YouTube2Text

Michael Dell on the 3,000-Pound Liquid-Cooled Racks Powering AI Data Centers — Transcript

by WSJ Podcasts · 1,229 words · 198 segments · language en · Watch on YouTube

Full transcript

  1. 0:00Elon is probably the most demanding
  2. 0:02customer. No, he's definitely He's
  3. 0:03definitely the most demanding customer
  4. 0:05there there is.
  5. 0:06>> Dell is best known as the king of
  6. 0:07direct-to-consumer PCs in the '90s and
  7. 0:10early 2000s.
  8. 0:11>> Dude, you're getting a Dell.
  9. 0:13>> The computer giant is now one of the
  10. 0:14most important manufacturers of the
  11. 0:16infrastructure behind the AI boom.
  12. 0:19>> Dell entered the server business in
  13. 0:211989.
  14. 0:22>> Wow.
  15. 0:23>> When I was 24 years old. So, we've been
  16. 0:25at it It didn't actually work the first
  17. 0:27time. Okay, it didn't It started working
  18. 0:29around the mid-'90s. You know, of course
  19. 0:31we got into storage shortly thereafter.
  20. 0:34Then we acquired EMC. And so, we've
  21. 0:37actually been in this business a long
  22. 0:38time and we are number one at it, right?
  23. 0:41And so,
  24. 0:42>> You You mean by
  25. 0:44like revenues, by volume?
  26. 0:46>> Exactly. And so, it shouldn't be too
  27. 0:47surprising that we're also number one in
  28. 0:50AI infrastructure.
  29. 0:51>> Dell did half of the servers in the
  30. 0:54Colossus one AI supercomputer for
  31. 0:57SpaceX. And you did it really fast,
  32. 0:59right? I think that was a record-fast
  33. 1:01build. It was like 6 weeks or something.
  34. 1:03>> Elon is probably the most demanding
  35. 1:05customer. No, he's definitely [laughter]
  36. 1:06He's definitely the most demanding
  37. 1:07customer there there is.
  38. 1:09>> Yeah. Was he like Was it Was there a
  39. 1:11bonus involved? Get this done in 6 weeks
  40. 1:13and
  41. 1:14>> I don't know if there was a bonus
  42. 1:15involved, but they have
  43. 1:18rewarded us with additional business
  44. 1:20because we've done a great job and
  45. 1:22that's pretty much the pattern.
  46. 1:24>> Dell has about 40% of the market for
  47. 1:26sales of powerful servers to companies
  48. 1:27in North America, but it also has stiff
  49. 1:30competition from number two Supermicro
  50. 1:32as well as HPE and Lenovo. And AI
  51. 1:36hyperscalers like Meta, Microsoft,
  52. 1:38Amazon, and Google design and build many
  53. 1:40of their own servers.
  54. 1:42>> A substantial portion of your business
  55. 1:44remains sales to I mean, companies like
  56. 1:47Dow Jones, where we are, right? Like We
  57. 1:49walked We toured the newsroom. You saw
  58. 1:50the Dell monitors, [laughter] the the
  59. 1:52Dell PCs, all the rest. That is still
  60. 1:54tens of billions in revenue for you
  61. 1:56every year. But the the growth driver is
  62. 2:00like to me this very peculiar almost
  63. 2:02alien object which is the I call it the
  64. 2:04AI supercomputer. 6 ft tall, it's 2 ft
  65. 2:08wide, it's 3 ft deep, but it weighs
  66. 2:093,000 lb and it draws as much power as a
  67. 2:14hundred typical US households or maybe
  68. 2:16two or three hundred if it's like a
  69. 2:18really high-end. And this is the brains
  70. 2:21of these new AI data centers.
  71. 2:23>> These systems are incredibly
  72. 2:24complicated. It's not just a rack with a
  73. 2:27bunch of chips in it, right?
  74. 2:29>> No, it's full of plumbing now.
  75. 2:31>> Well, it's you know, literal plumbing.
  76. 2:32Millions of miles of fiber optic cable
  77. 2:36and engineering the liquid cooling to be
  78. 2:40super power efficient, the storage, the
  79. 2:42networking, the deployment,
  80. 2:45you know, teams that come in and make
  81. 2:47these things all work and then keep them
  82. 2:48running all the time.
  83. 2:50>> Right. Because each one of these boxes
  84. 2:51is like six figures.
  85. 2:54>> They're they're they're pretty
  86. 2:55expensive, right? And and and uh we we
  87. 2:58have learned a lot about how to quickly
  88. 3:01deploy these because we have this
  89. 3:04demand.
  90. 3:05>> Let's talk about lessons from your
  91. 3:07[clears throat] big like neo cloud like
  92. 3:09hyper scalar customers, you know, the
  93. 3:11the folks that are standing up these
  94. 3:12these big data centers that we read
  95. 3:13often about in the news. You're
  96. 3:15transferring these lessons now because
  97. 3:17companies are coming to you and saying
  98. 3:19we want our own AI which means we have
  99. 3:22to have our own AI infrastructure,
  100. 3:24right?
  101. 3:24>> Companies want control of their data,
  102. 3:27their their economics, their kind of
  103. 3:30privacy, their security. So, let's say
  104. 3:32you have a hundred factories. You're a
  105. 3:34manufacturer.
  106. 3:36You actually want to have the AI inside
  107. 3:39each factory, right? Doing inference to
  108. 3:41help you be very effective. And so, yes,
