In this ‘Day 3 theCUBE Insights’ segment from SC25, theCUBE’s John Furrier and Jackie McGuire break down the top insights from the event, identifying the transition from traditional data centers to "AI factories" as the defining shift of the era. The duo explores how the industry is confronting physical limits, from the thermodynamics of chips to the sheer scale of facility construction and labor shortages. Furrier and McGuire analyze why liquid cooling has graduated from an experiment to a mandatory strategic layer and how physical and supply chain security are becoming paramount for this new class of critical infrastructure.
The conversation delves into the architectural bottlenecks reshaping the tech stack, specifically how storage has overtaken compute as a primary constraint and why networking must now be lossless, automated and open to function as the operating system for AI. They discuss the rise of software optimization and intelligent orchestration as necessary alternatives to brute-force hardware scaling, emphasizing that sustainability is now a business imperative that drives performance. The segment concludes with a forecast on the victory of open ecosystems, predicting that composable, vendor-agnostic architectures will define the next decade of computing innovation.
Forgot Password
Almost there!
We just sent you a verification email. Please verify your account to gain access to
SC25. If you don’t think you received an email check your
spam folder.
In order to sign in, enter the email address you used to registered for the event. Once completed, you will receive an email with a verification link. Open the link to automatically sign into the site.
Register for SC25
Please fill out the information below. You will receive an email with a verification link confirming your registration. Click the link to automatically sign into the site.
You’re almost there!
We just sent you a verification email. Please click the verification button in the email. Once your email address is verified, you will have full access to all event content for SC25.
I want my badge and interests to be visible to all attendees.
Checking this box will display your presense on the attendees list, view your profile and allow other attendees to contact you via 1-1 chat. Read the Privacy Policy. At any time, you can choose to disable this preference.
Select your Interests!
add
Upload your photo
Uploading..
OR
Connect via Twitter
Connect via Linkedin
EDIT PASSWORD
Share
Forgot Password
Almost there!
We just sent you a verification email. Please verify your account to gain access to
SC25. If you don’t think you received an email check your
spam folder.
In order to sign in, enter the email address you used to registered for the event. Once completed, you will receive an email with a verification link. Open the link to automatically sign into the site.
Sign in to gain access to SC25
Please sign in with LinkedIn to continue to SC25. Signing in with LinkedIn ensures a professional environment.
Are you sure you want to remove access rights for this user?
Details
Manage Access
email address
Community Invitation
Day 3 theCUBE Insights with John Furrier & Jackie McGuire
In this ‘Day 3 theCUBE Insights’ segment from SC25, theCUBE’s John Furrier and Jackie McGuire break down the top insights from the event, identifying the transition from traditional data centers to "AI factories" as the defining shift of the era. The duo explores how the industry is confronting physical limits, from the thermodynamics of chips to the sheer scale of facility construction and labor shortages. Furrier and McGuire analyze why liquid cooling has graduated from an experiment to a mandatory strategic layer and how physical and supply chain security are becoming paramount for this new class of critical infrastructure.
The conversation delves into the architectural bottlenecks reshaping the tech stack, specifically how storage has overtaken compute as a primary constraint and why networking must now be lossless, automated and open to function as the operating system for AI. They discuss the rise of software optimization and intelligent orchestration as necessary alternatives to brute-force hardware scaling, emphasizing that sustainability is now a business imperative that drives performance. The segment concludes with a forecast on the victory of open ecosystems, predicting that composable, vendor-agnostic architectures will define the next decade of computing innovation.
Day 3 theCUBE Insights with John Furrier & Jackie McGuire
John Furrier
Co-Founder & Co-CEOSiliconANGLE Media, Inc.
