1
00:00:00,200 --> 00:00:03,640
Speaker 1: Welcome to our Bloomberg television and radio audiences around the world.

2
00:00:03,680 --> 00:00:06,080
Speaker 1: We're live in Las Vegas. This is Money twenty twenty

3
00:00:06,320 --> 00:00:09,600
Speaker 1: and we're joined by open AI's CFO Sarah Fryer. And

4
00:00:09,600 --> 00:00:10,200
Speaker 1: this is interesting.

5
00:00:10,240 --> 00:00:11,480
Speaker 2: You have some history with.

6
00:00:11,520 --> 00:00:14,760
Speaker 1: Money twenty twenty, history with fintech, but that's kind of

7
00:00:14,800 --> 00:00:17,560
Speaker 1: the point. And so much of the focus to date

8
00:00:17,600 --> 00:00:20,800
Speaker 1: on open ai has been I guess, a personal use

9
00:00:20,800 --> 00:00:24,400
Speaker 1: of chat GPT, but let's start with the banks and finances.

10
00:00:24,480 --> 00:00:27,200
Speaker 1: How much of that is made up in your business?

11
00:00:27,280 --> 00:00:27,920
Speaker 2: Yeah, it's like you.

12
00:00:27,960 --> 00:00:29,720
Speaker 3: Ad, it's so great to be here, and you're right,

13
00:00:29,800 --> 00:00:32,680
Speaker 3: Money twenty twenty. I've seen it progress through the years

14
00:00:32,680 --> 00:00:35,600
Speaker 3: and what an incredible event it is today. And we're

15
00:00:35,600 --> 00:00:39,400
Speaker 3: here because our customers are here. AI is happening right now.

16
00:00:39,520 --> 00:00:42,360
Speaker 3: It's not experimental. It's not something that people are just

17
00:00:42,400 --> 00:00:47,600
Speaker 3: playing around with, banks, financial institutions, FinTechs. People are using

18
00:00:47,640 --> 00:00:49,400
Speaker 3: it today in their business.

19
00:00:49,600 --> 00:00:51,879
Speaker 1: There's the Morgan Stanley case study. Yeah, you know, this

20
00:00:52,000 --> 00:00:54,480
Speaker 1: week alone or past week, Bank of America's talks about

21
00:00:54,480 --> 00:00:57,880
Speaker 1: how many patents it's got in machine learning and artificial intelligence.

22
00:00:58,120 --> 00:00:59,800
Speaker 1: But do you actually have a tangible sense of what

23
00:00:59,880 --> 00:01:03,480
Speaker 1: is is those financial institutions are doing with your large

24
00:01:03,520 --> 00:01:04,240
Speaker 1: language models.

25
00:01:04,319 --> 00:01:05,240
Speaker 2: Yes, absolutely so.

26
00:01:05,280 --> 00:01:08,480
Speaker 3: Morgan Stanley is a great example in their wealth management business.

27
00:01:08,520 --> 00:01:11,480
Speaker 3: They're using our models both to help wealth advisors be

28
00:01:11,560 --> 00:01:14,360
Speaker 3: more productive. They're using it as a way to create

29
00:01:14,400 --> 00:01:18,479
Speaker 3: better financial advice and outcomes for customers. We're seeing folks

30
00:01:18,480 --> 00:01:21,840
Speaker 3: like Klarna use it in a customer service. CEO of

31
00:01:21,880 --> 00:01:23,920
Speaker 3: Clarina has been very loud and proud on this front.

32
00:01:24,120 --> 00:01:27,960
Speaker 3: That's another great case study in terms of productivity improvements.

33
00:01:28,240 --> 00:01:30,640
Speaker 3: And then we have banks like VBVA that are using

34
00:01:30,640 --> 00:01:33,720
Speaker 3: it all across their business. And that's really our message here.

35
00:01:34,000 --> 00:01:37,880
Speaker 3: It's just get started. Get enterprise chat, GPT, roll out

36
00:01:37,920 --> 00:01:40,479
Speaker 3: to your organization, see what your people do with it.

37
00:01:40,560 --> 00:01:41,640
Speaker 2: Help them just get started.

38
00:01:41,680 --> 00:01:45,560
Speaker 3: Whether they're in marketing, in finance and product everyone can

39
00:01:45,640 --> 00:01:46,880
Speaker 3: have really interesting even.

