1
00:00:02,520 --> 00:00:07,000
Speaker 1: Bloomberg Audio Studios, podcasts, radio news.

2
00:00:07,880 --> 00:00:11,000
Speaker 2: AMD has signed a definitive agreement with open ai to

3
00:00:11,080 --> 00:00:14,920
Speaker 2: deploy six gigawatts of AMD GPUs. AMD says it will

4
00:00:14,920 --> 00:00:17,920
Speaker 2: equate to tens of billions of dollars in revenue. Open

5
00:00:17,960 --> 00:00:20,280
Speaker 2: Ai will get up to one hundred and sixty million

6
00:00:20,320 --> 00:00:24,240
Speaker 2: am D shares in tranches and set against both operational

7
00:00:24,440 --> 00:00:28,640
Speaker 2: and financial milestones, the focus is inference. Let's bring in

8
00:00:28,720 --> 00:00:31,840
Speaker 2: AMD CO Lisa Sou and Open Ai president Greg Brockman.

9
00:00:32,120 --> 00:00:34,199
Speaker 2: Both of them join us on set here at Bloomberg

10
00:00:34,240 --> 00:00:36,840
Speaker 2: Tech in San Francisco. Good morning, Good morning, It's great

11
00:00:36,880 --> 00:00:40,280
Speaker 2: to see you here. Let's frame the opportunity. Lisa, you

12
00:00:40,320 --> 00:00:43,040
Speaker 2: know that the market reaction is very clear, But for

13
00:00:43,159 --> 00:00:46,080
Speaker 2: AMD and the AI industry at large, what do you

14
00:00:46,080 --> 00:00:47,120
Speaker 2: think this represents.

15
00:00:47,560 --> 00:00:50,600
Speaker 3: Well, look, this is a huge milestone for AMD. You know,

16
00:00:50,640 --> 00:00:53,360
Speaker 3: we are so thrilled with the partnership with the open

17
00:00:53,400 --> 00:00:55,760
Speaker 3: Ai team, and it's also you know, a huge moment

18
00:00:55,800 --> 00:00:58,200
Speaker 3: for the AI industry because you know, when you get

19
00:00:58,200 --> 00:01:01,040
Speaker 3: to the breakdown to it, you need more AI compute.

20
00:01:01,080 --> 00:01:02,920
Speaker 3: I mean, that's where we are today. Compute is a

21
00:01:02,960 --> 00:01:05,240
Speaker 3: foundation for all of the intelligence we can get from

22
00:01:05,240 --> 00:01:08,280
Speaker 3: AI and you know, we are a compute provider. We

23
00:01:08,360 --> 00:01:11,840
Speaker 3: have spent years on our roadmap. We've spent years working

24
00:01:11,880 --> 00:01:14,920
Speaker 3: with open Ai and the team and you know, together

25
00:01:15,080 --> 00:01:17,880
Speaker 3: now we're embarking on you know, a massive build out

26
00:01:18,240 --> 00:01:21,679
Speaker 3: of six gigawatts of AI compute, and it's it's a

27
00:01:21,680 --> 00:01:24,640
Speaker 3: big deal for us, for our shareholders, for our teams,

28
00:01:25,000 --> 00:01:27,560
Speaker 3: and for really you know, the partnership and the overall

29
00:01:27,600 --> 00:01:28,520
Speaker 3: AI ecosystem.

30
00:01:28,760 --> 00:01:31,200
Speaker 2: Greg I say that the top the focus is inference.

31
00:01:31,319 --> 00:01:34,680
Speaker 2: I think that's really important to be specific about what

32
00:01:34,720 --> 00:01:37,319
Speaker 2: you will do with this capacity. So so literally explain

33
00:01:37,440 --> 00:01:39,480
Speaker 2: that part. And I'm conscious that you know, in the

34
00:01:39,480 --> 00:01:42,560
Speaker 2: first instance, the first target is one gigawatt and then

35
00:01:42,600 --> 00:01:45,360
Speaker 2: eventually six gigawatts. But what will you use it for?

36
00:01:46,080 --> 00:01:49,440
Speaker 4: Well, I think that the world continues to underestimate the

37
00:01:49,480 --> 00:01:53,080
Speaker 4: amount of demand for AI compute, right that just we've

38
00:01:53,120 --> 00:01:55,960
Speaker 4: seen this explosion of demand with things like chat GBT.

