In The Arena by TechArena - OCP, Power Grids & Open Racks — with Dr. Andrew Chien

Episode Date: January 5, 2026

From OCP Summit San Jose, Allyson Klein and co-host Jeniece Wnorowski interview Dr. Andrew Chien (UChicago & Argonne) on grid interconnects, rack-scale standards, and how openness speeds innovation.

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 Alice Klein. Today, we are coming to you from OCP Summit in San Jose, California. And I am with Janice Norowski. Hey, Janice, how you doing? Hey, Allison. So, Janice, OCP, has been jam-packed, and you have brought with you a fantastic guest to talk to us. Tell me about the topic and who's sitting in between us. Yes, a fantastic guest.

You know, you come here and you do see most people that you've met previously in your career, but it's not too often that you get to meet a doctor. So today we actually have Dr. Chen with us, who is a professor at the University of Chicago. but you're also a senior data scientist with Argon Labs, is that correct? That's right. I'm a senior scientist there, again. Yes. So we are going to dive into all things AI today, which is also a hot topic for this conference. So welcome. Thank you. Great to be here. So Dr. Chen, let's just start out. Can you tell us a little bit about your role at the university

and what research do you study? Sure. I'm a full professor in computer science, and I've been doing research in really two areas. for the last four or five years. One is in scalable graph analytics, accelerators for that. And the second really is what brings me here. We've been studying how data centers interact with the power grid, how to make them more sustainable, how to make them more capable. You have a background that spans from industry all the way into academia. What does that bring you in terms of a unique focus in your current work? Yeah, it's kind of interesting because some people are surprise. I spent 15 years as an academic before I went to serve at Intel as a senior executive

leading research. And what I learned from that experience, actually, is that there's some things that academics can do, that problems are well suited for academics to attack, and then there are problems that are perfectly suited for an industry to solve. So I'd say, you know, what I've learned from that is maybe a little bit better choice of what to work on when I'm in the university to maybe potentially have an impact and try to avoid doing the things that industry is so good at, just let them solve this problem. Love that. Well said. So, So jump into it. Your work today spans sustainable computing, cloud architecture, high-performance systems. What trends or challenges are you seeing as data centers kind of evolve and support

increasingly energy-intensive workloads that they're running on AI? Oh, actually, I mean, the trends I think are sort of in everyone's face now. It's this trend towards higher power density. It's the trend towards hugely power-consuming kinds of computing. But I'm just going to put up by academic flag for a minute here. So we recognize these trends were happening as long as 10 years ago. So we started working on power grid meets sustainable computing with this zero carbon cloud project. Because in the years I'd spent that Intel, I'd figured out that Moore's Law was coming to an end. And when Moore's Law comes to an end and our scaling comes to an end, it means computing, which we seem to have an infinite appetite

for, was going to consume more and more power. So here we are. Well, your session at OCP Summit just jumped off the page for me. You're talking about AI data centers and microgrids. So But what does that actually mean and how does that relate to the challenge that you just described in terms of increased power? It's a great question. One of the things that people first think about when they think about data centers that consume a lot of power is they think hundreds of megawatts, gigawatts, it's just a lot. But actually, it's more complicated than that because our power grid is going through this transformation to become more renewable-based. And the challenge with renewable-based generation is it fluctuates. It's volatile.

Sometimes it's there. Sometimes it's not. So the key question, really, is how do we take these data centers, which we want to run flat out all the time, full load, make them as valuable as possible, and harmonize that with this fluctuating kind of power. So the idea of these microgrids are you want to have flexibility in how you use power inside the data center and how you actually manage a very interesting thing, cooling or cold inside the data center. So that's the two things I'm going to talk about. That's fascinating. And as data centers continue to grow and evolve, how do you see the change for partners, to be good stewards of the power grid, particularly as concerns rise about great capacity and just overall sustainability.

A great question. You know, I was talking before about this idea that data centers are these big loads, and the grids are increasingly these volatile fluctuating kinds of things. So to be a good citizen, right, in that kind of a grid, what you need is the data centers to be a little bit flexible because it turns out like just 1% of the time of the year, the grid is really stressed. And like, when you're having a bad day, you're like, I can't deal with this right now. That's how the grid feels.

Well, that's how the grid acts. So what they really need is for data centers to be able to back off a little bit, just a couple of hours a year. And then it turns out, we did some studies, some other people did some studies to show that you can dramatically increase the amount of power that's available on the grid

if you can just have that flexibility. So this is one of the big challenges and people are trying to work out how, like, these microgrids that I talk about can enable data centers to have the flexibility. Now, we'd like to do it in a way that the data centers can still run flat out almost all the time for their AI,

their training, their interest, whatever. So the trick is, how do you invent technology that people have their cake you need it to, if you will? Now, a microgrid enabled data center, how does it differ from data centers today? And what does this mean for infrastructure? We are at OCP, so we should talk about that. I think traditionally, data centers have been managed very statically. And they've been super optimized for efficiency and getting the right amount of power in the amount of cooling. And it was all based on the notion that things were kind of stable and we could tune them. We're talking about a world in which stuff is going to be fluctuating. We're seeing fluctuation from the grid side.

