1
00:00:02,520 --> 00:00:13,280
Speaker 1: Bloomberg Audio Studios, podcasts, radio news. Bloomberg Tech is a

2
00:00:13,360 --> 00:00:17,120
Speaker 1: live from coast to coast with Carolline Hide in New

3
00:00:17,200 --> 00:00:19,720
Speaker 1: York and Eva Low in San Francisco.

4
00:00:22,840 --> 00:00:25,480
Speaker 2: This is Bloomberg Tech coming up. Meta agrees to buy

5
00:00:25,520 --> 00:00:28,520
Speaker 2: AMD chips and data center gear and a deal worth

6
00:00:28,600 --> 00:00:30,600
Speaker 2: tens of billions of dollars.

7
00:00:30,240 --> 00:00:34,120
Speaker 3: Plus AI whiplash in markets, It's erupts again as investors

8
00:00:34,400 --> 00:00:37,760
Speaker 3: navigate the disruptive power of agentic tools and unveil by

9
00:00:37,800 --> 00:00:39,720
Speaker 3: Anthropic and Warner Brothers.

10
00:00:39,680 --> 00:00:43,200
Speaker 2: Discovery is considering a new takeover bid from paramounts Guidearce,

11
00:00:43,280 --> 00:00:46,080
Speaker 2: with its board set to review the proposal and respond.

12
00:00:46,680 --> 00:00:49,080
Speaker 3: First, we check in on these markets that bounce back

13
00:00:49,080 --> 00:00:52,600
Speaker 3: after yesterday's sell off again, Anthropic Front and Center, the

14
00:00:52,640 --> 00:00:55,800
Speaker 3: latest AI tools that have been battering software and now

15
00:00:56,040 --> 00:00:57,920
Speaker 3: adding some fuel to them. We'll get into that story

16
00:00:57,960 --> 00:00:59,320
Speaker 3: a little bit later, but at the moment, that's that

17
00:00:59,320 --> 00:01:02,400
Speaker 3: one hundred up, my nine tensven percent reprieve. Remember also

18
00:01:02,480 --> 00:01:06,200
Speaker 3: consumer confidence from a macro perspective coming in better than anticipated.

19
00:01:06,280 --> 00:01:09,399
Speaker 3: I give a bitcoin, though does nothing better than anticipated with.

20
00:01:09,319 --> 00:01:10,120
Speaker 4: This particular asset.

21
00:01:10,120 --> 00:01:12,360
Speaker 3: We're currently at sixty three, nine hundred and ninety three.

22
00:01:12,400 --> 00:01:14,759
Speaker 3: The pressure still there, D But what have you got?

23
00:01:15,600 --> 00:01:18,880
Speaker 2: Let's get to our top story. Shares of AMD pushing higher.

24
00:01:19,200 --> 00:01:22,039
Speaker 2: Meta flat had been slightly softer. At one point in

25
00:01:22,040 --> 00:01:24,800
Speaker 2: the session, AMD was up about nine percent and on

26
00:01:24,840 --> 00:01:27,120
Speaker 2: track for its best day since November. This is a

27
00:01:27,280 --> 00:01:30,920
Speaker 2: six gigawa deal. Metas agreed to buy AMD chips and

28
00:01:31,040 --> 00:01:34,440
Speaker 2: data center supplies in a deal worth double digit billions

29
00:01:34,640 --> 00:01:37,480
Speaker 2: per gigawat. This is the social media giant looks to

30
00:01:37,520 --> 00:01:40,800
Speaker 2: prioritize it's AI ambitions. Here to break it down, Bloomberg's

31
00:01:40,800 --> 00:01:43,360
Speaker 2: Ian King on the chip side, Riley Griffin on the

32
00:01:43,360 --> 00:01:45,840
Speaker 2: Meta side, around the table, Ian, I'm going to start

33
00:01:45,880 --> 00:01:48,320
Speaker 2: with you the basics of the deal. Six gigawatts over

34
00:01:48,400 --> 00:01:52,200
Speaker 2: time AMD accelerators, but also other gear. It's a big

35
00:01:52,240 --> 00:01:52,760
Speaker 2: deal for them.

36
00:01:53,000 --> 00:01:55,480
Speaker 5: Yeah, it's a real endorsement from one of the biggest

37
00:01:55,480 --> 00:01:58,240
Speaker 5: buyers of this kind of equipment. Obviously, they're playing catch

38
00:01:58,320 --> 00:02:00,520
Speaker 5: up with the in Video who had their own deal

39
00:02:00,560 --> 00:02:03,200
Speaker 5: with Meta last week. So this is a real strong

40
00:02:03,200 --> 00:02:07,080
Speaker 5: affirmation and it also includes a share warrants as well

41
00:02:07,080 --> 00:02:08,080
Speaker 5: that are going to Meta.

42
00:02:08,600 --> 00:02:12,240
Speaker 3: Riley, this is all about metacompute. Talk us through just

43
00:02:12,520 --> 00:02:14,840
Speaker 3: how large the scale is for demand for this sort

44
00:02:14,840 --> 00:02:15,840
Speaker 3: of infstructure right now.

45
00:02:16,880 --> 00:02:21,120
Speaker 6: It's insatiable demand from Meta. Mark Zuckerberg last month announced

46
00:02:21,200 --> 00:02:25,880
Speaker 6: Meta Compute, with ambitions to get hundreds of gigawatts to

47
00:02:26,000 --> 00:02:29,960
Speaker 6: fuel its data centers and ultimately reach that super intelligence

48
00:02:29,960 --> 00:02:33,200
Speaker 6: score where AI can outpace human intelligence. And this is

49
00:02:33,320 --> 00:02:36,760
Speaker 6: just the latest in a frenzy of deals. Again, we're

50
00:02:36,760 --> 00:02:39,600
Speaker 6: anticipating one hundred and thirty five billion in capex this year,

51
00:02:39,639 --> 00:02:41,840
Speaker 6: but now we know the spending won't stop.

52
00:02:42,880 --> 00:02:46,360
Speaker 2: The warrants part is very interesting. It's very similar to

53
00:02:46,440 --> 00:02:51,160
Speaker 2: what AMD did with open Ai. There are operational and

54
00:02:51,280 --> 00:02:56,000
Speaker 2: financial milestones in both directions in order for Meta to

55
00:02:56,000 --> 00:02:58,840
Speaker 2: potentially become a very big shareholder AMD. Could you explain

56
00:02:58,880 --> 00:02:59,240
Speaker 2: that in.

57
00:02:59,280 --> 00:03:01,799
Speaker 5: Yeah, the way to look at it, Lisas he said, Look,

58
00:03:01,840 --> 00:03:06,960
Speaker 5: this basically ties as closer to Meta. They get these

59
00:03:07,000 --> 00:03:09,040
Speaker 5: shares if we're in a very good position. I mean,

60
00:03:09,040 --> 00:03:11,840
Speaker 5: the last chalm I believe is about six hundred dollars there.

61
00:03:11,919 --> 00:03:14,520
Speaker 5: You know yesterday they closed at less than two hundred,

62
00:03:14,600 --> 00:03:16,160
Speaker 5: so we've got a room to go over before that

63
00:03:16,200 --> 00:03:18,560
Speaker 5: becomes a reality. But you know, her point was that

64
00:03:18,600 --> 00:03:22,320
Speaker 5: this kind of locks us together and tightens our relationship.

65
00:03:22,600 --> 00:03:25,040
Speaker 2: It was really interesting to read through the materials in

66
00:03:25,080 --> 00:03:28,920
Speaker 2: our reporting. Riley, like AMD has played this card before.

67
00:03:29,360 --> 00:03:32,280
Speaker 2: The emphasis is on inference. So m I four fifty

68
00:03:32,639 --> 00:03:36,040
Speaker 2: generation of accelerator. The engineering teams will work together, but

69
00:03:36,120 --> 00:03:39,440
Speaker 2: the inference part. Is there anything unique about this different

70
00:03:39,480 --> 00:03:42,080
Speaker 2: to what Meta is doing with nvidio because we had

71
00:03:42,120 --> 00:03:43,600
Speaker 2: a very similar deal very recently.

72
00:03:44,160 --> 00:03:46,680
Speaker 6: It's a really good question. Not only is it similar

73
00:03:46,680 --> 00:03:48,200
Speaker 6: to what Meta is doing with in a Video, but

74
00:03:48,320 --> 00:03:52,080
Speaker 6: Meta also has its own internal pipeline of custom chips

75
00:03:52,080 --> 00:03:55,600
Speaker 6: that it's building for AI purposes. What we heard yesterday

76
00:03:56,160 --> 00:03:59,920
Speaker 6: is that they see different applications, different workloads here being

77
00:04:00,000 --> 00:04:03,760
Speaker 6: supported by all three of those verticals, and so they're

78
00:04:03,800 --> 00:04:08,520
Speaker 6: trying to diversify as they pursue this massive scale in

79
00:04:08,600 --> 00:04:10,240
Speaker 6: terms of compute.

80
00:04:09,720 --> 00:04:12,560
Speaker 3: In take us back to in video because we've got

81
00:04:12,600 --> 00:04:16,120
Speaker 3: their earnings coming up tomorrow. And how much does this

82
00:04:16,600 --> 00:04:20,800
Speaker 3: show that there is anxiety that companies are diversifying looking

83
00:04:20,839 --> 00:04:23,680
Speaker 3: at MD, looking their own in house chips, or actually

84
00:04:23,680 --> 00:04:26,800
Speaker 3: does it still not a concerned in video is still dominant?

85
00:04:27,640 --> 00:04:27,839
Speaker 1: Yeah.

86
00:04:27,839 --> 00:04:29,480
Speaker 5: I mean at this point all we can do is say,

87
00:04:29,520 --> 00:04:32,200
Speaker 5: look at the numbers, and the numbers assuming in video

88
00:04:32,320 --> 00:04:34,960
Speaker 5: comes in and executes on what Wall Street thinks it's

89
00:04:34,960 --> 00:04:37,760
Speaker 5: going to do. The numbers say there's no anxiety here,

90
00:04:38,120 --> 00:04:41,160
Speaker 5: or boats are floating. Everybody is benefiting from the same

91
00:04:41,200 --> 00:04:44,960
Speaker 5: massive spending in equal measure. I guess what would change

92
00:04:45,000 --> 00:04:48,560
Speaker 5: that would be if in Vidia does not hit those targets,

93
00:04:48,600 --> 00:04:51,640
Speaker 5: does not exceed those targets, that I think would change

94
00:04:51,680 --> 00:04:53,320
Speaker 5: the conversation considerably.

95
00:04:53,720 --> 00:04:56,159
Speaker 3: And Riley, the conversation is nuanced when it comes to

96
00:04:56,160 --> 00:04:58,680
Speaker 3: capital expenditure. Many would say, yes, a lot of it

97
00:04:58,720 --> 00:05:00,359
Speaker 3: is about the chips, it is about the day centers,

98
00:05:00,400 --> 00:05:03,120
Speaker 3: but it's also about the power. Can we understand really

99
00:05:03,160 --> 00:05:07,200
Speaker 3: how much Meta is having to focus in on where

100
00:05:07,240 --> 00:05:09,839
Speaker 3: they can spend on GPUs or their own homegrown chips,

101
00:05:09,880 --> 00:05:11,360
Speaker 3: and where they do it globally as well.

102
00:05:12,360 --> 00:05:15,120
Speaker 6: So we asked yesterday where they thought they would deploy

103
00:05:15,240 --> 00:05:19,359
Speaker 6: these chips, and they couldn't specify which data centers. We

104
00:05:19,440 --> 00:05:21,960
Speaker 6: know that some of their biggest projects are targeting five

105
00:05:22,000 --> 00:05:25,719
Speaker 6: gigawatts right, but they need to get certain regulatory approvals

106
00:05:25,800 --> 00:05:28,479
Speaker 6: and they need energy companies to be able to deliver

107
00:05:29,440 --> 00:05:32,520
Speaker 6: there on the ground. So as for the merging of

108
00:05:32,600 --> 00:05:35,520
Speaker 6: the energy and the compute here time will tell as

109
00:05:35,560 --> 00:05:38,040
Speaker 6: we home in on where those chips will actually go.