HOST
Jackie McGuire
Principal AnalysttheCUBE Research
HOST
In this ‘Day 3 theCUBE Insights’ segment from SC25, theCUBE’s John Furrier and Jackie McGuire break down the top insights from the event, identifying the transition from traditional data centers to "AI factories" as the defining shift of the era. The duo explores how the industry is confronting physical limits, from the thermodynamics of chips to the sheer scale of facility construction and labor shortages. Furrier and McGuire analyze why liquid cooling has graduated from an experiment to a mandatory strategic layer and how physical and supply chain security ...Read more
exploreKeep Exploring
What impact is AI having on infrastructure and data centers, and what is the current state of the market for high-performance computing and supercomputing?add
What are some trends and challenges related to the implementation of liquid cooling technologies in data centers?add
What is the significance of storage in AI infrastructure and how does it relate to performance bottlenecks?add
What are the key considerations regarding software optimization in relation to hardware upgrades in AI development?add
What is the current priority for customers regarding software and mission-critical reliability?add
Day 3 theCUBE Insights with John Furrier & Jackie McGuire
search
John Furrier
>> Hello, welcome back to theCUBE third day of live coverage here at SuperCompute 2025. We had wall-to-wall team coverage. I'm John Furrier, your host of theCUBE here to wrap up the show and share with you our top 10 insights of what surfaced here this year. And then we have color commentary from Jackie McGuire, who's colorful. She's the lead analyst and always wearing good color. Jackie, color commentary as well as deep insights at theCUBE Research. Great to have you back.
Jackie McGuire
>> Thank you.
John Furrier
>> Dave Vellante and Savannah Peterson were here on day one as well. We had the whole team coverage and the crew here, just tsunami of data coming into theCUBE. As you know, we'd love to ingest that content and share that with you, but we get a lot of great insights. Jackie, every industry is being impacted. AI is rewriting the rules of infrastructure. This show very nerdy, high performance computing now full-blown supercomputing, AI factories, new software stacks that will comprise the AI factory. And all this is underpinned by the chips, NVIDIA, AMD, all the vendors that make the data centers, which again, data centers are back. NVIDIA's earnings came out last night. Their data center growth, massive, no bubble popping at any time in the near future. Yes, overinflation on price, but the market is clearly in massive build out mode. And we love data centers. We love the cloud, we love computing. This show is a bellwether for what will enable the agentic infrastructure. All the data, again, the themes of data. Let's get into it because you had some great commentary yesterday with some of our guests kind of teasing out what is our number one takeaway, which is the AI infrastructure is hitting physical limits and the word physical is in there because there's physics of the chips and the density, but there's also the physical aspect of the facilities. Physical AI is right around the corner on all the road maps. The top trend is the physical world, that's labor. I mean, you brought this up in multiple conversations around who installs the HVAC, who pulls the cable, real estate's involved. Physical is the big theme here.
Jackie McGuire
>> Yeah, well, I think there was an announcement by the administration yesterday that we just kind of rare earth minerals deal with Saudi Arabia. It literally starts at the actual raw materials and silicon it takes to make these things all the way through how do you install them? How do you secure the building once they're installed? We were talking before, I'm the cyber security person so I love all of this. I love all of the technology and the hardware, but literally the first thing I think of is, holy crap, there are so many different ways that you could attack this infrastructure. And we're building so much so fast that things like securing the physical supply chain are now coming into focus and we are innovating at a rate that I keep worrying we're not going to be able to scale labor fast enough to support. Yeah, I mean hardware's cool again, but hardware also has these very firm physical limits, not just in your ability to produce it, but then your ability to run it at temp. And so we heard from a lot of companies who were doing some really cool things to try to address this. Everything from being able to take a traditional data center and switch pieces over modularly rather than wholesale. But yeah, I mean the world has reality of physical limits and we can say we're going to do AGI, but without the servers to run it, how are we going to do AGI?
