CoreWeave and PwC: Partnering to Multiply Impact

1
00:00:00,040 --> 00:00:06,920
The pace of innovation is unprecedented, and alliances like yours, with CoreWeave and the other

2
00:00:06,920 --> 00:00:13,040
technology players is critical. It's not just a story about technology, it is a story about

3
00:00:13,240 --> 00:00:18,480
creating that trust dynamic. I think we're up to like, something incredible. 65 to 70% of total

4
00:00:18,480 --> 00:00:23,160
compute power is actually driven towards AI. It's really a natural fit to have organizations

5
00:00:23,160 --> 00:00:27,440
like ours working together. But without trust and safety, everything will fall apart.

6
00:00:39,160 --> 00:00:45,200
Hello everyone. Welcome to the AI Cloud Essentials, a podcast series brought to you by CoreWeave.

7
00:00:46,000 --> 00:00:52,560
Partnerships have always driven tech innovation, but in the AI era they are a fundamental necessity.

8
00:00:52,759 --> 00:00:58,000
Scaling enterprise agentic AI requires bridging world class infrastructure with high impact

9
00:00:58,000 --> 00:01:04,069
strategic execution and domain specific transformation. I know PwC has worked with CoreWeave

10
00:01:04,110 --> 00:01:09,509
on corporate readiness and IPO consulting. And today I'm super excited to be joined by Dallas.

11
00:01:09,550 --> 00:01:16,510
We'll be discussing the partnership between CoreWeave and PwC TMT. Welcome, Dallas. Ritu, great to

12
00:01:16,510 --> 00:01:23,510
be here. Thank you. Could you please introduce yourself and PwC TMT to our audience? I'll

13
00:01:23,550 --> 00:01:29,830
be happy to. So I'm Dallas Dolen. I lead the tech, media, and telco business for PwC in the US and

14
00:01:29,870 --> 00:01:35,230
Asia PAC. From a broad perspective, I've been involved with the tech community for well over two decades

15
00:01:35,389 --> 00:01:39,550
and have had the opportunity to work with everybody from the hyperscalers to the neoclouds

16
00:01:39,550 --> 00:01:43,750
and obviously software and hardware and everything in between. And we have a, you know, really

17
00:01:43,790 --> 00:01:49,070
I'd say wide-ranging practice in terms of capabilities, you know, whether it's, you know, the SI

18
00:01:49,070 --> 00:01:52,870
and engineering side of the business all the way to deals and tax and the core things that people

19
00:01:52,870 --> 00:01:59,070
would think about, you know, PwC. Awesome. Great to hear. I'm also an ex-PwC employee, so I'm super

20
00:01:59,070 --> 00:02:03,710
excited to talk to you. The first thought that comes to my mind today as I'm kicking off this

21
00:02:03,750 --> 00:02:09,349
conversation with you is that, you know, we have transitioned very rapidly from traditional gen AI

22
00:02:09,390 --> 00:02:14,509
to agentic AI and these systems that can independently plan, use tools and execute

23
00:02:14,509 --> 00:02:21,110
workflows. There is a tremendous amount of operational risk because things are going to be

24
00:02:21,110 --> 00:02:27,469
autonomous. So then the race stakes in the partnership between these

25
00:02:27,670 --> 00:02:34,549
companies and consulting firms is no longer just beneficial in my opinion. So I have to ask you

26
00:02:34,589 --> 00:02:40,309
that, is that what exact market-forcing function drove this partnership and brought you all

27
00:02:40,350 --> 00:02:45,950
together? Yeah. Well, I think there's a couple of things that that you touched on there. I'd say

28
00:02:45,990 --> 00:02:49,750
first and foremost, I remember, you know, when we first started playing with generative AI a few

29
00:02:49,750 --> 00:02:54,950
years ago, I was actually in an executive MBA classroom on the East Coast, actually in Boston,

30
00:02:54,950 --> 00:03:00,630
where you are. And the professor came in one day and said, hey, I want you to download this app and

31
00:03:00,630 --> 00:03:06,339
see if you can make a cool image with it. Of course, it turned out it was it was GPT 2.0 that

32
00:03:06,339 --> 00:03:11,019
we were playing with and and he wanted us to see like what was capable and like what was coming.

