The Information's TITV
00:32:15 · CJ Muse
[00:32:15] S4: I think they talked about 80% of their revenues in calendar twenty seven led by the data center. And so it highlights robust server CPU growth and robust AI accelerator GPU growth. And so I think the market was very concerned about what their business would look like over the next two to three quarters following Intel's miss, but they had so much pent up demand. They are the number three customer at TSMC. So NVIDIA, Apple, then AMD. They are an absolute priority. Right? TSMC views Intel as a competitor. They are going to enable AMD, so they're going to get them wafers. And I do think some of the three nanometer smartphone wafers are being pushed to AMD. AMD had the foresight, you know, to think that, you know, servers could grow at least a 100% next year. And so they got in line early for substrates. And lastly, would say what could drive incremental growth next year above that 100% is that the chiplets for logic tiles are fungible. So they can convert PC, you know, logic tiles to servers, and that will give them incrementally more upside if the weakness in PCs sustains. So that would be point one. Point two is really, you know, that they're doing well on the accelerator front. Obviously, you look at the revenue number, you know, versus what NVIDIA or Broadcom's doing, it's much smaller. But but that's kind of the path where, again, computers sold out. Whoever has access to wafers at TSM is selling them. And I think that that, you know, bodes well for their business. And the last note I would say is very interestingly, AMD's server CPU business in calendar twenty seven will exceed Intel's business for the first time in history. That is a huge change statement and really highlights, I think, Lisa Su's execution and strategy at AMD and the partnership with TSM and the foresight, their designs. You know, they're doing a great job.
00:34:23 · Akash Pasricha, CJ Muse
[00:34:23] S0: This is the this is the... So the server CPU business for AMD, the direct comp for Intel would be the data center CPU segment? [00:34:31] S4: Yeah. They're Deion data center CPU. Exactly. Right. So Right. You know, you're talking 36,000,000,000 AMD, 32,000,000,000 Intel What in calendar twenty [00:34:42] S0: about PC markets? What did we learn from it? I mean, Intel obviously gave us a hint there with their results. So did we get any kind of clarity on how long they are expecting the PC market to be down here? [00:34:58] S4: Yeah. I mean, they guided, call it down 22% sequentially Q4. They're guiding seasonal, which is like down 10, 12, 14% q one sequentially. And then talking about overall PC units, call it down 10 to 15% in calendar twenty seven, and that they would do better than that given their share gains, as they focus on the higher end mix. So I think what we learned I think is that there's more inventory downstream of Intel parts than AMD. And so the correction at Intel is gonna last longer. Whereas I think AMD has done a better job of kind of predicting the slowdown and focusing their efforts more on the data center than client. And I think, you know, for Intel, the challenge for them, they make all of their data center server CPUs internally. They produce PC CPUs both internally and at TSMC. And so I think they were slow to migrate wafers over the server world, and I think that's, you know, that's haunting them. [00:36:01] S0: Did they make any comments about power shortages last night on the call? [00:36:08] S4: No, not at all. I think...
00:36:08 · CJ Muse, Akash Pasricha
[00:36:08] S4: No, not at all. I think... [00:36:13] S0: Because One of the questions we've been wondering here is all these forecasts accelerating growth at AMD is forecasting NVIDIA. I don't know what it was, 70%, I think, growth they forecast last time for '27 or '28. I can't remember exactly. Anyway, the question is- [00:36:31] S4: It's now 110%. Right. [00:36:35] S0: Right. Whatever the... I mean, the forecasts are big anyway you slice it. So the question is, will they be able to meet the forecast if they can't get power to the chips at all, the customer end? Is that a concern at all? [00:36:49] S4: I mean, it's a market concern for sure. I would say to the question, AMD management highlighted that it's not a worry. We are in the realm of project finance. You are lining up obviously permitting Shell's power concurrent or even before the timing of the purchase of accelerators. Obviously, it's a concern, but I have not heard in my work that there's a hole there for them. But I know that the market is concerned about that. [00:37:27] S0: Great. Okay. Well, CJ, I want to thank you for coming on. That is CJ Muse, Senior Managing Director and Semiconductor Analyst at Cantor here on TITV. Our next segment is with our partner Radiant. Radiant is the AI infrastructure platform backed by Brookfield. Radiant is one partner among others on a new data center project in The UK For a conversation about that and more about this moment for Neo Clouds broadly, I want to bring on Mahdi Yahya, CEO of Radiant. Mahdi, welcome to the show. It's great to have you here.