John Furrier
>> Yeah, Elon Musk was talking about space, Jensen as well, there's a lot of water in space, but it's cool space. Space is an answer kind of on the North Star, but the AI industry is hitting these walls of physics. Energy is bounding all the upside that's got to be worked on. But in every interview here we heard the same theme, cooling, storage, networking, and compute. One message was unmistakable that they're all hitting this limit, energy and physics and everything is being redesigned. And that's a feature, not a bug. Okay, storage, different paradigm. KV cache, Dynamo from NVIDIA, we heard cooling, liquid cooling, but you can't just put liquid cooling everywhere. I mean all this is upside of storage.
Jackie McGuire
>> Well, and even so there's the original liquid cooling, but then you have the airless liquid cooling with the secondary liquid loops. And it's all this crazy kind of groundbreaking technology, but many of the existing data centers weren't built to hold this equipment. And this equipment is significantly heavier. It produces theoretically more heat. And also if you're... What we were talking about before is if I'm already using air cooling and then I want to bring in these immersion cooling technologies, how do you do that in a modular way where you're able to piece by piece, swap them out and you don't have to just shut down an entire data center and retrofit it for liquid cooling because not realistic, right?
John Furrier
>> But that's our number two trend. Liquid cooling is moving from experimental to instrumental and essential. And liquid cooling obviously is a hot trend. Precision liquid cooling came up. Immersion, advanced fluid networks are becoming mandatory infrastructure. The plumbing, because mostly people refer to plumbing as the networks is actual real plumbing.
Jackie McGuire
>> It's actual plumbing. Yeah.
John Furrier
>> It's actual plumbing.
Jackie McGuire
>> Like hoses and pipes and fittings and all of that. And I think we had some really interesting conversations, not just around, we always think about liquid cooling for the servers and things like that, but we're now starting to move to liquid cooling for the network cards, liquid cooling for power supplies. There's this whole kind of ecosystem of primary and secondary liquid cooling being built around every component because they all produce heat.
John Furrier
>> And Jensen had heat on his stack. You look at cooling is now a strategic layer in the AI infrastructure. Isotope, Hypertech, Midas, Valvoline, that was an amazing panel we had. They laid it all out. This is not an afterthought. I mean this is absolutely a strategic layer in the AI stack. This is back the physical builds on that. I mean, if we don't solve the cooling problem and get that density nailed down, density is a feature in these clusters, not a bug. Got to have the cooling. The question is standards. Well, what kind of standards? I mean yesterday you asked an open ecosystem question and there was a standard question and then what came out of the guess was, the experts was there's other standards. What's the hoses look like? What's the connections? You have now a standards aperture that's opened up big time.
Jackie McGuire
>> Yeah, I think it was Lauren Witter from the CoolFlow of Omni and Amanda Bryant with, oh goodness, it was Danfoss Power Solutions-
John Furrier
>> Danfoss, yes....
Jackie McGuire
>> yes. And that's the other thing that I took away from that too is that as we're trying to achieve these dramatic increases in efficiency and ability to cool, you're seeing significantly more really close co-innovation, co-development, and collaboration between components vendors, the data center vendors. I think it's bringing this level of collaboration that we haven't seen in decades probably since before the '90s. I mean intellectual property became the most important thing in being able to secure funding. And intellectual property is great, but walled gardens don't create new ecosystems. Walled gardens create silos. And so that was the other thing that I thought was awesome is that you're starting to hear over and over and over, collaboration, partnership and this kind of, we have to all succeed together.
John Furrier
>> Awesome. And we had a great conversation. Check out the videos. A lot of great conversations that we didn't expect on cooling and some of the issues there when you get the mechanical and the fluid networks of flow. That's really cool. Okay, number three, we had going in, certainly been covering networking as fundamental to AI. You got InfiniBand connects Blackwells that births the clusters, hence the AI factories. But this show, Jackie, was clear that storage, the new AI bottleneck we have Dan Noyan, Ace Stryker, they reinforced that AI isn't bounded by compute anymore, its IO bounded. Energy is bounding everything. But when you start to think, oh, compute, compute, compute. Yes, we're in accelerated computing, but the storage is where the AI speed gains will come from you. You heard from Solidigm the success they're having with offload a term that could be used as pejorative, but it's fundamental in these large scale infrastructures.