40
00:01:46,760 --> 00:01:48,920
Speaker 1: If we are just getting started. Can you help our

41
00:01:48,960 --> 00:01:54,240
Speaker 1: audience understand how meaningful contribution the financial services sector fintech

42
00:01:54,360 --> 00:01:56,080
Speaker 1: makes to open AIS revenue? Yes?

43
00:01:56,160 --> 00:01:58,800
Speaker 3: So today if you look at our largest verticals areas

44
00:01:58,880 --> 00:02:01,760
Speaker 3: like EEDU, education, healthcare.

45
00:02:01,320 --> 00:02:03,440
Speaker 2: But financials is probably third.

46
00:02:03,840 --> 00:02:06,960
Speaker 3: And again I think that's because they are typically early adopters.

47
00:02:07,200 --> 00:02:09,679
Speaker 3: They're often willing to take that risk because they see

48
00:02:09,720 --> 00:02:12,480
Speaker 3: the upside in terms of driving more revenue. But they

49
00:02:12,520 --> 00:02:15,440
Speaker 3: also are very good at managing their cost and efficiency.

50
00:02:16,320 --> 00:02:18,440
Speaker 3: And it's great when you get an early adopter like

51
00:02:18,440 --> 00:02:20,680
Speaker 3: Morgan Stanley because it tends to lead the way.

52
00:02:21,000 --> 00:02:22,239
Speaker 2: I can think of a wealth.

53
00:02:22,080 --> 00:02:24,560
Speaker 3: Management client today that it's not coming to us to

54
00:02:24,600 --> 00:02:25,960
Speaker 3: say what do we need to do?

55
00:02:26,200 --> 00:02:27,200
Speaker 2: How do we get started?

56
00:02:27,680 --> 00:02:31,200
Speaker 1: Banks in particular can be sensitive to pricing. There's a

57
00:02:31,200 --> 00:02:34,400
Speaker 1: lot of curiosity not just in as a consumer how

58
00:02:34,480 --> 00:02:38,520
Speaker 1: much I'm paying on a monthly basis for chat GPT access,

59
00:02:38,560 --> 00:02:41,440
Speaker 1: but at the enterprise level as well. Reports are for example,

60
00:02:41,520 --> 00:02:44,240
Speaker 1: at the corporate per user level, you're thinking it about

61
00:02:44,320 --> 00:02:47,320
Speaker 1: two thousand dollars per head? How are you managing us

62
00:02:47,360 --> 00:02:49,360
Speaker 1: give us insight into the pricing strategy.

63
00:02:49,440 --> 00:02:52,320
Speaker 3: Sure, so pricing is super interesting because we're really trying

64
00:02:52,360 --> 00:02:55,480
Speaker 3: to think about value and what is this person getting

65
00:02:55,520 --> 00:02:57,680
Speaker 3: from this tool? And I think I actually don't think

66
00:02:57,680 --> 00:03:00,560
Speaker 3: we've done a great job of that yet ourselves. Despite

67
00:03:00,600 --> 00:03:02,959
Speaker 3: that we have two hundred and fifty million weekly active

68
00:03:03,040 --> 00:03:06,000
Speaker 3: users and all of that's a you know, five six

69
00:03:06,040 --> 00:03:09,960
Speaker 3: percent actually converted to be plus customers, so they're paying

70
00:03:10,680 --> 00:03:13,200
Speaker 3: but if you look at the value. When we were

71
00:03:13,280 --> 00:03:15,799
Speaker 3: rolling out oh one, our reasoning model, and this is

72
00:03:15,840 --> 00:03:18,440
Speaker 3: a model that stops and thinks for you. It actually

73
00:03:18,440 --> 00:03:21,600
Speaker 3: does hard problems. I was watching a lawyer in action

74
00:03:21,840 --> 00:03:23,880
Speaker 3: using it to create a brief and at the end

75
00:03:23,919 --> 00:03:25,960
Speaker 3: we said to him, what would you have paid for that,

76
00:03:26,120 --> 00:03:28,800
Speaker 3: like if you had a paralegal doing that? He was like,

77
00:03:28,960 --> 00:03:32,160
Speaker 3: easily one thousand to two thousand dollars per hour. And

