WEBVTT - Cerebras CEO Andrew Feldman Talks IPO

0:00:02.520 --> 0:00:07.040
<v Speaker 1>Bloomberg Audio Studios, podcasts, radio news.

0:00:14.640 --> 0:00:17.360
<v Speaker 2>Welcome back to Bloomberg Tech to our TV viewers and

0:00:17.760 --> 0:00:21.000
<v Speaker 2>radio listeners around the world. Were joined by Andrew Feldman

0:00:21.160 --> 0:00:26.000
<v Speaker 2>Cerebres CEO Srebrest Priceless IPO. Andrew one hundred and eighty

0:00:26.000 --> 0:00:28.400
<v Speaker 2>five dollars a share above the top end of the

0:00:28.400 --> 0:00:32.600
<v Speaker 2>marketing range. I'm looking at the Bloomberg terminal shares indicated

0:00:32.600 --> 0:00:36.559
<v Speaker 2>to open three hundred and fifty dollars a share. Your

0:00:36.560 --> 0:00:37.319
<v Speaker 2>reaction to that.

0:00:39.000 --> 0:00:41.680
<v Speaker 1>Pretty good day? Huh?

0:00:41.760 --> 0:00:43.840
<v Speaker 2>How you feel? I mean, did you see this coming?

0:00:44.200 --> 0:00:46.199
<v Speaker 2>There's some mechanics to it, right, But you know, I

0:00:46.240 --> 0:00:48.479
<v Speaker 2>know you, I know you like to let's be honest,

0:00:48.800 --> 0:00:51.159
<v Speaker 2>you like to talk about what's happening in AI. The

0:00:51.200 --> 0:00:55.160
<v Speaker 2>big focus on inference Cerebress is placing it. But this

0:00:55.200 --> 0:00:57.880
<v Speaker 2>is the biggest IPO of the year so far. You know,

0:00:57.920 --> 0:01:00.160
<v Speaker 2>what does it mean to you and your employee in

0:01:00.200 --> 0:01:00.800
<v Speaker 2>your back.

0:01:00.720 --> 0:01:04.759
<v Speaker 1>Is it's the biggest tech one of the biggest tech

0:01:04.800 --> 0:01:08.880
<v Speaker 1>IPOs in history, and it's the biggest semi IPO in history.

0:01:10.280 --> 0:01:14.920
<v Speaker 1>We couldn't be more proud. This is the combination of

0:01:14.959 --> 0:01:21.399
<v Speaker 1>a decade, a work of countless late nights and long weekends.

0:01:21.720 --> 0:01:25.840
<v Speaker 1>We are enormously proud and excited and ready to get

0:01:25.880 --> 0:01:30.960
<v Speaker 1>back to work and start working on the on the

0:01:30.959 --> 0:01:31.720
<v Speaker 1>next great thing.

0:01:32.760 --> 0:01:35.920
<v Speaker 2>Again. The market is indicating shares to open at three

0:01:36.000 --> 0:01:37.720
<v Speaker 2>hundred and fifty dollars a share. Let's see where we're

0:01:37.760 --> 0:01:42.040
<v Speaker 2>at when trading starts. But you're being priced in as

0:01:42.040 --> 0:01:46.440
<v Speaker 2>a major player in this field. Now, what evidence would

0:01:46.480 --> 0:01:49.160
<v Speaker 2>you point to that you are in fact a major payer.

0:01:49.400 --> 0:01:52.360
<v Speaker 2>A lot of focus is on the concentration in the

0:01:52.400 --> 0:01:55.080
<v Speaker 2>relationship with open Ai. Tell me about some of the

0:01:55.120 --> 0:01:59.480
<v Speaker 2>other frontier labs or other names that you have some

0:01:59.560 --> 0:02:00.480
<v Speaker 2>concrete talks with.

0:02:00.840 --> 0:02:04.160
<v Speaker 1>Sure, in the last four months, we announced a deal

0:02:04.240 --> 0:02:08.239
<v Speaker 1>with open ai that's north of twenty billion dollars for

0:02:08.440 --> 0:02:13.080
<v Speaker 1>seven hundred and fifty megawatts of compute. We also announced

0:02:13.840 --> 0:02:19.120
<v Speaker 1>a major engagement with AWS where our equipment would be

0:02:19.120 --> 0:02:23.160
<v Speaker 1>deployed in their data centers. I think those are obviously

0:02:23.200 --> 0:02:26.160
<v Speaker 1>the largest, but there are dozens of others that are