33
00:03:11,020 --> 00:03:15,820
And the class was actually on the future of AI and future technology. Um, so I couldn't have been

34
00:03:15,820 --> 00:03:20,259
better positioned, I guess, to like, see the future from that, you know, vantage point sitting in a in

35
00:03:20,259 --> 00:03:24,419
the MBA classroom. What was notable at the time is, you know, people started thinking about, okay, well,

36
00:03:24,419 --> 00:03:28,979
what is it that that consultants and what is it that it, you know, engineers might be doing in the

37
00:03:28,979 --> 00:03:34,459
future? And this whole concept of a, a prompt engineer almost became a concept overnight. Of

38
00:03:34,460 --> 00:03:39,379
course, as quickly as it became a concept and sort of dissipated, right. And so now we've evolved

39
00:03:39,379 --> 00:03:43,339
probably 2 or 3 steps from there. This concept of forward-deployed engineers has become a really

40
00:03:43,340 --> 00:03:50,099
active, you know, dynamic. I think the real story as it relates to, you know, why, you know, do

41
00:03:50,140 --> 00:03:54,379
people need compute power, and why is it the partnerships between all the way at the the

42
00:03:54,380 --> 00:04:00,659
original origin point around energy? Right. So power generation through to, you know, the chips

43
00:04:00,659 --> 00:04:06,219
and the data centers through to the actual infrastructure, including the cloud capabilities

44
00:04:06,220 --> 00:04:12,379
that that CoreWeave delivers. And then also, you know, to the AI consumption that's happening. Each one

45
00:04:12,380 --> 00:04:17,019
of those steps along the way is some type of arrangement, some type of joint venture, some type

46
00:04:17,019 --> 00:04:21,899
of, you know, fundraising that's happened, certainly some type of engineering, you know, capacity that's

47
00:04:21,900 --> 00:04:26,899
going into play. And over the last three years we've been, to be quite honest, not necessarily

48
00:04:26,899 --> 00:04:31,980
learning as we go, but being really adaptive as we go, saying as that market changes, okay, how are we

49
00:04:32,019 --> 00:04:36,420
going to bring different capabilities and capacities together in order to ensure that our

50
00:04:36,420 --> 00:04:41,219
clients, right, our, our mutual clients are maximizing their opportunities from the things

51
00:04:41,219 --> 00:04:46,540
that can be possibly done, you know, even in and amongst the same. So I would look at the capacity

52
00:04:46,579 --> 00:04:51,500
dynamic that a CoreWeave offers to its customer base. You know, just from a pure compute point of

53
00:04:51,500 --> 00:04:56,099
view or storage or what have you. And then you look at it from our perspective of helping our

54
00:04:56,099 --> 00:05:00,739
clients maximize their outcomes, whether it's a transformation program that they're running or

55
00:05:00,899 --> 00:05:05,339
the implementation of AI. And it's really a natural fit to have organizations like ours

56
00:05:05,340 --> 00:05:09,689
working together. And what's really great about it is I think you're going to see further

57
00:05:09,690 --> 00:05:14,649
development of the capabilities, of competencies, and the things that we end up doing together. And

58
00:05:14,649 --> 00:05:18,689
that's also true amongst the hyperscalers, and it's certainly true amongst the AI native

59
00:05:18,689 --> 00:05:24,488
corporations as well. Yeah. Well said. Uh, Dallas, when you think about two different companies

60
00:05:24,489 --> 00:05:29,609
coming together, and especially in the AI era when you're integrating for velocity, the thought that

61
00:05:29,609 --> 00:05:35,170
comes to, you know, our mind is that how do you align the two distinct organizations, which have

62
00:05:35,170 --> 00:05:40,969
different culture, different decision velocity? Uh, you know, their code bases and stacks can be

63
00:05:40,970 --> 00:05:46,649
different, and they're all trying to remove the friction for the end user. So could you talk to us