78
00:03:32,200 --> 00:03:34,560
Speaker 3: this is someone who's using it. If it's an enterprise

79
00:03:34,600 --> 00:03:38,840
Speaker 3: per seeds, probably sixty bucks per month. And this is

80
00:03:38,880 --> 00:03:41,160
Speaker 3: someone who would have paid one thousand to two thousand

81
00:03:41,200 --> 00:03:43,560
Speaker 3: dollars per hour. So I think that there's a lot

82
00:03:43,560 --> 00:03:46,760
Speaker 3: of value that is in the product today, but we

83
00:03:46,800 --> 00:03:49,360
Speaker 3: are just trying to make sure people can get started,

84
00:03:49,760 --> 00:03:52,840
Speaker 3: can actually see the outcomes, and over time we believe

85
00:03:52,840 --> 00:03:55,119
Speaker 3: that value to price will come into alignment.

86
00:03:55,360 --> 00:03:58,280
Speaker 1: Is a balance right between what's of value to your customer,

87
00:03:58,320 --> 00:04:00,400
Speaker 1: but also you know you have to account for open

88
00:04:00,440 --> 00:04:03,880
Speaker 1: AIS spending. So we talked about endlessly, particularly on the

89
00:04:03,920 --> 00:04:07,520
Speaker 1: compute side, what is that balance in what works best

90
00:04:07,520 --> 00:04:09,280
Speaker 1: for you and your customer base.

91
00:04:09,520 --> 00:04:12,520
Speaker 3: Yeah, so our first and foremost, what's most important to

92
00:04:12,600 --> 00:04:16,000
Speaker 3: us is to stay on the frontier, building the frontier models,

93
00:04:16,080 --> 00:04:19,359
Speaker 3: making sure that we are bringing ultimately agi to the

94
00:04:19,360 --> 00:04:22,960
Speaker 3: benefit of humanity. To do that, it's expensive. We have

95
00:04:23,080 --> 00:04:26,159
Speaker 3: to invest in large scale compute and so to me,

96
00:04:26,240 --> 00:04:28,760
Speaker 3: there's really two ways to finance that.

97
00:04:28,800 --> 00:04:30,400
Speaker 2: It's either through the free cash flows of.

98
00:04:30,360 --> 00:04:33,800
Speaker 3: The business spoken like a good CFO, or it's through

99
00:04:34,080 --> 00:04:37,880
Speaker 3: raising equity and debt financing because investors can see the

100
00:04:37,960 --> 00:04:39,760
Speaker 3: long term potential of this business.

101
00:04:40,120 --> 00:04:41,919
Speaker 2: So on the former, on the cash.

102
00:04:41,720 --> 00:04:43,200
Speaker 3: Flows of the business, we want to make sure we

103
00:04:43,279 --> 00:04:45,719
Speaker 3: keep growing that business. I think we have been wowed

104
00:04:45,720 --> 00:04:49,120
Speaker 3: at just the pace of growth, particularly on the consumer side.

105
00:04:49,160 --> 00:04:51,560
Speaker 3: It's about seventy five percent of our business today. But

106
00:04:51,640 --> 00:04:54,440
Speaker 3: even our enterprise businesses, they are young, but they are

107
00:04:54,560 --> 00:05:00,080
Speaker 3: ready doing an incredible amount of annualized revenue and a

108
00:05:00,160 --> 00:05:01,919
Speaker 3: real excited by the potential there.

109
00:05:02,040 --> 00:05:05,080
Speaker 1: For a bluebig television and radio audience worldwide. We're in

110
00:05:05,120 --> 00:05:07,919
Speaker 1: Las Vegas at Money twenty twenty and we're speaking to

111
00:05:07,960 --> 00:05:11,080
Speaker 1: the Open AI CFO Sarah Fryar, and you talk there

112
00:05:11,080 --> 00:05:14,520
Speaker 1: about the consumer business. Something very interesting is the future

113
00:05:14,560 --> 00:05:20,160
Speaker 1: business model ADS supported tiers, for example, very specifically segmented pricing.

114
00:05:20,520 --> 00:05:21,359
Speaker 1: How do you think about that.

