1
00:00:02,720 --> 00:00:15,840
Speaker 1: Bloomberg Audio Studios, podcasts, radio news.

2
00:00:18,800 --> 00:00:21,280
Speaker 2: Hey, there are add lots listeners. You are about to

3
00:00:21,320 --> 00:00:25,600
Speaker 2: get a conversation with Don Wilson, founder and CEO of DRW,

4
00:00:25,840 --> 00:00:29,040
Speaker 2: sometimes called the smartest man in trading. This was recorded

5
00:00:29,080 --> 00:00:33,000
Speaker 2: live on stage at Chicago's Untitled Supper Club. We had

6
00:00:33,040 --> 00:00:37,760
Speaker 2: a blast and we hope you'll enjoy the show. All right, Don, Well,

7
00:00:38,080 --> 00:00:40,239
Speaker 2: thank you for being here. Really appreciate it.

8
00:00:40,360 --> 00:00:41,120
Speaker 3: Great to be here.

9
00:00:41,360 --> 00:00:45,320
Speaker 2: Truly the perfect guest to talk about what's next in trading.

10
00:00:45,760 --> 00:00:49,320
Speaker 2: But just to begin with why GPUs.

11
00:00:48,840 --> 00:00:53,800
Speaker 4: Well, obviously AI is becoming more and more useful, and

12
00:00:54,280 --> 00:00:57,360
Speaker 4: as it becomes more useful, people use more of it,

13
00:00:57,440 --> 00:01:00,000
Speaker 4: which means they need to use more GPUs to write

14
00:01:00,160 --> 00:01:03,720
Speaker 4: inference or trade new models. And I actually have this

15
00:01:03,880 --> 00:01:08,480
Speaker 4: theory that within the next ten years, the world will

16
00:01:08,480 --> 00:01:12,600
Speaker 4: spend more per year on GPUs than it does.

17
00:01:12,440 --> 00:01:17,560
Speaker 3: On crude oil, and that would of course make GPUs compute.

18
00:01:17,280 --> 00:01:21,120
Speaker 4: The largest commodity in the world. So it seems like

19
00:01:21,240 --> 00:01:22,960
Speaker 4: you would kind of need a market for that.

20
00:01:23,040 --> 00:01:26,160
Speaker 2: The very modest call just the largest market in the world.

21
00:01:27,000 --> 00:01:30,080
Speaker 5: Yeah, it's funny because you know, I associately oil often

22
00:01:30,160 --> 00:01:33,520
Speaker 5: coming out of you know, sandy deserts, but now they're

23
00:01:33,600 --> 00:01:38,840
Speaker 5: literally turning the sand via chips into the commodity itself,

24
00:01:38,959 --> 00:01:41,080
Speaker 5: or like breathing life into the sand. Just to back up,

25
00:01:41,120 --> 00:01:43,520
Speaker 5: I have a million questions about this for those who

26
00:01:43,560 --> 00:01:45,240
Speaker 5: don't know, why don't you give us the sort of

27
00:01:45,640 --> 00:01:48,520
Speaker 5: you know, the thirty second or the forty five second

28
00:01:49,000 --> 00:01:52,000
Speaker 5: description of what you do or what DRW is.

29
00:01:52,720 --> 00:01:53,040
Speaker 3: Yeah.

30
00:01:53,080 --> 00:01:56,320
Speaker 4: So I started off standing in the trading pit in

31
00:01:56,400 --> 00:01:59,200
Speaker 4: Chicago and the ear dollar option pit, yelling and screaming.

32
00:01:59,240 --> 00:02:01,920
Speaker 4: And then I go home and write code on my

33
00:02:02,000 --> 00:02:07,080
Speaker 4: Macintosh computer and build models, and essentially, you know, I

34
00:02:07,080 --> 00:02:08,880
Speaker 4: don't stand in the pit and yell and scream anymore.

35
00:02:08,919 --> 00:02:11,200
Speaker 4: Most of the pits are gone. But we kind of

36
00:02:11,240 --> 00:02:13,240
Speaker 4: do the same thing now with computers.

37
00:02:13,520 --> 00:02:15,520
Speaker 2: I heard a story that you were once on vacation

38
00:02:15,800 --> 00:02:18,799
Speaker 2: with your family and you were in Italy, I think

39
00:02:18,840 --> 00:02:21,720
Speaker 2: in Florence, and instead of I don't know, eating gelato

40
00:02:21,880 --> 00:02:25,480
Speaker 2: or something like that, you decided to invent a new

41
00:02:25,520 --> 00:02:27,400
Speaker 2: Greek letter for derivatives trading.

42
00:02:27,520 --> 00:02:29,000
Speaker 5: So this is cool.

43
00:02:29,200 --> 00:02:34,000
Speaker 4: Yeah, so here, I mean you're confusing two stories. So actually,

44
00:02:34,480 --> 00:02:38,480
Speaker 4: actually what happened was there was a new exchange that

45
00:02:38,560 --> 00:02:41,920
Speaker 4: had launched an interest rate swap futures contract. It was

46
00:02:41,960 --> 00:02:46,520
Speaker 4: called IDCG, and I looked at the contract and I

47
00:02:46,600 --> 00:02:50,280
Speaker 4: figured out that actually they had not designed the contracts properly,

48
00:02:50,880 --> 00:02:54,080
Speaker 4: and so although they were telling everybody that it was

49
00:02:54,160 --> 00:02:58,880
Speaker 4: economically equivalent to a regular interest rate swap, it wasn't

50
00:02:58,960 --> 00:03:03,600
Speaker 4: because it had the additional convexity bias in it, which

51
00:03:03,680 --> 00:03:07,919
Speaker 4: is we could talk about convexity bias. It goes even

52
00:03:07,919 --> 00:03:09,720
Speaker 4: more in the weeds than a lot of your podcasts

53
00:03:09,800 --> 00:03:12,800
Speaker 4: go into. But so when I was in Florence, I

54
00:03:12,840 --> 00:03:15,080
Speaker 4: had this idea of how you could create an interest

55
00:03:15,160 --> 00:03:19,200
Speaker 4: rate spap futures contract without this convexity bias problem, and

56
00:03:19,440 --> 00:03:21,440
Speaker 4: that is what I focused my time on there.

57
00:03:21,600 --> 00:03:24,120
Speaker 3: Yeah, so back to the letter.

58
00:03:24,440 --> 00:03:30,880
Speaker 4: The letter was about after a really unpleasant period in

59
00:03:30,919 --> 00:03:34,480
Speaker 4: the year dollar option pit where all the market makers

60
00:03:34,520 --> 00:03:37,040
Speaker 4: lost tons of money because the shape of the skew

61
00:03:37,200 --> 00:03:42,119
Speaker 4: shifted dramatically as the FED started hiking in a very

62
00:03:42,120 --> 00:03:46,880
Speaker 4: predictable manner, and nobody had really developed a measure for

63
00:03:47,360 --> 00:03:50,520
Speaker 4: linear skew. And so during the week I said, well,

64
00:03:51,000 --> 00:03:53,360
Speaker 4: this isn't that much fun. We're losing a lot of

65
00:03:53,400 --> 00:03:56,240
Speaker 4: money every day. But the good news is that that

66
00:03:56,280 --> 00:03:58,920
Speaker 4: means we have something to learn. And so I spent

67
00:03:58,960 --> 00:04:02,280
Speaker 4: the weekend working with the quants and we came up with,

68
00:04:02,480 --> 00:04:06,240
Speaker 4: you know, kind of a measure of the linear skew

69
00:04:06,360 --> 00:04:09,680
Speaker 4: between the calls and puts, and decided to use the

70
00:04:09,680 --> 00:04:13,960
Speaker 4: Greek letter ci to describe it. And you know, so

71
00:04:14,000 --> 00:04:17,239
Speaker 4: by Monday morning we had put it into the risk

72
00:04:17,520 --> 00:04:20,400
Speaker 4: and onto the sheets, and before the open I explained

73
00:04:20,400 --> 00:04:23,400
Speaker 4: to the traders how to talk about it, how to

74
00:04:23,600 --> 00:04:26,200
Speaker 4: use language around it, and before you know it, we

75
00:04:26,200 --> 00:04:28,719
Speaker 4: had made the money back because we were able to

76
00:04:28,880 --> 00:04:33,000
Speaker 4: trade manage this risk better than anybody else because we

77
00:04:33,080 --> 00:04:34,480
Speaker 4: had a whole language around it.

