WEBVTT - ASML CEO Christophe Fouquet Talks AI Demand, Chips

0:00:00.120 --> 0:00:09.600
<v Speaker 1>Bloomberg Audio Studios, podcasts, radio news. Well. When ASML introduced

0:00:09.680 --> 0:00:13.360
<v Speaker 1>its EUV machines, it helped customers move from seven nanometer

0:00:13.440 --> 0:00:16.680
<v Speaker 1>chips down to the three nanometer chips now used by

0:00:16.680 --> 0:00:20.080
<v Speaker 1>the likes of Nvidia and Apple. Its next big test

0:00:20.160 --> 0:00:22.640
<v Speaker 1>is whether it can transition to what's known as high

0:00:23.079 --> 0:00:28.760
<v Speaker 1>numerical aperture lithography or high NA. These newer EUV machines

0:00:28.880 --> 0:00:32.560
<v Speaker 1>raimed at pushing chips below two nanometers, giving them even

0:00:32.600 --> 0:00:37.360
<v Speaker 1>more capabilities considered crucial for future AI applications. The person

0:00:37.479 --> 0:00:42.120
<v Speaker 1>leading this effort is Chief executive Officer Christophe Fouke. In

0:00:42.159 --> 0:00:46.280
<v Speaker 1>an exclusive interview recorded on November fourteenth at ASML's headquarters,

0:00:46.560 --> 0:00:51.080
<v Speaker 1>Fuke explained why no other company on the planet can

0:00:51.159 --> 0:00:52.680
<v Speaker 1>do what his can.

0:00:55.920 --> 0:01:00.160
<v Speaker 2>Lithography will always be there, right because you can not

0:01:00.400 --> 0:01:04.440
<v Speaker 2>do any sake with lithography, and there will always be

0:01:04.880 --> 0:01:11.040
<v Speaker 2>the wish to get better lithography. Better lithography means better resolution,

0:01:11.959 --> 0:01:15.920
<v Speaker 2>better accuracy, and better productivity, and we're going to work

0:01:15.959 --> 0:01:19.800
<v Speaker 2>basically on those three axes for many, many years. If

0:01:19.800 --> 0:01:23.000
<v Speaker 2>we look at lithography in CML today, thanks to UV,

0:01:23.680 --> 0:01:25.920
<v Speaker 2>we most really know what to do for our customer

0:01:26.000 --> 0:01:29.040
<v Speaker 2>for the next ten to fifteen years and the next

0:01:29.160 --> 0:01:32.920
<v Speaker 2>ten to fifteen years, we'll still see major innovation in

0:01:33.000 --> 0:01:36.000
<v Speaker 2>lithography in order to continue to work with our customer

0:01:36.080 --> 0:01:40.160
<v Speaker 2>on cost and transistor density. So that's the first sake

0:01:41.280 --> 0:01:44.720
<v Speaker 2>we do lone today. We just talked about i NA.

0:01:45.560 --> 0:01:47.400
<v Speaker 2>We may at some point of time or so to

0:01:47.640 --> 0:01:51.880
<v Speaker 2>introduce even higher numerical aperture tool which we call like CURNA.

0:01:52.200 --> 0:01:57.360
<v Speaker 2>That's the point seventy five tool. Basically we hyper and so.

0:01:58.120 --> 0:02:00.400
<v Speaker 1>When does hyper and I come mostly mean over.

0:02:00.280 --> 0:02:02.600
<v Speaker 2>The next decade, So we still have time. We still

0:02:02.640 --> 0:02:05.720
<v Speaker 2>have time, but that's still something we are preparing for

0:02:06.040 --> 0:02:08.080
<v Speaker 2>because as you said, we have to look long term.

0:02:08.240 --> 0:02:11.919
<v Speaker 1>You're also pushing into advanced packaging or three D packaging,

0:02:11.960 --> 0:02:14.680
<v Speaker 1>and this is essentially where you get the components. Instead

0:02:14.680 --> 0:02:18.320
<v Speaker 1>of laying them out flat urban spool style, you build

0:02:18.360 --> 0:02:21.840
<v Speaker 1>them up a bit like a skyscraper. Cort. What is

0:02:21.880 --> 0:02:24.360
<v Speaker 1>the importance of that? How significant will will that be

0:02:24.680 --> 0:02:27.840
<v Speaker 1>for SML and why focus on that packaging?

