In this segment from theCUBE + NYSE Wired’s “AI Factories – Data Centers of the Future” series, theCUBE’s Dave Vellante sits down with Rob Biederman, managing partner at Asymmetric Capital, to unpack a disciplined approach to early-stage investing amid AI-scale infrastructure shifts. Biederman explains Asymmetric’s founder-first model: writing $1–$10M checks (often via SAFEs), joining boards as they form and helping operators with go-to-market, operations, finance and strategy (not product/engineering). He shares why the firm avoided 2021’s lofty SaaS multiples in favor of backing proven builders earlier (single-digit pre-money), and highlights portfolio execution such as a cash-efficient LATAM e-commerce company scaling from ~$1-2M to about $50M in revenue. The discussion also explores Asymmetric’s subscale buy-and-build plays (e.g., pool cleaning in San Diego, sleep apnea clinics in Houston), where density, tech-enabled services and platform ops expand margins and enterprise value.
Biederman weighs in on AI economics as enterprises race to “AI factories,” cautioning that not every AI workload creates ROI and that overbuilt compute assumptions could face a reckoning. He argues that winners will prove a clear 10× value equation and avoid scaling go-to-market before product-market fit. Additional insights include early liquidity discipline (returning $0.20 on the dollar before the fund’s third anniversary), portfolio survivability (34 of 35 companies still operating; three positive exits), and guidance to founders: make your value proposition relevant, credible and differentiated. Tune in for candid perspective on how capital efficiency, ownership discipline and anti-thematic sourcing intersect with a world where GPU-dense data centers and AI-scale software are reshaping enterprise infrastructure and economics.
Forgot Password
Almost there!
We just sent you a verification email. Please verify your account to gain access to
theCUBE + NYSE Wired: The AI Factory - Data Center of the Future. If you don’t think you received an email check your
spam folder.
Sign in to AI Factories - Data Centers of the Future.
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 AI Factories - Data Centers of the Future
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 AI Factories - Data Centers of the Future.
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
theCUBE + NYSE Wired: The AI Factory - Data Center of the Future. If you don’t think you received an email check your
spam folder.
Sign in to AI Factories - Data Centers of the Future.
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 theCUBE + NYSE Wired: The AI Factory - Data Center of the Future
Please sign in with LinkedIn to continue to theCUBE + NYSE Wired: The AI Factory - Data Center of the Future. 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
Sid Nag, Tekonyx
In this segment from theCUBE + NYSE Wired’s “AI Factories – Data Centers of the Future” series, theCUBE’s Dave Vellante sits down with Rob Biederman, managing partner at Asymmetric Capital, to unpack a disciplined approach to early-stage investing amid AI-scale infrastructure shifts. Biederman explains Asymmetric’s founder-first model: writing $1–$10M checks (often via SAFEs), joining boards as they form and helping operators with go-to-market, operations, finance and strategy (not product/engineering). He shares why the firm avoided 2021’s lofty SaaS multiples in favor of backing proven builders earlier (single-digit pre-money), and highlights portfolio execution such as a cash-efficient LATAM e-commerce company scaling from ~$1-2M to about $50M in revenue. The discussion also explores Asymmetric’s subscale buy-and-build plays (e.g., pool cleaning in San Diego, sleep apnea clinics in Houston), where density, tech-enabled services and platform ops expand margins and enterprise value.
Biederman weighs in on AI economics as enterprises race to “AI factories,” cautioning that not every AI workload creates ROI and that overbuilt compute assumptions could face a reckoning. He argues that winners will prove a clear 10× value equation and avoid scaling go-to-market before product-market fit. Additional insights include early liquidity discipline (returning $0.20 on the dollar before the fund’s third anniversary), portfolio survivability (34 of 35 companies still operating; three positive exits), and guidance to founders: make your value proposition relevant, credible and differentiated. Tune in for candid perspective on how capital efficiency, ownership discipline and anti-thematic sourcing intersect with a world where GPU-dense data centers and AI-scale software are reshaping enterprise infrastructure and economics.
In this theCUBE + NYSE Wired broadcast, Sid Nag, principal analyst and founder of Tekonyx, sits down with theCUBE’s Dave Vellante to deconstruct the concept of the "AI Factory" as the new cornerstone of enterprise infrastructure. Nag challenges the traditional view of hardware-centric data centers, defining the AI factory not as a physical asset, but as a logical, on-demand entity designed to deliver scalable intelligence. The conversation explores how enterprises must move beyond rigid plumbing to API-driven architectures, allowing them to spin up and tear d...Read more
exploreKeep Exploring
What is Sid Nag's perspective on the topic of AI factories and his contributions to the discussion?add
What are the current developments and roles of various industries in the advancement of AI technology?add
What is the concept of an AI factory, and how can it be utilized effectively by enterprises for deploying AI?add
What is the speaker's perspective on sourcing compute power for building an AI factory?add
What are the insights regarding the comparison between Google TPUs and Nvidia GPUs, particularly in the context of market reactions and the development of AI technologies?add
What challenges might a buyer face in utilizing AI effectively?add
>> Hi, everybody. Welcome back to New York Stock Exchange. My name is Dave Vellante and this is our AI Factories series, NYSE Wired + theCUBE. Sid Nag is here. He's the principal analyst, founder of Tekonyx, former Gartner analyst. Sid, good to see you.