  109. 3:44we're we're building these out. It took
  110. 3:46us eight quarters to have 3,200
  111. 3:49customers building these Dell AI
  112. 3:51factories.
  113. 3:52And in the last three quarters we've
  114. 3:54added 3,300 customers. So,
  115. 3:56>> So, it's doubled. It's it's more than
  116. 3:57doubled 6,500.
  117. 3:59>> And what's also interesting is when you
  118. 4:01look inside these companies and actually
  119. 4:03what they're doing with the
  120. 4:06with their data. They're just kind of at
  121. 4:08the beginning of figuring out how to do
  122. 4:10this.
  123. 4:11>> You're one of the few CEOs who was also
  124. 4:15a CEO of well, the same company
  125. 4:17obviously during the dot-com bubble.
  126. 4:20We're in the middle of an infrastructure
  127. 4:22buildout that according to economic
  128. 4:24historians is bigger than any in
  129. 4:26history. Like as a percentage of GDP
  130. 4:28it's bigger than the railroad boom. Do
  131. 4:30you have any concerns about
  132. 4:32overcapacity? You know, having seen a
  133. 4:34lot of these waves over the last 42
  134. 4:37years, we take a pretty measured view of
  135. 4:40this.
  136. 4:41And we think about the risks and the
  137. 4:44challenges and how we balance all that.
  138. 4:46You know, having seen the the dot-com
  139. 4:50period
  140. 4:51>> Which Dell survived, not just intact but
  141. 4:53pretty well.
  142. 4:54>> Sure. And and and you know, our plan is
  143. 4:56to survive this as well and and to come
  144. 4:58out stronger
  145. 4:59>> Are there similar strategies? I mean
  146. 5:01Dell famously you've always kept
  147. 5:02inventories low, you were building on
  148. 5:03demand. Are you still doing that?
  149. 5:05>> Sure. I mean we we're we're doing it to
  150. 5:07the best we can, right? There are some
  151. 5:09new challenges that this presents.
  152. 5:12But if you look at our cash conversion
  153. 5:14cycle, it's still negative and we, you
  154. 5:16know, have an incredible focus on a lot
  155. 5:19of the fundamentals.
  156. 5:21>> Well, let's talk about those challenges.
  157. 5:22I think on earnings calls some of the
  158. 5:24other executives at Dell have said, you
  159. 5:26know, our number one challenge is DRAM
  160. 5:28DRAM DRAM. is memory, memory chips.
  161. 5:31>> Right.
  162. 5:31>> And of course this is a problem that
  163. 5:33everybody has. Like if you're trying to
  164. 5:34buy a laptop, this is a problem.
  165. 5:36If you're trying to buy a PlayStation,
  166. 5:38this is a problem. If you're trying to
  167. 5:39buy phones, the price of those are going
  168. 5:40up.
  169. 5:41>> Cars, right? Because they need memory.
  170. 5:43Do you see that bottleneck easing at
  171. 5:45all?
  172. 5:45>> I think it's going to get worse in 2027
  173. 5:48based on what I see right now.
  174. 5:49>> crunch.
  175. 5:50>> Yes. And and the reason is that it takes
  176. 5:54about 4 years to build a new memory
  177. 5:56factory. There are a number of them
  178. 5:58under construction. You know, Micron for
  179. 6:00example, we're a big Micron customer.
  180. 6:02>> Right. Like they've pledged $30 billion
  181. 6:03dollars a
  182. 6:05construction here in the US.
  183. 6:06>> Exactly. And And those factories will
  184. 6:09start producing in 2029.
  185. 6:12Okay.
  186. 6:12>> And who knows where [laughter]
  187. 6:13everything else will be then.
  188. 6:14>> And so, maybe maybe 2029 everything will
  189. 6:17be fine, right?
  190. 6:19What I see is the way people are using
  191. 6:23this technology now is very different
  192. 6:25than it was even a year or two ago.
  193. 6:27>> With AI, you mean?
  194. 6:28>> With agents. Yes, exactly. And so, the
  195. 6:32demand is growing faster than supply. It
  196. 6:35takes 4 years to build these factories,
  197. 6:38so I don't think that the shortage is
  198. 6:41going to quickly resolve itself.

About this transcript

This page contains the full transcript of Michael Dell on the 3,000-Pound Liquid-Cooled Racks Powering AI Data Centers by WSJ Podcasts, generated from the public captions YouTube serves with the video. The transcript has 1,229 words across 198 segments, with the original timestamps preserved so you can click any line to jump to that moment in the embedded player.

What you can do with it

Use the transcript to take notes, quote the speaker, build a study guide, generate a summary with ChatGPT or Claude via the YouTube Summary tool, or export it as a timed subtitle file with YouTube to SRT. You can also re-open it in the transcriber to translate the transcript into 100+ languages.

Free YouTube transcript tool

YouTube2Text is a free YouTube transcript generator — no signup, no daily limit. Paste any YouTube link and get the full transcript instantly, with timestamps, click-to-jump, translation to 100+ languages, AI prompts for ChatGPT, Claude, and Gemini, and exports to TXT, SRT, VTT, or Markdown.