78
00:04:35,279 --> 00:04:38,240
Speaker 2: So we've established your street cred. When it comes to

79
00:04:38,839 --> 00:04:43,360
Speaker 2: solving problems in contracts for financial instruments. If I think

80
00:04:43,400 --> 00:04:47,560
Speaker 2: about a GPU future or something like that, the first

81
00:04:47,640 --> 00:04:51,239
Speaker 2: problem that comes to my mind is standardization, because of course,

82
00:04:51,360 --> 00:04:54,279
Speaker 2: you know all different types of chips, different types of memory,

83
00:04:54,400 --> 00:04:57,960
Speaker 2: different latency. I guess, how do you go about addressing that?

84
00:04:58,320 --> 00:04:59,400
Speaker 3: So that's a great question.

85
00:04:59,520 --> 00:05:02,080
Speaker 4: And right now, so what we've done is if we

86
00:05:02,160 --> 00:05:04,640
Speaker 4: set up two companies. One is called comput Exchange, not

87
00:05:04,760 --> 00:05:07,560
Speaker 4: very creatively named. We have a tendency to do that.

88
00:05:08,480 --> 00:05:10,559
Speaker 2: At DRW is your initials right?

89
00:05:10,640 --> 00:05:13,520
Speaker 3: That was my trading badge and yes, also my initials.

90
00:05:13,600 --> 00:05:14,919
Speaker 3: Yeah you know.

91
00:05:15,560 --> 00:05:18,000
Speaker 4: I mean we did better later on with Cumberland, our

92
00:05:18,200 --> 00:05:21,400
Speaker 4: cryptos trading arm. That was actually a reference to the

93
00:05:21,480 --> 00:05:23,960
Speaker 4: Grateful Dead song about the Cumberland minds.

94
00:05:24,320 --> 00:05:25,400
Speaker 5: I didn't know that. I didn't know that.

95
00:05:25,880 --> 00:05:26,080
Speaker 3: Yeah.

96
00:05:26,480 --> 00:05:29,880
Speaker 4: One of my partners, who does the more creative naming,

97
00:05:30,000 --> 00:05:30,839
Speaker 4: came up with that one.

98
00:05:30,920 --> 00:05:32,880
Speaker 3: He's a he's a dead fan anyway.

99
00:05:33,600 --> 00:05:37,960
Speaker 4: The other company is called Silicon Data, and Silicon Data's

100
00:05:38,120 --> 00:05:43,479
Speaker 4: job is to create indices that will become tradable, you know,

101
00:05:43,560 --> 00:05:47,080
Speaker 4: will be viable to have futures contracts listed on them.

102
00:05:47,440 --> 00:05:50,360
Speaker 4: And right now they've created a number of different ones.

103
00:05:50,400 --> 00:05:53,720
Speaker 4: But one is the H one hundred index. Another one

104
00:05:53,800 --> 00:05:57,240
Speaker 4: is the A one hundred indecks and believe it or not,

105
00:05:57,320 --> 00:05:59,599
Speaker 4: those indices are both available on Bloomberg.

106
00:06:00,200 --> 00:06:03,000
Speaker 5: Amazing. There's a love hearing about it. If we were

107
00:06:03,000 --> 00:06:05,960
Speaker 5: in the studio, I would already would already be looking

108
00:06:06,040 --> 00:06:08,880
Speaker 5: up the chart as you were talking about it. Who

109
00:06:08,920 --> 00:06:11,960
Speaker 5: are the natural participants? Because when I think about AI

110
00:06:12,120 --> 00:06:14,760
Speaker 5: or training, you know, imagine someone goes to one of

111
00:06:14,760 --> 00:06:17,320
Speaker 5: the big cloud vendors and they sign a long term

112
00:06:17,360 --> 00:06:21,280
Speaker 5: contract or whatever. Who are the participants who would be

113
00:06:21,279 --> 00:06:24,720
Speaker 5: better off in an environment where there was a liquid

114
00:06:24,839 --> 00:06:26,039
Speaker 5: market for compute.

115
00:06:26,279 --> 00:06:31,599
Speaker 4: So what we found and DRW actually uses a compute

116
00:06:31,600 --> 00:06:35,240
Speaker 4: exchange source Compute, and we find that because there are

117
00:06:35,279 --> 00:06:40,040
Speaker 4: something like seventy different cloud providers that participate, you can

118
00:06:40,120 --> 00:06:43,200
Speaker 4: often get better pricing. And one of the things that

119
00:06:43,240 --> 00:06:46,200
Speaker 4: you can do is you can specify if let's say

120
00:06:46,200 --> 00:06:50,440
Speaker 4: that you're an AI company and you know roughly what

121
00:06:50,560 --> 00:06:54,240
Speaker 4: kind of cluster you want, you can specify that. You

122
00:06:54,279 --> 00:06:58,160
Speaker 4: can even say, you know what, I'm indifferent between locations,

123
00:06:58,480 --> 00:07:01,640
Speaker 4: or you know, I fits in the Middle East, I'm

124
00:07:01,680 --> 00:07:03,560
Speaker 4: still okay with it, but I want to pay twenty

125
00:07:03,600 --> 00:07:06,640
Speaker 4: cents per gpuur less whatever it is. You can kind

126
00:07:06,640 --> 00:07:10,200
Speaker 4: of express your preference curve comput Exchange can conduct an

127
00:07:10,200 --> 00:07:13,480
Speaker 4: auction and then you know, find the kind of best

128
00:07:13,560 --> 00:07:17,000
Speaker 4: price compute that matches your needs. So that's that's kind

129
00:07:17,040 --> 00:07:19,440
Speaker 4: of the idea of how it works. And you know,

130
00:07:19,520 --> 00:07:23,240
Speaker 4: it probably doesn't work if you want a ten thousand

131
00:07:23,320 --> 00:07:27,800
Speaker 4: cluster monster for doing a huge training run, but for

132
00:07:27,880 --> 00:07:30,560
Speaker 4: inference it works great or for smaller training runs, it

133
00:07:30,600 --> 00:07:31,360
Speaker 4: works really well.

134
00:07:31,880 --> 00:07:34,680
Speaker 2: Is the broader impact The idea that once you establish

135
00:07:34,840 --> 00:07:38,440
Speaker 2: a liquid market where people can you know, presumably hedge

136
00:07:38,440 --> 00:07:41,200
Speaker 2: their exposure, that that would bring down the cost of capital.

137
00:07:41,960 --> 00:07:42,640
Speaker 3: So that's right.

138
00:07:42,720 --> 00:07:46,920
Speaker 4: So once you have a liquid market, then you have

139
00:07:47,040 --> 00:07:50,880
Speaker 4: much more confidence in the indices and you can then

140
00:07:50,960 --> 00:07:54,880
Speaker 4: list futures contracts. And so what does that do? It

141
00:07:55,000 --> 00:08:01,280
Speaker 4: enables the neo clouds that are going out capital buying

142
00:08:01,280 --> 00:08:03,360
Speaker 4: a bunch of GPUs, putting them in data centers and

143
00:08:03,440 --> 00:08:06,120
Speaker 4: kind of hoping that they can rent them out and

144
00:08:06,160 --> 00:08:08,600
Speaker 4: not really knowing what they're going to be able to

145
00:08:08,640 --> 00:08:11,320
Speaker 4: rent them out for six months from now, let alone

146
00:08:11,360 --> 00:08:12,240
Speaker 4: two years from now.

147
00:08:12,640 --> 00:08:14,120
Speaker 3: So a neil.

148
00:08:13,920 --> 00:08:18,560
Speaker 4: Cloud could buy the GPUs, sell a strip of futures contracts,

149
00:08:18,600 --> 00:08:21,160
Speaker 4: and I envision that these will be traded kind of

150
00:08:21,200 --> 00:08:23,800
Speaker 4: like electricity futures, where there's one for every month, and

151
00:08:24,280 --> 00:08:26,400
Speaker 4: if you want to hedge the next three years, you

152
00:08:26,440 --> 00:08:29,760
Speaker 4: sell thirty six of them and now you've locked in

153
00:08:29,800 --> 00:08:32,960
Speaker 4: your pricing. Obviously, their cost of capital is going to

154
00:08:33,000 --> 00:08:36,560
Speaker 4: go down, which in turn should make GPUs.

155
00:08:36,160 --> 00:08:37,480
Speaker 3: More readily available.