0:02:28.440 --> 0:02:32.640
<v Speaker 2>Yeah, to reason why this is becoming important. So the

0:02:32.680 --> 0:02:35.800
<v Speaker 2>first thing is our customer are continuing to drive for

0:02:35.919 --> 0:02:40.240
<v Speaker 2>more transistoror density. So the most law we usually refer

0:02:40.360 --> 0:02:45.040
<v Speaker 2>to is calling for doubling transition density every two years,

0:02:45.440 --> 0:02:47.839
<v Speaker 2>and this has been going on for many years. It's

0:02:47.880 --> 0:02:51.120
<v Speaker 2>still keep going. In fact, if you look at AI

0:02:51.240 --> 0:02:55.240
<v Speaker 2>customers you mentioned NVIDA before, they aren't even more than that.

0:02:55.919 --> 0:02:58.720
<v Speaker 2>So they don't want to get the transition density to

0:02:58.840 --> 0:03:01.640
<v Speaker 2>double every two years. They were like the number of

0:03:01.639 --> 0:03:04.800
<v Speaker 2>transistors to go up by a factor of sixteen every

0:03:04.800 --> 0:03:06.639
<v Speaker 2>two years. That's what we have seen happening in the

0:03:06.720 --> 0:03:09.720
<v Speaker 2>last two three years. So you're going completely off Moore's

0:03:09.760 --> 0:03:14.639
<v Speaker 2>law and you need even more transistor Now. We still

0:03:14.720 --> 0:03:18.000
<v Speaker 2>use lithography to try to put as many transistors as

0:03:18.000 --> 0:03:23.040
<v Speaker 2>possible per unit of area, but that's not enough. And

0:03:23.280 --> 0:03:26.160
<v Speaker 2>if you cannot get enough transistor as you show doing this,

0:03:26.960 --> 0:03:27.880
<v Speaker 2>then you also.

0:03:27.680 --> 0:03:30.200
<v Speaker 1>Have to do that. On AI. You made a big

0:03:30.440 --> 0:03:33.239
<v Speaker 1>step by investing almost one and a half billion US

0:03:33.280 --> 0:03:35.680
<v Speaker 1>dollars in Mistral, which is one of the leading large

0:03:35.720 --> 0:03:39.600
<v Speaker 1>language model companies in Europe. Certainly a competitor sort of

0:03:39.600 --> 0:03:43.480
<v Speaker 1>likes about pen Ai and Gemini and Anthropic. You're taking

0:03:43.520 --> 0:03:47.200
<v Speaker 1>about a ten percent stake with that Dale Why.

0:03:46.600 --> 0:03:50.720
<v Speaker 2>Well, we saw AI as also a huge opportunity for

0:03:51.000 --> 0:03:54.480
<v Speaker 2>SML if you look at a SML. We invest a

0:03:54.480 --> 0:03:57.160
<v Speaker 2>lot in R and D four and a half billion

0:03:57.240 --> 0:04:01.240
<v Speaker 2>euro every year to develop our product, and of course

0:04:01.280 --> 0:04:03.760
<v Speaker 2>we invest a lot in hardware. We also invest a

0:04:03.760 --> 0:04:06.520
<v Speaker 2>lot of software, and we saw that when it comes

0:04:06.520 --> 0:04:12.080
<v Speaker 2>to development cycle, AI could help us enormously with the software,

0:04:12.120 --> 0:04:16.320
<v Speaker 2>of course like many company, but for us to be

0:04:16.360 --> 0:04:20.400
<v Speaker 2>able to ship our machine to our customer and have

0:04:20.520 --> 0:04:23.200
<v Speaker 2>people to maintain those machines on the field, we're writing

0:04:23.400 --> 0:04:26.279
<v Speaker 2>a huge amount of procedure. We spend a lot of

0:04:26.320 --> 0:04:29.479
<v Speaker 2>time describing on paper our tool walk so that people

0:04:29.560 --> 0:04:32.320
<v Speaker 2>can deal with it, and they also AI can help

0:04:32.400 --> 0:04:35.200
<v Speaker 2>us to really reduce the time our engineers spend on

0:04:35.279 --> 0:04:38.200
<v Speaker 2>that so that we free them to do other things like,

0:04:38.360 --> 0:04:41.599
<v Speaker 2>for example, free the integration. So we saw a major

0:04:41.680 --> 0:04:49.440
<v Speaker 2>major efficiency opportunity with AI. More importantly, we also believe

0:04:49.520 --> 0:04:54.760
<v Speaker 2>that AI can improve our product because our product generates

0:04:54.920 --> 0:04:56.719
<v Speaker 2>tons of data.