39
00:01:56,280 --> 00:01:58,320
Speaker 4: You know, we're at eight hundred million weekly active users.

40
00:01:58,320 --> 00:02:00,800
Speaker 4: Now this probably didn't even exist three years go, and

41
00:02:00,840 --> 00:02:03,720
Speaker 4: we're in a position where we cannot launch futures. We

42
00:02:03,760 --> 00:02:06,840
Speaker 4: cannot launch new products simply because of lack of computational

43
00:02:06,880 --> 00:02:10,400
Speaker 4: power and we see these models continuing to get exponentially better,

44
00:02:10,840 --> 00:02:12,359
Speaker 4: and I think we're just heading to a world where

45
00:02:12,360 --> 00:02:13,960
Speaker 4: so much of the economy is going to be lifted

46
00:02:14,040 --> 00:02:16,359
Speaker 4: up and driven by progress and AI. And so we're

47
00:02:16,440 --> 00:02:18,120
Speaker 4: very much heading to a world by default that I

48
00:02:18,120 --> 00:02:20,320
Speaker 4: think looks like a compute desert, right that there's just

49
00:02:20,360 --> 00:02:22,360
Speaker 4: not enough compute to go around, and so we're trying

50
00:02:22,360 --> 00:02:24,560
Speaker 4: to build as much as possible, as quickly as possible.

51
00:02:24,840 --> 00:02:27,000
Speaker 4: So we're starting with one gigawatt simply because you've got

52
00:02:27,040 --> 00:02:29,920
Speaker 4: to start somewhere, But honestly, we're building as fast as

53
00:02:29,919 --> 00:02:32,600
Speaker 4: we possibly can and trying to bring as much computational

54
00:02:32,600 --> 00:02:35,240
Speaker 4: power to bear for the economy and for the world.

55
00:02:35,960 --> 00:02:39,520
Speaker 1: Lisa, this is such a big commitment to Instinct in

56
00:02:39,560 --> 00:02:42,720
Speaker 1: particular as a customer. Does it make open AI the

57
00:02:42,800 --> 00:02:44,880
Speaker 1: largest for that particular product.

58
00:02:45,680 --> 00:02:48,640
Speaker 3: Well, this is certainly the largest deployment that we have

59
00:02:48,639 --> 00:02:51,880
Speaker 3: announced by far. I mean, you know, six gigawatts of compute.

60
00:02:52,280 --> 00:02:53,839
Speaker 3: As Greg said, we're going to start with the first

61
00:02:53,880 --> 00:02:56,560
Speaker 3: gigawatt in the second half of twenty twenty six on

62
00:02:56,639 --> 00:03:01,240
Speaker 3: our new next generation four fifty chip. I think the

63
00:03:01,280 --> 00:03:04,120
Speaker 3: thing to understand is, you know, these types of partnerships

64
00:03:04,200 --> 00:03:07,480
Speaker 3: actually take you know, years to really get comfortable with

65
00:03:07,560 --> 00:03:09,320
Speaker 3: the idea that we're going to you know, go all

66
00:03:09,360 --> 00:03:12,679
Speaker 3: in together. And this isn't all in partnership in terms

67
00:03:12,720 --> 00:03:16,800
Speaker 3: of building out you know, the AI compute that open

68
00:03:16,840 --> 00:03:19,680
Speaker 3: ai needs for everything that they're offering to the world. So, yes,

69
00:03:19,720 --> 00:03:22,720
Speaker 3: it's a huge deal, and it also says a lot

70
00:03:22,840 --> 00:03:25,200
Speaker 3: about you know, how much needs to come together for

71
00:03:25,480 --> 00:03:27,800
Speaker 3: you know, this entire ecosystem to operate. So you know,

72
00:03:27,840 --> 00:03:29,720
Speaker 3: we are setting up you know, certainly there's a lot

73
00:03:29,760 --> 00:03:34,040
Speaker 3: of engineering work, but our teams are working together on hardware, software,

74
00:03:34,120 --> 00:03:37,120
Speaker 3: We're ensuring the supply chain, all of those elements are

75
00:03:37,160 --> 00:03:40,960
Speaker 3: set up and ready to deliver on this massive commitment.