156
00:08:37,520 --> 00:08:40,320
Speaker 4: And then on the flip side, if you're running an

157
00:08:40,360 --> 00:08:43,760
Speaker 4: AI company and you raise a finite amount of dollars

158
00:08:43,800 --> 00:08:46,680
Speaker 4: and you kind of know how much training you're going

159
00:08:46,760 --> 00:08:50,360
Speaker 4: to do, but you don't know exactly what configuration. You

160
00:08:50,400 --> 00:08:54,360
Speaker 4: can go ahead buy the compute in the derivatives market,

161
00:08:54,760 --> 00:08:58,559
Speaker 4: and then once you have a clear view on exactly

162
00:08:58,600 --> 00:09:03,120
Speaker 4: what configuration you want, then you can swap those derivatives

163
00:09:03,120 --> 00:09:04,760
Speaker 4: for actual computing.

164
00:09:05,200 --> 00:09:09,600
Speaker 5: Talk to us a little bit more about the cell side,

165
00:09:09,640 --> 00:09:12,160
Speaker 5: So like we have these like big clouds, right, the

166
00:09:12,160 --> 00:09:15,120
Speaker 5: ones that everybody knows, and then you mentioned the neo clouds.

167
00:09:15,280 --> 00:09:18,320
Speaker 5: Do you see that changing? Like what do you see

168
00:09:18,320 --> 00:09:22,360
Speaker 5: as the future mix of cloud vendors in the future.

169
00:09:22,880 --> 00:09:24,920
Speaker 3: So that is a great question.

170
00:09:25,160 --> 00:09:27,760
Speaker 4: I think that the whole space is going to grow,

171
00:09:28,440 --> 00:09:33,720
Speaker 4: but that the aws gcps of the world will make

172
00:09:33,800 --> 00:09:35,480
Speaker 4: up a smaller percentage of the whole.

173
00:09:35,679 --> 00:09:37,280
Speaker 5: Okay, that's my guess.

174
00:09:37,320 --> 00:09:42,559
Speaker 3: But how come because there's such proliferation of other companies

175
00:09:42,679 --> 00:09:44,280
Speaker 3: buying GPUs and deploying them.

176
00:09:44,600 --> 00:09:46,120
Speaker 5: Okay, that's a good answer.

177
00:09:46,679 --> 00:09:48,720
Speaker 2: You know, Joe asked you who would be the natural

178
00:09:48,760 --> 00:09:51,120
Speaker 2: market participants for this? I'm going to ask you the

179
00:09:51,120 --> 00:09:55,600
Speaker 2: opposite question, who wouldn't want this? Because I think of

180
00:09:55,640 --> 00:09:59,280
Speaker 2: some of the hyperscalers they seem to like controlling the

181
00:09:59,320 --> 00:10:03,040
Speaker 2: GPU supply and maybe squeezing some of their competitors. Would

182
00:10:03,080 --> 00:10:04,360
Speaker 2: you expect resistance from.

183
00:10:04,200 --> 00:10:07,880
Speaker 4: Them, Yeah, I mean I think the hyperscalers benefit from

184
00:10:08,000 --> 00:10:11,520
Speaker 4: opaque pricing and kind of bundled pricing, and of course

185
00:10:11,559 --> 00:10:13,960
Speaker 4: they would prefer to have all the GPUs could but

186
00:10:14,720 --> 00:10:15,479
Speaker 4: in video.

187
00:10:15,320 --> 00:10:16,960
Speaker 2: I would also prefer to have all the GPUs.

188
00:10:17,040 --> 00:10:18,480
Speaker 3: Yeah, yeah, that's always a good thing.

189
00:10:18,559 --> 00:10:21,640
Speaker 4: But I think in Vidia wants the GPUs to be

190
00:10:21,760 --> 00:10:25,080
Speaker 4: widely distributed, and they're really the ones that make the call.

191
00:10:25,679 --> 00:10:28,760
Speaker 5: This isn't the first time that there's been an attempt

192
00:10:28,840 --> 00:10:32,440
Speaker 5: to create futures markets out of technology. I think there's

193
00:10:32,480 --> 00:10:36,679
Speaker 5: been multiple efforts decades ago to Like d RAM futures

194
00:10:37,120 --> 00:10:41,440
Speaker 5: doesn't seem that fundamentally different, although maybe it is. Why

195
00:10:41,480 --> 00:10:44,320
Speaker 5: did those fail? Like when you think about, like what's

196
00:10:44,360 --> 00:10:46,360
Speaker 5: going to be different at this time, what was the

197
00:10:46,360 --> 00:10:49,280
Speaker 5: failure that caused? Like why didn't RAM futures take off?

198
00:10:49,440 --> 00:10:52,240
Speaker 4: So the thing about d RAM was that the price

199
00:10:52,320 --> 00:10:55,679
Speaker 4: just kept on going down so in a very predictable way,

200
00:10:56,360 --> 00:10:58,720
Speaker 4: And so why would you want to buy a futures

201
00:10:58,760 --> 00:11:00,560
Speaker 4: contract if you know the price and future is going

202
00:11:00,600 --> 00:11:01,160
Speaker 4: to be lower?

203
00:11:01,480 --> 00:11:02,760
Speaker 3: Whereas GPUs.

204
00:11:03,160 --> 00:11:06,040
Speaker 4: You know, we've certainly gone through periods where GPU demand

205
00:11:06,200 --> 00:11:09,800
Speaker 4: was super high, and then we've gone through period where

206
00:11:10,240 --> 00:11:12,040
Speaker 4: you know, there was kind of some excess to.

207
00:11:12,600 --> 00:11:14,840
Speaker 5: Not a consistent trajectory of pricing.

208
00:11:15,080 --> 00:11:19,000
Speaker 4: I think that there will be a consistent trajectory lower

209
00:11:19,160 --> 00:11:21,679
Speaker 4: in terms of I don't know, however you want to

210
00:11:21,720 --> 00:11:25,720
Speaker 4: measure it dollars per flop put for dollars per token.

211
00:11:25,960 --> 00:11:29,040
Speaker 4: I think that that's going to continue to decline. But

212
00:11:29,520 --> 00:11:31,400
Speaker 4: you know, in each one hundred is going to be

213
00:11:31,400 --> 00:11:34,760
Speaker 4: a useful GPU for a very long time, and over

214
00:11:34,800 --> 00:11:37,400
Speaker 4: its life, I think there will be periods where there's

215
00:11:37,440 --> 00:11:39,760
Speaker 4: more demand, less demand, and you know, a little bit

216
00:11:39,840 --> 00:11:43,040
Speaker 4: more cyclicality and less predictability.

217
00:11:43,640 --> 00:11:46,160
Speaker 2: So I know that the Trump administration has said that

218
00:11:46,240 --> 00:11:49,040
Speaker 2: they want this market to happen, right, So you seem

219
00:11:49,080 --> 00:11:52,880
Speaker 2: to have some regulatory I guess tailwind behind you.

220
00:11:53,480 --> 00:11:55,480
Speaker 3: Yeah. I mean, I don't think that this is a

221
00:11:55,520 --> 00:11:56,760
Speaker 3: controversial thing.

222
00:11:57,000 --> 00:12:01,520
Speaker 4: I think that it's pretty clear that once we figure

223
00:12:01,520 --> 00:12:04,840
Speaker 4: out the right index construction and have kind of sufficient

224
00:12:04,920 --> 00:12:08,400
Speaker 4: data that I don't think the CFTC would complain about

225
00:12:08,440 --> 00:12:08,959
Speaker 4: the product.

226
00:12:24,960 --> 00:12:27,240
Speaker 5: This is a little bit of a sideways question from

227
00:12:27,320 --> 00:12:30,280
Speaker 5: your attempt to build this market. But speaking of the

228
00:12:30,320 --> 00:12:34,360
Speaker 5: cloud in your main business at DRW, I assume you're

229
00:12:34,440 --> 00:12:38,400
Speaker 5: sort of major customers or users of the CME. Are

230
00:12:38,440 --> 00:12:42,600
Speaker 5: you excited about the CME's migration of its back end

231
00:12:42,679 --> 00:12:44,920
Speaker 5: to Google Cloud because they tout it, they talk about

232
00:12:44,920 --> 00:12:48,200
Speaker 5: their partnership with Google, et cetera. As a client or customer,

233
00:12:49,480 --> 00:12:51,120
Speaker 5: are you enthusiastic about this move?

234
00:12:51,280 --> 00:12:54,959
Speaker 2: We interviewed Terry earlier today and he was excited for sure.

235
00:12:55,280 --> 00:12:55,520
Speaker 3: Yeah.