Jackie McGuire
>> And I think when you think about, we talked a lot yesterday about KV cache, and this is kind of like the memory layer of your say agentic deployment. And so if you think about it like a relay race, you either have a relay race where everybody has to carry one bucket of water somewhere or where the last guy ends up with 10 buckets. And so the way we've been doing it with agents is the last agent has this insane amount of memory they're trying to carry and they end up dropping tokens. And so having KV cache and this memory layer enables you to more efficiently share the right data with the right agents and then store what they're producing. And if you can't do that at the speed that you expect your agents to be running, and that was what we talked about was kind of the in, out of storage and optimizing storage. The storage is also not cheap. Being able to figure out what memory should go into that high-performance storage versus what stuff just needs to be stored for a potential later date is something I think we're still working on optimizing. And we've seen some really big leaps in the last year that-
John Furrier
>> I love the analogy. You're great with analogies. In fact all scale up, scale out, scale across metaphors could be described as playground games. What about the sack with two people in the sack? Tie your legs together-
Jackie McGuire
>> Potato sack....
John Furrier
>> tossing the egg. Is that cross? That is the bottleneck if you are not in sync, this is why I think the storage is an important conversation. It's not your grandfather's storage anymore. It's tightly engineered as a memory tier because-
Jackie McGuire
>> Yeah, exactly....
John Furrier
>> training hours are going to come in. Huge point there. And again, we've all struggled in those playground games, face plants, that's called failure.
Jackie McGuire
>> I'd be remiss if I didn't talk about security. I did a great interview a couple weeks ago with Dell and AMD talking about security at the silicon level. And that's the other thing with this kind of high-performance memory and compute is that then you have to move the security of that as close to that as possible. And that security and detections need to be as performant as your agents in your memory. Because if you wait until something goes wrong and it shows up in your SIM, that's like seconds later and millions of things could have executed. And when you're talking about nanoseconds, this is one of the things that Dell and AMD are working on is bringing the security way closer to the chip level so that we can secure these things as we're doing this high-velocity edge computing and things like that. I have to plug the security part-
John Furrier
>> No, security is... Again, energy is down there. I'd say security's probably next to energy in terms of the fundamental Maslow hierarchy of needs of the stack. Okay, number four, we said going in, networking is hot, networking is the operating system, KV cache those things. But networking this year, you start to get into the nuances as you start to unpack networking and why it matters. Networking, we heard this week must be open, automated, and lossless. We had experts on talking about AI frameworks required lossless ethernet, networking is becoming open source. All networks have to be automated end-to-end, not hand-built. Multivendor silicon-independent fabrics are the new reality and networking is triggered from a component to a full AI operating system. That builds on our thesis that networking is the OS for these AI factories. But now it's like they're getting into the weeds. Okay, what does that mean, lossless, automated, end-to-end? Again, networking's still super important. The storage, I think bumps above it because that's jumping out this year in the architecture. But again, networking is still fundamental.
Jackie McGuire
>> One of the things that we talked about yesterday that I, when I was walking around the show floor, I had actually had a couple conversations with different vendors about this and then it was brought up in our interviews is how you move the data across those networks. If you're using TCP/IP versus some of the newer technologies, and I made the analogy of Amazon, you can't be using TCP/IP, which really requires you to unpack the data, send it, and then repack it. And we've developed all these great new networking technologies that allow you to just send the data. You don't have to process it. And it's kind of like Amazon, you just send your products to Amazon and they ship them back out versus sending them to a retail store where they put them on display and there's all this extra cost. And so the lossless part, I feel like that's coming from these advancements in how we send data across the network.