0:02:27.600 --> 0:02:30.520
<v Speaker 1>in what used to be a big deal in the

0:02:30.560 --> 0:02:33.600
<v Speaker 1>ten to fifty million dollar category, and so there's just

0:02:34.200 --> 0:02:38.680
<v Speaker 1>an extraordinary demand right now for fast inference. We're the fastest,

0:02:38.760 --> 0:02:40.720
<v Speaker 1>not by a little bit, but by more than an

0:02:40.840 --> 0:02:44.120
<v Speaker 1>order of magnitude. You know, we're fifteen times faster than

0:02:44.160 --> 0:02:48.079
<v Speaker 1>the next nearest competitor, and as AI has become useful,

0:02:48.360 --> 0:02:50.920
<v Speaker 1>everybody wants it to be fast. Nobody wants to wait.

0:02:51.880 --> 0:02:54.920
<v Speaker 2>On that, you know, twenty one x on you know

0:02:55.040 --> 0:02:57.040
<v Speaker 2>performance and also like a lot of emphasis on the

0:02:57.080 --> 0:03:00.200
<v Speaker 2>dollar PA token, right that that is the metric that

0:03:00.240 --> 0:03:03.760
<v Speaker 2>the field cares about. But that's on paper. You know,

0:03:03.840 --> 0:03:06.799
<v Speaker 2>you just talked about the AWS relationship as an engagement.

0:03:07.400 --> 0:03:10.640
<v Speaker 2>When does the word engagement end and real revenues start

0:03:10.680 --> 0:03:13.679
<v Speaker 2>to show up for you from those relationships.

0:03:14.000 --> 0:03:17.600
<v Speaker 1>Sure, we signed a binding term sheet with AWS as

0:03:17.639 --> 0:03:21.720
<v Speaker 1>described in RS one, and we're working through the master agreement.

0:03:23.200 --> 0:03:26.280
<v Speaker 1>I think in dealing with organizations of that size, it

0:03:26.320 --> 0:03:29.800
<v Speaker 1>takes a little time to dot all the eyes across

0:03:29.880 --> 0:03:33.079
<v Speaker 1>the t's, but we are extremely confident that they will

0:03:33.120 --> 0:03:37.120
<v Speaker 1>be an enormous channel for us and a partner in

0:03:37.200 --> 0:03:44.120
<v Speaker 1>delivering our technology to large enterprises and medium sized enterprises

0:03:44.160 --> 0:03:47.880
<v Speaker 1>around the world. I think they are one of the

0:03:47.920 --> 0:03:51.400
<v Speaker 1>preferred cloud providers for just about every enterprise on Earth,

0:03:52.000 --> 0:03:56.560
<v Speaker 1>and so an opportunity to have your solution embedded in

0:03:56.600 --> 0:04:01.520
<v Speaker 1>their offering as part of their bedrock offering. That's a

0:04:01.600 --> 0:04:02.480
<v Speaker 1>huge win for us.

0:04:03.920 --> 0:04:06.240
<v Speaker 2>We're live on Bloomberg Television and Radio. This is a

0:04:06.280 --> 0:04:09.280
<v Speaker 2>Bloomberg Tech takeover, and we're speaking to Andrew Feldman, the

0:04:09.320 --> 0:04:12.440
<v Speaker 2>CEO of Cerebrus, whose company just priced the IPO at

0:04:12.440 --> 0:04:14.320
<v Speaker 2>one hundred and eighty five dollars a share and is

0:04:14.360 --> 0:04:17.599
<v Speaker 2>indicated right now, I'm looking on the Bloomberg terminal maybe

0:04:17.640 --> 0:04:20.560
<v Speaker 2>to start trade at three hundred and fifty dollars a share. Andrew,

0:04:20.640 --> 0:04:22.600
<v Speaker 2>I think you know this about me. Whenever we have

0:04:23.080 --> 0:04:25.280
<v Speaker 2>a big moment to have a conversation like this, I

0:04:25.320 --> 0:04:27.840
<v Speaker 2>always go to the Bloomberg Tech audience. What do they

0:04:27.880 --> 0:04:32.640
<v Speaker 2>want to know? And actually the first question is what

0:04:32.760 --> 0:04:37.279
<v Speaker 2>was your attitude toward retail investors? Why did you not

0:04:37.400 --> 0:04:40.240
<v Speaker 2>do more for the retail investor in this big moment

0:04:40.320 --> 0:04:41.800
<v Speaker 2>for your company?