00:37:27 · Akash Pasricha, Mahdi Yahya, CJ Muse
[00:37:27] S0: Great. Okay. Well, CJ, I want to thank you for coming on. That is CJ Muse, Senior Managing Director and Semiconductor Analyst at Cantor here on TITV. Our next segment is with our partner Radiant. Radiant is the AI infrastructure platform backed by Brookfield. Radiant is one partner among others on a new data center project in The UK For a conversation about that and more about this moment for Neo Clouds broadly, I want to bring on Mahdi Yahya, CEO of Radiant. Mahdi, welcome to the show. It's great to have you here. [00:38:01] S3: Thanks, Akash. Good to be here. [00:38:03] S0: Always an exciting time to talk about Neo Clouds. I want to talk about the new data center project that you guys have embarked on, but tell us a little bit about your offering at Radiant and how you differentiate from all the other cloud offerings. [00:38:17] S3: So Radiant, we are an AI infrastructure builder and provider, and we are basically operated 100% owned by Brookfield. Radiant was formed through the acquisition of my previous company, Ori Industries, And the core thesis of Radiant was to bring together the four pillars of what we believe to make AI infrastructure successful under one roof, and that is powered land, capital, software, and compute. And at Radiant, we vertically integrate all these four pillars as we build our projects to effectively deliver and deploy AI infrastructure for some of the largest customers in the world today. [00:38:57] S0: And so how do you differentiate yourself then from all these other AI infrastructure companies? [00:39:04] S4: Look,
00:39:04 · CJ Muse, Mahdi Yahya, Akash Pasricha
[00:39:04] S4: Look, [00:39:07] S3: the conversations over the last few months, I would say, it's becoming very apparent that the bottleneck for delivering AI infrastructure today is the ability to execute and the ability to source power and data centers globally. There is no shortage of demand. There is no shortage of capital even. And the differentiation is happening today across the ability to deliver these projects on time and to highest standards. And we, as Radiant, have the capability to actually source power, it's part of our strategy around going power first, around site selections we do, and around how we build our data centers. But also, we have a very long operating history, both as Radiant and previously as Ori, but also through the partnership we have with Brookfield. So together as organizations, we have a longstanding operational and execution history that we can point to, to actually deliver large scale infrastructure at scale today. That's what The we're [00:40:07] S0: Brookfield part of this is I think really interesting because we have talked on this show about power being the bottleneck here. Brookfield, of course, is a leader in that space. Can you give us an example here of specifically how your relationship with Brookfield has helped you access power in a cleaner way, perhaps in an expedited way, perhaps in ways that other facilities may not have access to power?
00:40:07 · Akash Pasricha, Mahdi Yahya
[00:40:07] S0: Brookfield part of this is I think really interesting because we have talked on this show about power being the bottleneck here. Brookfield, of course, is a leader in that space. Can you give us an example here of specifically how your relationship with Brookfield has helped you access power in a cleaner way, perhaps in an expedited way, perhaps in ways that other facilities may not have access to power? [00:40:36] S3: Brookfield is one of the largest asset managers in the world. They're also one of the largest investor in power and renewables. I mean, they have a large platform for renewables investments in the world. And this alone gives us effectively access to a very large portfolio and a large network of power sources and power operations that we can leverage to effectively deliver the compute infrastructure that we are delivering. And as I said earlier, whenever we start with any project, we start with the principle of PowerFest. And our starting point is always around the availability of power, the type of power, and the strategy around power in each site. And from there, we build everything accordingly from that regard. And being part of the Brookfield ecosystem allows us to basically employ this strategy much more effectively than anyone else. [00:41:35] S0: Okay. And I mean, is the power cleaner at all?