0:04:55.880 --> 0:04:58.920
<v Speaker 1>Are we in an AI bubble?

0:05:00.120 --> 0:05:03.440
<v Speaker 2>Well, you know, when people talk about AI bubble, I

0:05:03.480 --> 0:05:06.000
<v Speaker 2>think I don't know exactly what they mean, and usually

0:05:06.040 --> 0:05:07.560
<v Speaker 2>I say, there's two ways to look at it. If

0:05:07.560 --> 0:05:10.160
<v Speaker 2>you look at the industry, I don't think there is

0:05:10.200 --> 0:05:13.640
<v Speaker 2>a bubble. So the impact of AI in industry is

0:05:13.680 --> 0:05:18.040
<v Speaker 2>just starting and the impact on the industry will be

0:05:18.560 --> 0:05:21.880
<v Speaker 2>positive for many, many years to come. So there there

0:05:21.920 --> 0:05:24.120
<v Speaker 2>is no bubble. I think that the value of AI,

0:05:24.480 --> 0:05:27.839
<v Speaker 2>the industrial value of AI, is extremely high, and this

0:05:27.960 --> 0:05:31.039
<v Speaker 2>will be developing over time. There's no bubble because, like

0:05:31.080 --> 0:05:35.000
<v Speaker 2>I said, we're just starting. Sometimes people talk about a

0:05:35.040 --> 0:05:37.479
<v Speaker 2>bubble in a reference to the stock market because we

0:05:37.520 --> 0:05:40.880
<v Speaker 2>have seen some company getting extremely high valuation as a

0:05:40.920 --> 0:05:43.320
<v Speaker 2>resource of the excitement of AI.

0:05:44.520 --> 0:05:44.719
<v Speaker 1>There.

0:05:44.760 --> 0:05:48.360
<v Speaker 2>I think what you will see is more players coming

0:05:48.480 --> 0:05:52.560
<v Speaker 2>over time. So initially a few company where I would

0:05:52.600 --> 0:05:57.400
<v Speaker 2>say very much the only winner out of AI because

0:05:57.400 --> 0:06:01.320
<v Speaker 2>of the demand. Because industry, the industrial demand will be

0:06:01.320 --> 0:06:03.640
<v Speaker 2>so high, we need more players, and you will see

0:06:03.640 --> 0:06:07.919
<v Speaker 2>more and more company designing chips, designing AI, product manufacturing chips,

0:06:07.920 --> 0:06:11.760
<v Speaker 2>et cetera, et cetera. That could create, of course some

0:06:11.839 --> 0:06:15.920
<v Speaker 2>change on the stock market, but you know the two

0:06:15.960 --> 0:06:17.960
<v Speaker 2>are a bit unrelated.

0:06:18.040 --> 0:06:20.240
<v Speaker 1>Because I can make me three hundred million dollars just

0:06:20.279 --> 0:06:22.320
<v Speaker 1>this year along from hyper scales, maybe four hundred million

0:06:22.360 --> 0:06:26.000
<v Speaker 1>next year. Yeah, that translates into real orders for your

0:06:26.120 --> 0:06:28.560
<v Speaker 1>kids in the years ahead. You start to see it over.

0:06:28.480 --> 0:06:30.960
<v Speaker 2>Time, and sometimes you know the last few months. I

0:06:31.040 --> 0:06:32.919
<v Speaker 2>used to joke with some of our investors because they

0:06:32.960 --> 0:06:35.160
<v Speaker 2>told us, well, you know, two ndred billion here, five

0:06:35.279 --> 0:06:37.760
<v Speaker 2>ulred billion here. I say, well, I still don't have

0:06:37.880 --> 0:06:42.440
<v Speaker 2>the equation to translate doorse order into order for us,

0:06:43.200 --> 0:06:45.040
<v Speaker 2>and it takes a bit of time, but you're right.

0:06:45.640 --> 0:06:50.880
<v Speaker 2>Over time that demand translate into cheap demand to our customer.

0:06:51.640 --> 0:06:55.000
<v Speaker 2>This translate into a need for more capacity, and this

0:06:55.160 --> 0:06:59.919
<v Speaker 2>translate into demand for ISML and our peers in the industry.