0:04:42.960 --> 0:04:46.120
<v Speaker 1>We were more than twenty five xover subscribed. There was

0:04:46.800 --> 0:04:49.560
<v Speaker 1>a lot of hard decisions that needed to be made,

0:04:50.279 --> 0:04:54.040
<v Speaker 1>and nobody got what they wanted. And you know, we

0:04:54.080 --> 0:04:57.360
<v Speaker 1>did our best. It's about all you can do. When

0:04:57.400 --> 0:05:00.280
<v Speaker 1>there's twenty five times more demand for your stock at

0:05:00.279 --> 0:05:04.400
<v Speaker 1>the institutional level and at the retail level, then there

0:05:04.440 --> 0:05:06.840
<v Speaker 1>are shares to be sold. So we're really proud of

0:05:07.120 --> 0:05:08.840
<v Speaker 1>the way we chose to do it. We thought we

0:05:08.920 --> 0:05:13.360
<v Speaker 1>did it with integrity, and I think we are very

0:05:13.360 --> 0:05:15.880
<v Speaker 1>comfortable with the ending allocation.

0:05:17.040 --> 0:05:20.400
<v Speaker 2>Full stack, fully vertically integrated. And what I mean by

0:05:20.400 --> 0:05:23.680
<v Speaker 2>that is you literally build the supercomputer top to tail, right,

0:05:23.720 --> 0:05:26.240
<v Speaker 2>so in Video will do the tray, not just GPU,

0:05:26.320 --> 0:05:30.240
<v Speaker 2>but then Dell super Micro will assemble it. Dell's margins

0:05:30.279 --> 0:05:35.960
<v Speaker 2>low teens, in Video's margins mid seventies, Your margins forty

0:05:36.080 --> 0:05:38.360
<v Speaker 2>forty one percent. And what I'm trying to help the

0:05:38.400 --> 0:05:42.920
<v Speaker 2>audience understand is why that strategy of owning everything top

0:05:42.960 --> 0:05:44.839
<v Speaker 2>to sail is going to pay off in the long run.

0:05:44.960 --> 0:05:46.560
<v Speaker 2>And I would say that The other question I get

0:05:46.600 --> 0:05:49.760
<v Speaker 2>for you is what's the future outlook for margins based

0:05:49.760 --> 0:05:50.440
<v Speaker 2>on your plans?

0:05:52.120 --> 0:05:54.599
<v Speaker 1>Well, I think a couple of things. I think we

0:05:54.720 --> 0:06:00.560
<v Speaker 1>have obviously opportunities to improve our cost structure. You know,

0:06:00.600 --> 0:06:04.160
<v Speaker 1>we did half a billion last year in sales. That

0:06:04.240 --> 0:06:06.520
<v Speaker 1>means we put two hundred and fifty million in the

0:06:06.520 --> 0:06:10.719
<v Speaker 1>supply chain. Obviously that's not an efficient spot. As we grow,

0:06:11.120 --> 0:06:13.680
<v Speaker 1>we will have more leverage in the supply chain and

0:06:13.720 --> 0:06:16.920
<v Speaker 1>our cost of goods will come down. I think we

0:06:17.000 --> 0:06:23.520
<v Speaker 1>have an opportunity to increase prices. I think the demand

0:06:23.640 --> 0:06:27.600
<v Speaker 1>for fast inference is overwhelming this minute, and so I

0:06:27.920 --> 0:06:30.080
<v Speaker 1>think in the in the long run, we'll be really

0:06:30.120 --> 0:06:33.240
<v Speaker 1>proud of our of our gross margins and where they

0:06:33.279 --> 0:06:35.880
<v Speaker 1>will sort of wash out as we hit scale.

0:06:36.880 --> 0:06:38.520
<v Speaker 2>I've got a question for you from a from a

0:06:38.600 --> 0:06:42.040
<v Speaker 2>terminal client via IV. Thank you for the question. Everyone's

0:06:42.080 --> 0:06:44.120
<v Speaker 2>tuned in right now, Andrew. They want they want, they

0:06:44.120 --> 0:06:46.840
<v Speaker 2>want the detail. And so one way that people look

0:06:46.839 --> 0:06:48.960
<v Speaker 2>at it is that you know, this is custom silicon,

0:06:49.480 --> 0:06:52.240
<v Speaker 2>but Cerebrius isn't. Isn't a chip per se. It's like

0:06:52.320 --> 0:06:56.000
<v Speaker 2>wafer level, right. Is there a reason why you can't

0:06:56.040 --> 0:06:59.920
<v Speaker 2>just sell that versus the whole server? Yeah?