115
00:05:21,440 --> 00:05:23,600
Speaker 3: Sarah, Yeah, so I think you always want to stay

116
00:05:23,680 --> 00:05:26,560
Speaker 3: open to alternate business models or ways that you can

117
00:05:26,720 --> 00:05:29,360
Speaker 3: layer a new business model on top. Now, the key

118
00:05:29,400 --> 00:05:31,480
Speaker 3: for us is always access. How do we make sure

119
00:05:31,800 --> 00:05:34,440
Speaker 3: as many people as possible get access to this tool?

120
00:05:35,160 --> 00:05:37,960
Speaker 3: And that's on a global stage, by the way, And

121
00:05:38,040 --> 00:05:40,520
Speaker 3: so to do that sometimes you do pivot away from

122
00:05:40,560 --> 00:05:44,160
Speaker 3: just pure subscription models to models like ads. My last

123
00:05:44,160 --> 00:05:47,720
Speaker 3: company was all ads, so I've definitely experienced this. I

124
00:05:47,760 --> 00:05:49,800
Speaker 3: think ads have their place, but you have to be

125
00:05:49,880 --> 00:05:52,920
Speaker 3: really mindful of were I think in areas like commerce,

126
00:05:53,000 --> 00:05:55,800
Speaker 3: ads are great. Right if I do a chat GPT

127
00:05:56,240 --> 00:05:59,880
Speaker 3: prompt for black high heels shoes, something I probably would do,

128
00:06:00,040 --> 00:06:02,440
Speaker 3: so I actually don't want a history of the black

129
00:06:02,480 --> 00:06:05,320
Speaker 3: high heeled chew. I want five stores I could buy

130
00:06:05,360 --> 00:06:08,400
Speaker 3: from right now, probably e commerce. So that's why companies

131
00:06:08,440 --> 00:06:11,479
Speaker 3: like Shopify are great customers of ours as well. But

132
00:06:11,520 --> 00:06:14,240
Speaker 3: there are other places where actually the AD model doesn't

133
00:06:14,240 --> 00:06:16,520
Speaker 3: make as much sense. You want to get the consumer

134
00:06:16,800 --> 00:06:19,599
Speaker 3: to the answer they need as fast as possible, and

135
00:06:19,680 --> 00:06:22,040
Speaker 3: I think that's where chat Gibt is a really very

136
00:06:22,080 --> 00:06:24,440
Speaker 3: different platform from say something like Google Search.

137
00:06:24,680 --> 00:06:28,599
Speaker 1: You are still relatively early in this role, but it's

138
00:06:28,640 --> 00:06:33,600
Speaker 1: been two years since the original deal with Microsoft was negotiated,

139
00:06:33,640 --> 00:06:37,520
Speaker 1: and that compute relationship is critically important. Yes, all those

140
00:06:37,640 --> 00:06:41,080
Speaker 1: terms change, are they fluid or are you just sticking

141
00:06:41,120 --> 00:06:44,040
Speaker 1: to the contract that was on your desk when you arrived.

142
00:06:43,839 --> 00:06:46,400
Speaker 3: Now, and it's actually longer than two years. Microsoft and

143
00:06:46,400 --> 00:06:49,360
Speaker 3: OpenAI have been partners, I think for actually almost five years,

144
00:06:49,400 --> 00:06:51,440
Speaker 3: so they really came to us when we were a

145
00:06:51,480 --> 00:06:54,400
Speaker 3: research lab, and the deal that we've worked with them

146
00:06:54,560 --> 00:06:58,240
Speaker 3: is they do provide compute exclusively and we give them

147
00:06:58,480 --> 00:07:01,360
Speaker 3: the IP around artificial intelligence.

148
00:07:01,400 --> 00:07:02,160
Speaker 2: So it's incredible.

149
00:07:02,240 --> 00:07:04,960
Speaker 3: The products they're rolling out today are really built a

150
00:07:05,080 --> 00:07:09,039
Speaker 3: top of open AI's AI. As we go forward and

151
00:07:09,120 --> 00:07:11,760
Speaker 3: as we get bigger, we absolutely see a maturing in

152
00:07:11,800 --> 00:07:15,080
Speaker 3: that relationship, and so for consumer benefit, we want to

153
00:07:15,080 --> 00:07:17,960
Speaker 3: make sure consumers always get access to what they need.