110
00:05:38,400 --> 00:05:40,680
Speaker 2: It's probably time for a bit of a reality check in,

111
00:05:40,800 --> 00:05:42,800
Speaker 2: and I'm hoping that you'll provide it for us. Like

112
00:05:42,920 --> 00:05:45,560
Speaker 2: in any given quarter, right now, AMD is going to

113
00:05:45,560 --> 00:05:48,599
Speaker 2: do ten billion dollars of revenue all told, every single

114
00:05:48,600 --> 00:05:52,040
Speaker 2: segment in Vidia's data center business? Is it more than

115
00:05:52,080 --> 00:05:54,960
Speaker 2: fifty billion dollars a quarter? Yes, that includes networking. But

116
00:05:55,000 --> 00:05:57,000
Speaker 2: they're now one and the same, right, you can't have

117
00:05:57,120 --> 00:06:00,440
Speaker 2: chips without the other. How do we know if this

118
00:06:00,480 --> 00:06:02,200
Speaker 2: is evidence that MD's catching up.

119
00:06:02,680 --> 00:06:05,159
Speaker 5: I mean again, you have to go back to the percentages,

120
00:06:05,240 --> 00:06:09,200
Speaker 5: the growth percentages. At the moment, both of them would say,

121
00:06:09,560 --> 00:06:12,240
Speaker 5: you know, this isn't about competition. This is about there's

122
00:06:12,279 --> 00:06:14,960
Speaker 5: so much demand a company like Meta, which makes his

123
00:06:15,000 --> 00:06:17,120
Speaker 5: own chips, is buying from both of us. That would

124
00:06:17,160 --> 00:06:20,920
Speaker 5: be the answer. But ultimately, when the market slows down,

125
00:06:21,000 --> 00:06:23,160
Speaker 5: and you can be the judge of when that happens,

126
00:06:23,440 --> 00:06:26,239
Speaker 5: then we'll see the market share shift, because then it'll

127
00:06:26,279 --> 00:06:26,880
Speaker 5: really matter.

128
00:06:27,720 --> 00:06:30,640
Speaker 2: Meta is not a hyperscaler, but it operates its own

129
00:06:30,720 --> 00:06:33,480
Speaker 2: data centers at hyper scale, if you know what I mean.

130
00:06:33,760 --> 00:06:35,760
Speaker 2: I'm thinking back to when Mark Zackobog was sat next

131
00:06:35,760 --> 00:06:38,440
Speaker 2: to the president six hundred billion dollars over the next

132
00:06:38,440 --> 00:06:41,039
Speaker 2: few years, and then he's also talked about this idea

133
00:06:41,080 --> 00:06:44,719
Speaker 2: that they might misspend two hundred billion dollars here or there,

134
00:06:44,839 --> 00:06:48,599
Speaker 2: but that's no bad thing. He's front loading purchases. Just

135
00:06:48,839 --> 00:06:50,280
Speaker 2: explain his thesis a bit.

136
00:06:50,400 --> 00:06:53,200
Speaker 6: Yeah, His strategy is just that he's used the language

137
00:06:53,200 --> 00:06:55,160
Speaker 6: of front loading capacity. They're going to try to get

138
00:06:55,200 --> 00:06:57,480
Speaker 6: as much as they can gobble it up while there's

139
00:06:57,560 --> 00:07:01,960
Speaker 6: still availability. And this, he says, could be applied not

140
00:07:02,040 --> 00:07:05,159
Speaker 6: just for AI purposes, but for that core social media business,

141
00:07:05,200 --> 00:07:09,520
Speaker 6: which still drives more than ninety eight percent of their revenue.

142
00:07:09,600 --> 00:07:12,720
Speaker 6: Right so they see applications across the board and they're

143
00:07:12,760 --> 00:07:16,040
Speaker 6: not worried about getting too much in the interim.

144
00:07:16,560 --> 00:07:18,680
Speaker 3: In why do is AM do you have to keep

145
00:07:18,720 --> 00:07:20,720
Speaker 3: giving it shares away?

146
00:07:21,560 --> 00:07:24,400
Speaker 5: Again, we asked Lisa about this and she said that

147
00:07:25,000 --> 00:07:26,600
Speaker 5: this is Lisa to the CEO, and she said, this

148
00:07:26,640 --> 00:07:28,520
Speaker 5: is different than the open air idea. The open air

149
00:07:28,560 --> 00:07:32,520
Speaker 5: ideal is a slightly different arrangement. That is a company

150
00:07:32,560 --> 00:07:37,560
Speaker 5: obviously that is seeking liquidity. Meta clearly does not need liquidity.

151
00:07:37,800 --> 00:07:41,760
Speaker 5: This is just a way of showing Meta's commitment to

152
00:07:42,000 --> 00:07:43,800
Speaker 5: what they're doing. You've got to remember that, you know,

153
00:07:43,880 --> 00:07:46,800
Speaker 5: AMD is only a couple of generations into being any

154
00:07:46,840 --> 00:07:49,960
Speaker 5: type of presence at all in this market. So if

155
00:07:49,960 --> 00:07:52,160
Speaker 5: a big buyer of this stuff comes along and says, oh,

156
00:07:52,200 --> 00:07:54,160
Speaker 5: we like your gear, oh guess what, we also like

157
00:07:54,200 --> 00:07:56,520
Speaker 5: you as a company, like your prospects, and that is

158
00:07:56,560 --> 00:08:00,440
Speaker 5: somehow a bigger validation than just a straight purchase agreement.

159
00:08:00,480 --> 00:08:03,600
Speaker 5: And that really was what they were pushing as an idea. Look,

160
00:08:03,640 --> 00:08:05,280
Speaker 5: this is how closely we are tied.

161
00:08:05,080 --> 00:08:07,280
Speaker 4: Together, beneffecting from the upside.

162
00:08:07,280 --> 00:08:10,360
Speaker 3: Bloomberg's in King and Riley Griffin, thank you so much

163
00:08:10,400 --> 00:08:11,120
Speaker 3: for joining.

164
00:08:10,880 --> 00:08:12,080
Speaker 4: Us on that roundtable.

165
00:08:12,160 --> 00:08:15,000
Speaker 3: Meanwhile, coming up software earnings, they take center stage two.

166
00:08:15,280 --> 00:08:18,160
Speaker 4: What to expect as AI fears enter the picture. This

167
00:08:18,280 --> 00:08:19,040
Speaker 4: is Bloomberg Tech.

168
00:08:33,360 --> 00:08:35,520
Speaker 2: Shares of IBM had the worst day in more than

169
00:08:35,559 --> 00:08:39,400
Speaker 2: twenty five years Monday, after Mthropic announced that its clawed

170
00:08:39,400 --> 00:08:43,600
Speaker 2: code could help to modernize the dated coding language Cobol,

171
00:08:43,920 --> 00:08:47,240
Speaker 2: largely run on IBM computers. This is in Fropit continues

172
00:08:47,280 --> 00:08:50,800
Speaker 2: to unveil new AI tools for its cowork agent software

173
00:08:50,800 --> 00:08:54,600
Speaker 2: across human resources, investment, banking, and design. Bloomberg's Brady Foyd

174
00:08:54,640 --> 00:08:57,559
Speaker 2: joins us with the latest Cobol was invented in nineteen

175
00:08:57,679 --> 00:09:00,280
Speaker 2: fifty nine. There are two hundred billion lines of OBO

176
00:09:00,360 --> 00:09:04,440
Speaker 2: called the underpin financial systems, banking payments around the world.

177
00:09:04,480 --> 00:09:07,600
Speaker 2: And basically anthropics said it would be really useful to

178
00:09:07,679 --> 00:09:11,360
Speaker 2: use Claude to help maybe investigate looking to change that,

179
00:09:11,760 --> 00:09:14,160
Speaker 2: and then one of the most story technology companies in

180
00:09:14,240 --> 00:09:17,199
Speaker 2: history had its worst day in twenty five years. Take

181
00:09:17,240 --> 00:09:17,560
Speaker 2: it from that.

182
00:09:18,920 --> 00:09:21,559
Speaker 7: You've got to feel powerful if you're anthropic right now,

183
00:09:21,640 --> 00:09:24,040
Speaker 7: right if you just say the name of a product

184
00:09:24,559 --> 00:09:28,400
Speaker 7: the company associated with that project tanks. I mean, IBM

185
00:09:28,480 --> 00:09:32,720
Speaker 7: has been talking about using AI for cobal modernization and

186
00:09:32,840 --> 00:09:35,679
Speaker 7: it seems successfully offering that tool for a couple of

187
00:09:35,760 --> 00:09:40,280
Speaker 7: years now. And yeah, markets are really jumpy, right, I mean,

188
00:09:40,320 --> 00:09:45,680
Speaker 7: the potential that Claude or another AI tool can disrupt

189
00:09:45,720 --> 00:09:50,040
Speaker 7: the leaders in a given software category is really frightening

190
00:09:50,080 --> 00:09:53,439
Speaker 7: investors right now. And it's just it's an incredibly jumpy market.

191
00:09:53,559 --> 00:09:57,480
Speaker 3: Ibmci A VP Rob Thomas was pushing back against this

192
00:09:57,679 --> 00:09:59,320
Speaker 3: writing in a blog, and you quote him in your

193
00:09:59,320 --> 00:10:02,880
Speaker 3: story that the value of IBM mainframe delivers has nothing

194
00:10:02,880 --> 00:10:04,720
Speaker 3: to do with cobol is trying to talk about the

195
00:10:04,760 --> 00:10:07,520
Speaker 3: platform more broadly brody, but what you say is so brilliant.

196
00:10:07,640 --> 00:10:11,480
Speaker 3: How powerful anthropic must feel. Also to the upside, because look,

197
00:10:11,720 --> 00:10:15,160
Speaker 3: they do a partnership with Intuit today and that fuels

198
00:10:15,200 --> 00:10:17,120
Speaker 3: into it shares a little bit on the upside, so

199
00:10:17,200 --> 00:10:18,760
Speaker 3: it can be make or break in either direction.

200
00:10:20,080 --> 00:10:21,960
Speaker 7: It fuels into it a bit on the upside. But

201
00:10:22,040 --> 00:10:25,360
Speaker 7: the company's still down what forty percent this year. I

202
00:10:25,400 --> 00:10:27,880
Speaker 7: don't know if it's quite that dramatic, but it's a

203
00:10:27,920 --> 00:10:30,880
Speaker 7: scary time to be an application software vendor, right. I

204
00:10:30,920 --> 00:10:34,640
Speaker 7: mean the most extreme idea that companies are just going

205
00:10:34,679 --> 00:10:37,040
Speaker 7: to vibe code their own solutions. I don't think anybody

206
00:10:37,120 --> 00:10:40,679
Speaker 7: really believes that, But the idea that software vendors lose

207
00:10:40,760 --> 00:10:43,720
Speaker 7: the kind of pricing leverage they've enjoyed for so long

208
00:10:43,840 --> 00:10:48,160
Speaker 7: because of that potential disruption. It's really something that markets

209
00:10:48,160 --> 00:10:50,400
Speaker 7: are struggling to grapple with right now, and it's going

210
00:10:50,440 --> 00:10:52,760
Speaker 7: to be a really interesting earning cycle.

211
00:10:55,120 --> 00:10:57,440
Speaker 2: It's basically pitching itself now as a platform. Right. So

212
00:10:57,640 --> 00:11:01,600
Speaker 2: historically Claude is focused on code and making engineering more efficient,

213
00:11:02,000 --> 00:11:03,960
Speaker 2: and like if you're an into it customer, you're aready

214
00:11:04,000 --> 00:11:06,640
Speaker 2: used of into its platforms and data. You just have

215
00:11:06,679 --> 00:11:08,840
Speaker 2: an agent inside of it. If you're a small business

216
00:11:09,200 --> 00:11:12,839
Speaker 2: or or a consumer. It's the same with DocuSign. We're

217
00:11:12,880 --> 00:11:16,280
Speaker 2: seeing all these names move. How is it making these

218
00:11:16,320 --> 00:11:19,480
Speaker 2: companies better, Brodie, you cover this beat inside and out.