236
00:12:55,600 --> 00:12:58,160
Speaker 4: So it depends on what you put into the cloud.

237
00:12:58,559 --> 00:13:01,080
Speaker 4: And it's totally fine to put a lot of things

238
00:13:01,120 --> 00:13:03,040
Speaker 4: into the cloud. But the thing that you don't want

239
00:13:03,080 --> 00:13:05,480
Speaker 4: to put into the cloud is a matching engine. And

240
00:13:05,520 --> 00:13:09,360
Speaker 4: the reason for that is you want the matching engine

241
00:13:09,400 --> 00:13:14,360
Speaker 4: to be as deterministic as possible. So that means that

242
00:13:15,040 --> 00:13:20,520
Speaker 4: if you send two orders into the matching engine, one

243
00:13:21,040 --> 00:13:24,640
Speaker 4: let's say a couple of microseconds behind the other one,

244
00:13:25,280 --> 00:13:27,640
Speaker 4: you want the one that gets there first to be

245
00:13:27,720 --> 00:13:32,440
Speaker 4: filled every time. Yeah, and if you put stuff into

246
00:13:32,440 --> 00:13:34,640
Speaker 4: the cloud, it's very hard to make that happen. You

247
00:13:34,679 --> 00:13:40,040
Speaker 4: wind up getting a wide distribution around which order will

248
00:13:40,040 --> 00:13:42,760
Speaker 4: be filled first. And even as you kind of stretch

249
00:13:42,840 --> 00:13:46,319
Speaker 4: those times out, you could have an order that comes

250
00:13:46,320 --> 00:13:51,560
Speaker 4: in maybe a couple milliseconds later be filled first. That

251
00:13:52,040 --> 00:13:57,440
Speaker 4: is super disruptive for liquidity providers and it means that

252
00:13:58,000 --> 00:13:59,720
Speaker 4: the liquidity in the market's going to suffer.

253
00:14:00,200 --> 00:14:03,040
Speaker 5: Humhm. But this is you say, it's not ideal for

254
00:14:03,080 --> 00:14:05,320
Speaker 5: them to have a matching engine in the cloud, but

255
00:14:05,400 --> 00:14:06,599
Speaker 5: this is the direction it's going in.

256
00:14:06,960 --> 00:14:11,080
Speaker 4: Yeah, and it's unclear exactly which part of the matching

257
00:14:11,120 --> 00:14:13,720
Speaker 4: engine moving the cloud. Is it some kind of a

258
00:14:13,880 --> 00:14:17,599
Speaker 4: dual structure. I don't know, But that's what matters, is

259
00:14:18,240 --> 00:14:21,440
Speaker 4: a deterministic matching engine. I mean, if Google can figure

260
00:14:21,440 --> 00:14:24,640
Speaker 4: out how to make matching engine and the cloud deterministic,

261
00:14:25,280 --> 00:14:27,720
Speaker 4: go for it. I'm very skeptical that that's even possible.

262
00:14:27,760 --> 00:14:31,080
Speaker 5: Can you just describe the sort of theoretical problem. What

263
00:14:31,240 --> 00:14:35,280
Speaker 5: is it about cloud computing that makes this particular problem,

264
00:14:35,440 --> 00:14:39,840
Speaker 5: the deterministic aspect difficult as opposed to traditional infrastructure.

265
00:14:40,040 --> 00:14:44,400
Speaker 4: Well, when you have on prem computers, you can it's

266
00:14:44,440 --> 00:14:47,600
Speaker 4: all right there, you can control where the wires go,

267
00:14:47,760 --> 00:14:50,360
Speaker 4: and so when it's in the cloud, it's a little

268
00:14:50,360 --> 00:14:54,240
Speaker 4: bit more well nebulous. I guess it's just harder to do.

269
00:14:54,600 --> 00:14:59,160
Speaker 2: That's a good pun I admire it. So you mentioned

270
00:14:59,160 --> 00:15:01,680
Speaker 2: that you have this law and storied career in the

271
00:15:01,720 --> 00:15:04,600
Speaker 2: trading industry, starting from old school trading, and now we're

272
00:15:04,600 --> 00:15:08,280
Speaker 2: here talking about GPU trading and what's in the cloud

273
00:15:08,320 --> 00:15:11,880
Speaker 2: and what works and what doesn't tell us what your company,

274
00:15:11,880 --> 00:15:14,720
Speaker 2: what DRW is actually doing when it comes to practical

275
00:15:14,760 --> 00:15:19,000
Speaker 2: application of AI, this is a question we're asking everyone.

276
00:15:19,040 --> 00:15:22,600
Speaker 2: We asked all companies to spill all their proprietary secrets

277
00:15:23,120 --> 00:15:23,880
Speaker 2: about AI.

278
00:15:23,880 --> 00:15:27,000
Speaker 5: Excluding the engineers. We know that, we know, we know

279
00:15:27,040 --> 00:15:30,560
Speaker 5: people are yes, we know that they're using clog code

280
00:15:30,600 --> 00:15:30,920
Speaker 5: or whatever.

281
00:15:30,960 --> 00:15:33,120
Speaker 4: So besides the yeah, yeah, yeah, you're right, that's that's

282
00:15:33,200 --> 00:15:34,280
Speaker 4: kind of the boring answer.

283
00:15:34,760 --> 00:15:37,160
Speaker 5: Yeah, And then the other thing is, then when we

284
00:15:37,200 --> 00:15:40,280
Speaker 5: ask this question, people cite a bunch of machine learning things,

285
00:15:40,760 --> 00:15:42,720
Speaker 5: which has been here for a while. So let's talk.

286
00:15:42,600 --> 00:15:43,640
Speaker 3: About actual Ah.

287
00:15:44,200 --> 00:15:47,720
Speaker 4: Yeah, So I think that the way that we make

288
00:15:47,760 --> 00:15:52,400
Speaker 4: trading decisions is going to change dramatically, and it already is.

289
00:15:52,600 --> 00:15:58,840
Speaker 4: You can use AI to interact with your proprietary data,

290
00:15:59,400 --> 00:16:03,520
Speaker 4: your propriety terry models, and suggest trades.

291
00:16:04,240 --> 00:16:05,080
Speaker 3: That's pretty cool.

292
00:16:05,360 --> 00:16:06,360
Speaker 5: Are you doing that right now?

293
00:16:06,600 --> 00:16:09,720
Speaker 4: Yeah, so we're starting to do that, but we have

294
00:16:09,840 --> 00:16:12,440
Speaker 4: some tools that kind of do that now. And the

295
00:16:12,520 --> 00:16:15,280
Speaker 4: other thing that's really interesting is to fiddle around with

296
00:16:15,440 --> 00:16:18,560
Speaker 4: agents and have different agents interact, and so you could

297
00:16:18,600 --> 00:16:22,120
Speaker 4: kind of think about maybe you have a couple different

298
00:16:22,160 --> 00:16:27,720
Speaker 4: analysts AI analysts that both work on some stock, and

299
00:16:27,760 --> 00:16:31,720
Speaker 4: then you have kind of a risk taking agent or

300
00:16:31,720 --> 00:16:34,280
Speaker 4: maybe a couple different risk taking agents that interact with

301
00:16:34,320 --> 00:16:38,160
Speaker 4: those analysts and then come up with trades based on that. So,

302
00:16:38,480 --> 00:16:39,920
Speaker 4: I mean, these are you know, that's a little bit

303
00:16:39,920 --> 00:16:42,040
Speaker 4: of a theoretical concept, but I don't think we're that

304
00:16:42,120 --> 00:16:43,720
Speaker 4: far away from things like that.

305
00:16:44,520 --> 00:16:48,200
Speaker 5: Just on the cloud trading a little bit more. I

306
00:16:48,240 --> 00:16:51,440
Speaker 5: am really interested in this topic. What is the current

307
00:16:51,480 --> 00:16:53,440
Speaker 5: stay today, just so that we understand where you're at,

308
00:16:53,440 --> 00:16:57,000
Speaker 5: Like what is today's snapshot of usage of the platforms?

309
00:16:57,320 --> 00:17:00,080
Speaker 3: I mean as far as where the matching engines are or.

310
00:17:00,120 --> 00:17:03,320
Speaker 5: Oh sorry, on the on the GPU trading is it

311
00:17:03,400 --> 00:17:03,720
Speaker 5: right now?

312
00:17:03,960 --> 00:17:04,280
Speaker 3: Oh?

313
00:17:04,320 --> 00:17:05,600
Speaker 5: Like where is the state of the business?