0:07:00.279 --> 0:07:02.920
<v Speaker 1>The infrastructure span from the hyper scales and the kind

0:07:02.920 --> 0:07:05.280
<v Speaker 1>of deal making. The open AI has done almost a

0:07:05.400 --> 0:07:08.200
<v Speaker 1>trillion dollars worth of deals committed just by that one

0:07:08.240 --> 0:07:11.400
<v Speaker 1>company alone, which isn't making a profit this year.

0:07:11.520 --> 0:07:15.640
<v Speaker 2>That makes sense to you, Well, I think the investment

0:07:15.840 --> 0:07:20.640
<v Speaker 2>makes sense overhaul, because you cannot play in AI without

0:07:20.760 --> 0:07:23.360
<v Speaker 2>investing in hyperscaler. And there may be even a few

0:07:23.480 --> 0:07:27.680
<v Speaker 2>cycle of investment because the investment today is done on

0:07:27.880 --> 0:07:32.120
<v Speaker 2>certain chips. The chips in two three years from now,

0:07:32.240 --> 0:07:34.680
<v Speaker 2>if you look at Nvidia announcement, there will be a

0:07:34.720 --> 0:07:38.240
<v Speaker 2>lot more powerful and people may be tempted again in

0:07:38.280 --> 0:07:42.280
<v Speaker 2>twenty twenty seven to invest again in Hyperscaler to make

0:07:42.440 --> 0:07:46.120
<v Speaker 2>use of those chips, so you have different cycle and

0:07:46.880 --> 0:07:48.760
<v Speaker 2>you have a bit of a harm rest because if

0:07:48.800 --> 0:07:52.720
<v Speaker 2>you don't invest today, you're out right. It's true for company.

0:07:53.240 --> 0:07:56.680
<v Speaker 2>I think it's true for government. You cannot not play

0:07:56.720 --> 0:07:58.320
<v Speaker 2>in AI and I think that's why you have this

0:07:58.560 --> 0:07:59.920
<v Speaker 2>huge backlog of demand.

0:08:00.240 --> 0:08:02.480
<v Speaker 1>What you touched on on one part of this, which

0:08:02.560 --> 0:08:05.120
<v Speaker 1>is that the pace of innovation, and it makes me

0:08:05.160 --> 0:08:08.160
<v Speaker 1>think of depreciation around some of these assets, particularly when

0:08:08.160 --> 0:08:10.600
<v Speaker 1>you spend heavily on Blackwell and then we've got Rubin

0:08:10.640 --> 0:08:13.480
<v Speaker 1>coming up for it as one example. Good news for

0:08:13.560 --> 0:08:16.160
<v Speaker 1>you because you benefit from that increased investment that's gonna

0:08:16.160 --> 0:08:18.400
<v Speaker 1>have to follow that, but some are gonna get burned

0:08:18.440 --> 0:08:20.840
<v Speaker 1>by that depreciating asset, presumably.

0:08:20.800 --> 0:08:23.080
<v Speaker 2>Could be I think, you know, it's a bit what

0:08:23.160 --> 0:08:25.679
<v Speaker 2>I said. You have a certain landscape when you start.

0:08:25.760 --> 0:08:29.560
<v Speaker 2>The portity is big. You would see more and more players.

0:08:30.560 --> 0:08:35.000
<v Speaker 2>Will everyone win, I don't know, but the importity is

0:08:35.120 --> 0:08:37.520
<v Speaker 2>such that there would be more and more people entering

0:08:37.600 --> 0:08:41.240
<v Speaker 2>the business, and time would say makes the right turn,

0:08:41.320 --> 0:08:43.800
<v Speaker 2>or makes the right decision, the right investment, et cetera,

0:08:43.840 --> 0:08:47.319
<v Speaker 2>et cetera. But the opportunity will I think on the

0:08:47.360 --> 0:08:50.160
<v Speaker 2>short term you will see first more player, and then

0:08:50.160 --> 0:08:53.000
<v Speaker 2>at some point you may see a consolidation or a change.

0:08:53.240 --> 0:08:55.880
<v Speaker 2>This is very naturally in this scale of industry, but

0:08:56.280 --> 0:08:58.520
<v Speaker 2>it's too big for people to not want to play

0:08:58.559 --> 0:08:59.040
<v Speaker 2>in today.

0:09:01.040 --> 0:09:02.560
<v Speaker 1>As a malcio, Christopher Fouquet