219
00:11:21,320 --> 00:11:24,520
Speaker 7: Well, it's forcing them to really try to drive adoption

220
00:11:24,640 --> 00:11:28,480
Speaker 7: and use among customers. Like we all know that a

221
00:11:28,480 --> 00:11:32,080
Speaker 7: lot of the AI height was a bit premature in

222
00:11:32,160 --> 00:11:34,920
Speaker 7: terms of how mature these tools were actually and how

223
00:11:34,960 --> 00:11:37,240
Speaker 7: ready they were to be used in the enterprise. And

224
00:11:37,320 --> 00:11:40,640
Speaker 7: it's really holding toes to the fire that hey, it's

225
00:11:40,679 --> 00:11:42,960
Speaker 7: time to make sure your customers are actually using your

226
00:11:43,000 --> 00:11:45,400
Speaker 7: AI tools or else the market's going to sell your

227
00:11:45,440 --> 00:11:47,360
Speaker 7: stock and act like you are going to go poof

228
00:11:47,400 --> 00:11:48,439
Speaker 7: in a couple of years.

229
00:11:49,040 --> 00:11:51,800
Speaker 3: Plombo's Brodie Ford always a perfect way with words.

230
00:11:51,920 --> 00:11:52,520
Speaker 4: We thank you.

231
00:11:52,920 --> 00:11:55,480
Speaker 3: Let's get more on the wider AI impact on tech

232
00:11:55,600 --> 00:11:58,920
Speaker 3: and software with Clearbridge Investments senior research analysts for Software

233
00:11:59,040 --> 00:12:02,240
Speaker 3: for IT services. Anyway, Fresh Henry the perfect person. I

234
00:12:02,280 --> 00:12:04,640
Speaker 3: go back to in many ways what Brady was just saying.

235
00:12:04,920 --> 00:12:07,960
Speaker 3: We're questioning just how good the tools are. How have

236
00:12:08,040 --> 00:12:12,240
Speaker 3: we seen a sudden shift in how brilliant these AI

237
00:12:12,320 --> 00:12:13,000
Speaker 3: agents are.

238
00:12:13,320 --> 00:12:15,360
Speaker 4: In the last few months we have.

239
00:12:16,800 --> 00:12:20,439
Speaker 8: We're seeing this sellof and software in particularly SaaS because

240
00:12:21,320 --> 00:12:24,240
Speaker 8: so much is changing so rapidly, and these tools are

241
00:12:24,280 --> 00:12:29,840
Speaker 8: evolving and tremendously. There's probably a pretty big disparity between

242
00:12:30,320 --> 00:12:33,720
Speaker 8: some of the capabilities of the tools and what's happening

243
00:12:33,800 --> 00:12:37,360
Speaker 8: and likely to happen, probably even intermediate term in the marketplace,

244
00:12:37,440 --> 00:12:43,400
Speaker 8: particularly the enterprise marketplace. But investors shoot first and ask

245
00:12:43,520 --> 00:12:46,640
Speaker 8: questions at some very distant dates. So we're seeing a

246
00:12:46,640 --> 00:12:50,200
Speaker 8: pretty dramatic compression in terminal multiples of many of the names.

247
00:12:51,360 --> 00:12:53,640
Speaker 3: At what point are they going to become discerning? Because

248
00:12:53,679 --> 00:12:55,880
Speaker 3: at the moment, it's very hard for a CEO, for

249
00:12:55,920 --> 00:13:00,000
Speaker 3: Aarvin krishnover IBM, for whether it's service now, Bill McDermott,

250
00:13:00,120 --> 00:13:04,240
Speaker 3: to disprove that they're not going to be impacted, that's right.

251
00:13:04,320 --> 00:13:06,000
Speaker 3: So how do they prove a negative? How do you

252
00:13:06,080 --> 00:13:08,120
Speaker 3: start to see people willing to catch falling knife and

253
00:13:08,120 --> 00:13:09,880
Speaker 3: be like, oh, these are platform companies are not going

254
00:13:09,880 --> 00:13:10,920
Speaker 3: to be disrupted.

255
00:13:10,559 --> 00:13:12,800
Speaker 8: Or maybe they will be sure sure, It's so hard

256
00:13:12,840 --> 00:13:15,760
Speaker 8: to tell the near term reaction to any kind of

257
00:13:15,760 --> 00:13:19,840
Speaker 8: event it's been asymmetrically skewed to the downside. I think

258
00:13:19,880 --> 00:13:21,760
Speaker 8: what the vendors really have to do is show that

259
00:13:21,800 --> 00:13:24,400
Speaker 8: AI is additive to the business, to beat and raise

260
00:13:24,920 --> 00:13:28,680
Speaker 8: and show that AI isn't just off setting into clients

261
00:13:28,679 --> 00:13:31,040
Speaker 8: on the core business, but it's actually adding to the business,

262
00:13:31,080 --> 00:13:33,160
Speaker 8: and that could really change the narrative. We actually saw

263
00:13:33,160 --> 00:13:35,200
Speaker 8: that with the Snowflakes of the World two years ago,

264
00:13:35,600 --> 00:13:39,640
Speaker 8: with MAMO d B last year into this year. But

265
00:13:39,960 --> 00:13:44,120
Speaker 8: that's typically what's required for investors to realize it's not

266
00:13:44,160 --> 00:13:45,040
Speaker 8: just a zero sum game.

267
00:13:47,000 --> 00:13:50,280
Speaker 2: On the Bloomberg terminal, there are let's say a dozen

268
00:13:50,360 --> 00:13:53,559
Speaker 2: reports of single name stocks that are moving because of

269
00:13:53,600 --> 00:13:56,360
Speaker 2: the association with Anthropic, who held an event this morning.

270
00:13:56,760 --> 00:14:00,640
Speaker 2: You studied the software sector right very closely. Any evidence

271
00:14:00,720 --> 00:14:04,360
Speaker 2: those names are moving because people genuinely believe there's a

272
00:14:04,360 --> 00:14:07,000
Speaker 2: fundamental change to how they do business, or is it

273
00:14:07,080 --> 00:14:10,800
Speaker 2: just simply name association At this point.

274
00:14:10,440 --> 00:14:13,200
Speaker 8: I think it has to do with the idea that

275
00:14:13,320 --> 00:14:16,359
Speaker 8: if Anthropic sees the need to partner with these companies,

276
00:14:16,480 --> 00:14:20,120
Speaker 8: and Thropics sees that they provide something that Anthropic doesn't,

277
00:14:20,280 --> 00:14:25,160
Speaker 8: and they understand these are incumbent vendors with installed customer

278
00:14:25,200 --> 00:14:27,840
Speaker 8: bases and loyalty and processes and things like that.

279
00:14:28,440 --> 00:14:29,200
Speaker 4: It doesn't mean.

280
00:14:29,080 --> 00:14:33,560
Speaker 8: There isn't future disintermediation disintermediation risk, we all know that,

281
00:14:34,520 --> 00:14:37,600
Speaker 8: but I think it's a signal to investors at these

282
00:14:38,280 --> 00:14:42,000
Speaker 8: companies are viable and important to the ecosystem, and investors

283
00:14:42,080 --> 00:14:43,480
Speaker 8: have not been assuming that.

284
00:14:43,560 --> 00:14:44,960
Speaker 4: Over the prior few weeks.

285
00:14:45,920 --> 00:14:48,760
Speaker 2: Hiary, there was a research report, a bearish research report

286
00:14:48,760 --> 00:14:52,040
Speaker 2: from a little known outfit called Catrini that basically outlined

287
00:14:52,120 --> 00:14:55,480
Speaker 2: various risks to different sectors from AI. One of the

288
00:14:55,520 --> 00:14:58,080
Speaker 2: co authors joined Bloomberg Television. I just want to play

289
00:14:58,080 --> 00:14:59,040
Speaker 2: you some of that conversation.

290
00:15:00,040 --> 00:15:01,760
Speaker 9: I thought there was going to be a small reaction.

291
00:15:02,840 --> 00:15:05,800
Speaker 9: It was definitely larger than we expected. But I think

292
00:15:05,880 --> 00:15:07,720
Speaker 9: it's not that surprising when you take a step back

293
00:15:07,720 --> 00:15:09,680
Speaker 9: and consider kind of where the markets are in the US.

294
00:15:10,360 --> 00:15:11,880
Speaker 9: You know, the AI trade has been going on for

295
00:15:11,920 --> 00:15:13,720
Speaker 9: three and a half years. It's been more or less

296
00:15:13,720 --> 00:15:17,680
Speaker 9: a straight line up, and essentially like everyone is max

297
00:15:17,720 --> 00:15:20,560
Speaker 9: long today and so there really aren't many incremental buyers left,

298
00:15:20,960 --> 00:15:23,840
Speaker 9: and so it on the one hand, you know, spookspeople

299
00:15:23,840 --> 00:15:26,800
Speaker 9: when you do consider what is negative about this, But

300
00:15:26,800 --> 00:15:29,440
Speaker 9: I think specifically the market right now is trying to

301
00:15:29,440 --> 00:15:32,240
Speaker 9: digest this idea that AI has gotten a lot more

302
00:15:32,280 --> 00:15:33,520
Speaker 9: powerful in the last six months.

303
00:15:34,520 --> 00:15:36,080
Speaker 2: You know, in that he's kind of saying, well, this

304
00:15:36,120 --> 00:15:38,040
Speaker 2: is kind of the function of the market of the

305
00:15:38,120 --> 00:15:41,360
Speaker 2: last three years, but still poses the central question, which

306
00:15:41,440 --> 00:15:44,960
Speaker 2: is is AI rendering existing software obsolete or is it

307
00:15:45,000 --> 00:15:48,320
Speaker 2: making it better? Just your reaction to that research report

308
00:15:48,320 --> 00:15:50,160
Speaker 2: and the co author's thoughts there.

309
00:15:50,200 --> 00:15:51,600
Speaker 4: Sure, I think it's both.

310
00:15:51,640 --> 00:15:54,560
Speaker 8: I think AI represents represents a risk and it also

311
00:15:54,600 --> 00:15:57,640
Speaker 8: represents an opportunity for many of the incombent vendors. We

312
00:15:57,640 --> 00:16:01,200
Speaker 8: could take sas separately, but we've been talking internally about

313
00:16:01,200 --> 00:16:04,520
Speaker 8: what I call the great catchdown to sas. Every other sector,

314
00:16:04,760 --> 00:16:08,120
Speaker 8: every other part of software selling down to SAS on

315
00:16:08,200 --> 00:16:11,480
Speaker 8: disintermediation risks. But there are reasons why a lot of

316
00:16:11,560 --> 00:16:15,920
Speaker 8: subsectors should be relatively more defensible, and we could talk

317
00:16:15,920 --> 00:16:16,520
Speaker 8: about those a.

318
00:16:16,440 --> 00:16:20,240
Speaker 3: Few like, Okay, let's talk about what's just spent defensible

319
00:16:20,720 --> 00:16:24,160
Speaker 3: Because there's this drip feed of AI dooomerism, shall we say,

320
00:16:24,200 --> 00:16:27,360
Speaker 3: in these various reports and blog posts coming from the

321
00:16:27,400 --> 00:16:30,280
Speaker 3: CEO of Anthropic himself to over at Szuctrini. We've also

322
00:16:30,320 --> 00:16:33,400
Speaker 3: had Harvard coming out with their piece that we're seeing

323
00:16:33,440 --> 00:16:37,880
Speaker 3: it being incredibly effective when it comes to financial analysis

324
00:16:38,160 --> 00:16:41,560
Speaker 3: and portfolio decision making. The Harvard led study is talking

325
00:16:41,560 --> 00:16:44,840
Speaker 3: about how trading mutual fund training decisions. Seventy one percent

326
00:16:44,920 --> 00:16:48,600
Speaker 3: it predicted right the model was trained over a five

327
00:16:48,680 --> 00:16:51,480
Speaker 3: year window. But they're also talking about maybe the outperformance,

328
00:16:51,480 --> 00:16:54,360
Speaker 3: but is what it misses? So from your perspective, where's

329
00:16:54,360 --> 00:16:56,600
Speaker 3: the outperformance going to come from? Some of the companies

330
00:16:56,640 --> 00:16:59,600
Speaker 3: that can withstand this, that can show that they're really

331
00:16:59,600 --> 00:17:02,160
Speaker 3: effective while using the AI tools.