314
00:17:05,600 --> 00:17:10,800
Speaker 4: Oh, you know, I think last month we conducted five

315
00:17:10,880 --> 00:17:14,680
Speaker 4: or six auctions, so it's early, but it's happening.

316
00:17:15,240 --> 00:17:18,600
Speaker 2: So when I think about how like futures contracts are born,

317
00:17:19,040 --> 00:17:23,720
Speaker 2: it's usually bespoke options and then you get the index.

318
00:17:23,800 --> 00:17:26,000
Speaker 2: I guess, and then you get a forward and then

319
00:17:26,040 --> 00:17:27,919
Speaker 2: a future. That's kind of how I think about it

320
00:17:27,920 --> 00:17:30,480
Speaker 2: in my head, is that the process that you imagine.

321
00:17:30,119 --> 00:17:34,640
Speaker 3: For this h not necessarily. I think that the simplest.

322
00:17:35,440 --> 00:17:38,520
Speaker 4: I mean, yeah, I suppose you could do some privately

323
00:17:38,560 --> 00:17:42,280
Speaker 4: negotiated compute swap or something, and maybe that will happen first,

324
00:17:42,960 --> 00:17:46,880
Speaker 4: but no, I think the first thing is a futures

325
00:17:46,880 --> 00:17:49,760
Speaker 4: contract that settles to an index. If the spot market

326
00:17:49,800 --> 00:17:54,199
Speaker 4: becomes really liquid and you have very standardized auctions, and

327
00:17:54,280 --> 00:17:56,240
Speaker 4: you know, one of the things that you asked about was, well,

328
00:17:56,240 --> 00:17:59,919
Speaker 4: how do you deal with the lack of standardized you know?

329
00:18:00,080 --> 00:18:02,760
Speaker 4: And so one thing is you go to a certain

330
00:18:03,200 --> 00:18:06,200
Speaker 4: type of GPU, you know, H one hundred for instance.

331
00:18:06,720 --> 00:18:10,600
Speaker 4: But even within that, you can configure them in different ways.

332
00:18:11,080 --> 00:18:13,040
Speaker 4: You could use in finiband, you can use some other

333
00:18:13,080 --> 00:18:16,320
Speaker 4: way of connecting them. And so what's important is you

334
00:18:16,400 --> 00:18:20,040
Speaker 4: need to decide on some benchmark. And one of the

335
00:18:20,080 --> 00:18:23,600
Speaker 4: things that Silicon Data has done is they've actually built

336
00:18:23,640 --> 00:18:29,159
Speaker 4: some measurement tools that measure how fast a GPU cluster is,

337
00:18:29,359 --> 00:18:32,760
Speaker 4: and so you can then say, okay, well, in order

338
00:18:32,840 --> 00:18:36,480
Speaker 4: for this GPU to be kind of eligible to be

339
00:18:36,520 --> 00:18:39,480
Speaker 4: in the index it needs to meet a certain standard,

340
00:18:40,040 --> 00:18:41,920
Speaker 4: and you can there are a couple different vectors you

341
00:18:41,960 --> 00:18:44,560
Speaker 4: can measure by, so I think that that's kind of

342
00:18:45,000 --> 00:18:47,760
Speaker 4: how you would do it. And then if you got

343
00:18:48,880 --> 00:18:53,119
Speaker 4: very liquid auctions, you could actually have a futures contract

344
00:18:53,160 --> 00:18:57,399
Speaker 4: that cash settles to the auction price, and then people

345
00:18:57,400 --> 00:19:01,400
Speaker 4: can have the option of either essentially just cash settling

346
00:19:01,440 --> 00:19:05,680
Speaker 4: their derivative and walking away, or cash selling their derivative

347
00:19:05,760 --> 00:19:09,720
Speaker 4: and participating in the auction, and they would know that

348
00:19:09,880 --> 00:19:12,920
Speaker 4: price would transfer from one thing to another. That might

349
00:19:12,960 --> 00:19:14,880
Speaker 4: be a future state of the world, and the initial

350
00:19:14,920 --> 00:19:17,879
Speaker 4: state is probably just a generic index, and the future

351
00:19:17,960 --> 00:19:18,800
Speaker 4: is cash settled.

352
00:19:18,560 --> 00:19:19,040
Speaker 3: To the index.

353
00:19:19,720 --> 00:19:23,439
Speaker 2: What would a market failure look like in GPU trading,

354
00:19:23,440 --> 00:19:25,760
Speaker 2: because your analogy is the oil market, and you know,

355
00:19:25,840 --> 00:19:28,520
Speaker 2: weird stuff happens in the oil market. Could we get

356
00:19:28,840 --> 00:19:32,439
Speaker 2: negative GPU prices? Or if everyone wakes up one day

357
00:19:32,480 --> 00:19:34,960
Speaker 2: and decides they want to use chat GPT as their

358
00:19:35,000 --> 00:19:38,280
Speaker 2: psychotherapist or whatever some people are doing, could you have

359
00:19:38,320 --> 00:19:42,840
Speaker 2: a GPU shortage where maybe people can't deliver into the contract.

360
00:19:43,400 --> 00:19:45,560
Speaker 4: There are lots of ways that markets can break and

361
00:19:45,600 --> 00:19:48,640
Speaker 4: go wrong. And I remember to this day that when

362
00:19:48,960 --> 00:19:50,600
Speaker 4: oil futures went negative.

363
00:19:50,680 --> 00:19:52,880
Speaker 3: It was during COVID. I was sitting at home.

364
00:19:52,920 --> 00:19:55,879
Speaker 4: I was trading oil futures and I bought oil futures

365
00:19:55,880 --> 00:19:56,800
Speaker 4: for negative prices.

366
00:19:57,320 --> 00:20:00,480
Speaker 5: You were one of the one team made money.

367
00:19:59,359 --> 00:19:59,600
Speaker 3: Yeah.

368
00:20:00,600 --> 00:20:05,280
Speaker 4: My, Uh then what was that twenty twenty one? So yeah,

369
00:20:05,359 --> 00:20:09,400
Speaker 4: my then fourteen year old said to me, please, please please,

370
00:20:10,160 --> 00:20:14,080
Speaker 4: I want to buy negative priced futures contracts. And I said, well,

371
00:20:14,119 --> 00:20:14,920
Speaker 4: you have no way.

372
00:20:14,720 --> 00:20:16,960
Speaker 3: Of taking delivery the oil.

373
00:20:17,040 --> 00:20:19,679
Speaker 4: And he said, I will go to Cushing, Oklahoma and

374
00:20:19,840 --> 00:20:21,000
Speaker 4: figure out how to do it.

375
00:20:21,520 --> 00:20:27,680
Speaker 5: You've really raised a son, daughter, son, son. You've really uh,

376
00:20:27,359 --> 00:20:28,520
Speaker 5: he's been learning.

377
00:20:29,119 --> 00:20:32,160
Speaker 2: We have we have an episode about taking physical possession

378
00:20:32,160 --> 00:20:34,560
Speaker 2: of oil. I do not recommend. It. Turns out if

379
00:20:34,600 --> 00:20:36,320
Speaker 2: you keep it on your desk for long enough, it

380
00:20:36,359 --> 00:20:39,000
Speaker 2: evaporates into the atmosphere and poisons your colleagues.

381
00:20:39,800 --> 00:20:42,480
Speaker 3: Yeah, anyway, a little bit of a tangent.

382
00:20:43,040 --> 00:20:47,240
Speaker 4: So I think on the upwards rejectory, if there's tons

383
00:20:47,280 --> 00:20:49,760
Speaker 4: of demand, you know, that's something that commodity markets are

384
00:20:49,760 --> 00:20:52,080
Speaker 4: really good at dealing with. The price will go up

385
00:20:52,119 --> 00:20:54,400
Speaker 4: and more supply will come in, and I think that's

386
00:20:54,400 --> 00:20:54,879
Speaker 4: all good.

387
00:20:55,440 --> 00:20:57,480
Speaker 3: On the downward side, you know.

388
00:20:57,480 --> 00:21:00,320
Speaker 4: You can always just turn the GPUs off, so I

389
00:21:00,320 --> 00:21:01,520
Speaker 4: don't think they trade negative.

390
00:21:01,960 --> 00:21:05,040
Speaker 5: How much of the volatility that do you when you

391
00:21:05,080 --> 00:21:08,240
Speaker 5: anticipate market volatility in the price of GPUs? How much

392
00:21:08,560 --> 00:21:12,760
Speaker 5: is that like embedded electricity costs, So when you buy compute, right,

393
00:21:12,800 --> 00:21:15,360
Speaker 5: you're buying the chip, but also the power, Like, how

394
00:21:15,440 --> 00:21:17,320
Speaker 5: much of that volatility will be the power?