Sid Nag
>> Good to see you, Dave.
Dave Vellante
>> Thanks for coming in.
Sid Nag
>> Thanks for having me here.
Dave Vellante
>> How have you been? I love the talk about AI factories. You've been prolific, writing a lot of stuff. I appreciate the text, "Hey, I just dropped another piece." It's thoughtful. You spent a lot of time thinking about infrastructure, what's riding on top of the infrastructure. It's funny, Sid. Our industry, it's better than watching sports right now. You know what I mean?
Sid Nag
>> Yeah.
Dave Vellante
>> You watch the news channel, it's like one day this company is out of favor. The next day, their stock is exploding. Five days later, oh, they're in the tank because this has circular references. You've seen these waves before. People get confused. But we know when the new stuff is not really big enough, it's growing fast, but it's not big enough to offset the decline in the old stuff, people get really confused and things ebb and flow. So what do you think of this wave? Having fun?
Sid Nag
>> Oh, yeah. It's fascinating. I mean, you talked about the TV analogy. I feel like you've got to have all these picture and picture screens of all the different things that are going on in one single screen. So yeah, I mean it's fast and furious, that's for sure. AI has really taken off. I think it's definitely matured beyond the invention phase. People are actually starting to use it, so that's good. I mean, we keep hearing about the MIT study, but I think that's like the lack of tooling that's causing that, but the technology is solid. And I think you're going to see more and more of AI taking off the role of the hyperscalers, the role of the model builders, now the role of the telecom providers is extremely interesting with all the networking piece, the role of the switch vendors. Sudden networking is sexy again. It's cool. Connecting to GPU clusters and neoclouds, I mean, my mind is blowing.
Dave Vellante
>> Semiconductors is the hottest thing on the planet. Infrastructure is cool again. We're just coming off a reinvent. Amazon announced AI factories and was like, "Oh, where'd you get that name?" So that's cool. Of course, Dell was the first with AI factories. I think HP has got a version of that as well. But what is an AI factory to you?
Sid Nag
>> That's a good question. I've been thinking about this for a while now, as you pointed out, I posted a few LinkedIn posts on AI factory. To me, the AI factory is not a physical thing, it's a logical thing. And it's a logical thing that enterprises can use to deploy AI effectively for delivering, and I talked to John about this last time we spoke on this show, is delivering what I call scalable intelligence. So it's important that the data gets ingested into some sort of a pipeline. Then you've got the inferencing, the models, the robotics, the tools, the software, all of that coming together. And to me, it's a logical thing, and it's not a hard logical thing. It's a thing that is spun up on demand. So I'm an enterprise in the financial industry. I want to build an AI factory to get predictive models for a certain stock. Tomorrow I want to lose something in terms of bonds. I want to build an AI factory on demand that helps me deliver the intelligence to drive and make decisions. And unlike the normal factory that people talk about, either hardware like GPUs and tools and models and other things, it's really much more logical than physical. And it's important that these entities that drive building the AI factory are connected in some sort of a manner that is API driven. So it can't be hard plumbing every time. So let's take, we're sitting here in New York City, so you think about all the different infrastructure pieces that drive the city. You've got the sanitation system, you've got the subway, you've got the fire department, you've got all the sewer system, and the roadways, of course, the grid. But what happens if something breaks? There's sinkhole on Broadway. I got to dig the sinkhole, but I got to make sure the sewer system doesn't get broken. I got to make sure that the subway system doesn't get impacted and I got to shut down everything. So that's not a scalable model in the AI world. So we got to make sure that we do it in a logical fashion, connect with things with each other, so I don't have to go digging things underneath the plumbing. I can connect at the API layer. I can stand it up on as demand, as needed, and get that scalable in terms of the output. And that is going to be driving the growth of AI in my opinion.
Dave Vellante
>> So that's interesting. Scalable intelligence is the output. That's what you're creating out of the factory. Versus what do data centers produce? Produce data. And I guess workflows, automated workflows. So what happens to all that general purpose infrastructure out there? Will it transition in your mind to so called accelerated computing or will they coexist like the mainframe has coexisted for decades after the PCs? How do you see that playing?
Sid Nag
>> I think all of these things will be there. The physical things.
Dave Vellante
>> Yeah.