76
00:03:41,520 --> 00:03:44,480
Speaker 1: Greg, talk us through a little bit about the players

77
00:03:44,480 --> 00:03:46,080
Speaker 1: that you need to also lean on. This has been

78
00:03:46,200 --> 00:03:47,680
Speaker 1: years in the making, as you say.

79
00:03:47,520 --> 00:03:48,080
Speaker 4: With a m D.

80
00:03:48,280 --> 00:03:51,280
Speaker 1: But what other cloud providers were involved? How you thinking

81
00:03:51,320 --> 00:03:54,240
Speaker 1: about this working with an Oracle or others out there.

82
00:03:55,120 --> 00:03:57,640
Speaker 4: Yeah, we really think of this as an industry wide effort,

83
00:03:57,960 --> 00:04:00,200
Speaker 4: and in general, we think that compute is something that

84
00:04:00,560 --> 00:04:03,080
Speaker 4: does require the entire supply chain to really wake up

85
00:04:03,120 --> 00:04:05,760
Speaker 4: and to really to start building much more than people

86
00:04:05,760 --> 00:04:08,680
Speaker 4: we're planning on. I think this starts from energy to

87
00:04:08,720 --> 00:04:12,080
Speaker 4: try to get far more power to be built. Things

88
00:04:12,120 --> 00:04:13,880
Speaker 4: like nuclear I think are going to be very important

89
00:04:13,880 --> 00:04:16,120
Speaker 4: to come online. The cloud providers are an important part

90
00:04:16,120 --> 00:04:17,640
Speaker 4: of this as well. So we're going to be deploying

91
00:04:17,680 --> 00:04:20,000
Speaker 4: AMD in our own data centers. We'll be deploying them

92
00:04:20,279 --> 00:04:22,560
Speaker 4: together with cloud providers. You know, we have a deal

93
00:04:22,600 --> 00:04:25,919
Speaker 4: with Oracle, lots of other cloud providers out there. You

94
00:04:25,920 --> 00:04:28,320
Speaker 4: can really see that we're very much in the We

95
00:04:28,440 --> 00:04:30,720
Speaker 4: just want compute as much compute as possible. We think

96
00:04:30,760 --> 00:04:32,640
Speaker 4: this is important for the economy, we think this is

97
00:04:32,640 --> 00:04:35,120
Speaker 4: important for the nation, we think this is important for humanity.

98
00:04:35,440 --> 00:04:37,600
Speaker 4: And so really we're working with everyone in this whole

99
00:04:37,640 --> 00:04:40,640
Speaker 4: industry in order to get as much compute power online

100
00:04:40,720 --> 00:04:41,680
Speaker 4: as quickly as we can.

101
00:04:42,279 --> 00:04:46,440
Speaker 2: Lisa, I'm sorry specifics where is this data center going

102
00:04:46,480 --> 00:04:50,960
Speaker 2: to be? Is it one single site? Is it Oracle

103
00:04:51,800 --> 00:04:53,039
Speaker 2: that we'll partner with you on this?

104
00:04:53,520 --> 00:04:56,680
Speaker 3: Well, actually, what this really is is an announcement of

105
00:04:56,720 --> 00:04:58,440
Speaker 3: what you know, AMD and open a are going to

106
00:04:58,480 --> 00:05:00,800
Speaker 3: do together. You know, open a I has a lot

107
00:05:00,800 --> 00:05:03,280
Speaker 3: of partners in terms of you know, where they deploy

108
00:05:03,680 --> 00:05:06,159
Speaker 3: I imagine a lot of it will be in cloud

109
00:05:06,200 --> 00:05:08,800
Speaker 3: service providers. It's really up to you know, open Ai

110
00:05:08,920 --> 00:05:12,120
Speaker 3: and Greg and Sam and the team. But the way

111
00:05:12,160 --> 00:05:15,760
Speaker 3: to think about it is, for this amount of compute,

112
00:05:15,800 --> 00:05:17,160
Speaker 3: it's going to have to be in a lot of

113
00:05:17,160 --> 00:05:18,320
Speaker 3: different places.

114
00:05:18,360 --> 00:05:20,560
Speaker 2: It's a massive amount, multiple locations.

115
00:05:20,120 --> 00:05:24,279
Speaker 3: Multiple locations, I would imagine, you know, multiple providers to

116
00:05:24,360 --> 00:05:26,160
Speaker 3: really get this online as fast as possible.