395
00:21:17,600 --> 00:21:22,680
Speaker 4: So the industry lingo that's used is total cost of ownership,

396
00:21:22,800 --> 00:21:25,399
Speaker 4: And you know what percentage of the total cost of

397
00:21:25,440 --> 00:21:28,359
Speaker 4: ownership is the power price? And for an H one hundred,

398
00:21:28,520 --> 00:21:29,880
Speaker 4: it's less than fifteen.

399
00:21:29,560 --> 00:21:34,080
Speaker 2: Percent less than fifteen Okay, So GPU trading obviously one

400
00:21:34,080 --> 00:21:36,000
Speaker 2: of the things you're working on, but you're a busy

401
00:21:36,000 --> 00:21:38,800
Speaker 2: guy and you've got other stuff up your sleeve. What

402
00:21:38,840 --> 00:21:41,639
Speaker 2: are you doing in the realms of tokenized trading?

403
00:21:42,920 --> 00:21:46,040
Speaker 4: So that is an area that we're super excited about,

404
00:21:46,080 --> 00:21:48,080
Speaker 4: and we've been thinking about this for a very long time.

405
00:21:48,720 --> 00:21:53,639
Speaker 4: So in twenty twelve, when we started talking about bitcoin

406
00:21:53,960 --> 00:21:56,760
Speaker 4: at Dow and there were a number of traders at

407
00:21:56,760 --> 00:21:58,720
Speaker 4: DORW that were very excited about.

408
00:21:58,520 --> 00:22:01,280
Speaker 2: Bitcoin, you were very early into it. Twenty twelve was

409
00:22:01,280 --> 00:22:01,960
Speaker 2: still pretty.

410
00:22:01,720 --> 00:22:05,160
Speaker 4: Early, very early, Yeah, So we were having these discussions

411
00:22:05,160 --> 00:22:08,040
Speaker 4: of why is this interesting. Is it interesting? What about

412
00:22:08,040 --> 00:22:11,560
Speaker 4: it is interesting? And we came away with the following thesis,

413
00:22:11,760 --> 00:22:14,280
Speaker 4: there's some small chance that bitcoin could be digital gold.

414
00:22:14,760 --> 00:22:17,399
Speaker 4: I don't know, you know, call it one percent. It's

415
00:22:17,440 --> 00:22:19,840
Speaker 4: kind of an interesting product, so we should probably make

416
00:22:19,880 --> 00:22:22,600
Speaker 4: markets in it. So we set up Cumberland as the

417
00:22:23,440 --> 00:22:25,280
Speaker 4: and you know, we didn't call it DRW because at

418
00:22:25,320 --> 00:22:29,400
Speaker 4: the time, everybody knew that anybody trading crypto was obviously

419
00:22:29,440 --> 00:22:32,200
Speaker 4: a crook, so so you know, we wanted to kind

420
00:22:32,200 --> 00:22:34,480
Speaker 4: of separate the brand a little bit. But you know,

421
00:22:34,560 --> 00:22:38,120
Speaker 4: the other thing was this idea that you could move

422
00:22:38,560 --> 00:22:44,600
Speaker 4: value instantaneously in a trustless ecosystem was super interesting to me,

423
00:22:44,680 --> 00:22:47,119
Speaker 4: and I said, wow, if you could do that in

424
00:22:47,200 --> 00:22:50,359
Speaker 4: traditional financial markets, that would make the market so much better,

425
00:22:50,440 --> 00:22:53,320
Speaker 4: so much more resilient, and so we should really figure

426
00:22:53,320 --> 00:22:55,240
Speaker 4: out how to do that. So we started a company called,

427
00:22:55,359 --> 00:22:59,880
Speaker 4: again not very creatively named Digital Asset Holdings, which created

428
00:23:00,000 --> 00:23:04,680
Speaker 4: the Canton blockchain. Initially, the Canton blockchain was a private,

429
00:23:04,720 --> 00:23:09,040
Speaker 4: permissioned chain, but last summer it actually became a public chain,

430
00:23:10,080 --> 00:23:15,560
Speaker 4: and that chain was designed specifically with tokenization of traditional

431
00:23:15,600 --> 00:23:19,280
Speaker 4: financial instruments in mind. So it has a couple of characteristics.

432
00:23:19,280 --> 00:23:22,919
Speaker 4: One is it has configurable privacy, and believe it or not,

433
00:23:23,920 --> 00:23:28,080
Speaker 4: for people who are in the finance business, they don't

434
00:23:28,119 --> 00:23:30,760
Speaker 4: want to broadcast to the entire world when they are

435
00:23:30,800 --> 00:23:33,600
Speaker 4: buying or selling something. I mean, obviously, if it's above

436
00:23:33,640 --> 00:23:36,280
Speaker 4: the reporting thresholds, you do. So that was kind of

437
00:23:36,320 --> 00:23:40,600
Speaker 4: a fundamental characteristic of this chain. It's different than Ethereum

438
00:23:40,720 --> 00:23:42,919
Speaker 4: or Solona or any of these other things where if

439
00:23:42,960 --> 00:23:45,560
Speaker 4: you tokenize something and put it on top and you

440
00:23:45,640 --> 00:23:48,880
Speaker 4: move it around, everybody sees it move around. So that's

441
00:23:48,960 --> 00:23:52,000
Speaker 4: kind of something we've been working on for quite a while.

442
00:23:52,400 --> 00:23:55,760
Speaker 5: How big could this get? Like? Could it swallow everything?

443
00:23:55,800 --> 00:23:59,520
Speaker 5: Could you imagine a world in which, given any financial instrument,

444
00:24:00,119 --> 00:24:04,000
Speaker 5: do bond et cetera, that it all sort of ends

445
00:24:04,080 --> 00:24:04,560
Speaker 5: up on chin.

446
00:24:04,960 --> 00:24:07,440
Speaker 3: Yeah, I think that everything will be on chain.

447
00:24:07,480 --> 00:24:10,040
Speaker 5: Wow by when give us a year?

448
00:24:10,119 --> 00:24:12,119
Speaker 3: No, I'm always way too early on this stuff.

449
00:24:12,119 --> 00:24:15,399
Speaker 4: But I think in the next five years all of

450
00:24:15,400 --> 00:24:17,240
Speaker 4: these instruments will be on chain.

451
00:24:17,520 --> 00:24:20,960
Speaker 5: Okay, that's good. Primarily we will have a live episode

452
00:24:21,000 --> 00:24:23,680
Speaker 5: in twenty we'll come back to Chicago, US that question.

453
00:24:23,760 --> 00:24:42,439
Speaker 2: Yeah, is the ideal with tokenized assets also that you

454
00:24:42,440 --> 00:24:45,760
Speaker 2: could use that for collateral management and use it as

455
00:24:45,800 --> 00:24:46,840
Speaker 2: a way to move collateral.

456
00:24:48,160 --> 00:24:51,480
Speaker 4: And so everybody's talking about moving to twenty four five

457
00:24:51,560 --> 00:24:54,760
Speaker 4: or twenty four to seven markets, and if you want

458
00:24:54,800 --> 00:24:58,600
Speaker 4: to do that, it's really important to be able to

459
00:24:58,600 --> 00:25:01,800
Speaker 4: move collateral twenty four five or twenty four seven and

460
00:25:02,040 --> 00:25:05,480
Speaker 4: move variation margin twenty four or five or twenty four seven.

461
00:25:06,000 --> 00:25:10,200
Speaker 3: And so yes, that is a very important use case.

462
00:25:10,640 --> 00:25:14,520
Speaker 5: So speaking of very exciting sexy topics in trading, right

463
00:25:14,600 --> 00:25:18,000
Speaker 5: after you, we're going to be speaking with Terrek Mansur

464
00:25:18,200 --> 00:25:21,560
Speaker 5: of Kelshi And so prediction markets are super hot. Where

465
00:25:21,600 --> 00:25:24,600
Speaker 5: are you at with them? Is dorw making markets in

466
00:25:24,640 --> 00:25:26,960
Speaker 5: any of these in any of the spaces right now?

467
00:25:27,400 --> 00:25:30,960
Speaker 4: So a million years ago we actually made markets and

468
00:25:31,040 --> 00:25:34,560
Speaker 4: prediction markets. I think it was, I don't know, in

469
00:25:34,600 --> 00:25:37,159
Speaker 4: trade or something, yeah, and it never went anywhere.