64
00:05:46,649 --> 00:05:52,809
about a couple of, you know, challenges that you guys kind of were confronted with or how are you're

65
00:05:52,849 --> 00:05:59,329
navigating these kind of, you know, scenarios and bringing in synergies? Yeah. Well, look, one of

66
00:05:59,329 --> 00:06:05,010
my favorite. Phrases is metrics drive behavior, right? And ultimately, if you have salespeople

67
00:06:05,049 --> 00:06:09,209
being driven by one metric and then you have, you know, an engineering team or a transformation team

68
00:06:09,209 --> 00:06:13,679
or what have you driven by another, the challenge always becomes, you know, the likelihood that

69
00:06:13,680 --> 00:06:18,279
friction will be introduced to the system because a salesperson might be focused on a short term

70
00:06:18,279 --> 00:06:23,520
metric, right? Getting, you know, one more increment of of whatever the unit is that they're selling

71
00:06:23,559 --> 00:06:28,119
sold. And a transformation team may be promising to a customer that they're going to have a

72
00:06:28,119 --> 00:06:33,839
certain outcome. I think so aligning at the at the like, what is our ultimate goal is really critical.

73
00:06:34,000 --> 00:06:39,639
If you think about, you know, the origin points of selling, you know, selling cloud, right. Going back a

74
00:06:39,640 --> 00:06:44,399
few years or even just selling SaaS in general, you know, there was a lot of focus on seat

75
00:06:44,400 --> 00:06:49,799
licenses. There's a lot of focus on selling consumption. It's how much throughput can I get?

76
00:06:50,160 --> 00:06:55,079
What we're seeing is that that largely does not create a sticky customer. Now, you can have an

77
00:06:55,079 --> 00:06:59,799
organization who might be wed to a specific cloud because they have so much volume going through it,

78
00:06:59,799 --> 00:07:03,519
but it doesn't necessarily make them sticky unless there's something about the customer

79
00:07:03,520 --> 00:07:08,279
relationship, you know, whether, again, it's cloud or SaaS, that people say, oh, I really want to be a

80
00:07:08,279 --> 00:07:14,119
part of that. And oh, I need all 1000 of my people, 10,000 of my people were, in the case of PwC. All

81
00:07:14,119 --> 00:07:20,238
375,000 of our people, you know, to be using a specific stack or a specific unit or a specific

82
00:07:20,239 --> 00:07:26,040
capability. And so what, you know, what I find is the first thing first is like the the end goal is

83
00:07:26,040 --> 00:07:29,959
having a satisfied customer. And if you want to have a satisfied customer, you're going to have to

84
00:07:29,959 --> 00:07:32,999
ask them, hey, what are they trying to get out of the transformation? Are they just looking for

85
00:07:32,999 --> 00:07:36,519
compute? Or are they actually looking for some sort of outcome as it relates to the AI that

86
00:07:36,519 --> 00:07:40,519
they're implementing? You know, they're trying to build a customized stack with a, you know, harness

87
00:07:40,559 --> 00:07:45,359
operating on, you know, CoreWeave, you know, componentry, and then, you know, building an open

88
00:07:45,359 --> 00:07:50,119
model on top of that. Great. But also. What are they trying to do with that? Is it going to be a speed

89
00:07:50,119 --> 00:07:54,719
oriented dynamic that they're going for? Is it going to be quality. Is it. Is it reliability

90
00:07:54,720 --> 00:07:59,920
reliance? Is it security? All those things have to be lined up and it still is a challenge today. I

91
00:07:59,920 --> 00:08:04,519
think every time we go into a room, you know, we're always thinking with the metric that we have. And

92
00:08:04,519 --> 00:08:09,119
what I've been really coaching my teams around is what is the outcome we want to have. We're not

93
00:08:09,120 --> 00:08:13,949
playing a finite game. We're not trying to sell something in the minute. We're trying to sell

94
00:08:13,950 --> 00:08:18,309
something in the long run and really get to that, that infinite dynamic, right? Like, we want

95
00:08:18,350 --> 00:08:23,429
customers to be customers forever. It even goes back to our own brand when we think about PwC. You

96
00:08:23,429 --> 00:08:27,190
know, we're a trust brand, right? It's the audit and tax, you know, relationships we've had with some

97
00:08:27,190 --> 00:08:32,349
companies for upwards of well over 100 years. Well, the reason they come back is not because we solve

98
00:08:32,349 --> 00:08:36,389
a problem that happens in the quarter. It's because we're solving an enterprise level problem.