154
00:07:18,000 --> 00:07:20,679
Speaker 3: That will probably mean compute for more parties over time.

155
00:07:20,920 --> 00:07:21,600
Speaker 2: Where we did.

156
00:07:21,480 --> 00:07:24,440
Speaker 3: Discuss the Oracle deal, or Oracle discussed it a few

157
00:07:24,520 --> 00:07:27,239
Speaker 3: quarters back, and I think that's a good starting point

158
00:07:27,440 --> 00:07:30,400
Speaker 3: for just how do we maximize compute so we maximize

159
00:07:30,400 --> 00:07:31,720
Speaker 3: the impact for consumers.

160
00:07:32,240 --> 00:07:35,880
Speaker 1: The catchphrase or buzz word of this year, I think

161
00:07:35,960 --> 00:07:39,440
Speaker 1: has been ship products. You're the CFO, and so there's

162
00:07:39,440 --> 00:07:42,320
Speaker 1: a tension between the cost of doing so and the

163
00:07:42,360 --> 00:07:45,600
Speaker 1: need to move quickly. I think Sam has denied recently

164
00:07:45,680 --> 00:07:48,440
Speaker 1: reports that the next model will be out by year end.

165
00:07:48,760 --> 00:07:50,840
Speaker 1: What can you tell us about the path forward their

166
00:07:51,080 --> 00:07:52,640
Speaker 1: the cadence of new models to come.

167
00:07:52,880 --> 00:07:54,360
Speaker 3: Yeah, I mean I think you hit the nail on

168
00:07:54,400 --> 00:07:58,560
Speaker 3: the head ship product product velocity. That is the mantra

169
00:07:58,760 --> 00:08:01,480
Speaker 3: internally to open an eye. And it's something I've just

170
00:08:01,520 --> 00:08:04,000
Speaker 3: been so wowed by since I started as just.

171
00:08:03,960 --> 00:08:05,640
Speaker 2: How fast we do ship products right.

172
00:08:05,680 --> 00:08:09,280
Speaker 3: Even in my short tenure, I've seen one come out

173
00:08:09,280 --> 00:08:12,880
Speaker 3: our Reasoning model, that huge step forward from kind of

174
00:08:12,880 --> 00:08:16,360
Speaker 3: what has been more the Chat ChiPT model series. We've

175
00:08:16,440 --> 00:08:20,960
Speaker 3: launched things like advanced speech talk to the model itself.

176
00:08:21,680 --> 00:08:23,080
Speaker 2: The O series four.

177
00:08:23,040 --> 00:08:26,200
Speaker 3: O enough four Mini for example, four O Mini, which

178
00:08:26,200 --> 00:08:30,640
Speaker 3: is our distilled model, is one hundredth the cost of

179
00:08:30,680 --> 00:08:34,120
Speaker 3: what the original Chat Ept four model was. That is

180
00:08:34,160 --> 00:08:37,240
Speaker 3: incredible for developers, and that's why you see the API

181
00:08:37,360 --> 00:08:40,680
Speaker 3: products be so successful. And again it goes back to

182
00:08:40,720 --> 00:08:42,480
Speaker 3: how do we get this into the hands of many

183
00:08:42,880 --> 00:08:46,160
Speaker 3: developers are a force multiplier and today I'm super proud.

184
00:08:46,160 --> 00:08:49,240
Speaker 3: I think every single Ai unicorn is built atop of

185
00:08:49,360 --> 00:08:52,280
Speaker 3: open AI's API, and that will tell you how we

186
00:08:52,440 --> 00:08:53,559
Speaker 3: are the frontier model.

187
00:08:53,760 --> 00:08:55,880
Speaker 1: I think when I started covering open Ai, there was

188
00:08:55,920 --> 00:08:58,000
Speaker 1: around five hundred people at the company. It's probably near

189
00:08:58,080 --> 00:08:58,960
Speaker 1: two thousand now.

190
00:08:59,160 --> 00:08:59,400
Speaker 2: Yeah.

191
00:09:00,160 --> 00:09:03,480
Speaker 1: That as the CFOs to keep talent long serving talent happy.