332
00:17:02,200 --> 00:17:05,840
Speaker 8: Sure Sure, I think of security as an example of that.

333
00:17:05,960 --> 00:17:09,960
Speaker 8: Anthropic introduced a vulnerability management tool, not even management, just

334
00:17:10,359 --> 00:17:14,040
Speaker 8: a scanner for vulnerabilities and code that's produced on the platform,

335
00:17:14,840 --> 00:17:18,000
Speaker 8: and that's a build time phenomenon. Most security vendors are

336
00:17:18,119 --> 00:17:20,679
Speaker 8: run time, but they bring a lot to bear that

337
00:17:20,720 --> 00:17:24,520
Speaker 8: we could talk about from a technological perspective that it's

338
00:17:24,560 --> 00:17:24,920
Speaker 8: going to.

339
00:17:24,840 --> 00:17:27,520
Speaker 4: Be hard for an l ELAM or an agent to have.

340
00:17:27,960 --> 00:17:31,240
Speaker 8: They have the physical infrastructure, they cover all the enforcement points.

341
00:17:31,280 --> 00:17:34,760
Speaker 8: It's partly a physical phenomenon, not just a digital phenomenon

342
00:17:34,760 --> 00:17:37,080
Speaker 8: per se. But on top of that, when you think

343
00:17:37,119 --> 00:17:43,520
Speaker 8: about it. I don't know how entities, and especially regulated

344
00:17:43,680 --> 00:17:46,679
Speaker 8: entities can have the fox guarding the henhouse when it

345
00:17:46,720 --> 00:17:50,320
Speaker 8: comes to security, and kind of the same existed in

346
00:17:50,400 --> 00:17:54,679
Speaker 8: cloud as well. Because AI creates so many vulnerabilities, so

347
00:17:54,720 --> 00:17:58,760
Speaker 8: many new vulnerabilities that we've never had, that it's a

348
00:17:58,800 --> 00:18:02,120
Speaker 8: major issue when you're looking to what's becoming the greatest

349
00:18:02,160 --> 00:18:06,520
Speaker 8: attack surface to be the solution to the problem. Similarly,

350
00:18:06,600 --> 00:18:11,159
Speaker 8: the data platform vendors are really the platforms for the

351
00:18:11,200 --> 00:18:14,000
Speaker 8: next generation of AI applications. We're seeing them build on

352
00:18:14,000 --> 00:18:17,560
Speaker 8: top of the snowflakes and Mango dbs, even potentially the

353
00:18:17,640 --> 00:18:21,320
Speaker 8: Oracles to some degree data dog that's slightly different, but

354
00:18:22,680 --> 00:18:26,440
Speaker 8: what customers are looking for is higher levels of automation

355
00:18:26,680 --> 00:18:30,159
Speaker 8: to make their lives easier when there's a tsunami.

356
00:18:29,560 --> 00:18:30,960
Speaker 4: Of data coming at them.

357
00:18:31,359 --> 00:18:34,800
Speaker 8: These infrastructures and oftentimes are quite different. The data models

358
00:18:34,800 --> 00:18:38,480
Speaker 8: are different from llms, the infrastructure on which they run

359
00:18:38,840 --> 00:18:42,359
Speaker 8: are far more efficient than what lmms do and seem

360
00:18:42,400 --> 00:18:45,480
Speaker 8: to be able to do. And so I believe the

361
00:18:45,520 --> 00:18:48,399
Speaker 8: greatest source of automation is going to come from these

362
00:18:48,440 --> 00:18:50,640
Speaker 8: incumbent vendors. I should have mentioned data bricks of course,

363
00:18:50,640 --> 00:18:55,359
Speaker 8: in which Clearbridge has a private investment, and so solving

364
00:18:55,400 --> 00:18:57,640
Speaker 8: and headaches for customers in the face of a ton

365
00:18:57,680 --> 00:19:00,800
Speaker 8: of complexity is going to be a Paramount Imports.

366
00:19:01,240 --> 00:19:04,880
Speaker 2: Hillary Fresh clear Bridge Investments, thank you very much. Now

367
00:19:04,880 --> 00:19:07,000
Speaker 2: coming up on the program, Paramount is said to have

368
00:19:07,080 --> 00:19:09,640
Speaker 2: raised its offer for Warner Brothers More on the Drama

369
00:19:10,040 --> 00:19:13,119
Speaker 2: to buy one of Hollywood's most story brands, the snackt

370
00:19:13,320 --> 00:19:14,520
Speaker 2: This is Bloomberg Tech.

371
00:19:21,440 --> 00:19:23,080
Speaker 4: Time now for Talking Tech and First Up.

372
00:19:23,080 --> 00:19:28,080
Speaker 3: Athropic has accused Deep Seat, Minimax and Moonshot of distilling.

373
00:19:27,560 --> 00:19:29,240
Speaker 4: Its clawed models to bolster their.

374
00:19:29,200 --> 00:19:32,240
Speaker 3: Own AI capabilities, adding to concerns in the US about

375
00:19:32,320 --> 00:19:34,960
Speaker 3: Chinese firms improperly gaining an edge.

376
00:19:35,160 --> 00:19:38,200
Speaker 4: Representatives for the three firms didn't respond to requests for comment.

377
00:19:38,520 --> 00:19:41,320
Speaker 3: Sticking with Anthropic look, sources say the company will let

378
00:19:41,480 --> 00:19:44,520
Speaker 3: some current and former employees sell their shares at evaluation

379
00:19:44,520 --> 00:19:45,920
Speaker 3: for about three hundred and fifty billion.

380
00:19:45,760 --> 00:19:48,080
Speaker 4: Dollars, the level reached during a recent fund raise.

381
00:19:48,359 --> 00:19:51,040
Speaker 3: Anthropic has lined up five to six billion dollars from

382
00:19:51,040 --> 00:19:54,000
Speaker 3: outside investors, though the amount and details have not yet

383
00:19:54,040 --> 00:19:58,119
Speaker 3: been finalized, and Canada Well It's summoned open AI executives

384
00:19:58,359 --> 00:20:01,760
Speaker 3: after it was revealed the company debate but ultimately didn't

385
00:20:01,920 --> 00:20:04,840
Speaker 3: refer a chat GBT user to the police. That person

386
00:20:04,960 --> 00:20:07,440
Speaker 3: became the sole suspect in one of Canada's worst ever

387
00:20:07,520 --> 00:20:10,760
Speaker 3: mass shootings. Opening I bound the use of months before

388
00:20:10,800 --> 00:20:13,560
Speaker 3: the attack, but said it found no credible or imminent

389
00:20:13,720 --> 00:20:14,639
Speaker 3: threat ed.

390
00:20:15,000 --> 00:20:17,920
Speaker 2: Okay, let's take a look at shares Warner Brothers, Discoveries, Paramount,

391
00:20:17,960 --> 00:20:19,560
Speaker 2: and Netflix and get the latest.

392
00:20:20,160 --> 00:20:21,600
Speaker 10: Paramount has raised.

393
00:20:21,320 --> 00:20:23,720
Speaker 2: Its offer for Warner Brothers from its thirty dollars a

394
00:20:23,760 --> 00:20:28,000
Speaker 2: share all cash proposal. That's according to Bloomberg Reporting. Now

395
00:20:28,160 --> 00:20:31,399
Speaker 2: Warner Brothers Discovery and its board is reviewing that revised proposal.

396
00:20:31,440 --> 00:20:34,680
Speaker 2: Bloomberg's managing editor for Media and Entertainment leading the screen

397
00:20:34,760 --> 00:20:38,080
Speaker 2: Time team because sure, we actually don't have the specifics.

398
00:20:38,080 --> 00:20:40,600
Speaker 2: I don't think yet of how improved this offer is.

399
00:20:40,640 --> 00:20:43,240
Speaker 2: But it's on the desk of Warner Brothers Discovery, It's

400
00:20:43,320 --> 00:20:45,240
Speaker 2: leaders and its board. Take it from that.

401
00:20:46,080 --> 00:20:49,920
Speaker 11: Yeah, I mean, you see both sides taking this more

402
00:20:50,080 --> 00:20:52,360
Speaker 11: seriously and being quiet. I mean there's been a kind

403
00:20:52,359 --> 00:20:55,800
Speaker 11: of a progression with these talks, where back when the

404
00:20:55,840 --> 00:20:58,680
Speaker 11: first bidding war was happening, you had Warner Brothers reviewing

405
00:20:58,760 --> 00:21:02,280
Speaker 11: offers from Comcast Netflix in Paramount, it was very quiet, right,

406
00:21:02,280 --> 00:21:04,239
Speaker 11: people didn't say a lot because everyone thought they were

407
00:21:04,240 --> 00:21:06,960
Speaker 11: in it. Then Warner Brothers picked Netflix, and I got

408
00:21:07,000 --> 00:21:09,680
Speaker 11: really loud, Paramount accusing Warner brother of doing things that

409
00:21:09,720 --> 00:21:12,760
Speaker 11: weren't right. Netflix and Warner Brothers often firing back. And

410
00:21:12,760 --> 00:21:14,960
Speaker 11: now it's gone quiet again, which to me is clearly

411
00:21:15,000 --> 00:21:18,000
Speaker 11: a sign that the board is seriously reviewing the Paramount offer.

412
00:21:19,119 --> 00:21:21,600
Speaker 11: If Paramount has increased it five a dollar to a share,

413
00:21:21,640 --> 00:21:24,720
Speaker 11: which we assume they have, it seems hard to believe

414
00:21:24,720 --> 00:21:26,679
Speaker 11: that they would outright reject it as they have the

415
00:21:26,720 --> 00:21:28,800
Speaker 11: recent ones, which means that they would have to go

416
00:21:28,840 --> 00:21:30,360
Speaker 11: to Netflix and sort of say, what do you got

417
00:21:30,400 --> 00:21:30,639
Speaker 11: for me?

418
00:21:31,920 --> 00:21:34,280
Speaker 3: Well, exactly, we'll get a drip feed of what it

419
00:21:34,359 --> 00:21:39,040
Speaker 3: is exactly that they've got. But if it is markedly better,

420
00:21:39,520 --> 00:21:41,320
Speaker 3: do we expect Netflix to respond in kind?

421
00:21:42,400 --> 00:21:46,240
Speaker 11: You know, it's a tough question to answered. Netflix has

422
00:21:46,280 --> 00:21:49,960
Speaker 11: communicated to some of its shareholders. I've heard from people

423
00:21:49,960 --> 00:21:52,320
Speaker 11: at Netflix that they feel they have the balance sheet

424
00:21:52,359 --> 00:21:52,840
Speaker 11: to go up.

425
00:21:52,920 --> 00:21:53,080
Speaker 9: Right.

426
00:21:53,119 --> 00:21:56,919
Speaker 11: We're talking about a growing business with a really significant

427
00:21:56,920 --> 00:21:58,840
Speaker 11: market cap. Even if this stalk's taken a little bit

428
00:21:58,840 --> 00:22:02,640
Speaker 11: of a beating during these negotiations. So relative to paramount

429
00:22:02,680 --> 00:22:05,080
Speaker 11: and even relative to the Ellison family, Netflix should have

430
00:22:05,119 --> 00:22:07,000
Speaker 11: the money to go there. It's a question of whether

431
00:22:07,040 --> 00:22:09,160
Speaker 11: they want to. You know, their stock is down. I've

432
00:22:09,200 --> 00:22:11,560
Speaker 11: lost track at this point. It's more than thirty percent

433
00:22:11,600 --> 00:22:15,440
Speaker 11: since these negotiations started, and they may decide that there's

434
00:22:15,480 --> 00:22:17,159
Speaker 11: a number that is too high for them. We just

435
00:22:17,200 --> 00:22:18,639
Speaker 11: don't know what that threshold is.

436
00:22:19,600 --> 00:22:21,760
Speaker 3: It continues to evolve and we know that you always

437
00:22:21,800 --> 00:22:24,040
Speaker 3: have the story, but mostly Beka, sure, thank you so much.