0:07:00.240 --> 0:07:06.320
<v Speaker 1>Yeah, that's a good question. For the entire seventy year

0:07:06.400 --> 0:07:09.880
<v Speaker 1>history of the computer industry, every previous effort had failed

0:07:09.880 --> 0:07:13.800
<v Speaker 1>to build a chip of the size. So for your audience,

0:07:13.880 --> 0:07:17.000
<v Speaker 1>this chip is the size of a dinner plate, while

0:07:17.120 --> 0:07:22.320
<v Speaker 1>traditional chips are the size of a postage stamping. And

0:07:22.360 --> 0:07:25.560
<v Speaker 1>not only had everybody failed until we succeeded, but several

0:07:25.600 --> 0:07:28.880
<v Speaker 1>tried to copy us and have since failed as well.

0:07:29.920 --> 0:07:32.679
<v Speaker 1>Part of what we were capable of doing and able

0:07:32.760 --> 0:07:36.200
<v Speaker 1>to do was that we were able to use not

0:07:36.400 --> 0:07:39.360
<v Speaker 1>just the chip, but also our expertise in packaging and

0:07:39.400 --> 0:07:43.200
<v Speaker 1>in system design to solve some of the problems and

0:07:43.280 --> 0:07:46.760
<v Speaker 1>build a truly compelling solution. Right, you don't just get

0:07:46.920 --> 0:07:49.960
<v Speaker 1>fifteen or eighteen or twenty times faster than the competition

0:07:50.640 --> 0:07:53.280
<v Speaker 1>because you built a good chip. That's one of a

0:07:53.360 --> 0:07:57.440
<v Speaker 1>collection of different things that enable that sort of performance.

0:07:57.840 --> 0:08:01.360
<v Speaker 1>You can build a great chip, and the system vendor,

0:08:01.960 --> 0:08:06.040
<v Speaker 1>the ODM, can nibble away at your performance by not

0:08:06.080 --> 0:08:08.440
<v Speaker 1>delivering the right amount of power or the right amount

0:08:08.480 --> 0:08:12.400
<v Speaker 1>of IO. Right. Right, there's there's a reason why in

0:08:12.560 --> 0:08:15.640
<v Speaker 1>Video then sought to control the IO the Envy link

0:08:16.080 --> 0:08:20.600
<v Speaker 1>because they didn't want others to nibble away at their performance.

0:08:20.680 --> 0:08:24.080
<v Speaker 1>And so by building the system, we were able to

0:08:25.040 --> 0:08:27.280
<v Speaker 1>optimize all parts. I mean, you could ask the same

0:08:27.360 --> 0:08:30.800
<v Speaker 1>thing about Porscho. You know, why don't you sell just engines? Right?

0:08:30.840 --> 0:08:34.560
<v Speaker 1>And it turns out that that's a nine to eleven

0:08:34.679 --> 0:08:37.199
<v Speaker 1>is a is a beautiful car because of the engine

0:08:37.559 --> 0:08:39.080
<v Speaker 1>and everything else they put in it.

0:08:39.320 --> 0:08:41.560
<v Speaker 2>Andrew the team's telling me we've got to go, But

0:08:41.600 --> 0:08:44.319
<v Speaker 2>I got to ask you five point five billion dollars,

0:08:44.600 --> 0:08:47.439
<v Speaker 2>what do you use the proceeds for and actually, how

0:08:47.480 --> 0:08:50.839
<v Speaker 2>flexible can you now be an allocating capacity to new customers.

0:08:50.960 --> 0:08:52.120
<v Speaker 2>We just have thirty seconds.

0:08:53.000 --> 0:08:57.479
<v Speaker 1>Oh, I think we will use it to increase capacity.

0:08:57.679 --> 0:09:01.040
<v Speaker 1>We are excited to bring many new customers on board.

0:09:01.080 --> 0:09:04.679
<v Speaker 1>There's a tremendous demand for what we're doing. I think

0:09:04.720 --> 0:09:06.120
<v Speaker 1>we can be aggressive on that front.

0:09:07.679 --> 0:09:12.000
<v Speaker 2>Andrew Feldman Cerebras CEO pricing the IPO one eighty five

0:09:12.679 --> 0:09:15.079
<v Speaker 2>indicated to open three hundred and fifty dollars a share.

0:09:15.440 --> 0:09:16.320
<v Speaker 2>Thank you for your time,