Sid Nag
>> So if I want a unit of compute, it doesn't matter where I get it from. Whether I get it from a GPU sitting in CoreWeave or Atrainium sitting in Amazon or a mainframe sitting from IBM whether Z or power or whatever. I want a unit of compute. And these are distributed systems that people have, own, are going to buy, and they're all available. So when I build an AI factory and I'm ingesting a volume of data into it, I want to process that with some compute power at scale to drive these large language models for my benefit. I think it doesn't matter to me how I get that one unit of compute. That's the way I see it.
Dave Vellante
>> What do you make, you wrote about this. I want to pull up one of your awesome posts. You had this chart here, Google TPU versus Nvidia, TPU, Google purpose built, AIA6, MatrixMath, Extreme Scale, deep integration with JAX and XLA. GPU invented flexible accelerator, universal for all AI tax, massive CUDA tooling ecosystem. What was your point of that? How do you see this playing out? Of course, the stock market went crazy when TPUs, it's been around forever. People started to realize, "Oh, well, Google might sell to outside of Google. Oh, Nvidia is in big trouble." And so how do you see that playing out? Was the reaction in the market, was it an overreaction? Was it appropriate? How do you see it?
Sid Nag
>> I think it's a little bit of both, honestly, to be candid with you. I mean, to your point, the TPUs have been there forever. I mean, this whole transformer model was a paper that was written by Google brain guys right back in the day sat on the shelves. And that became the underpinning of what we call Gen AI today. NLP, sequence transduction, all that crazy technology. So the TPU, in a way, I guess is more geared towards doing large matrix multiplication and that's what I'm talking about in that post.
Dave Vellante
>> So is that inference? Is that inference?
Sid Nag
>> It's more trading and inference. Anything that requires that heavy duty compute, depending on the size of the model, of course. That's really the key here. But I think as you were talking before the show, you're going to have some silicon vendors that are going to be the Switzerlands of the world, and some silicon vendors that are going to attach their future to the other way around. The hyperscalers going to build their own silicon that's going to be tied to that hyperscaler itself and will not be available commercially for anybody else to consume. Unless you strike some specialized deals. So in the case of Google is TPU, in the case of AWS Trainium. Now Google, interestingly enough, Meta decided to use TPUs and that sparked this whole storm in the discussion. Now, we shall see how generically available TPU will be, how much will Google do that. If Google does that, I mean, that's the inflection point I'm looking at. That's what's going to cause a lot of heartache for the Nvidia's of the world. But now I think, given Nvidia's position, and especially with all the surround technologies on CUDA and other things, they're sitting in a very interesting position and they're definitely the Switzerlands of the world today. Whether the TPU will become the Switzerland of the world, we shall see, maybe Google will go in that direction. That remains to be seen.
Dave Vellante
>> The way I look at it-
Sid Nag
>> Yeah, what's your thought?
Dave Vellante
>> Your thoughts on this . I think volume is a real driver in semiconductors because of experience curves. You hear Morris Wang talk about it or Morris Chang rather talk about it with TSM. Of course, that's in semiconductor manufacturing, but I think it also applies in the design. And I think volume, which we've seen with ARM, ARM wafer volumes are 10x those of X86 and they took over the world from an economic standpoint. That's where I see Nvidia having the advantage. In addition to the things, you're talking about, CUDA, the libraries, the ecosystem. So I don't see them... I mean, margins may come down some, but I don't see any of these other, whether it's Trainium, whatever, Microsoft's doing. TPU, I like. I mean, I think it's very impressive, but I don't see it getting the volume that Nvidia gets. And as a result, I think the lowest performance watt advantage accrues to Nvidia. Do you buy that?
Sid Nag
>> Yeah, I do. I do. I mean, today, yes. And the volume is because they are making themselves available to every possible vendor from an ecosystem perspective.
Dave Vellante
>> The Switzerland point that you made.
Sid Nag
>> Yeah. So if Google decides to do that, then they can certainly achieve that volume. They can give Nvidia run for their money. But right now, the whole hype was about their partnership with Meta, and I think that's part of the whole conversation. Are we going to see more Meta start to using TPUs? Then you'll see that volume discussion become more prominent.
Dave Vellante
>> I was talking to Dev Ittycheria a couple of years ago actually now when this whole thing started. And we were talking about Nvidia, and he brought up, well, is it like Cisco, your old company, which remember they were the, when you were there, actually the most valuable company in the world or is it moderated, came back to the mean, or is it more like Google where they came out of the post.com era and dominated. And Crawford Del Prete from IDC said it's neither, it's Wintel. Because they got the hardware and the software, which I thought was interesting. It was interesting. And some others have picked up on that, subsequent to that. But what do you think about GPU cycles? Michael Burry came out and shorted some of these AI stocks saying that they're cooking the books, which I laugh at because you're talking about a CPU depreciation cycle or server depreciation cycle of whatever, moving from three years to four years to five years. Some even do six years now. But in the grand scheme of thing, it's billions of dollars, but the grand scheme of thing, it's not that much. But my question to you is, will AI, because of the faster cycles, compress those depreciation cycles? Like let's call it useful life, forget about how the accounting is done. Will the training GPU today be useful three, four, five years down the road, maybe as an inference infrastructure? What do you think?