117
00:05:26,520 --> 00:05:30,679
Speaker 2: Greg, there is a lot of focus on where open

118
00:05:30,760 --> 00:05:32,919
Speaker 2: ai is going to get the money from to fund

119
00:05:32,920 --> 00:05:38,280
Speaker 2: all of this. Sam Altman's big picture commitment is well documented, right,

120
00:05:38,320 --> 00:05:41,000
Speaker 2: and the numbers to his mind are in the trillions.

121
00:05:41,520 --> 00:05:45,000
Speaker 2: But have you specifically thought about debt financing for this

122
00:05:45,080 --> 00:05:49,200
Speaker 2: relationship with a MD? Have you thought about doing a

123
00:05:49,240 --> 00:05:53,640
Speaker 2: specific equity raise? You are very committed across multiple projects.

124
00:05:53,880 --> 00:05:54,080
Speaker 3: Yeah.

125
00:05:54,120 --> 00:05:55,520
Speaker 4: Look, the way that I would the way that I

126
00:05:55,520 --> 00:05:58,640
Speaker 4: would look at this is that AI revenue is growing

127
00:05:58,680 --> 00:06:02,360
Speaker 4: faster than I think almost any product in history, and

128
00:06:02,600 --> 00:06:04,719
Speaker 4: that ultimately, at the end of the day, the reason

129
00:06:04,720 --> 00:06:07,359
Speaker 4: this compute power is so important and is so worthwhile

130
00:06:07,440 --> 00:06:11,040
Speaker 4: for everyone to build is because the revenue ultimately will

131
00:06:11,040 --> 00:06:13,760
Speaker 4: be there. Now as a company that is trying to

132
00:06:13,800 --> 00:06:16,080
Speaker 4: move as fast as we can, we look at everything right,

133
00:06:16,080 --> 00:06:19,680
Speaker 4: we look at equity debt, we look at trying to

134
00:06:19,680 --> 00:06:23,120
Speaker 4: find creative ways of financing all of this. That's been

135
00:06:23,160 --> 00:06:25,000
Speaker 4: actually a huge focus of us for the past couple

136
00:06:25,040 --> 00:06:27,840
Speaker 4: of years as thinking about how can we possibly build

137
00:06:27,920 --> 00:06:30,280
Speaker 4: the amount of compute that is required in order to

138
00:06:30,400 --> 00:06:33,359
Speaker 4: really transform this whole economy into an aipowered economy. And

139
00:06:33,440 --> 00:06:37,200
Speaker 4: so I think you'll see lots of creative ideas, but fundamentally,

140
00:06:37,360 --> 00:06:38,680
Speaker 4: I think at the end of the day, it is

141
00:06:38,720 --> 00:06:39,359
Speaker 4: because we believe.

142
00:06:39,839 --> 00:06:41,919
Speaker 2: Sorry to jump in an interrupt and carriage, just forgive

143
00:06:41,960 --> 00:06:46,839
Speaker 2: me on this one. The condition of AMD issuing the

144
00:06:46,920 --> 00:06:50,560
Speaker 2: stock to open Ai requires you to spend money basically

145
00:06:50,600 --> 00:06:55,200
Speaker 2: because you have to deliver that gig awad of capacity first. Lisa,

146
00:06:55,240 --> 00:06:58,039
Speaker 2: I have to ask you if you have assurances that

147
00:06:58,120 --> 00:06:59,240
Speaker 2: open Ai is good for it.

148
00:06:59,640 --> 00:07:02,200
Speaker 3: Well, let me be clear. I mean, this deal is

149
00:07:02,240 --> 00:07:05,160
Speaker 3: a win for am D, it's a win for open Ai,

150
00:07:05,360 --> 00:07:07,600
Speaker 3: and it's a win for our shareholders. And that's kind

151
00:07:07,600 --> 00:07:10,160
Speaker 3: of the way we put this together. I have full

152
00:07:10,200 --> 00:07:14,120
Speaker 3: confidence in you know, open Ai, Sam, Greg Sarah. I mean,

153
00:07:14,160 --> 00:07:17,480
Speaker 3: this is a massive opportunity for us right now. Right here,

154
00:07:17,600 --> 00:07:20,840
Speaker 3: it's about who has the most compute and how fast

155
00:07:20,840 --> 00:07:23,200
Speaker 3: can we get it online? And we're committing to doing

156
00:07:23,240 --> 00:07:27,040
Speaker 3: this together. And the fact is as open ai buys chips,

157
00:07:27,080 --> 00:07:29,880
Speaker 3: that's great for AMD. Our revenue goes up, our earnings

158
00:07:29,920 --> 00:07:32,600
Speaker 3: go up. You know, we expect that it will also

159
00:07:32,760 --> 00:07:37,240
Speaker 3: be very very accretive to our shareholders from day one.