99
00:08:36,389 --> 00:08:41,429
And they say, oh, I like that. I want to be back. And, you know, with them in the future, I know they'll

100
00:08:41,469 --> 00:08:46,270
be there for us. The same thing is absolutely true, I think now for the CoreWeaves of the world. It's

101
00:08:46,270 --> 00:08:50,390
absolutely true for the hyperscalers and certainly true for the AI companies as well.

102
00:08:50,430 --> 00:08:54,949
Otherwise, people will engage in short term behaviors which, you know, ultimately will not

103
00:08:54,949 --> 00:09:00,869
result in stickiness with customers. Yeah. You know, I spend a lot of time with the end users and the

104
00:09:00,909 --> 00:09:06,109
C-suite there. And one of the things that has come to my attention consistently is that, you know,

105
00:09:06,150 --> 00:09:11,349
historically, a major barrier to AI adoption was that people were buying tools exactly the way

106
00:09:11,349 --> 00:09:16,059
they bought their standard software licenses. And then they realized that without deep

107
00:09:16,060 --> 00:09:21,739
organizational alignment, nothing changes. And, you know, the Technology Consulting Alliance is really

108
00:09:21,780 --> 00:09:27,139
kind of, you know, targeting this, breaking the pilot purgatory. Right. So you need workflow

109
00:09:27,139 --> 00:09:32,339
redesigns and change management. So I would love to ask you that, you know, when the two companies

110
00:09:32,340 --> 00:09:38,340
come together like you and CoreWeave, what is the multiplier effect? What is the joint good work

111
00:09:38,379 --> 00:09:44,019
that you deliver to the end user? Yeah. Great question. I mean, you know, speaking from a, you know,

112
00:09:44,060 --> 00:09:47,579
from a broader business perspective, you know, whether it's satisfying the, you know, the

113
00:09:47,580 --> 00:09:50,979
investors of the organization or customers of the organization or even the employees of the

114
00:09:50,979 --> 00:09:57,260
organization. I mean, the underlying goal, right, is, is that, you know, whomever we're trying to ensure

115
00:09:57,299 --> 00:10:02,299
gets the output that they want, they're getting it. And just to go back to the prior commentary, right.

116
00:10:02,300 --> 00:10:07,459
So that's that's number one. Um, in order to get to that point though, like I, I really stop and focus

117
00:10:07,460 --> 00:10:14,059
on the, you know, how do you line up the one, two, three order effects? Like, what's the objective? You

118
00:10:14,059 --> 00:10:19,459
know, one week out, four weeks out, you know, three months out. You know, how do we make sure that when

119
00:10:19,460 --> 00:10:24,219
you get to those points in time, you're looking at, you know, what you've done over that period of

120
00:10:24,219 --> 00:10:30,020
time and saying, aha, we've we've achieved the outcome we wanted to achieve. And so as it relates

121
00:10:30,020 --> 00:10:35,579
to, you know, the neoclouds in particular, it's really interesting. Right? I think the the biggest

122
00:10:35,580 --> 00:10:42,299
thing is that flexibility in supplying incredible compute capacity and capabilities, right,

123
00:10:42,340 --> 00:10:47,499
especially on the AI front. But again other areas too, to their end customers. And we're looking at it

124
00:10:47,499 --> 00:10:52,459
saying, okay, great. Like what is it that needs to happen in the in the short to medium term in

125
00:10:52,460 --> 00:10:57,099
order to get there? Because that is I think, the sweet spot for, you know, the neocloud business.

