.420
<v Speaker 15>controversial on this project. Still, to this day, as we

0:42:14.480 --> 0:42:17.440
<v Speaker 15>are speaking, October 14th, it is still a very controversial

0:42:17.480 --> 0:42:20.200
<v Speaker 15>policy at Rockstar. Not just because people have to go

0:42:20.239 --> 0:42:23.130
<v Speaker 15>back in five days a week and have no possibility

0:42:23.190 --> 0:42:26.150
<v Speaker 15>of hybrid work schedules, but because they can't even log

0:42:26.290 --> 0:42:29.650
<v Speaker 15>into their work remotely, which means if you have a

0:42:29.690 --> 0:42:32.190
<v Speaker 15>child care emergency or you want to go home and

0:42:32.230 --> 0:42:34.689
<v Speaker 15>get a couple hours of work done at night after

0:42:34.730 --> 0:42:36.950
<v Speaker 15>the kids go to bed or something like that, that

0:42:36.969 --> 0:42:39.250
<v Speaker 15>just isn't an option. And that, I think, has been

0:42:39.290 --> 0:42:41.190
<v Speaker 15>very polarizing among Rockstar staff.

0:42:42.090 --> 0:42:44.989
<v Speaker 1>Uh inside the making of grand theft auto 6 for

0:42:45.030 --> 0:42:47.989
<v Speaker 1>business week it's a must read from jason bloomberg terminal

0:42:48.250 --> 0:42:51.399
<v Speaker 1>bloomberg.com bloomberg's jason stride thank you so much for being

0:42:51.440 --> 0:42:53.380
<v Speaker 1>with us that does it for this edition of bloomberg

0:42:53.420 --> 0:42:56.200
<v Speaker 1>tech let's get back to semis um the chip route's

0:42:56.239 --> 0:43:00.360
<v Speaker 1>deepening right so yes uh asml is higher but actually

0:43:00.480 --> 0:43:04.450
<v Speaker 1>in the session uh the philadelphia semiconductor index down for

0:43:04.469 --> 0:43:06.350
<v Speaker 1>a seventh straight day which i think is the longest

0:43:06.390 --> 0:43:10.230
<v Speaker 1>run of decline since may of 2025 Although, as we're

0:43:10.270 --> 0:43:13.689
<v Speaker 1>showing on a year-to-date basis, the index is up handsomely.

0:43:13.969 --> 0:43:16.780
<v Speaker 1>Some short-term pressure. Check out the pod. Recap the show.

0:43:17.060 --> 0:43:17.760
<v Speaker 1>This is Bloomberg.