438
00:22:24,560 --> 00:22:28,479
Speaker 3: I'm coming up more on Meta and AMD's multi billion

439
00:22:28,520 --> 00:22:32,320
Speaker 3: dollar deal as the race to power AI intensifies and

440
00:22:32,560 --> 00:22:35,760
Speaker 3: AMD up coming off of those previous highs.

441
00:22:36,560 --> 00:22:39,760
Speaker 2: Yep, and then Meta flat as a pancake. Much more

442
00:22:39,800 --> 00:22:50,359
Speaker 2: to come. It is halftime. This is Bloomberg Tech. Welcome

443
00:22:50,359 --> 00:22:53,040
Speaker 2: back to Bloomberg Tech. We're continuing to track the impact

444
00:22:53,119 --> 00:22:57,439
Speaker 2: of AI on legacy software. Essentially lots of names actually

445
00:22:57,440 --> 00:22:58,400
Speaker 2: moving to the upside.

446
00:22:58,440 --> 00:22:59,080
Speaker 10: The story is.

447
00:22:59,080 --> 00:23:03,320
Speaker 2: An anthropic and basically taking its clawed AI agent and

448
00:23:03,400 --> 00:23:06,639
Speaker 2: linking it in a platform for everything from HR investment

449
00:23:06,720 --> 00:23:10,359
Speaker 2: banking documents, and that is pushing a lot of names higher.

450
00:23:10,680 --> 00:23:13,960
Speaker 2: IBM is up four percent, but remember yesterday Monday it

451
00:23:14,080 --> 00:23:17,000
Speaker 2: fell by the most since two thousand when it said

452
00:23:17,040 --> 00:23:21,479
Speaker 2: that Claude could be used basically to improve archive change

453
00:23:21,600 --> 00:23:25,000
Speaker 2: update cobol coding bases, a program that was invented in

454
00:23:25,119 --> 00:23:28,560
Speaker 2: nineteen fifty nine. Otherwise, our top story is a deal

455
00:23:28,880 --> 00:23:33,399
Speaker 2: between AMD and Meta six gigawatts of capacity over the

456
00:23:33,400 --> 00:23:36,720
Speaker 2: balance of the decade, largely focused in the first instance

457
00:23:36,760 --> 00:23:40,200
Speaker 2: on AMD's latest accelerator and the server that is based

458
00:23:40,240 --> 00:23:43,439
Speaker 2: on mi I four fifty, but this is basically sixty

459
00:23:43,480 --> 00:23:46,640
Speaker 2: billion dollars of potential revenue for AMD over that time period.

460
00:23:46,840 --> 00:23:49,440
Speaker 2: The stock up more than seven percent. MEATA has traded flat,

461
00:23:49,720 --> 00:23:52,359
Speaker 2: but I guess that it's hard to gauge if this

462
00:23:52,480 --> 00:23:55,920
Speaker 2: makes any different Semara at all given its existing spending commitments.

463
00:23:56,040 --> 00:23:57,720
Speaker 4: Karrac Well, someone's been writing on just that.

464
00:23:58,240 --> 00:24:01,240
Speaker 3: We meg Intelligence senior analyst Manday saying says that Meta's

465
00:24:01,359 --> 00:24:04,240
Speaker 3: six giga what IMD deal actually lowest chip spending is

466
00:24:04,240 --> 00:24:07,560
Speaker 3: a share of capex and secure supply tailored to AI inference.

467
00:24:08,240 --> 00:24:09,600
Speaker 4: Also right, so the stock.

468
00:24:09,400 --> 00:24:13,080
Speaker 3: Link structure could give Meta added leverage against Nvidia, Mande

469
00:24:13,119 --> 00:24:16,119
Speaker 3: saying joints us now, why does any leverage against Innvideo

470
00:24:16,200 --> 00:24:18,359
Speaker 3: It struck a deal last week with them. Is this

471
00:24:18,440 --> 00:24:20,240
Speaker 3: more like we've got other options?

472
00:24:20,640 --> 00:24:21,000
Speaker 9: Yeah?

473
00:24:21,040 --> 00:24:23,840
Speaker 12: And look, I mean the fact that they are planning

474
00:24:23,880 --> 00:24:27,639
Speaker 12: to use Nvidia's CPUs, to me, that was a sign that,

475
00:24:27,760 --> 00:24:30,200
Speaker 12: you know, if you had to get the GPU capacity

476
00:24:30,240 --> 00:24:33,199
Speaker 12: from Nvidia, you pretty much had to use everything they

477
00:24:33,240 --> 00:24:36,800
Speaker 12: had to offer. With this deal, I think AMD sort

478
00:24:36,800 --> 00:24:41,040
Speaker 12: of comes across as the more desperate partner in terms of,

479
00:24:41,119 --> 00:24:45,600
Speaker 12: you know, giving Meta their stock warrants and making sure

480
00:24:45,680 --> 00:24:48,159
Speaker 12: that you know, they are at Meta, which is going

481
00:24:48,240 --> 00:24:50,880
Speaker 12: to spend one hundred and thirty five billion dollars. Out

482
00:24:50,880 --> 00:24:53,880
Speaker 12: of that, half of that will be on chips approximately,

483
00:24:54,240 --> 00:24:56,760
Speaker 12: So think of it this way. They currently are about

484
00:24:56,760 --> 00:24:59,840
Speaker 12: a ten to twelve billion dollar rund rate. This could

485
00:25:00,040 --> 00:25:02,480
Speaker 12: potentially double, you know, every year in terms of just

486
00:25:02,880 --> 00:25:06,280
Speaker 12: by adding one customer at the scale. Right, you know

487
00:25:06,320 --> 00:25:08,439
Speaker 12: they plan to buy the chips from AMD.

488
00:25:09,359 --> 00:25:11,719
Speaker 2: You know the way that that AMD puts say right,

489
00:25:11,800 --> 00:25:15,240
Speaker 2: single double digit sorry, billions of dollars per gigawa, but

490
00:25:15,280 --> 00:25:19,400
Speaker 2: the gigawatts need to get built, right, There are operational milestones.

491
00:25:19,760 --> 00:25:22,000
Speaker 2: Every time I go and cee AMD or we speak

492
00:25:22,040 --> 00:25:25,280
Speaker 2: to Lisasus she puts a lot of emphasis on inference,

493
00:25:25,600 --> 00:25:27,680
Speaker 2: you know, m I for fifty in particular, and what

494
00:25:27,760 --> 00:25:31,199
Speaker 2: those systems can do in the inference phase. Do you

495
00:25:31,320 --> 00:25:34,199
Speaker 2: in this deal see a point of differentiation for AMD

496
00:25:34,640 --> 00:25:36,600
Speaker 2: if they do go after the inference piece.

497
00:25:37,440 --> 00:25:40,760
Speaker 12: I do, and simply because all these companies are trying

498
00:25:40,800 --> 00:25:43,600
Speaker 12: to emulate what Google has been doing with TPUs. And

499
00:25:43,640 --> 00:25:47,399
Speaker 12: remember TPUs are used for both training and inferencing, not

500
00:25:47,640 --> 00:25:51,200
Speaker 12: just for Gemini but also for Entropics. So the fact

501
00:25:51,200 --> 00:25:53,320
Speaker 12: that you know they have signed this deal and are

502
00:25:53,840 --> 00:25:56,720
Speaker 12: willing to, you know, go from being a merchant silicon

503
00:25:56,760 --> 00:26:01,280
Speaker 12: provider to a custom silicon provider for Meta, and Meta's

504
00:26:01,320 --> 00:26:04,840
Speaker 12: workloads are more skew towards you know video and you

505
00:26:04,880 --> 00:26:08,280
Speaker 12: know the family of apps they have similar to Alphabet.

506
00:26:08,320 --> 00:26:11,400
Speaker 12: To me, this is a sign that's what Meta fields

507
00:26:11,440 --> 00:26:14,280
Speaker 12: they can accomplish, you know with the AMD chips. Yes,

508
00:26:14,600 --> 00:26:18,000
Speaker 12: they will continue to use in video for training, but

509
00:26:18,160 --> 00:26:22,080
Speaker 12: for inferencing, Meta's workloads are different from you know, any

510
00:26:22,119 --> 00:26:24,840
Speaker 12: other provider when it comes to the chat boards or

511
00:26:24,880 --> 00:26:26,600
Speaker 12: anything else that they're doing.

512
00:26:27,359 --> 00:26:29,320
Speaker 2: You know, AMD was it pains to point out they'll

513
00:26:29,400 --> 00:26:32,720
Speaker 2: kind of co engineer the roadmap to future server designs

514
00:26:32,720 --> 00:26:34,600
Speaker 2: starting m I four fifteen and go beyond that man

515
00:26:34,640 --> 00:26:37,359
Speaker 2: deep seeing a Boomberg intelligence with the React Thank you

516
00:26:37,400 --> 00:26:40,000
Speaker 2: so much. Now coming up on the program, Rayna Pope

517
00:26:40,040 --> 00:26:43,400
Speaker 2: from matt X joins us to discuss his chip startup's

518
00:26:43,480 --> 00:26:47,440
Speaker 2: latest fundraise as it aims to take on in Nvidia

519
00:26:47,680 --> 00:26:48,600
Speaker 2: that's coming up next.

520
00:26:48,720 --> 00:27:06,200
Speaker 10: This is Bloomberg Tech. AI startup Matt Eggs.

521
00:27:05,920 --> 00:27:08,520
Speaker 2: Has raised more than five hundred million dollars to build

522
00:27:08,600 --> 00:27:11,879
Speaker 2: hardware to compete with Nvidia. The chip company was founded

523
00:27:11,880 --> 00:27:15,440
Speaker 2: by two alumni of Google semiconductive business and its aim

524
00:27:16,000 --> 00:27:18,560
Speaker 2: is to make a product specifically designed to run large

525
00:27:18,600 --> 00:27:22,080
Speaker 2: language models. Co founder and CEO Ryana Pope joins us, Now,

526
00:27:22,480 --> 00:27:24,680
Speaker 2: I think we should get into the specifics, but my goodness,

527
00:27:24,680 --> 00:27:27,199
Speaker 2: five hundred million dollars is quite a large series B.

528
00:27:28,200 --> 00:27:31,080
Speaker 2: What do you need that level of capital for? And

529
00:27:31,080 --> 00:27:31,960
Speaker 2: what does it reflect?

530
00:27:32,600 --> 00:27:34,879
Speaker 13: Yeah, so, I mean I would say I'm very happy

531
00:27:34,880 --> 00:27:38,440
Speaker 13: to be here and thank you one of the Really

532
00:27:38,440 --> 00:27:41,200
Speaker 13: what it reflects is, on the one hand, very strong

533
00:27:41,200 --> 00:27:44,040
Speaker 13: confidence from some of our lead investors in the product.

534
00:27:45,440 --> 00:27:49,359
Speaker 13: This is from Jane Streets and situational awareness have have

535
00:27:49,480 --> 00:27:53,920
Speaker 13: led our around very strong on Jane Street sides. They're

536
00:27:53,920 --> 00:27:56,440
Speaker 13: absolute technical experts. They understand the kind of product we're doing.

537
00:27:56,760 --> 00:28:00,400
Speaker 13: And then situational awareness. That's Leopold ashen Renner's fund. He

538
00:28:00,480 --> 00:28:05,440
Speaker 13: wrote the book on AGI brand. He really understands where

539
00:28:05,440 --> 00:28:08,760
Speaker 13: this old space is going, what we like, what we

540
00:28:08,800 --> 00:28:11,240
Speaker 13: are looking to do with with this product and with

541
00:28:11,280 --> 00:28:15,680
Speaker 13: this money. Firstly, I would say the demand for l

542
00:28:15,840 --> 00:28:18,400
Speaker 13: M computer is just insatiable. All of the frontier labs

543
00:28:18,400 --> 00:28:20,840
Speaker 13: are looking at where this space is going, and they're

544
00:28:20,880 --> 00:28:22,479
Speaker 13: all concerned I'm going to run out of silicon.

545
00:28:22,520 --> 00:28:24,520
Speaker 4: I won't be able to solve all the demand I've got.