126
00:10:57,099 --> 00:11:01,620
And so it's really aligning on that. If we can find that like sweet spot for customers looking

127
00:11:01,620 --> 00:11:06,659
for X and the ability to deliver the technology at Y, you know, that's going to give us the outcome

128
00:11:06,659 --> 00:11:10,579
that the customers are happy with. But it is like finding that alignment spot. If somebody's

129
00:11:10,580 --> 00:11:15,219
saying, I'm going on a ten year journey, okay, fine, there might be some. But like, let's find the short

130
00:11:15,260 --> 00:11:19,369
term things in there. Like if we just want to do it a ten year journey thing. It's going to be

131
00:11:19,369 --> 00:11:24,449
harder, probably to find those immediate, you know, solve-fors that a PwC and CoreWeave are going to be, you

132
00:11:24,449 --> 00:11:28,889
know, as successful at. So we really think about it from that that medium to short term basis as we

133
00:11:28,890 --> 00:11:33,450
look at these, these opportunities and try to solve for the same. That's a very important, uh, you

134
00:11:33,450 --> 00:11:39,049
know, approach to kind of deal with the complexities that we are confronted with, right? Um,

135
00:11:39,090 --> 00:11:44,009
I want to go back to your comment earlier that you know why people choose PwC because of the

136
00:11:44,010 --> 00:11:50,129
trust aspect, right? And when you think about, you know, the AI workloads today, trust and safety is

137
00:11:50,130 --> 00:11:55,729
becoming more complex, more interconnected and more central to business resilience. Right. People

138
00:11:55,729 --> 00:12:01,009
are talking about AI playing a critical role in the business transformation. But without trust and

139
00:12:01,010 --> 00:12:07,809
safety, everything will fall apart. So when you think about advising the end users, as we wrap up

140
00:12:07,809 --> 00:12:14,369
this conversation, your core innovation playbook and your, you know, what is your future look like

141
00:12:14,409 --> 00:12:21,400
in terms of you guys working together to kind of transform the end user, you know, business, and

142
00:12:21,400 --> 00:12:27,359
their journey towards AI. Yeah, well, I'll tell you right. Statistics don't lie as it relates to what

143
00:12:27,360 --> 00:12:32,120
the concerns are. And why is trust important? I mean, look at right now, I think recent studies

144
00:12:32,120 --> 00:12:36,599
have have demonstrated that well over half the internet traffic that's out there is actually

145
00:12:36,599 --> 00:12:43,359
automated bot traffic. Yeah. And if you look at like what AI is doing, AI is really enhancing the

146
00:12:43,360 --> 00:12:47,439
amount of compute that's going to be going into that space. I think we're up to like something

147
00:12:47,440 --> 00:12:53,759
incredible. 65 to 70% of total AI compute, uh, total compute power is actually driven towards AI.

148
00:12:54,320 --> 00:12:59,040
The problem they're in is, you know, do you know what your agent is doing? Or do you know what

149
00:12:59,040 --> 00:13:03,359
other agents are doing within your ecosystem? If you're selling something through a platform or,

150
00:13:03,599 --> 00:13:08,199
you know, even if you're just doing your your mundane tasks and your job, you know, are you able

151
00:13:08,199 --> 00:13:13,399
to control the things that that that agent or other AI you know, agent might be doing for you?

152
00:13:13,479 --> 00:13:19,199
And so and we sort of look at that as the premise point to say, okay, there's a whole, you know, a

153
00:13:19,200 --> 00:13:23,519
whole lot of people out there who are beginning to play with this technology, who probably aren't

154
00:13:23,520 --> 00:13:27,919
realizing how much risk they are inducing into the system, notwithstanding the headlines that

155
00:13:27,919 --> 00:13:32,839
tell you there's a lot of risk, because the amount of the amount of, you know, cyber, you know, hacking

156
00:13:32,840 --> 00:13:36,999
and the amount of, you know, unfortunate results that we've seen. There was a famous case two weeks

157
00:13:36,999 --> 00:13:41,599
ago where the French tax authority disclosed that they had had a significant breach. You know, those

158
00:13:41,600 --> 00:13:46,399
types of situations are obviously very concerning. So we could look at it from a couple of different