470
00:25:37,200 --> 00:25:37,840
Speaker 3: Nobody cared.

471
00:25:37,960 --> 00:25:40,320
Speaker 4: And I always thought, you know, prediction markets should be

472
00:25:40,359 --> 00:25:42,280
Speaker 4: a thing everybody should care.

473
00:25:42,800 --> 00:25:45,120
Speaker 3: But nobody did. And then Auger came out.

474
00:25:45,200 --> 00:25:46,800
Speaker 4: I was like, oh, this is really cool, this is

475
00:25:46,800 --> 00:25:49,919
Speaker 4: going to take off, and nobody cared, and so it's it's.

476
00:25:49,800 --> 00:25:50,560
Speaker 3: Taken a long time.

477
00:25:50,600 --> 00:25:53,840
Speaker 4: So at this point we use it as a reference price,

478
00:25:54,320 --> 00:25:56,600
Speaker 4: you know, obviously during the election, it was super helpful

479
00:25:56,640 --> 00:25:58,800
Speaker 4: to use that as a gauge.

480
00:25:58,480 --> 00:26:01,199
Speaker 2: Of So you were actually using that because you know,

481
00:26:01,200 --> 00:26:05,200
Speaker 2: we hear stories about institutional investors maybe finding prediction markets

482
00:26:05,280 --> 00:26:07,800
Speaker 2: useful perhaps, but you were looking at it.

483
00:26:07,880 --> 00:26:09,720
Speaker 4: We were definitely looking at it. We were not using

484
00:26:09,800 --> 00:26:13,640
Speaker 4: it as a hedge. And it was funny. Shane messaged

485
00:26:13,640 --> 00:26:16,679
Speaker 4: me and said, hey, you know it's it's it's up

486
00:26:16,680 --> 00:26:19,600
Speaker 4: on Bloomberg now and I was like, oh, that's awesome, Shane.

487
00:26:19,600 --> 00:26:23,439
Speaker 5: Shane Coplin from Yeah, that's right. Yeah, But currently, like

488
00:26:23,520 --> 00:26:25,560
Speaker 5: do you foresee, like are you going to enter not

489
00:26:26,000 --> 00:26:30,120
Speaker 5: either making markets on some of these exchanges, and would

490
00:26:30,119 --> 00:26:32,080
Speaker 5: you get into the sports contracts?

491
00:26:32,680 --> 00:26:34,040
Speaker 3: I mean, so we're not here.

492
00:26:34,320 --> 00:26:36,959
Speaker 4: I think it's highly likely that we'll start trading some

493
00:26:37,000 --> 00:26:40,760
Speaker 4: of the prediction markets. Some of our competitors already trade

494
00:26:41,040 --> 00:26:44,360
Speaker 4: in the sports markets pretty actively. We don't, so it's

495
00:26:44,400 --> 00:26:47,480
Speaker 4: not necessarily a natural fit. But I don't have like

496
00:26:47,560 --> 00:26:49,400
Speaker 4: a religious opposition to it.

497
00:26:50,320 --> 00:26:53,399
Speaker 2: Would there be different considerations for trading in a prediction

498
00:26:53,480 --> 00:26:56,920
Speaker 2: market versus a traditional financial asset? Are there different things

499
00:26:56,960 --> 00:26:59,639
Speaker 2: you have to think about, either in terms of like

500
00:26:59,720 --> 00:27:02,280
Speaker 2: price seeing the trade or maybe risk management.

501
00:27:02,119 --> 00:27:04,600
Speaker 3: Well, I think it depends on what the prediction market is.

502
00:27:04,760 --> 00:27:08,359
Speaker 4: I mean, if you're trading a prediction market on I

503
00:27:08,400 --> 00:27:14,199
Speaker 4: don't know whether somebody will throw a rubber object onto

504
00:27:14,359 --> 00:27:19,760
Speaker 4: a WNBA court, then I mean that's something that people

505
00:27:19,840 --> 00:27:22,520
Speaker 4: in the audience can control. And so it seems like

506
00:27:23,040 --> 00:27:26,560
Speaker 4: providing liquidity in that you would be at a disadvantage.

507
00:27:26,720 --> 00:27:30,679
Speaker 2: That was a very particular example. By the way, I

508
00:27:30,720 --> 00:27:32,960
Speaker 2: was going to go with Taylor Swift getting married.

509
00:27:32,840 --> 00:27:34,439
Speaker 3: But you went with that one.

510
00:27:34,960 --> 00:27:39,080
Speaker 5: Well, these are markets that people can directly intervene.

511
00:27:38,960 --> 00:27:42,880
Speaker 3: Directly as opposed to, for instance, their.

512
00:27:42,800 --> 00:27:45,720
Speaker 5: Own yeah, antisocial behavior, that's.

513
00:27:45,400 --> 00:27:45,919
Speaker 3: Right, and.

514
00:27:47,480 --> 00:27:51,560
Speaker 4: As opposed to will the Fed cut twenty five or

515
00:27:51,600 --> 00:27:54,840
Speaker 4: fifty or stay on hold? I mean you can trade

516
00:27:54,840 --> 00:27:58,159
Speaker 4: that in sofur, you can trade that in in the

517
00:27:58,200 --> 00:28:01,840
Speaker 4: FED funds futures. There some binaries you can trade. And

518
00:28:01,880 --> 00:28:06,600
Speaker 4: so the prediction market version of that is totally fits

519
00:28:06,640 --> 00:28:08,640
Speaker 4: in with the risk that we already trade.

520
00:28:09,000 --> 00:28:11,679
Speaker 5: So we mentioned in the intro there's gonna be this

521
00:28:11,720 --> 00:28:14,680
Speaker 5: big meeting in DC next week, and we just happened

522
00:28:14,680 --> 00:28:17,640
Speaker 5: to sort of catch a bunch of the participants. When

523
00:28:17,680 --> 00:28:21,040
Speaker 5: you look at the landscape for these new futures, platforms

524
00:28:21,040 --> 00:28:24,959
Speaker 5: because that's what they are, right. The CME has regulation

525
00:28:25,359 --> 00:28:29,600
Speaker 5: been part of their dominance. Has regulation made it harder

526
00:28:30,000 --> 00:28:34,959
Speaker 5: for other entrants to cut into CME margins or volumes.

527
00:28:35,520 --> 00:28:39,120
Speaker 5: So I'm trying to ask questions that are going to

528
00:28:39,120 --> 00:28:40,560
Speaker 5: create some tension around the table.

529
00:28:40,600 --> 00:28:43,440
Speaker 3: Next week, Yeah still here, I mean, oh yeah, you.

530
00:28:43,400 --> 00:28:45,520
Speaker 2: Should hear what Terry said about Howard Lutnick.

531
00:28:47,040 --> 00:28:48,000
Speaker 5: It'll be on the podcast.

532
00:28:48,040 --> 00:28:49,360
Speaker 3: I'm sure I can.

533
00:28:49,480 --> 00:28:53,000
Speaker 4: I could probably repeat it without having heard it. So

534
00:28:55,280 --> 00:28:58,120
Speaker 4: once you have a liquid market in something, it becomes

535
00:28:58,120 --> 00:29:01,280
Speaker 4: a natural monopoly. It's very hard to move that to

536
00:29:01,840 --> 00:29:05,520
Speaker 4: a different venue. It's happened before. I was living in

537
00:29:05,560 --> 00:29:10,680
Speaker 4: London in the mid nineties and the bond futures were

538
00:29:11,560 --> 00:29:13,520
Speaker 4: on the floor of the Life. It was this huge

539
00:29:13,600 --> 00:29:16,920
Speaker 4: trading pit with a bunch of guys pushing and shoving,

540
00:29:17,120 --> 00:29:22,240
Speaker 4: and over the course of twelve months, the DTV now

541
00:29:22,240 --> 00:29:25,680
Speaker 4: called the UX was able to move the entire bond

542
00:29:25,720 --> 00:29:30,040
Speaker 4: futures complex onto the computer on a different exchange. Now,

543
00:29:30,280 --> 00:29:34,040
Speaker 4: I mean they gave hefty incentives to people. I think

544
00:29:34,080 --> 00:29:36,240
Speaker 4: they went to all the German banks and they said,

545
00:29:36,560 --> 00:29:39,640
Speaker 4: don't you dare trade on Life anymore. So it's possible,

546
00:29:40,120 --> 00:29:43,680
Speaker 4: but I think that these things are generally I don't

547
00:29:43,720 --> 00:29:47,640
Speaker 4: think that it's really a regulatory issue that causes them

548
00:29:47,680 --> 00:29:49,800
Speaker 4: to be sticky. I think it's more just kind of

549
00:29:50,040 --> 00:29:52,360
Speaker 4: a natural state of affairs.