546
00:28:24,760 --> 00:28:28,000
Speaker 13: So our goal overall has been to make the highest

547
00:28:28,040 --> 00:28:31,520
Speaker 13: throughput per per square milorde of silicon that any product.

548
00:28:31,600 --> 00:28:34,400
Speaker 2: Right, this is about computational density, that's right. That's right.

549
00:28:34,440 --> 00:28:37,320
Speaker 2: And so you guys are basically saying through in terms

550
00:28:37,320 --> 00:28:40,560
Speaker 2: of flops per minimeter squared, like, this is something you

551
00:28:40,560 --> 00:28:42,880
Speaker 2: can own. What was the breakthrough?

552
00:28:43,040 --> 00:28:43,240
Speaker 14: You know?

553
00:28:43,240 --> 00:28:45,720
Speaker 2: What is it that you're so good at to achieve this?

554
00:28:46,320 --> 00:28:48,640
Speaker 13: Yeah, So there's really a combination of two things. If

555
00:28:48,680 --> 00:28:51,640
Speaker 13: you look at the products in the market. Previously, there's

556
00:28:51,680 --> 00:28:55,640
Speaker 13: been the HBM based family, which is in video Google, Amazon,

557
00:28:55,840 --> 00:28:57,720
Speaker 13: and then there's been the s RAM based family and

558
00:28:57,800 --> 00:29:01,720
Speaker 13: you are we are both actually uniquely so it's kind

559
00:29:01,720 --> 00:29:03,760
Speaker 13: of taking two good ideas and putting them together. It

560
00:29:03,840 --> 00:29:07,280
Speaker 13: is possible to do both very high throughput as you

561
00:29:07,320 --> 00:29:09,800
Speaker 13: get from HBM, but also very low latency as you

562
00:29:09,800 --> 00:29:12,040
Speaker 13: get from ASTRAAM, and do that in the same product.

563
00:29:12,360 --> 00:29:14,400
Speaker 13: What it gives you is actually a product that is

564
00:29:14,400 --> 00:29:16,720
Speaker 13: better than any other product in the market at through put.

565
00:29:17,200 --> 00:29:19,760
Speaker 13: This is exactly the flop Spisco milimeter that you described,

566
00:29:20,520 --> 00:29:22,760
Speaker 13: while also matching some of the best like the Cerebras

567
00:29:22,760 --> 00:29:23,880
Speaker 13: and the groc at latency.

568
00:29:24,560 --> 00:29:27,840
Speaker 3: Let's talk about how quickly people can start deploying this miner,

569
00:29:28,000 --> 00:29:28,880
Speaker 3: because what your.

570
00:29:28,760 --> 00:29:30,840
Speaker 4: Aim is to complete the final.

571
00:29:30,560 --> 00:29:34,320
Speaker 3: Design this year, you hope to start manufacturing shipping even

572
00:29:34,320 --> 00:29:36,320
Speaker 3: in twenty twenty seven. Who do you need to partner

573
00:29:36,320 --> 00:29:38,440
Speaker 3: with on that? How do you expect to be manufacturing

574
00:29:38,480 --> 00:29:38,800
Speaker 3: him in the.

575
00:29:38,840 --> 00:29:39,640
Speaker 4: US or abroad?

576
00:29:41,120 --> 00:29:41,400
Speaker 2: Yeah?

577
00:29:41,440 --> 00:29:43,040
Speaker 13: So, I mean there's a few big parts of the

578
00:29:43,040 --> 00:29:45,200
Speaker 13: supply chain, and this is common for US as well

579
00:29:45,240 --> 00:29:49,680
Speaker 13: as many other semiconductor firms in this space. Really, you

580
00:29:49,720 --> 00:29:53,520
Speaker 13: need logic wafers, memory wafers, which is HBM, and then

581
00:29:53,560 --> 00:29:55,640
Speaker 13: you need rack buildouts, and so those are the big

582
00:29:55,680 --> 00:29:56,080
Speaker 13: parts of.

583
00:29:56,040 --> 00:29:56,760
Speaker 4: Our supply chain.

584
00:29:57,800 --> 00:30:00,680
Speaker 13: TSMC is well recognized as a the best provider of

585
00:30:00,680 --> 00:30:04,880
Speaker 13: logic wafers, and then the memory wafers. There's the big

586
00:30:04,920 --> 00:30:08,280
Speaker 13: three which are Skhiinix, Samsung, and Micron, and then there's

587
00:30:08,280 --> 00:30:13,720
Speaker 13: a whole range of providers across the rack and manufacturing side.

588
00:30:14,120 --> 00:30:16,320
Speaker 13: One of the big things that if you want to

589
00:30:16,400 --> 00:30:19,160
Speaker 13: manufacturing very large volumes we hear about these you know,

590
00:30:19,400 --> 00:30:23,480
Speaker 13: multi gig or what deals that are coming out. These

591
00:30:23,560 --> 00:30:27,600
Speaker 13: require you know, billions of dollars of manufacturing and then

592
00:30:27,600 --> 00:30:30,160
Speaker 13: actually setting up like hundreds of millions of dollars put

593
00:30:30,200 --> 00:30:32,720
Speaker 13: into setting up supply chains in advance of delivering that.

594
00:30:32,760 --> 00:30:35,320
Speaker 13: And so that's a big part of what we're excited.

595
00:30:35,000 --> 00:30:35,640
Speaker 4: To be able to do.

596
00:30:36,080 --> 00:30:36,320
Speaker 11: Now.

597
00:30:36,920 --> 00:30:39,760
Speaker 3: You left Google in twenty twenty two and the goal

598
00:30:39,840 --> 00:30:42,520
Speaker 3: was creating a better chit from scratch, Ryanan, But have

599
00:30:42,640 --> 00:30:43,800
Speaker 3: you been impressed by.

600
00:30:44,120 --> 00:30:46,320
Speaker 4: The leaps at TPU has taken. It seems to have

601
00:30:46,480 --> 00:30:47,360
Speaker 4: impressed the market.

602
00:30:47,440 --> 00:30:50,040
Speaker 3: What is it that you felt wasn't at Google for

603
00:30:50,120 --> 00:30:51,840
Speaker 3: you that you now can bild better?

604
00:30:53,400 --> 00:30:56,040
Speaker 13: Yeah, So I think what is really required is if

605
00:30:56,040 --> 00:30:58,240
Speaker 13: you want to absolutely nail the LLM workload, you have

606
00:30:58,280 --> 00:31:01,440
Speaker 13: to be willing to break compatibility with previous chips. And

607
00:31:01,480 --> 00:31:04,120
Speaker 13: so one of the strong guarantees you see all of

608
00:31:04,160 --> 00:31:06,760
Speaker 13: the existing players providing is you can take a program

609
00:31:06,760 --> 00:31:08,640
Speaker 13: that was written on my previous generation ship or my

610
00:31:08,880 --> 00:31:11,440
Speaker 13: generation of chips five years ago, and it will run

611
00:31:11,480 --> 00:31:14,360
Speaker 13: on my next generation chip. And so a lot of

612
00:31:14,360 --> 00:31:16,800
Speaker 13: what that means is there are constraints on my chip

613
00:31:16,840 --> 00:31:19,440
Speaker 13: has to support all of the previous number formats I supported.

614
00:31:19,520 --> 00:31:22,280
Speaker 13: It has to support all of the different programming model

615
00:31:22,280 --> 00:31:24,440
Speaker 13: the way I communicate between cores on the chip, but

616
00:31:24,560 --> 00:31:26,160
Speaker 13: all of those have to be the same as each other.

617
00:31:26,920 --> 00:31:28,680
Speaker 13: We felt that it would be necessary, like if you

618
00:31:28,760 --> 00:31:31,640
Speaker 13: really want to just absolutely nail this workload without regards

619
00:31:31,640 --> 00:31:34,680
Speaker 13: for backwards compatibility or other workloads or anything like that,

620
00:31:34,880 --> 00:31:38,920
Speaker 13: you need to something of a blank slate design is required.

621
00:31:39,160 --> 00:31:43,040
Speaker 13: For us, this means very large matrices, very low precision support,

622
00:31:43,560 --> 00:31:46,000
Speaker 13: and then in fact an ability to split your very

623
00:31:46,040 --> 00:31:46,960
Speaker 13: large cystolic.

624
00:31:46,720 --> 00:31:49,880
Speaker 2: Ray into small pieces. You name checked Grok, I think

625
00:31:49,920 --> 00:31:52,680
Speaker 2: with some admiration a Sesson Ago. I mean like when

626
00:31:53,200 --> 00:31:56,719
Speaker 2: in video acquired Grock Jensen. Wong's view is that they

627
00:31:56,720 --> 00:31:58,880
Speaker 2: were struggling to find their place in the world. In

628
00:31:59,080 --> 00:32:02,440
Speaker 2: markere forswere Sarah Brasse you named it as well, filed

629
00:32:02,440 --> 00:32:07,320
Speaker 2: confidentially for IPO yesterday. Why might you succeed where CROC

630
00:32:08,120 --> 00:32:10,800
Speaker 2: had to go to Nvidia And I guess they're working

631
00:32:10,800 --> 00:32:14,560
Speaker 2: on something you know, and the public markets you know

632
00:32:15,160 --> 00:32:16,720
Speaker 2: are needed for capital going forward.

633
00:32:17,160 --> 00:32:20,880
Speaker 13: Yeah, so I would say this is like historically the

634
00:32:20,920 --> 00:32:23,760
Speaker 13: market has been won by the HBM place based players,

635
00:32:23,800 --> 00:32:26,640
Speaker 13: that's the Google Amazon in VideA, and not by Grock

636
00:32:26,720 --> 00:32:31,960
Speaker 13: and SERIBRUSSTRAM only chips are very good for latency, but

637
00:32:32,840 --> 00:32:34,600
Speaker 13: when you want to run very long context models you

638
00:32:34,680 --> 00:32:35,640
Speaker 13: run out of memory capacity.

639
00:32:35,720 --> 00:32:36,560
Speaker 4: SRAM is too small.

640
00:32:36,600 --> 00:32:40,200
Speaker 13: It's fast, but too small. Really, the hybrid of doing

641
00:32:40,880 --> 00:32:43,640
Speaker 13: weights INSTRAM so you get the low latency as well

642
00:32:43,640 --> 00:32:46,240
Speaker 13: as having the HBM for for very long context support

643
00:32:46,360 --> 00:32:49,480
Speaker 13: is we believe that's what enables the low latency without

644
00:32:49,480 --> 00:32:51,640
Speaker 13: all of the compromises that you would get otherwise.

645
00:32:51,680 --> 00:32:54,160
Speaker 3: Minor Pope, thank you so much for joining us today,

646
00:32:54,280 --> 00:32:57,200
Speaker 3: Matt x CEO, Look we gotta stit with funding. News

647
00:32:57,480 --> 00:33:00,800
Speaker 3: Basis startup making AI tools for account unting, has raised

648
00:33:00,800 --> 00:33:03,320
Speaker 3: one hundred million dollars and an over one million dollar valuation.

649
00:33:03,520 --> 00:33:05,680
Speaker 4: This, of course is investors we're putting.

650
00:33:05,360 --> 00:33:08,360
Speaker 3: Wary of disruption from agentic AI here with us is

651
00:33:08,440 --> 00:33:12,040
Speaker 3: Matt Hart based as CEO. I go to the point,

652
00:33:12,080 --> 00:33:15,800
Speaker 3: how is BASIS different from anthropics cowork plug in for

653
00:33:15,840 --> 00:33:16,800
Speaker 3: an accounting business.

