In The Arena by TechArena - Kelley Mullick of Avayla on Rack-Scale Design & Hybrid Cooling

Episode Date: December 15, 2025

From OCP Summit 2025, Kelley Mullick joins Allyson Klein and co-host Jeniece Wnorowski for a Data Insights episode on rack-scale design, hybrid cooling (incl. immersion heat recapture), and open stand...ards.

Transcript
Discussion (0)
Starting point is 00:00:00 Welcome to Tech Arena, featuring authentic discussions between tech's leading innovators and our host, Alison Klein. Now, let's step into the arena. Welcome to Tech Arena Data Insights. My name's Allison Cline. It is a Data Insights episode coming to you from the OCP Summit in San Jose, California. And that means I'm here with Janice Norowski. Hey, Janice. How's it go? Hello, I'm. Hey, Allison. So, Janice, what have you thought about OCP Summit this far? It is amazing. It's so much bigger than last year. I know. And the topics are similar to the topic we're going to talk about today, but it's double what it was last year. So let's get into it. What is the topic for today? And who is this sitting in between us? I am very excited about who's sitting in between us. We're honored here today to have Kelly Mullick, who is CEO and founder of Avala. Welcome to the program.
Starting point is 00:00:57 Thank you so much. I'm so happy to be here. on the arena with you two amazing ladies. And we're going to talk today about all things cooling. This has obviously been the number one topic as it was last year into this year. Kelly, we're excited to just get your perspective on some things. Yes, I'm excited to share that. So before we get there, we were talking before we started recording, tell us about your company's name and where it comes from.
Starting point is 00:01:22 I would love to share that. So, Janice, you commented that my company has a beautiful name. The reason is because it was named after. a partner at the time, they had a daughter named Ava, and my daughter, her name is Ila. My oldest daughter, her name is Ila. So we, at the time when we were making this company, we decided to combine the girls' names and we called it Avila. So that's the kind of the rationale of the company's name is something that's meaningful to us. That's fantastic. Love that. We love that. Let's just jump right into it, Kelly. So, you know, the Dana Center industry is rapidly evolving to
Starting point is 00:01:55 support high-density AI workloads. From your point of view, What are some of the biggest challenges that operators face today, especially around thermal management, scalability, and sustainability? Yeah, so really, I can break that down. So from an infrastructure perspective, the number one challenge is absolutely how do we scale these clusters? We heard in the keynotes this morning and from last night that scaling these clusters is something that we've never seen in our industry before. In fact, Dan Rabinovitz from Meta talked about how he could,
Starting point is 00:02:30 compared how we're scaling these huge racks of compute, which normally would have taken, you know, we would do that every couple of years. We're now on a cadence of every year, and it's comparable to, like, consumer electronics. That was a big insight for me. I mean, I know that because I live and breathe in this space, but to have that comparison from someone so senior at Meta, the head of infrastructure at Meta, is a really interesting insight to make that comparison. So scalability at the infrastructure level is one of the key challenges and probably the biggest challenge.
Starting point is 00:03:04 And with that comes smaller challenges. And another one obviously is getting the infrastructure ready for liquid cooling. And so we're seeing such a huge influx of coolant distribution units or CDUs into the market as a result because you have to have a CDU to be able to deliver coolant from the chip to the coolant that then distributes the fluid. That's part of one of the loops that's being built as part of the infrastructure. And so another announcement at OCP was the new work stream as part of cooling environments for the CDU working group and best specifications and best practices. You know, it's funny. I was thinking about this interview, and I know how long you've been in the cooling space. Liquid cooling is something that you've spent a tremendous amount of time and effort at different stops in your career on.
Starting point is 00:03:52 very engaged in the cooling efforts in OCP. Seems like last year, everybody was talking about, yeah, we're going to make this transition to liquid and we need to do some standardization on dielectrics. This year, it feels like we are starting to see that adoption curve and just take flight. How do you see the market advancement and where do you think we are in terms of full transition
Starting point is 00:04:18 of AI infrastructure to liquid? Yeah, so let's talk. a little bit with numbers. So I would say before 2022 and large language models and the advent of generative AI, we were still more than 90% and traditional air-cooled facilities. In roughly two years, we've seen that 10% of liquid cooling grow to about 30%. That is huge. So we're, I would say, almost 30% now because of all of the construction, the builds of new data centers, and also the retrofitting of existing data centers to support AI. So just talking with numbers, that amount of growth into the infrastructure and change rapidly evolving, thinking about how we are trying to
Starting point is 00:05:02 future-proof these AI data centers, and that's all happening, and that's happened within the last couple of years. Yeah, that's amazing. One thing I love about you, Kelly, is your just vast experience, right? You have so much experience. And at OCP, you're actually the chair of an industry liaison team, right, which OCP connects external standards like ASTM, ASHRAE, AshRAE, and then, of course, I-T-E. Yeah. So can you tell us a little bit more about why this collaboration is so critical right now? And how are you seeing the alignment in liquid cooling and standards across this overall ecosystem?