And now everyone knows that AI training loads cause fluctuating power loads. So you see these spiky kinds of waveforms. So the difference is now we have to learn how to design and manage these systems that are gigantic and hugely finicky with dynamic, responsive, flexible control. And there's two challenges with that.

First, it's just hard. Second, it scares the heck out of people that have a reliability culture, which if data center people don't have reliability culture, I don't know what is their primary culture. So it's a little bit scary to move to a dynamically managed world. Sure. Now, one of the things that you really research is kind of this notion around zero carbon, right?

Zero carbon cloud computing. So can you tell us a little bit about how you see microgrids contributing to that goal and how the new technical operational models might emerge as a result? Sure. So we started this zero carbon cloud project, as I mentioned, almost 10 years ago. with kind of a crazy idea. The crazy idea was sometimes the grid

has too much renewable energy. So as you could somehow use that energy, you could actually operate a cloud data center, and this was before AI was such a big deal, with zero carbon, zero operational carbon kind of impact. So that was the idea. And being an academic, you know, we took this extreme view.

The data center would go off when there wasn't this excess renewable energy. You can't really do that. So the modern view of this is you have a data center that slightly slows down or consumes the whole less power when there's less renewable energy available.

Or maybe you have a little microgrid in there that fills the gap between what the grid can provide and what the data center needs. Like I said, you get the best of both worlds. The data center gets what it wants, and the grid can flex and be low carbon and all of those good things. Now, when you think about this vision,

it's very enchanting. What is gaining us from achieving it? Is it the technology? Is it operational habits? Is it cost? What is the biggest barrier to that? Wow, that's a good question. We're just writing some research papers now.

Just plug my research papers for a minute. That actually show that you can build these power microgrids in such a way that they can be a small fraction of data center cost. So there's technology available. It's lightweight generators. It's small-scale storage and so on, even for these large data centers. And you can deploy it in such a way that it's cheap relative to the cost of the data center. So I don't think it's so much a technological barrier at this.

point, but the barriers are a little bit more difficult than that. The first is who's going to pay? Always a challenge. The second is who is really responsible for this. So right now, there's this debate going on between the data center companies and the power grants, like how much should I have to pay to be connected? How much is your responsibility? How much is my responsibility? And it's new ground, I think. I think we're breaking new ground for what has to be done. So I think establishing clear responsibility and expectations, which I learned in industry, actually, It's not that industry won't pay. It's they want everyone to pay fairly.

They don't want to be disadvantaged. So we need to establish clear standards for that. And then all of this stuff can go. And we can have, like I said, have her cake and eat it too. Well, I mean, collaboration clearly plays a major role, right, in the type of research you do. How important our initiatives like OCP, we mentioned we are at OCP, but how is that bringing in academia, industry, and standard communities to kind of accelerate? Oh, the CP is this amazing phenomenon. I mean, I've known about OCP for over 10 years now.

For some reason, this is my first time to be able to be here. Well, welcome. Well, thank you. But when I look at what it's done in terms of enabling the speed of industry and the speed of integration and the speed of scaling, right, what we heard about in all the keynotes. It's just extraordinary. It enables all those things for industry.

As an academic, it makes a world of complexity much easier to study. You know, as you said, I study hardware architecture, large-scale systems design, and so on. And instead of having to look at things that are just, you know, just gratuitously different because it came for vendor A or vendor B, we could look at things in a uniform setting. We can evaluate them directly and fairly and so on. And then when we want to build things, we can build based on all of that leverage. So I think it provides us many of the same advantages that provides to industry. Obviously, we're not operating at quite that budget or scale. Yeah. And you're based on your history of doing research at Intel, running a startup,

you're now working across the fields of academia and research within Argonne. What do you? What are you most excited about as you look out at the future? Oh. I'm actually, and I should have said this before about OCP, it also makes it possible for us to create innovations and deliver them in a framework that the industry can understand. You know, I say it's an open rack-wide format. They know exactly what I mean. I don't have to explain to them anymore.

So that's a really great thing. But back to the question about what I'm most excited about. I'm not going to say AI because everyone's excited about AI, but I'm going to give a plug for other things. I think that the scale of infrastructure we have is not only great for AI, but it's going to be great for all kinds of other applications and different amazing uses of computing. And let's not forget, there are other kinds of computing that are going to be enriching our lives, creating commercial opportunity,

hopefully leading to exciting research for many more years to come. Thank you so much for being on the show today. Janice, that wraps another edition of Data Insights. Andrew, one final question for you. Where can folks engage with you if they want to continue the dialogue? Oh, that's a great question. Obviously, you know, you can find my information on the UChicago website. There's a wealth of papers and other information about what we're doing, but they can just reach out. I'd be happy to interact with folks through email or chat or something like that.

Thanks so much for being here. And thanks so much, Janice, for partnering on this great episode from OCPi. Thank you. Thank you, Andrew. Thank you, Allison. Thank you for having me here. Thanks for joining Tech Arena. Subscribe and engage at our website, Techorena.com. All content is copyright by tuckering.