654
00:33:17,320 --> 00:33:19,200
Speaker 15: Yeah, well, first of all, great to be here, thanks

655
00:33:19,240 --> 00:33:22,960
Speaker 15: for having me, Caroline. Look, we have an excellent relationship

656
00:33:23,000 --> 00:33:26,400
Speaker 15: with Open Eye and with Anthropic, and I think Cowork

657
00:33:26,560 --> 00:33:28,640
Speaker 15: cha attribut the whole range of tools that they have

658
00:33:28,960 --> 00:33:32,200
Speaker 15: are absolutely fantastic. I think though you can only sort

659
00:33:32,200 --> 00:33:34,440
Speaker 15: of serve so many masters, and when it comes to

660
00:33:34,520 --> 00:33:38,320
Speaker 15: domain specific work specifically in our situation accounting, there's a

661
00:33:38,400 --> 00:33:41,000
Speaker 15: need to build with that domain specificity in mind. I

662
00:33:41,000 --> 00:33:42,840
Speaker 15: think comes down to a couple of things. First of all,

663
00:33:42,840 --> 00:33:45,520
Speaker 15: you want domain specific capabilities. So, for instance, we announced

664
00:33:45,520 --> 00:33:48,560
Speaker 15: today the first example of a long running agent completing

665
00:33:48,560 --> 00:33:52,040
Speaker 15: an entire business tax return workbook. That's something that you

666
00:33:52,080 --> 00:33:54,240
Speaker 15: can't really do in any other AI tool that exists

667
00:33:54,280 --> 00:33:57,120
Speaker 15: out there. You need domain specific accuracy. So we are

668
00:33:57,120 --> 00:33:59,400
Speaker 15: able to guarantee to the firms that we work with

669
00:33:59,600 --> 00:34:02,200
Speaker 15: that the I will meet the requisite level of accuracy

670
00:34:02,240 --> 00:34:04,440
Speaker 15: for it to be used in a real manner and

671
00:34:04,480 --> 00:34:06,920
Speaker 15: we can sort of guarantee that performance. You need demain

672
00:34:06,960 --> 00:34:10,480
Speaker 15: specific user experience so that it's fluid and it works

673
00:34:10,520 --> 00:34:13,080
Speaker 15: well for people who are experts in their field as

674
00:34:13,080 --> 00:34:15,480
Speaker 15: they go about doing their work. You can't just have

675
00:34:15,520 --> 00:34:17,439
Speaker 15: a chat pot or you know, the variety of other

676
00:34:17,480 --> 00:34:20,239
Speaker 15: sort of basic user experiences that you need need to

677
00:34:20,239 --> 00:34:25,080
Speaker 15: build enterprise sort of grade features. You know, collaboration, audit trails,

678
00:34:25,360 --> 00:34:26,239
Speaker 15: you know, audibility etc.

679
00:34:26,520 --> 00:34:27,320
Speaker 16: Things of this nature.

680
00:34:27,400 --> 00:34:31,040
Speaker 15: And finally you need demain specific deployment approaches. It's going

681
00:34:31,120 --> 00:34:33,200
Speaker 15: to you know, the capacity of a lot of these

682
00:34:33,560 --> 00:34:36,520
Speaker 15: AI tools to be extremely useful in workflows probably far

683
00:34:36,600 --> 00:34:38,960
Speaker 15: exceeds the adoption today and there's going to be a

684
00:34:39,000 --> 00:34:41,200
Speaker 15: great challenge over the next decade of figuring out how

685
00:34:41,200 --> 00:34:42,920
Speaker 15: to get AI into all the places it needs to

686
00:34:42,920 --> 00:34:44,799
Speaker 15: get into. And so we think for all those reasons,

687
00:34:44,840 --> 00:34:46,960
Speaker 15: being demained specific is very important. I think that's pretty

688
00:34:46,960 --> 00:34:47,440
Speaker 15: evident to this.

689
00:34:47,320 --> 00:34:50,280
Speaker 3: Stat And so is that one hundred million to ensure

690
00:34:50,400 --> 00:34:52,799
Speaker 3: the adoption curve is where you need it to be.

691
00:34:52,840 --> 00:34:54,480
Speaker 3: Where does that money get deployed first and.

692
00:34:54,400 --> 00:34:56,840
Speaker 15: Foremost, Yeah, it's a combination of things. I think for

693
00:34:56,960 --> 00:34:59,600
Speaker 15: us first and foremers who are always focused on building

694
00:34:59,600 --> 00:35:02,440
Speaker 15: the most capable and the most accurate AI for accounting,

695
00:35:02,480 --> 00:35:05,920
Speaker 15: So we are growing our engineering and mL teams dramatically.

696
00:35:06,239 --> 00:35:08,480
Speaker 15: We're a fully New York based company, and we essentially

697
00:35:08,480 --> 00:35:11,080
Speaker 15: want to be the home for applied AI work in

698
00:35:11,440 --> 00:35:13,640
Speaker 15: New York City, so that's always number one priority. But

699
00:35:13,719 --> 00:35:15,359
Speaker 15: the other aspect of it is we are now starting

700
00:35:15,360 --> 00:35:17,760
Speaker 15: to serve accounting firms across all of their practices CAS

701
00:35:18,000 --> 00:35:20,200
Speaker 15: and core accounting being the initial segment, but now also

702
00:35:20,600 --> 00:35:24,040
Speaker 15: tax audit, broader advisory, and over time. Figuring out how

703
00:35:24,040 --> 00:35:26,360
Speaker 15: we can get AI into all these various places is

704
00:35:26,400 --> 00:35:29,080
Speaker 15: extremely important. Demand has exceeded our capacity to serve it,

705
00:35:29,120 --> 00:35:30,800
Speaker 15: and so we felt it was important to bring additional

706
00:35:30,840 --> 00:35:32,120
Speaker 15: capital to make that puzsible.

707
00:35:32,920 --> 00:35:35,879
Speaker 2: So now I'm going through my taxes right now, and

708
00:35:35,920 --> 00:35:38,080
Speaker 2: through the poor too of my accountant, who I won't name.

709
00:35:38,200 --> 00:35:39,040
Speaker 4: They do a good job.

710
00:35:39,120 --> 00:35:42,439
Speaker 2: I upload all of the documents W two, ten ninety nine,

711
00:35:42,480 --> 00:35:45,080
Speaker 2: ten ninety eight whatever, but there is an element of

712
00:35:45,160 --> 00:35:48,120
Speaker 2: observability because I still have to go in and double

713
00:35:48,239 --> 00:35:50,239
Speaker 2: check all of the automated part of it. I know

714
00:35:50,280 --> 00:35:52,399
Speaker 2: that's not the same, but the case study you gave

715
00:35:53,040 --> 00:35:56,640
Speaker 2: of a business tax filing start to finish. The observability

716
00:35:56,680 --> 00:36:01,640
Speaker 2: piece must be critical if you're handling that for businesses

717
00:36:01,640 --> 00:36:02,200
Speaker 2: of scale.

718
00:36:03,840 --> 00:36:07,279
Speaker 15: Yeah, absolutely, good question. You know, we think that the

719
00:36:07,880 --> 00:36:10,920
Speaker 15: ability for an accountant to understand what's happening at a

720
00:36:10,920 --> 00:36:13,879
Speaker 15: granular level as the AI is helping them get work

721
00:36:13,920 --> 00:36:15,960
Speaker 15: done is extremely important. That goes back to a little

722
00:36:16,000 --> 00:36:18,160
Speaker 15: bit about what Caroline was referring to in terms of

723
00:36:18,200 --> 00:36:21,200
Speaker 15: the you know, what makes us different than generic tools.

724
00:36:21,440 --> 00:36:26,400
Speaker 15: Having demain specific user experiences and demain specific auditibility functionality

725
00:36:26,440 --> 00:36:29,160
Speaker 15: is extremely important. And then we also think that accounting

726
00:36:29,160 --> 00:36:30,920
Speaker 15: firms have a very important role to play at the

727
00:36:31,000 --> 00:36:32,920
Speaker 15: end of the day. Where people like yourself and what

728
00:36:32,960 --> 00:36:35,719
Speaker 15: businesses around America probably want is a human that they

729
00:36:35,760 --> 00:36:37,960
Speaker 15: have a relationship with, that they trust to do this

730
00:36:38,000 --> 00:36:40,360
Speaker 15: in chreatly important work. And so having the AI collaborate

731
00:36:40,360 --> 00:36:42,600
Speaker 15: with humans to make that possible in an even better

732
00:36:42,600 --> 00:36:45,000
Speaker 15: fashion maybe will improve the experience that you have on

733
00:36:45,040 --> 00:36:47,440
Speaker 15: your taxes in future years as firms figure out how

734
00:36:47,440 --> 00:36:49,680
Speaker 15: to bring these two things together is extremely important.

735
00:36:49,960 --> 00:36:52,760
Speaker 2: So here in the central question is is it an

736
00:36:52,880 --> 00:36:55,799
Speaker 2: aid to an existing job or does it displace an

737
00:36:55,840 --> 00:37:00,360
Speaker 2: existing role in accounting going forward, we just have thirty seconds' not.

738
00:37:00,360 --> 00:37:02,080
Speaker 15: A thirty second question, but I'll give you my best

739
00:37:02,080 --> 00:37:04,160
Speaker 15: answer on it. So I think, you know, if you

740
00:37:04,160 --> 00:37:06,160
Speaker 15: look at what's happened with software engineering over the course

741
00:37:06,160 --> 00:37:07,279
Speaker 15: of the last year, and you can look at this

742
00:37:07,360 --> 00:37:10,440
Speaker 15: internal into basis, pretty much today, no engineer at the

743
00:37:10,440 --> 00:37:13,360
Speaker 15: company should be writing any meaningful amount of code, and

744
00:37:13,440 --> 00:37:15,839
Speaker 15: yet our engineering team, our machine learning team is as

745
00:37:15,880 --> 00:37:18,200
Speaker 15: busy as its ever been. We're hiring as aggressively as

746
00:37:18,200 --> 00:37:20,960
Speaker 15: we possibly can be in those areas, and that's because

747
00:37:20,960 --> 00:37:22,920
Speaker 15: there's so much additional engineering work that we want to

748
00:37:22,960 --> 00:37:24,600
Speaker 15: be able to do. That when you free up time

749
00:37:24,640 --> 00:37:26,799
Speaker 15: to focus on more things, you can do more ambitious things.

750
00:37:26,840 --> 00:37:28,600
Speaker 15: You can build things you never thought you'd want, you

751
00:37:28,719 --> 00:37:30,319
Speaker 15: never thought you'd be able to do. And the same

752
00:37:30,320 --> 00:37:32,160
Speaker 15: thing is true for accounting. There's tons of accounting work

753
00:37:32,200 --> 00:37:33,920
Speaker 15: in the world that doesn't get done today. The Pentagon

754
00:37:34,120 --> 00:37:37,000
Speaker 15: just failed its eighth consecutive audit. You know, companies are

755
00:37:37,040 --> 00:37:40,520
Speaker 15: misstating financials because they don't have enough accounting resources. Beyond that,

756
00:37:40,600 --> 00:37:42,400
Speaker 15: you know the accounting. You know, if you go to

757
00:37:42,440 --> 00:37:44,640
Speaker 15: any hospital system in America, they can't tell you how

758
00:37:44,680 --> 00:37:47,080
Speaker 15: much it costs them to provide a simple procedure like

759
00:37:47,120 --> 00:37:49,640
Speaker 15: a knee surgery. These are all things that accounting can

760
00:37:49,640 --> 00:37:51,520
Speaker 15: make possible. It is sort of the fundamental way that

761
00:37:51,560 --> 00:37:54,479
Speaker 15: we understand economic activity that goes on in and around

762
00:37:54,520 --> 00:37:57,719
Speaker 15: our organizations, and so we think there's a huge opportunity

763
00:37:57,760 --> 00:38:00,719
Speaker 15: to do more accounting work. Most accounting firms the opportunity

764
00:38:00,760 --> 00:38:02,400
Speaker 15: as well, and need the capacity to be able to

765
00:38:02,400 --> 00:38:04,319
Speaker 15: do that. And so from our perspective, this is going

766
00:38:04,400 --> 00:38:06,440
Speaker 15: to allow firms and the accountants at those firms to

767
00:38:06,480 --> 00:38:08,640
Speaker 15: take on even more work and get things more done

768
00:38:08,680 --> 00:38:10,600
Speaker 15: and get more things done in the same way that

769
00:38:11,719 --> 00:38:13,560
Speaker 15: sort of software engineering has been able to take off

770
00:38:13,600 --> 00:38:14,880
Speaker 15: over the course of the last twelve months.