160
00:07:37,720 --> 00:07:39,800
Speaker 3: And as we do that, you know, we're very happy

161
00:07:39,840 --> 00:07:41,720
Speaker 3: to have open Ai as a deep partner and we

162
00:07:41,800 --> 00:07:45,120
Speaker 3: win together. So it's like a virtuous positive cycle in

163
00:07:45,160 --> 00:07:47,480
Speaker 3: how we build out. You know, this big vision for

164
00:07:47,560 --> 00:07:50,440
Speaker 3: having all this compute out there, and yet we.

165
00:07:50,480 --> 00:07:53,200
Speaker 1: Still question as you were just talking about greg some

166
00:07:53,240 --> 00:07:55,080
Speaker 1: of the other supply chain elements. You're talking about the

167
00:07:55,160 --> 00:07:58,080
Speaker 1: need for nuclear for power. What's really interesting is we

168
00:07:58,480 --> 00:08:02,320
Speaker 1: are you feeling confident enough about the rest of the compute,

169
00:08:02,320 --> 00:08:04,040
Speaker 1: the supply chain is there? Is this going to be

170
00:08:04,200 --> 00:08:07,560
Speaker 1: US manufactured? From your perspective, were you looking and also

171
00:08:07,600 --> 00:08:09,320
Speaker 1: building out internationally with MD.

172
00:08:10,360 --> 00:08:13,920
Speaker 4: Yeah, we've been looking at really all options our preference

173
00:08:13,920 --> 00:08:15,880
Speaker 4: and really the core thing that we try to do

174
00:08:15,920 --> 00:08:17,600
Speaker 4: is build as much as possible in the US. And

175
00:08:17,600 --> 00:08:19,280
Speaker 4: you can see the commitments that we've made over the

176
00:08:19,280 --> 00:08:21,920
Speaker 4: past year, you know, five hundred billion dollars of investment

177
00:08:22,000 --> 00:08:25,040
Speaker 4: in the US, and that's not stopping. We're continuing to build.

178
00:08:25,240 --> 00:08:28,520
Speaker 4: I do think that international that there it is also

179
00:08:28,520 --> 00:08:30,320
Speaker 4: going to be important for the world to have compute.

180
00:08:30,840 --> 00:08:32,560
Speaker 4: I think that computer is going to become this like

181
00:08:32,679 --> 00:08:36,240
Speaker 4: national security strategic resource, and every country is going to

182
00:08:36,280 --> 00:08:39,440
Speaker 4: need computational power, and so that we are really not

183
00:08:39,679 --> 00:08:43,000
Speaker 4: limiting our sort of sites in terms of where to build.

184
00:08:43,040 --> 00:08:44,600
Speaker 4: But we do think it is important that the US

185
00:08:44,720 --> 00:08:47,439
Speaker 4: leads in this technology, leads in computational power, and we're

186
00:08:47,440 --> 00:08:49,559
Speaker 4: expanding the supply chain. But you can see that we've

187
00:08:49,559 --> 00:08:51,920
Speaker 4: really been working with partners across the globe in order

188
00:08:51,960 --> 00:08:54,360
Speaker 4: to actually meet the demand that we expect to becoming

189
00:08:54,360 --> 00:08:55,240
Speaker 4: in upcoming years.

190
00:08:55,720 --> 00:08:58,960
Speaker 1: Lisa, the manufacturing of these chips, will you look to

191
00:08:59,000 --> 00:09:00,840
Speaker 1: Intel at all for it? Do you think of the future?