192
00:09:04,000 --> 00:09:06,280
Speaker 1: My understanding is that tenders will be a big part

193
00:09:06,320 --> 00:09:09,679
Speaker 1: of that, giving employees liquidity. What will be the cadence

194
00:09:09,720 --> 00:09:12,560
Speaker 1: and sort of increments of that going forward, Sarah, Yeah.

195
00:09:12,360 --> 00:09:14,560
Speaker 3: So we are a company that has done tenders to date,

196
00:09:14,720 --> 00:09:16,560
Speaker 3: and part of that is because we are in a

197
00:09:16,640 --> 00:09:19,839
Speaker 3: competitive market, particularly for research talent. When I think about

198
00:09:19,840 --> 00:09:22,640
Speaker 3: what keeps you on that front edge of the best

199
00:09:22,679 --> 00:09:26,640
Speaker 3: models out there, it is compute, but more importantly, it's people.

200
00:09:26,840 --> 00:09:28,239
Speaker 2: It's great researchers.

201
00:09:28,679 --> 00:09:31,040
Speaker 3: And so in order to compete with companies that have

202
00:09:31,160 --> 00:09:34,360
Speaker 3: liquidity already in their stock public companies, we have taken

203
00:09:34,400 --> 00:09:36,480
Speaker 3: this approach to tenders a little bit like others in

204
00:09:36,520 --> 00:09:36,840
Speaker 3: the space.

205
00:09:36,880 --> 00:09:39,720
Speaker 1: SpaceX is a great bax one we've covered closer, yes.

206
00:09:39,760 --> 00:09:42,240
Speaker 3: And so we want to be able to keep doing that.

207
00:09:42,840 --> 00:09:45,120
Speaker 3: We want to do it thoughtfully and mindfully, knowing that.

208
00:09:45,240 --> 00:09:47,360
Speaker 3: The other rule one is to keep it on the field,

209
00:09:47,720 --> 00:09:50,199
Speaker 3: make sure we have money for computees. It's always a balance,

210
00:09:50,480 --> 00:09:52,800
Speaker 3: but we do think it's important to give our researchers

211
00:09:52,880 --> 00:09:53,800
Speaker 3: that access for.

212
00:09:53,840 --> 00:09:56,800
Speaker 1: Rob Bloomberg television and radio audience all around the globe.

213
00:09:56,840 --> 00:09:58,760
Speaker 1: We're in Las Vegas and we're speaking to the open

214
00:09:58,800 --> 00:10:04,120
Speaker 1: Ai CFO, Sarah. Open Ai is a software company, or

215
00:10:04,120 --> 00:10:06,839
Speaker 1: it was. Now A lot of the focus is on

216
00:10:06,960 --> 00:10:11,800
Speaker 1: Sam and the team's ambitions with safeguarding infrastructure. How involved

217
00:10:11,800 --> 00:10:16,000
Speaker 1: are you in that talking about the sort of construction

218
00:10:16,080 --> 00:10:18,520
Speaker 1: of five gig what data centers and the financing of

219
00:10:18,520 --> 00:10:21,040
Speaker 1: such things. Yeah, was that a surprise to come in

220
00:10:21,480 --> 00:10:22,960
Speaker 1: and sort of think, Okay, I need to get a

221
00:10:22,960 --> 00:10:24,040
Speaker 1: handle on that project.

222
00:10:24,520 --> 00:10:27,079
Speaker 2: Not a surprise, but definitely a stretch.

223
00:10:27,720 --> 00:10:30,760
Speaker 1: It's a new territory, stretch from a capital perspective, stretch.

224
00:10:30,480 --> 00:10:32,920
Speaker 3: From a capital and also just my own learning. Frankly,

225
00:10:32,960 --> 00:10:36,200
Speaker 3: I think we're all learning in this space. Infrastructure is destiny.