771
00:38:15,600 --> 00:38:17,880
Speaker 2: Matthew Hart, CEO and co founder of Basis. That was

772
00:38:17,880 --> 00:38:21,480
Speaker 2: a pretty good, pretty good summary. Actually a little more

773
00:38:21,480 --> 00:38:22,720
Speaker 2: than a pay seconds, so we'll.

774
00:38:22,560 --> 00:38:23,680
Speaker 4: Give you it. Thank you very much.

775
00:38:23,840 --> 00:38:25,920
Speaker 2: Now, coming up, we've got to look forward to what

776
00:38:26,040 --> 00:38:29,760
Speaker 2: to expect from President Trump's State of the Union address,

777
00:38:29,800 --> 00:38:30,640
Speaker 2: which is later tonight.

778
00:38:30,680 --> 00:38:31,480
Speaker 4: We have the preview.

779
00:38:31,760 --> 00:38:46,960
Speaker 2: This is Bloomberg Tech. All eyes are on President Trump's

780
00:38:47,040 --> 00:38:50,879
Speaker 2: State of the Union speech tonight. This only days after

781
00:38:50,920 --> 00:38:53,920
Speaker 2: the Supreme Court decided to strike down his tariff policies.

782
00:38:53,960 --> 00:38:57,840
Speaker 2: Bloomberg TV's Washington correspondent Tyler Kendall joins us, what do

783
00:38:57,840 --> 00:38:59,360
Speaker 2: we need to expect?

784
00:39:00,160 --> 00:39:00,920
Speaker 4: Via Hey Edwar.

785
00:39:01,000 --> 00:39:03,960
Speaker 14: President Trump is expected to tout his economic policies that

786
00:39:04,040 --> 00:39:08,040
Speaker 14: are already enacted, but also push ahead some policy proposals

787
00:39:08,040 --> 00:39:10,719
Speaker 14: that he would like to see enacted related to affordability,

788
00:39:10,760 --> 00:39:13,360
Speaker 14: as the White House really plays defense on the issue

789
00:39:13,400 --> 00:39:15,920
Speaker 14: according to recent polling. In fact, the Wall Street Journal

790
00:39:15,960 --> 00:39:18,720
Speaker 14: is now reporting that President Trump is set to announce

791
00:39:18,760 --> 00:39:21,920
Speaker 14: a negotiated commitment from big tech companies to pay more

792
00:39:21,960 --> 00:39:25,800
Speaker 14: when it comes to electricity costs related to AI data centers,

793
00:39:25,840 --> 00:39:27,919
Speaker 14: in a bid to remove some of that burden from

794
00:39:28,000 --> 00:39:28,759
Speaker 14: US consumers.

795
00:39:28,800 --> 00:39:29,720
Speaker 4: I was here at the White.

796
00:39:29,560 --> 00:39:32,440
Speaker 14: House just last month speaking to the US Energy Secretary,

797
00:39:32,520 --> 00:39:36,360
Speaker 14: Christopher Wright, as the administration urged the nation's largest power

798
00:39:36,440 --> 00:39:40,239
Speaker 14: grid to hold an emergency power auction specifically for these

799
00:39:40,280 --> 00:39:43,000
Speaker 14: big tech companies. So perhaps we'll get some firmer details

800
00:39:43,280 --> 00:39:45,359
Speaker 14: on that tonight, because, as you well know, this has

801
00:39:45,400 --> 00:39:47,560
Speaker 14: been a White House trying to thread the needle between

802
00:39:47,600 --> 00:39:51,239
Speaker 14: bolstering artificial intelligence but also trying to quell some of

803
00:39:51,239 --> 00:39:54,000
Speaker 14: those concerns from Americans. When it comes to jobs, and

804
00:39:54,040 --> 00:39:57,360
Speaker 14: of course the cost of living related to electricity, because

805
00:39:57,360 --> 00:39:59,360
Speaker 14: that is going to be the number one issue as

806
00:39:59,360 --> 00:40:00,680
Speaker 14: we head into the MA in terms of a new

807
00:40:00,719 --> 00:40:03,480
Speaker 14: poll out this week from Ipsos finds that fifty seven

808
00:40:03,560 --> 00:40:07,359
Speaker 14: percent of US adults disapprove of the President's handling of

809
00:40:07,360 --> 00:40:10,120
Speaker 14: the economy. Front and centered, as you mentioned, is going

810
00:40:10,160 --> 00:40:12,359
Speaker 14: to be that tariff policy, and sitting there in front

811
00:40:12,360 --> 00:40:14,920
Speaker 14: of the presidents tonight is going to be those Supreme

812
00:40:15,000 --> 00:40:17,719
Speaker 14: Court justices that struck down those IPA teriffs.

813
00:40:17,440 --> 00:40:20,520
Speaker 3: Last week, Mex Tyler Kendall with a rundown, thank you

814
00:40:20,640 --> 00:40:21,120
Speaker 3: very much.

815
00:40:21,239 --> 00:40:22,879
Speaker 4: And look, the state of the Union.

816
00:40:22,719 --> 00:40:26,000
Speaker 3: Is also the target for bets on prediction markets like Cawshi,

817
00:40:26,080 --> 00:40:28,480
Speaker 3: like poly markets. Look a market that was once a

818
00:40:28,480 --> 00:40:30,280
Speaker 3: fringe obsession of economists.

819
00:40:29,840 --> 00:40:30,680
Speaker 4: And election longks.

820
00:40:30,760 --> 00:40:34,040
Speaker 3: Now traders are wagering on just about everything. Critics call

821
00:40:34,040 --> 00:40:37,359
Speaker 3: it unregulated gambling, and its today's big take deep dive

822
00:40:37,360 --> 00:40:39,600
Speaker 3: and please Save Blumberg contributor to Chris Beams here to

823
00:40:39,600 --> 00:40:40,520
Speaker 3: talk us through it.

824
00:40:40,520 --> 00:40:41,640
Speaker 4: It is a wonderful deep dive.

825
00:40:41,680 --> 00:40:43,440
Speaker 3: And just as we think about the state of the Union,

826
00:40:43,800 --> 00:40:47,400
Speaker 3: how has the idea of regulating this changed in the years.

827
00:40:48,760 --> 00:40:53,000
Speaker 16: So currently the status quo is that prediction markets are

828
00:40:53,040 --> 00:40:57,600
Speaker 16: regulated by the CFTC, and the reason is that the

829
00:40:58,880 --> 00:41:03,640
Speaker 16: technically offer what are called event contracts, and event contracts

830
00:41:03,680 --> 00:41:07,040
Speaker 16: are derivatives, and so the CFTC argues that falls under

831
00:41:07,080 --> 00:41:09,720
Speaker 16: their purview. Now, there's a lot of people who disagree

832
00:41:09,719 --> 00:41:14,960
Speaker 16: with the CFTC state gambling regulators. A lot of casinos

833
00:41:15,200 --> 00:41:19,800
Speaker 16: and other members of the gambling industry argue, yes, that

834
00:41:21,520 --> 00:41:24,600
Speaker 16: prediction markets should actually count as gambling and therefore be

835
00:41:24,640 --> 00:41:25,720
Speaker 16: regulated by the states.

836
00:41:27,000 --> 00:41:31,960
Speaker 2: How good are prediction markets at predicting the future.

837
00:41:32,880 --> 00:41:35,720
Speaker 16: So it depends what kind of market you're talking about.

838
00:41:36,600 --> 00:41:40,200
Speaker 16: A lot of scholars have looked at political markets and

839
00:41:40,280 --> 00:41:44,239
Speaker 16: prediction markets to predict elections and have overall found that

840
00:41:44,280 --> 00:41:48,080
Speaker 16: prediction markets are more accurate than polling, and certainly more

841
00:41:48,080 --> 00:41:51,680
Speaker 16: accurate than you or me or any individual trying to

842
00:41:51,840 --> 00:41:56,640
Speaker 16: prognosticate about elections. What's interesting, though, is that those markets

843
00:41:56,680 --> 00:42:00,880
Speaker 16: become less accurate the smaller they are. So, you know,

844
00:42:00,920 --> 00:42:04,000
Speaker 16: a presidential election market could be quite accurate, but then

845
00:42:04,040 --> 00:42:06,319
Speaker 16: once you get down to more you know, state and

846
00:42:06,400 --> 00:42:10,440
Speaker 16: local races, when there's less liquidity, that's going to be

847
00:42:10,520 --> 00:42:11,160
Speaker 16: less accurate.

848
00:42:11,600 --> 00:42:14,520
Speaker 4: I mean some of the bets that we see your extraordinary.

849
00:42:14,560 --> 00:42:16,360
Speaker 3: I mean, I don't know how much liquidity there is

850
00:42:16,400 --> 00:42:18,040
Speaker 3: on whether Jesus Christ is going to return in the

851
00:42:18,080 --> 00:42:20,160
Speaker 3: next couple of years, but that's literally something you can

852
00:42:20,200 --> 00:42:22,640
Speaker 3: go and place a wager or at least a prediction

853
00:42:22,760 --> 00:42:24,280
Speaker 3: on using these markets.

854
00:42:25,280 --> 00:42:27,839
Speaker 4: How have the companies themselves, CALCI.

855
00:42:27,560 --> 00:42:32,000
Speaker 3: And Polymarket navigated what has been them thrust into really

856
00:42:32,040 --> 00:42:34,840
Speaker 3: now success but also backlash at the same time and

857
00:42:34,920 --> 00:42:36,239
Speaker 3: worries about insider trading in.

858
00:42:36,200 --> 00:42:42,399
Speaker 16: Thus, Yeah, the companies have taken different approaches. CALSHI has

859
00:42:42,440 --> 00:42:45,840
Speaker 16: really tried to position itself as the adult in the room.

860
00:42:46,440 --> 00:42:52,000
Speaker 16: They emphasize the internal rules that they have around insider trading,

861
00:42:52,080 --> 00:42:57,239
Speaker 16: around market manipulation. They have whole teams and software algorithms

862
00:42:57,239 --> 00:43:01,400
Speaker 16: designed to detect trading patterns that might set off red flags,

863
00:43:01,920 --> 00:43:05,920
Speaker 16: which they then report to regulators. Polymarket is in a

864
00:43:05,960 --> 00:43:09,000
Speaker 16: slightly different position because so much of their trading happens

865
00:43:09,040 --> 00:43:12,520
Speaker 16: overseas and is not regulated by the CFTC, and they

866
00:43:12,520 --> 00:43:16,279
Speaker 16: also just haven't talked about it as much. So I'd

867
00:43:16,320 --> 00:43:21,040
Speaker 16: say CALSHI, particularly because it was the earlier entrant in

868
00:43:21,080 --> 00:43:26,680
Speaker 16: the US, has emphasized cooperating with regulators, trying to be

869
00:43:26,840 --> 00:43:29,920
Speaker 16: above board about the markets that they provide.

870
00:43:31,080 --> 00:43:32,840
Speaker 2: It's the latest Bloomberg, a big take, and it's a

871
00:43:32,920 --> 00:43:35,799
Speaker 2: must read. Bloomberg contributes its crispy thank you very much.

872
00:43:36,280 --> 00:43:38,600
Speaker 2: That does it for this edition of Bloomberg Tech. It's

873
00:43:38,600 --> 00:43:40,200
Speaker 2: not as if there's a shortage of things to come.

874
00:43:40,200 --> 00:43:42,120
Speaker 2: This week State of the Union and then in video

875
00:43:42,200 --> 00:43:43,280
Speaker 2: on Wednesday, Oh.

876
00:43:43,200 --> 00:43:45,160
Speaker 3: And earnings after the Bell today as well. Look, just

877
00:43:45,200 --> 00:43:47,160
Speaker 3: don't forget to check out our podcast. You can find

878
00:43:47,200 --> 00:43:49,280
Speaker 3: it on the terminal as well as online on Apple

879
00:43:49,400 --> 00:43:50,360
Speaker 3: or Spotify.

880
00:43:49,960 --> 00:43:53,160
Speaker 4: On iHeart. The tech news keeps coming as as a disruption.

881
00:43:53,440 --> 00:43:54,600
Speaker 4: This is Bloomberg Tech