192
00:09:02,240 --> 00:09:04,800
Speaker 3: Well, as you know, the supply chain is something that

193
00:09:04,840 --> 00:09:08,920
Speaker 3: we work on, you know, very very meticulously. I think

194
00:09:08,960 --> 00:09:11,960
Speaker 3: we have a very strong supply chain. We're certainly deeply

195
00:09:12,040 --> 00:09:15,680
Speaker 3: partnered with you know, TSMC across the supply chain. You know,

196
00:09:15,800 --> 00:09:19,599
Speaker 3: just to that earlier question, we're absolutely prioritizing building in

197
00:09:19,600 --> 00:09:22,800
Speaker 3: the United States because I think that's super important. This

198
00:09:22,880 --> 00:09:26,160
Speaker 3: is the US AI stack. We want to have as

199
00:09:26,200 --> 00:09:28,320
Speaker 3: much of it in the US as possible, and you know,

200
00:09:28,400 --> 00:09:30,480
Speaker 3: we continue to really look at, you know, how do

201
00:09:30,559 --> 00:09:32,800
Speaker 3: we ensure that there will be a strong supply chain,

202
00:09:32,840 --> 00:09:33,640
Speaker 3: you know, going forward.

203
00:09:33,960 --> 00:09:37,120
Speaker 2: Greg Sam posted on x that this deal with a

204
00:09:37,240 --> 00:09:40,079
Speaker 2: m D is incremental to what's already being done with Nvidia.

205
00:09:40,720 --> 00:09:42,560
Speaker 2: But as least know so, I spent quite a lot

206
00:09:42,600 --> 00:09:44,840
Speaker 2: of time looking at them I family and the newer

207
00:09:44,880 --> 00:09:48,040
Speaker 2: generations of products to come. Is there a very clear

208
00:09:48,160 --> 00:09:52,679
Speaker 2: specific benefit to using a m D technology for inference

209
00:09:52,840 --> 00:09:56,080
Speaker 2: relative to the capabilities of Nvidia, or do you just

210
00:09:56,120 --> 00:09:59,840
Speaker 2: see it broadly as some sort of diversifying factor.

211
00:10:00,240 --> 00:10:01,920
Speaker 4: Well, I would look at it this way, that there's

212
00:10:01,960 --> 00:10:05,080
Speaker 4: a huge fixed cost to getting AI models running on

213
00:10:05,120 --> 00:10:08,400
Speaker 4: any platform, and so that when we look at what's

214
00:10:08,440 --> 00:10:11,240
Speaker 4: out there, that actually getting AI training to work is

215
00:10:11,640 --> 00:10:14,160
Speaker 4: a huge, huge amount of lift. That's something we've really

216
00:10:14,200 --> 00:10:16,679
Speaker 4: only done the work for in Vidia, but for inference,

217
00:10:16,720 --> 00:10:20,679
Speaker 4: that that's something that's much more that there's an easier

218
00:10:20,840 --> 00:10:23,080
Speaker 4: barrier to entry there. And one thing we found is

219
00:10:23,120 --> 00:10:25,680
Speaker 4: that I think that the work that Lisa and team

220
00:10:25,720 --> 00:10:27,560
Speaker 4: have been doing on the M four to fifty series.

221
00:10:27,960 --> 00:10:30,000
Speaker 4: It's looking like it's going to be a really incredible chip.

222
00:10:30,200 --> 00:10:32,679
Speaker 4: I think that there's the way that these things work

223
00:10:32,760 --> 00:10:35,839
Speaker 4: is that there's niches for different balances of memory and

224
00:10:36,440 --> 00:10:40,040
Speaker 4: computational power, and so as we have a diversity of workloads,

225
00:10:40,040 --> 00:10:42,760
Speaker 4: we're finding that having a diversity of chips also really

226
00:10:42,800 --> 00:10:44,079
Speaker 4: accelerates what we're able to do.

227
00:10:44,400 --> 00:10:46,280
Speaker 2: Lisa. At the beginning of this conversation, I said, there

228
00:10:46,280 --> 00:10:49,640
Speaker 2: are both operational and financial milestones to be met, and

229
00:10:49,880 --> 00:10:52,000
Speaker 2: Greg explained, you've got to start somewhere. So in the

230
00:10:52,000 --> 00:10:55,120
Speaker 2: first instance, one giga what But would you just sort

231
00:10:55,120 --> 00:10:57,280
Speaker 2: of draw out the pathway to that first giga what?