226
00:10:36,320 --> 00:10:39,400
Speaker 3: It's this wonderful phrase that Chris Lane has managed to

227
00:10:39,400 --> 00:10:41,680
Speaker 3: get up there in the world, and we do think

228
00:10:41,760 --> 00:10:46,320
Speaker 3: that this build is important. It's important for us competitiveness,

229
00:10:46,440 --> 00:10:49,800
Speaker 3: it's important for world productivity. It's important even with a

230
00:10:49,880 --> 00:10:52,760
Speaker 3: national security lens. And so you are right. One of

231
00:10:52,800 --> 00:10:55,199
Speaker 3: the key jobs I need to do is to figure

232
00:10:55,240 --> 00:10:58,520
Speaker 3: out that capital allocation story. It's going to be both

233
00:10:58,640 --> 00:11:01,920
Speaker 3: a working with part. It's going to be raising financing,

234
00:11:02,280 --> 00:11:05,280
Speaker 3: but it's always making sure that we are staying ahead

235
00:11:05,400 --> 00:11:08,560
Speaker 3: of what will be required. Call it two three years out,

236
00:11:08,679 --> 00:11:12,280
Speaker 3: because you can't just turn on compute today if you

237
00:11:12,360 --> 00:11:14,880
Speaker 3: need it. You actually have to have thought about it,

238
00:11:14,920 --> 00:11:17,160
Speaker 3: probably about three years ahead.

239
00:11:17,040 --> 00:11:17,800
Speaker 2: On that, if I may.

240
00:11:17,920 --> 00:11:19,360
Speaker 1: One of the things I heard from some of your

241
00:11:19,360 --> 00:11:21,520
Speaker 1: investors is you did a very good job early in

242
00:11:21,679 --> 00:11:24,520
Speaker 1: explaining the basics of the plan, but the ultimate goal

243
00:11:24,640 --> 00:11:25,200
Speaker 1: is AGI.

244
00:11:25,679 --> 00:11:26,400
Speaker 2: That's correct.

245
00:11:27,280 --> 00:11:29,360
Speaker 1: How confident do you feel you have that sort of

246
00:11:29,360 --> 00:11:32,080
Speaker 1: into infrastructure in place or a plan to have it

247
00:11:32,160 --> 00:11:32,800
Speaker 1: for AGI?

248
00:11:33,160 --> 00:11:35,520
Speaker 3: I think we have the plan in place. I think

249
00:11:35,520 --> 00:11:37,320
Speaker 3: if Sam we're sitting on the seat, he would tell

250
00:11:37,360 --> 00:11:39,120
Speaker 3: you AGI is closer.

251
00:11:38,640 --> 00:11:39,400
Speaker 2: Than most think.

252
00:11:40,679 --> 00:11:42,079
Speaker 1: But I what would you say?

253
00:11:42,440 --> 00:11:44,640
Speaker 3: I would agree based on what I'm seeing. Like one

254
00:11:44,640 --> 00:11:46,680
Speaker 3: of the best meetings I get to go to once

255
00:11:46,679 --> 00:11:49,040
Speaker 3: in a while is the research meeting, and it would

256
00:11:49,080 --> 00:11:52,440
Speaker 3: blow your mind to see what's already coming and what

257
00:11:52,760 --> 00:11:55,640
Speaker 3: as we have learned how to take reasoning models like

258
00:11:55,760 --> 00:12:00,440
Speaker 3: oh one preview yes on top of GPT models and

259
00:12:00,679 --> 00:12:04,600
Speaker 3: the interplay between those. You're now really starting to see

260
00:12:04,640 --> 00:12:09,880
Speaker 3: some incredible outcomes. PhD level outcomes where you have in

261
00:12:09,920 --> 00:12:13,760
Speaker 3: your pocket human intelligence that is PhD level and physics

262
00:12:13,760 --> 00:12:17,839
Speaker 3: and biology and chemistry in English literature, like whatever the

263
00:12:17,960 --> 00:12:20,439
Speaker 3: job is you need to do if you're a healthcare researcher,

264
00:12:20,520 --> 00:12:24,360
Speaker 3: if you're a banker, if you're in education. The tool

265
00:12:24,440 --> 00:12:27,800
Speaker 3: that you are now caring the power there blows my mind.

266
00:12:28,080 --> 00:12:30,400
Speaker 1: Sarah, you touched on it a moment ago, financing the

267
00:12:30,480 --> 00:12:33,880
Speaker 1: needs to raise capital. You've just done a pretty astonishing

268
00:12:33,960 --> 00:12:36,760
Speaker 1: large round. But a follow on, I mean you must

269
00:12:36,840 --> 00:12:39,200
Speaker 1: have a good sense of the cadence of needing to

270
00:12:39,280 --> 00:12:41,360
Speaker 1: raise funds on an annual basis. I don't know how

271
00:12:41,360 --> 00:12:41,959
Speaker 1: you plan it.