John Furrier
>> And again, what came out of that too, good point about Amazon, just send the payload in one package, TCP/IP, hey, here's the package. Take the package, take the handoff. Now it's like direct access to the GPU. They were talking about direct access from memory or storage right to the GPU, very efficient, not a losing part. Again, it must be treated as a key part of the stack. Very key point. Number five, networking. Number six, you're going to love this because you brought up a couple of software comments on some of the comments. One was the engineer's going to get low level and you are like, "I'm a Python developer." It's kind of like well Python developers-
Jackie McGuire
>> Somebody started talking smack about Python and I was like, "Hey," as a data scientist-
John Furrier
>> I know. We'll get into it. Number six. Number six, software optimization is as important as hardware upgrades. Okay, what came out of here was the bigger models or delivering diminishing returns. Smaller models on use cases, whether it's proprietary data or however models will be used. I think we're going to hear a lot about half-baked models in the future where models are going to have frontier capabilities but open for customization. I think there's going to be a major software wave where the developers will have that capability. And if you look at what's happening here, just some tea leaves to read, KV cache reuse, model compression, token reuse, pipeline optimization, real-time inference tuning, all customers and development faster time to value, not just more compute. AI optimized will be a definition that will be unpacked because what does that even mean? Brute force, you're going to have to go to orchestration because agents want to orchestrate, data wants to orchestrate. Orchestration in the OS is coming out.
Jackie McGuire
>> And one of the things we talked about as part of this is with these new AI enabled coding assistance, if they are not programmed to optimize the software they're writing. To take a big step back, I was part of the problem because as a data scientist, when you do a bootcamp, which is how I got into data science, I didn't do a four-year college program, they teach you open your Jupiter notebook, which is how you can kind of execute Python in different segments and you import every package you could possibly think of that you might want to use in your model. And that's a problem. And that's why you see companies like Minimus and Chainguard and those kind of companies that do these kind of minimize containers because we usually just end up deploying them like that. And you have all these dependencies in there that you really don't need that not only slows down the AI but also introduces potential security issues. And so we have not gotten to a point yet where these vibe coding platforms are optimizing for container efficiency or for code efficiency. And I think that's part of what we were talking about in this is we need to get to a place where we're not just importing 100 dependencies and we only need four of them.
John Furrier
>> And I think this shift from brute force or purpose-built to intelligent orchestration is a really key point because if you think about security, DevOps, infrastructure management, there's a lot of stuff where agents will come in and optimize so software layers will emerge around the infrastructure. I think we're going to hear a lot more about how AI factors we're incorporating agentic into it. Software, again, software is eating the world. We'll see. Number seven, mission-critical reliability is now the number one customer priority. We heard from Vavoline, Omni, Danfoss downtime and clusters are expensive. We heard that building your own... We heard from Vast, they're selling their platform because they can keep the Neo clouds clocking at above 70, 80% GP utilization. The Wild West engineering fluid networks are replacing... Hey, call the plumber, let's get some water. These are the things that are now mission-critical environments that tie into those other themes. Jackie, this is where the reliability concept shifts is now a holistic view.
Jackie McGuire
>> And it's one thing if a data center has some type of issue and it's supporting, say Instagram and like, "Oh no, your photo failed to upload to Instagram." It's completely different if it's running the power plant down the street from your house. And so what we've heard is essentially because we were running so fast and had so few standards, it's been kind of acceptable to just make it work. Try this, try that. Let's roll liquid racks in next to the air racks. We'll figure out how to manage the cooling on each side as we go. And now we've started to see large enterprises make mission-critical things dependent on AI. That level of, I say resiliency, redundancy, all needs to increase because a couple seconds of downtime in one of these applications could be massive. And the other thing with data centers is that it's kind of a domino effect. If one thing in the physical part of a data center goes wrong, it can cause this cascading wave of physical failures that can take out massive amounts of racks, not just one server. We were talking before we started filming and I was like, "People think about hacking into servers or doing things with software." I'm like, "If you know infrastructure and physical infrastructure and data centers, you can mess with a couple sensor readings and say trigger the fire suppression foam." And that stuff takes days to... You're talking about shutting an entire data center down because you just hacked a couple readings and sensors-
John Furrier
>> Don't give anyone ideas, Jackie, come on.