232
00:10:58,200 --> 00:11:01,040
Speaker 2: You know, it seems like you're prepared to move quickly here.

233
00:11:01,200 --> 00:11:01,440
Speaker 1: Yeah.

234
00:11:01,440 --> 00:11:03,679
Speaker 3: Absolutely, And and maybe ed, if I can just build

235
00:11:03,720 --> 00:11:06,920
Speaker 3: on something that Greg said, I think he's absolutely right.

236
00:11:07,000 --> 00:11:09,560
Speaker 3: You know, we're a believer in there's a diversity of

237
00:11:09,559 --> 00:11:12,360
Speaker 3: workloads and there will be a diversity of workloads across

238
00:11:12,720 --> 00:11:17,040
Speaker 3: you know, customers, models, use cases, and from that standpoint,

239
00:11:17,480 --> 00:11:19,720
Speaker 3: you know, we feel really good about how we're positioned.

240
00:11:19,760 --> 00:11:22,800
Speaker 3: You know, we we love the work here because you know, frankly,

241
00:11:23,320 --> 00:11:25,880
Speaker 3: you know, open Ai is the ultimate power user of

242
00:11:25,920 --> 00:11:28,920
Speaker 3: our chips and and and test us in very good ways.

243
00:11:29,360 --> 00:11:32,280
Speaker 3: So I think that's that's what gives us confidence that

244
00:11:32,360 --> 00:11:34,280
Speaker 3: you know, the technology is there. And then to your

245
00:11:34,360 --> 00:11:37,080
Speaker 3: point about milestones, Yes, I mean this is you know,

246
00:11:37,120 --> 00:11:41,559
Speaker 3: clearly a case where we are tied to each other. Uh,

247
00:11:41,600 --> 00:11:44,440
Speaker 3: the first gigawatt of deployment is super important. We're going

248
00:11:44,520 --> 00:11:47,000
Speaker 3: to start that, you know, second half of next year,

249
00:11:47,240 --> 00:11:49,520
Speaker 3: and we're going to build on from there. And it

250
00:11:49,600 --> 00:11:53,640
Speaker 3: really is not just the technology, but you know, commercial milestones,

251
00:11:53,679 --> 00:11:57,080
Speaker 3: adoption milestones, and and just how we proliferate the capability

252
00:11:57,120 --> 00:11:59,280
Speaker 3: going forward. But I'm looking forward to building this as

253
00:11:59,280 --> 00:12:01,400
Speaker 3: fast as possible. Well, you know, we're already working with

254
00:12:01,480 --> 00:12:05,040
Speaker 3: a number of cloud service providers who are also very

255
00:12:05,120 --> 00:12:07,720
Speaker 3: active on our technology, and I think this is a

256
00:12:07,760 --> 00:12:10,680
Speaker 3: great catalyst to get the industry to build faster.

257
00:12:11,679 --> 00:12:14,640
Speaker 1: Tied to each other is such an interesting turn of phrase.

258
00:12:14,720 --> 00:12:19,199
Speaker 1: And Greg, look, you are seeing more AI users and

259
00:12:19,320 --> 00:12:22,880
Speaker 1: chip makers and designers becoming more financially tied to each other.

260
00:12:23,360 --> 00:12:26,000
Speaker 1: Is this going to continue? Is this the step forward

261
00:12:26,080 --> 00:12:27,920
Speaker 1: for how you see this financing going forward?

262
00:12:28,840 --> 00:12:32,480
Speaker 4: Well, I really see the world transitioning to this AI

263
00:12:32,480 --> 00:12:35,240
Speaker 4: powered economy and The interesting thing is within open AI

264
00:12:35,520 --> 00:12:39,199
Speaker 4: that we've really seen what it's like when your progress

265
00:12:39,520 --> 00:12:42,480
Speaker 4: is limited and accelerated as true sides of the coin

266
00:12:43,400 --> 00:12:46,199
Speaker 4: by computational power, like teams within open EI, that their

267
00:12:46,280 --> 00:12:49,040
Speaker 4: ability to deliver really is tied to the amount of

268
00:12:49,040 --> 00:12:51,000
Speaker 4: compute that they get. And I think we're heading to

269
00:12:51,040 --> 00:12:53,720
Speaker 4: a world where that is how the whole economy will function.