550
00:29:52,040 --> 00:29:55,640
Speaker 2: Network effect, I guess. So our theme for this evening

551
00:29:55,720 --> 00:29:57,840
Speaker 2: is obviously the future of trading, and one of the

552
00:29:57,840 --> 00:30:00,080
Speaker 2: things that seems to be happening is the sort of

553
00:30:00,120 --> 00:30:04,760
Speaker 2: intermingling of professional and retail trading. And we again talked

554
00:30:04,800 --> 00:30:06,760
Speaker 2: about that with Terry. I'm sure we're about to talk

555
00:30:06,800 --> 00:30:11,160
Speaker 2: about it with the Calhi CEO. But from your perspective,

556
00:30:11,320 --> 00:30:14,560
Speaker 2: and again, you started this career back when I don't

557
00:30:14,560 --> 00:30:17,920
Speaker 2: think there were any retail traders doing day trading. Really,

558
00:30:18,600 --> 00:30:21,200
Speaker 2: how has that changed the way you think about trading?

559
00:30:21,200 --> 00:30:23,719
Speaker 2: And can you envision a future where I don't know,

560
00:30:24,040 --> 00:30:26,360
Speaker 2: AI fires all of us and we're all going to

561
00:30:26,360 --> 00:30:29,280
Speaker 2: be just day trading from home as an insurance policy,

562
00:30:29,520 --> 00:30:32,080
Speaker 2: Robinhood is really a full employment program.

563
00:30:31,800 --> 00:30:33,280
Speaker 5: Maybe for us workers.

564
00:30:34,160 --> 00:30:37,280
Speaker 4: Yeah, So, I mean that is a thesis that I

565
00:30:37,400 --> 00:30:41,760
Speaker 4: have heard, is that what's happening is a bunch of

566
00:30:42,440 --> 00:30:48,080
Speaker 4: relatively successful people are losing their jobs and they're retiring,

567
00:30:48,520 --> 00:30:52,160
Speaker 4: but in their retirement they decide to just manage their

568
00:30:52,200 --> 00:30:56,520
Speaker 4: portfolios on robinhood And so there's this surge in trading

569
00:30:56,560 --> 00:31:00,960
Speaker 4: activity that wouldn't have happened ten years ago, and it's

570
00:31:01,000 --> 00:31:02,800
Speaker 4: only going to grow from here.

571
00:31:03,320 --> 00:31:04,680
Speaker 3: And I don't know, maybe that's right.

572
00:31:05,200 --> 00:31:08,160
Speaker 5: It feels to me like culturally, because you're talking about, right,

573
00:31:08,160 --> 00:31:10,800
Speaker 5: why have prediction markets taken off when they've been around

574
00:31:10,800 --> 00:31:13,400
Speaker 5: for over twenty years. I think I first heard about

575
00:31:13,400 --> 00:31:15,440
Speaker 5: them in like two thousand and two or two thousand

576
00:31:15,440 --> 00:31:17,680
Speaker 5: and three. They've suddenly taken off. There was never a

577
00:31:17,680 --> 00:31:20,880
Speaker 5: bright line between what's gambling and what's sort of hedging

578
00:31:20,960 --> 00:31:23,840
Speaker 5: or what's trading, but there's clearly whatever line that is

579
00:31:23,960 --> 00:31:28,080
Speaker 5: just feels like it's completely collapsing. Is this good? Do

580
00:31:28,120 --> 00:31:31,600
Speaker 5: you have an opinion? Like should? Is there is? And

581
00:31:31,680 --> 00:31:33,440
Speaker 5: I don't know if any of our opinions matter on

582
00:31:33,480 --> 00:31:36,360
Speaker 5: the question, because it feels like culturally we're entering that

583
00:31:36,560 --> 00:31:39,720
Speaker 5: world where everything will be tradeable on any app and

584
00:31:39,760 --> 00:31:42,080
Speaker 5: there's you know, you're gonna see a price for gold

585
00:31:42,120 --> 00:31:44,920
Speaker 5: futures right next to one day, the line on a

586
00:31:44,920 --> 00:31:46,080
Speaker 5: football match, et cetera.

587
00:31:46,480 --> 00:31:46,760
Speaker 1: Is it?

588
00:31:47,360 --> 00:31:48,360
Speaker 5: Do we want this world?

589
00:31:49,080 --> 00:31:49,160
Speaker 3: So?

590
00:31:50,840 --> 00:31:55,200
Speaker 4: I don't think there's anything particularly wrong with it, But

591
00:31:56,400 --> 00:32:02,320
Speaker 4: I am a little bit confused about whether prediction markets

592
00:32:02,320 --> 00:32:06,480
Speaker 4: and sports are actually consistent with what the Commodity Exchange

593
00:32:06,520 --> 00:32:08,200
Speaker 4: Act says is permissible.

594
00:32:08,800 --> 00:32:11,560
Speaker 3: And so I know that your next your.

595
00:32:11,400 --> 00:32:16,480
Speaker 4: Next guest is perfect is benefits from his ability to

596
00:32:16,560 --> 00:32:19,400
Speaker 4: list these contracts. And and I don't know if you know,

597
00:32:19,480 --> 00:32:22,440
Speaker 4: the CFTC is just kind of asleep, and I know

598
00:32:22,480 --> 00:32:31,040
Speaker 4: they're kind of understaffed now or uh or or maybe

599
00:32:31,120 --> 00:32:35,720
Speaker 4: they've decided that actually these are economically important transactions that

600
00:32:36,000 --> 00:32:37,480
Speaker 4: are consistent with the CEA.

601
00:32:37,880 --> 00:32:38,720
Speaker 3: It's unclear to me.

602
00:32:39,120 --> 00:32:41,640
Speaker 2: All Right, well, we're going to have to leave it there.

603
00:32:41,800 --> 00:32:45,720
Speaker 2: But Don Wilson, founder and CEO of dr W, thank

604
00:32:45,720 --> 00:32:46,720
Speaker 2: you so much for being here.

605
00:32:47,000 --> 00:32:48,880
Speaker 3: Really appreciated, Thank you for having me.

606
00:33:02,880 --> 00:33:07,040
Speaker 2: That was our conversation with DRW founder and CEO Don Wilson,

607
00:33:07,080 --> 00:33:11,160
Speaker 2: recorded live on stage in Chicago. I'm Tracy Alloway. You

608
00:33:11,160 --> 00:33:13,320
Speaker 2: can follow me at Tracy Alloway.

609
00:33:13,000 --> 00:33:15,880
Speaker 5: And I'm Jill Wisenthal. You can follow me at the Stalwart.

610
00:33:16,280 --> 00:33:19,440
Speaker 5: Follow our producers Carmen Rodriguez at Carmen armand dash Ol

611
00:33:19,440 --> 00:33:22,440
Speaker 5: Bennett at Dashbot and kil Brooks at Kilbrooks. Form more

612
00:33:22,480 --> 00:33:24,960
Speaker 5: odd lots content go to Bloomberg dot com slash odd

613
00:33:25,000 --> 00:33:27,840
Speaker 5: Lots with the daily newsletter and all of our episodes.

614
00:33:28,040 --> 00:33:29,960
Speaker 5: You can chat about all of these topics twenty four

615
00:33:29,960 --> 00:33:33,920
Speaker 5: to seven in our discord Discord dot gg slash online.

616
00:33:33,880 --> 00:33:35,880
Speaker 2: And if you enjoy odd Lots, if you like it

617
00:33:35,920 --> 00:33:38,120
Speaker 2: when we do these live shows, then please leave us

618
00:33:38,240 --> 00:33:41,640
Speaker 2: a positive review on your favorite podcast platform. And remember,

619
00:33:41,720 --> 00:33:44,239
Speaker 2: if you are a Bloomberg subscriber, you can listen to

620
00:33:44,360 --> 00:33:47,480
Speaker 2: all of our episodes absolutely ad free. All you need

621
00:33:47,520 --> 00:33:50,520
Speaker 2: to do is find the Bloomberg channel on Apple Podcasts

622
00:33:50,520 --> 00:34:10,000
Speaker 2: and follow the instructions there. Thanks for listening in