Jackie McGuire
>> I know. I know.
John Furrier
>> You're giving all the hackers all the good ideas.
Jackie McGuire
>> Yeah, I know. But that's the kind of thing that we need to be thinking about. And if you think about Department of Defense or some of these higher level universities that are doing these really advanced things, and we're talking about things like quantum physics, sorry, lost the word for a second. You have to have reliable, consistent standards-based delivery.
John Furrier
>> And you're talking about the Department of Energy, we heard Horizon, all these big data centers that have hundreds of thousands of GPUs. These aren't labs. No, they're mission-critical infrastructure. They're critical infrastructure. And obviously with the theme towards national sovereignty, another theme that popped up, not consistently, but it definitely was relevant. I put that in this category, national critical infrastructure is more than the grid, more than the nuclear plant. It's the data center. And you're starting to see sovereignty come out. I think that's a huge point. If you're not thinking critical infrastructure, then if you're not in security, get out of the way because all critical infrastructure has to be locked down solid as a rock.
Jackie McGuire
>> Especially with the size of these data centers. These buildings are like three football fields long each. And to be able to secure that amount of physical space, and that's another place where AI may come in and we were talking to Chris Sullivan about using AI to grade video presentations of student tests. I also sat down with a company at RSA this year that's using AI to review security camera footage. That may be a place where AI can actually-
John Furrier
>> Absolutely....
Jackie McGuire
>> where software AI can help secure physical AI because there's just the more of these we build, I think one of the new campuses is like 12 buildings, and that's across hundreds and hundreds of acres of land. How do you secure that?
John Furrier
>> And this is why I love the whole simulation, AI and quantum coming down the pike, because that's just going to add more value. Okay, number eight, open ecosystems are winning. Of course, we knew this. We say it all the time. We love open source. We're very much bias. Disclosure, we love open source. But let's go into the reality. It's been repeated across Dell and NVIDIA and the integrators, including SONiC, which is going to open source project. Everyone wants openness, not lock in. Even NVIDIA, even though they could be viewed as monolithic in their stack by some, they're putting out libraries in every vertical. They're open sourcing everything. NVIDIA is not a closed architecture and the market favors momentum. OCP, moving the needle. I was schooled on one interview because I did not know that they were doing work in cooling, not just rack standards, but they're actually doing some other work around the physical. Again, OCP.
Jackie McGuire
>> Yeah. Yeah. And the OCP conference was just a handful of weeks ago and did see significant. The size was bigger, the buzz was bigger. And I almost think I talk a lot about the whole economic shift to Millennials and how Millennials and Gen Z and Gen Alpha really prefer collaboration over competition. I think that coupled with this, we all need to be able to work on the same things, is really pushing people towards the open source. We're huge supporters of KubeCon and Kubernetes, and so we have always been really deep in the open source community, but I think we're starting to see previously walled garden business strategy open up.
John Furrier
>> And we're seeing that same strategy just to put a plug in for other areas we're covering in crypto, you got Ethereum, Solana, they're all open. You have decentralized on that side. Also bounded by energy. The open is huge for the culture. But you mentioned collaboration as a choice of mode of operation and work. Composability and vendor-agnostic architectures feed into choice and choice is huge. When you look at... You were just talking about cooling, you can't have a retrofit. I mean, you got to have the puzzle pieces to put it all together.
Jackie McGuire
>> That modularity is something that I have seen mature quite a bit over the last two years.
John Furrier
>> The number nine, we're almost at 10, sustainability is a business imperative not a PR line. I put that in there because actually there's two sides of that coin. Sustainability was not viewed as check the ESG box we heard from people. They're almost dogmatic about like, "We'll get to that later. We want get sustainability because we get performance advantage."
Jackie McGuire
>> It's cheaper.