Starting point is 00:05:38 Yeah, so I'm actually very excited. I do have a big announcement that I'll be sharing on Thursday when I give my talk about some of the standards and working groups with one of the. of those bodies that you mentioned and OCP. So I'm excited to share that. And I'll tell you a little bit of more about that after Thursday. But let's talk about a little bit with Ashray and OCP. Last year, I was part of a panel discussion that was talking about the need to have more closer collaboration between OCP, Ashray, as well as some of the external standards bodies. And so today, my colleagues from the cooling environment gave an update about where we stand with Ashray and OCP. And so one important
Starting point is 00:06:17 area was that there was an MOU signed between Ashtray and OCP about how we're going to work together and ensuring that there's no duplication of efforts, as well as a liaison role was created for OCP into Ashray. And so those are some really important areas of progress as we move forward to make standards more readily available and to really help allow growth within the ecosystem. Now, you know, I think that one thing that was really notable this week is OCP's letter to the industry that talked about the need to go faster. One of those areas to go faster is in interoperable open cooling solutions. How does interoperability really fit within this from a full cooling system standpoint? Yeah, so I think a great example of where the industry is
Starting point is 00:07:08 looking to address interoperability, specifically as it relates to liquid cooling, is the UQD specification 2.0. So the specification 1.4.0. So the specification 1.4. really focused, and a UQD is the quick disconnects that connect within the cold plate cooling loop from the chip to the coolant distribution unit. Looking at either side of where these quick disconnects adhere, we had many different technologies, valving technologies, many different vendors, and so you may have a quick disconnect from component A that may not work with component B. There was also a lot of proprietary designs, and this was creating a lot of problems and challenges within the industry. And so really
Starting point is 00:07:47 dispestication 2.0 was one of the first within liquid cooling to come out and really address and make standards more readily available to address the challenge of interoperability and heterogeneity within the data center. Okay, so let's take a step back. So looking at the overall big picture, right, how do you see AI infrastructure scaling? Do you anticipate new technologies, materials, or design principles kind of emerging in the next few years? Yeah, I think it's all be based on rack scale design. And this was a concept, actually, that Intel really looked at almost a decade ago, and the industry wasn't quite ready for this idea of rack scale design and disaggregated compute. That's exactly where the industry has headed today, where the whole
Starting point is 00:08:27 rack is essentially the computer. And being able to have disaggregation and being able to have intelligent compute is absolutely where the industry is heading as we think about AI, shaping AI infrastructures. The other key area is, I think, again, on the same theme of heterogeneity within the computer, I think when it comes to power and cooling, it's going to be a hybrid world. I think we're going to have, we're going to see an uptick of immersion cooling solutions, just like we've seen with direct-to-chip. I know the industry really wants to, you know, have a long runway with direct-a-chip, and I do see that, but at some point, you're going to need to be able to cool the entire compute stack. That's the networking, the memory dims, the storage,
Starting point is 00:09:07 in addition to the CPUs and GPUs. And I think the 100% heat recapture that you have with immersion cooling is a real key benefit. And I think being able to adapt to different infrastructures and having multiple advanced cooling solutions within the data center is key.
Starting point is 00:09:23 Now, technology is important, but so is business process and partnership. How do you see operators starting to make the decisions that they need to make and the connections that they need to make to actually deploy this? scale. I think we're starting to see some of the key technologies in terms of if we look at it from a cooling lens, a lot of the key startups are being acquired by bigger corporations. So like I think
Starting point is 00:09:50 a great example is flexes acquisition of jet coal. Jet coal is a very great innovative coal plate design. And now that collaboration is really starting to take off. And there's multiple types of those kinds of collaborations across the industry that we can look to. It's awesome. love that. And, you know, speaking of collaboration, where can listeners go to learn more about your organization and the collaboration that you are driving through OCP? Yeah, so for any information related to my company of Viola, you can go check out our website, www.com.com. For more information on how to become involved with OCP, you can check out the cooling environment's homepage wiki and the industry liaison team. We have a wiki as well. And I would invite, if you're here at the
Starting point is 00:10:36 conference, come and check out our talk on Thursday, 825, where we're going to be having a big announcement related to standards and collaboration with an OCP. That's awesome, Kelly. Thank you. You're a frequent guest on Tech Arena, and I always appreciate it. Thanks for spending time with us today. Thanks, Allison. Thanks to be you. Thank you, Kelly. Thanks for joining Tech Arena. Subscribe and engage at our website, Techorina.com. All content is copyright by TechRena. Thank you.

There aren't comments yet for this episode. Click on any sentence in the transcript to leave a comment.