159
00:13:46,400 --> 00:13:49,960
ways. And one of them is sort of like the natural thing that we've always done, which is, you know,

160
00:13:50,000 --> 00:13:53,479
hey, we're they're trying to help people think about like, what is it they should be doing to

161
00:13:53,479 --> 00:13:58,560
ensure they have secure, you know, systems and secure networks and, you know, just just the normal

162
00:13:58,560 --> 00:14:02,718
run of the mill process and technology front. But where I think it gets really interesting for us

163
00:14:02,719 --> 00:14:06,919
from a relationship point of view with the hyperscalers, with the neoclouds, with the likes

164
00:14:06,919 --> 00:14:13,879
of CoreWeave is, you know, it's not just a story about technology, it is a story about creating

165
00:14:13,880 --> 00:14:19,159
that trust dynamic for you to say, okay, if you're going to go put something into my ecosystem to

166
00:14:19,159 --> 00:14:24,589
run, you know, my applications on, on your domain. How do I trust it? Like, what's going on is

167
00:14:24,590 --> 00:14:28,709
actually something that my customers are going to get benefit from. And am I putting too much risk

168
00:14:28,709 --> 00:14:33,669
or inducing too much risk into the system by allowing this to happen? So, you know, it is really

169
00:14:33,669 --> 00:14:38,309
a soup to nuts conversation that we're beginning to have with the CISOs, the CIOs and CTOs of these

170
00:14:38,309 --> 00:14:43,349
organizations saying, oh, I should say, by the way, at the board level two, because the boards really

171
00:14:43,350 --> 00:14:47,149
care about this, right? Like the worst call you can get as a board member is you've had a cyber

172
00:14:47,149 --> 00:14:52,229
breach. And oh, by the way, all of your customer data or all of your employee data is now out on

173
00:14:52,229 --> 00:14:56,590
the dark web, or it's with some hacker who's extorting you for money. So when you start

174
00:14:56,590 --> 00:15:01,229
thinking about, like where we come into the to the fray and looking at the risk, the regulatory, the

175
00:15:01,229 --> 00:15:05,589
cyber response, you know, what are the things you could do right now to mitigate those things? And

176
00:15:05,590 --> 00:15:10,749
how do we work with the providers? Right, both from an infrastructure perspective and otherwise, you

177
00:15:10,750 --> 00:15:15,309
know, to ensure that that risk is as mitigated as possible. And combining and this is the most

178
00:15:15,309 --> 00:15:20,669
important part, combining the critical software and people features, right, including process

179
00:15:20,670 --> 00:15:26,429
controls, etc. that allow you to mitigate risk in the system. It's not just turning something on that

180
00:15:26,469 --> 00:15:30,869
Causes it to work. And for us, working with the likes of CoreWeave and others, you know, it's that coordination

181
00:15:30,869 --> 00:15:35,589
feature. It's that high degree of capability and the high reliance that people are willing to put

182
00:15:35,589 --> 00:15:39,468
in us because we've gone through it and we've done it to ourselves. Right. That's a real critical

183
00:15:39,469 --> 00:15:44,549
piece as well, you know, doing it to ourselves, using technology from CoreWeave and others to ensure

184
00:15:44,549 --> 00:15:49,509
that we offer the best experience for our customers, you know, and for our, our, our employees,

185
00:15:49,510 --> 00:15:54,310
too, because we care about both of them equally. Yeah. Thank you. Dallas, I'm sure our audience is

186
00:15:54,310 --> 00:15:59,349
going to find this discussion very, very helpful and fascinating because we all know that the pace

187
00:15:59,349 --> 00:16:05,829
of innovation is unprecedented, and alliances like yours with CoreWeave and the other

188
00:16:05,830 --> 00:16:11,909
technology players, is critical to kind of protect users, govern the emerging technologies, and build

189
00:16:11,909 --> 00:16:17,909
that trust and foundation that is needed to scale AI adoption. Thank you so much for joining us

190
00:16:17,909 --> 00:16:20,908
today. Absolutely. Thank you very much.

‍