John Furrier
>> It's cheaper and performance advantages and from a mission-critical standpoint, the more sustainable these data centers are, the more reliable and they get more performance. That was a surprise to me. I never looked at it that way. But you get good power and cooling, you get more performance.
Jackie McGuire
>> Well, you also open up the number of places who will be willing to let you build data centers there. I live in Tucson, Arizona. There was a massive pushback against an AWS data center because of the water usage. The closer that these... And not from a environmental standpoint, but just from a, hey, we're not going to come in and require you to build 20 miles of water line because we don't need it. And I think that, the ability to open up real estate is going to drive this as well.
John Furrier
>> And the precision cooling conversation eliminates AC loads. Again, another savings. SSDs cooling efficiency's all good at scale. At scale requires a sustainable architecture just on pure uptime and performance. Yeah, I think they'll tally up the ESG gains later. By that time, maybe we'll be space data centers. We'll see.
Jackie McGuire
>> Yeah.
John Furrier
>> Okay. Finally, number 10, the outlook going forward. It's clear to me, Jackie, that the AI factories are here. The era is here. It's a thing. It's not just data centers. We're hearing things like hyperconverged edge with factories of the edge. We're hearing micro factories. We're seeing AI factories almost like sovereign cloud centers. I mean, what's a cloud? It's a data center. Amazon, Azure got data center. Data center-
Jackie McGuire
>> Somebody else's computer.
John Furrier
>> It's someone else's computer.
Jackie McGuire
>> The cloud is somebody else's computer.
John Furrier
>> Data centers are transforming into AI factories, and that is the new architecture and everything's being redesigned.
Jackie McGuire
>> Well, if you think about what happened through the '90s with the whole PC revolution, we went from these kind of modular PCs that you built piece by piece. I still remember picking out the RAM sticks for my tower, but then as we started to move to laptops, you did, you started to see this more cohesive ecosystem of I'm not buying a hard drive and a CD-ROM. And it started to be a package. And again, I have to shout out economic trends. Millennials buy packages and experiences. They don't buy individual products and features. Yes, AI factories, these kind of already pre-built systems that you don't have to interface with nine different vendors, they've already collaborated together before you even got there. That is what's going to lower the barrier to entry for a lot of mid to large size enterprises who don't have the technical talent to build their own AI factory.
John Furrier
>> And the customers that are here at SC25, the builders, the data center providers, all the people that are looking at the future are moving away from experimenting and having small proprietary data centers to AI powered operations. The traditional data center is ending, and the AI factory is the next Industrial Revolution. It's here. That's the show. Summary?
Jackie McGuire
>> Yeah, it's interesting you mentioned Industrial Revolution. I was just reading an opinion piece, and I'm so sorry. I think it was The Economist, but it may not have been, about how super computing needs its own industrial revolution. And it was essentially talking about how if we're going to put these things in houses and offices and things like that, we're going to have to completely rethink how we build them because they take up so much footprint right now. Yeah, I think we are on the absolute bleeding edge of what will end up being this little box that sits on your desk in 10 years that can do these crazy throughput and flops. And it's in a closet right now, but I honestly think it'll be the size of a Mac mini in five years.
John Furrier
>> Jackie, great show. As always sporting the color and the color commentary.
Jackie McGuire
>> I wish people could see my shoes because the shoes match too. I feel like that's important.
John Furrier
>> As always, always a surprise. Jackie, the show was awesome this year. Again, we're starting to see the tea leaves, we're reading the tea leaves. The companies that build the composable open and optimized systems will define the next decade in computing. As it accelerates, you're going to start to see a lot more movement. And as this continues to get built out, that will usher in a massive tsunami of software innovation. You'll start to see agents working well. You'll start to see the data layers coming, as everything will be redesigned, the data center era is over. The AI factory era is upon us. That's going to be a distributed community cloud, core, edge, and ultimately the best user experience, hopefully change the society. I'm John Furrier, Jackie McGuire, and our CUBE team. Thanks for watching theCUBE at SC25.