270
00:12:54,000 --> 00:12:56,520
Speaker 4: And we're starting to see it right that people having

271
00:12:56,559 --> 00:12:59,079
Speaker 4: access to better AI tools. If you're a coder, you're

272
00:12:59,120 --> 00:13:01,320
Speaker 4: able to do far more, or if you have access

273
00:13:01,480 --> 00:13:04,360
Speaker 4: to better AI models, And we're heading to a world

274
00:13:04,360 --> 00:13:06,640
Speaker 4: where if you can have ten times as much AI

275
00:13:06,760 --> 00:13:10,080
Speaker 4: power behind you, you will probably be ten times more productive.

276
00:13:10,360 --> 00:13:11,840
Speaker 4: And so I think that we're moving to a world

277
00:13:11,880 --> 00:13:14,200
Speaker 4: where the whole industry is waking up to the fact

278
00:13:14,200 --> 00:13:17,000
Speaker 4: that we have just not planned. We have not planned

279
00:13:17,040 --> 00:13:20,559
Speaker 4: for this moment where this explosion in AI demand is happening.

280
00:13:20,720 --> 00:13:22,960
Speaker 4: So it's happening all the way from the power to

281
00:13:23,000 --> 00:13:25,560
Speaker 4: the silicon, and I think this whole industry has to

282
00:13:25,600 --> 00:13:28,400
Speaker 4: find a way to actually rise to meet the occasion.

283
00:13:28,960 --> 00:13:31,800
Speaker 2: Lisa, you have given us a look into the future

284
00:13:31,800 --> 00:13:34,800
Speaker 2: before about how you see the total addressable market the

285
00:13:34,880 --> 00:13:38,800
Speaker 2: industry now that the ink is dry with open AI

286
00:13:38,920 --> 00:13:43,280
Speaker 2: and Greg, are you rethinking either your bigger picture analysis

287
00:13:43,760 --> 00:13:46,720
Speaker 2: of the market for AI accelerators and GPUs or do

288
00:13:46,800 --> 00:13:49,880
Speaker 2: you see AMD now having an improved position in that

289
00:13:49,960 --> 00:13:52,760
Speaker 2: market relative of course to your friends at Nvidia.

290
00:13:53,280 --> 00:13:56,360
Speaker 3: Well, again, I think, and I've told you before, I

291
00:13:56,400 --> 00:13:59,880
Speaker 3: believe that this is a huge market we have saw

292
00:14:00,200 --> 00:14:04,440
Speaker 3: is just the AI accelerator TAM being you know, over

293
00:14:04,480 --> 00:14:07,720
Speaker 3: five hundred billion dollars in TAM over the next few years.

294
00:14:07,960 --> 00:14:09,760
Speaker 3: I think some might say, you know, maybe I was

295
00:14:09,800 --> 00:14:13,080
Speaker 3: a little conservative in that TAM analysis, but the way

296
00:14:13,080 --> 00:14:16,160
Speaker 3: to think about it is there's so much need from compute.

297
00:14:16,200 --> 00:14:18,080
Speaker 3: I mean, you just heard it from Greg, so you

298
00:14:18,080 --> 00:14:20,800
Speaker 3: know this is a huge pie and you're going to

299
00:14:20,880 --> 00:14:23,680
Speaker 3: see the need for you know, more players coming into it.

300
00:14:23,760 --> 00:14:27,280
Speaker 3: And you know, from my standpoint, this is a big

301
00:14:27,360 --> 00:14:30,680
Speaker 3: validation of our technology and our capability. You know, as

302
00:14:30,760 --> 00:14:32,720
Speaker 3: much as we love the work with open AI, we're

303
00:14:32,760 --> 00:14:35,120
Speaker 3: working with a lot of other customers as well. There's

304
00:14:35,160 --> 00:14:37,040
Speaker 3: a lot of excitement in the industry around m I

305
00:14:37,080 --> 00:14:38,760
Speaker 3: four fifty, so we're ready for it.

306
00:14:39,520 --> 00:14:42,840
Speaker 1: M d C Lisa s f and AI President Greg Brockman,

307
00:14:43,040 --> 00:14:44,840
Speaker 1: it's been a joy having you on the show. Thank

308
00:14:44,880 --> 00:14:45,760
Speaker 1: you both very much.