272
00:12:42,160 --> 00:12:43,200
Speaker 2: Yeah, so it goes back.

273
00:12:43,120 --> 00:12:45,280
Speaker 3: To what you said, which is really understanding in your plan,

274
00:12:45,400 --> 00:12:47,840
Speaker 3: what are your big expenses? Compute is the biggest, but

275
00:12:47,880 --> 00:12:49,520
Speaker 3: we also need to run a company, so we have

276
00:12:49,559 --> 00:12:52,480
Speaker 3: real operating expenses. We're right in the guts right now

277
00:12:52,480 --> 00:12:55,040
Speaker 3: of FY twenty five planning that is usually a three

278
00:12:55,120 --> 00:12:57,439
Speaker 3: year outlook in most companies and for us, it needs

279
00:12:57,480 --> 00:12:59,719
Speaker 3: to be because we have to make those compute decisions.

280
00:13:00,120 --> 00:13:02,559
Speaker 3: And with that comes then Okay, what is our balance

281
00:13:02,559 --> 00:13:04,920
Speaker 3: sheet going to look like? What's our cash burn rate?

282
00:13:05,679 --> 00:13:08,040
Speaker 3: At what point can we generate enough free cash flow

283
00:13:08,120 --> 00:13:10,760
Speaker 3: to actually help the business? Not ready to tell you

284
00:13:10,800 --> 00:13:13,600
Speaker 3: that today, that's for the next time we talk. And

285
00:13:13,640 --> 00:13:15,600
Speaker 3: then on top of that, how do I help keep

286
00:13:15,640 --> 00:13:17,720
Speaker 3: bringing our syndicative investors along?

287
00:13:17,800 --> 00:13:22,079
Speaker 2: I called them with us.

288
00:13:22,600 --> 00:13:25,160
Speaker 1: We are less than a week from the election, and

289
00:13:25,240 --> 00:13:27,120
Speaker 1: I think about my own use of four to Oh,

290
00:13:28,160 --> 00:13:31,280
Speaker 1: I've not used it related to the elections searching for information,

291
00:13:31,440 --> 00:13:34,680
Speaker 1: but are you preparing for that election and what impact

292
00:13:34,800 --> 00:13:35,960
Speaker 1: might it have on open AI?

293
00:13:36,120 --> 00:13:37,360
Speaker 2: Yeah, we absolutely are.

294
00:13:37,720 --> 00:13:41,160
Speaker 3: We have to be very mindful from a safety perspective today.

295
00:13:41,160 --> 00:13:41,600
Speaker 2: If you do.

296
00:13:41,600 --> 00:13:44,760
Speaker 3: Searching around the election, you'll actually see that often we

297
00:13:44,880 --> 00:13:47,960
Speaker 3: will not return a response or will return with a caveat.

298
00:13:48,000 --> 00:13:50,480
Speaker 3: That says to be mindful of your sources, and I

299
00:13:50,480 --> 00:13:53,040
Speaker 3: think I learned a lot of that. Frankly, Yet nextdoor, right,

300
00:13:53,320 --> 00:13:56,720
Speaker 3: you cannot need to be careful of not being paternalistic

301
00:13:56,800 --> 00:13:59,320
Speaker 3: or paternalistic around people. People need to be able to

302
00:13:59,320 --> 00:14:03,600
Speaker 3: find information make their own educated decisions. But at the

303
00:14:03,600 --> 00:14:05,880
Speaker 3: same time, you also need to be very aware when

304
00:14:05,920 --> 00:14:08,640
Speaker 3: you have a platform that today services two hundred and

305
00:14:08,679 --> 00:14:12,199
Speaker 3: fifty million people every single week, we have to recognize

306
00:14:12,200 --> 00:14:13,960
Speaker 3: that they're going to want to do things, and we

307
00:14:14,040 --> 00:14:17,000
Speaker 3: need to provide avenues, but in a safe and secure way.

308
00:14:17,480 --> 00:14:20,640
Speaker 1: Open AI CFO Sarah Friar here live in Las Vegas.

309
00:14:20,640 --> 00:14:22,720
Speaker 1: Thank you very much, exactly to catch up