Sid Nag
>> That's an interesting thing to ponder, I suppose. But I'm reminded of the days of the Intel Pentium chips. Remember those days?
Dave Vellante
>> Yeah, of course.
Sid Nag
>> Where people used to go buy a new laptop every year.
Dave Vellante
>> Right. 286, 386, 486. I want a 486.
Sid Nag
>> People used to buy a new because they wanted the latest chip. But what happens today? Nobody buys a laptop-
Dave Vellante
>> Same with iPhone.
Sid Nag
>> iPhone, yeah, exactly. So I think in the world of GPUs, you're probably going to see that Pentium cycle emerge and last for a while, and then you're going to see the flattening of the curve. That's my short answer. If that makes sense.
Dave Vellante
>> Let's talk about enterprises. So you mentioned the MIT study before, which basically you've been under a rock and haven't seen it, said 95% of the POCs don't return significant ROI when they go to production, something along those lines. Why not? What's missing? First of all, every conference you go to, they trot out the anti MIT study example. Their customers are like, "Oh, our customers are seeing ROI." Okay. So this is always a handful of examples, but what are the challenges that they face? Data obviously is one of them. Some of the narrow use cases, maybe they're not going after big ideas, big NPVs at this point in time. What are the challenges that organizations, the companies that used to work with face?
Sid Nag
>> From a vendor perspective or-
Dave Vellante
>> No, specifically the buyer.
Sid Nag
>> Yeah. I mean, I think we talked about this briefly earlier, is back to the factory. If you don't know how to use AI, not don't know. If you don't have the right set of tools and right logical paths and pipelines to get a actionable output or outcome, like a scalable intelligence. I know it sounds like repetitive. But that if I can hold onto something that I get out of the AI pipeline or the AI factory, which I can then go to my people and say, "Action these things for your business." So whether it's operations, whether it's sales, whether it's IT, whether supply chain, procurement, sales, productivity. So once you start seeing folks take that output from that AI factory, that logical thing I talked about earlier. And drive actionable results for each of the subdivisions within the enterprise, I think that's what's going to be the winner. And in order to get there, again, spinning up these logical AI factories without having to do a lot of hand stitching and connecting this technology to this door, this hardware to this software, that's going to be not the answer. Or to this model. If I'm using Vertex AI and Bedrock and I've got Novant and Bedrock, I've got Gemini underneath Vertex, I'm a multi-cloud user. How do I federate across these layers? Who's going to do it for me? If I bring an AI workload and my vendor tells me, "Oh, I've got 15,000 flavors of silicon that can service AI workload." Okay, that's great. My head is spinning, tell me which silicon applies to this workload. So like the serverless idea, call it silicon less. I mean, you heard it here first. That's what is going to drive the adoption of AI in a much faster way. So things like that, like factories, logical spin ups of these things, things like silicon less, things like how do I federate across multiple cloud environments. Those are the things that it's not about the next frontier model that's going to drive the adoption of AI.
Dave Vellante
>> Oh, there'll be plenty of frontier models. They'll take care of that. What about energy? You've written about it. Everybody talks about it as a big limiting factor. I feel like the public policy suggests that we have to have plentiful energy if we want to win this AI race. It seems to me it's a mandate. We're going to figure it out. Do you buy that, that we will figure it out as an industry? Is it overstated as a blocker? How do you see that playing out? Or is it understated as a blocker where the US is at a significant disadvantage, for instance, to China, which gets plenty of power?
Sid Nag
>> Yeah. So I think the energy discussion is very different than everything else. I'll tell you why. Because energy companies today are monopolies.
Dave Vellante
>> Yep, right.
Sid Nag
>> And the energy companies are regulated by the government, the state government, federal government, what have you. I mean, if I'm a consumer of electricity in my town, I can only buy electricity from one company. And that company, if they do something wrong, I have no recourse, unless I go to the board of public BPU, public utilities, and file a complaint against them.
Dave Vellante
>> And then you get the-
Sid Nag
>> Then they show up really fast.
Dave Vellante
>> And then you have the one called the plug compatible energy company that's actually running off the monopolies grid.
Sid Nag
>> Correct. Right. So my point is, I think that's where the role of the government is going to matter. I hate to see government. I'm a free economy, free market lace kind of a guy. But that's where I think the government has to take some steps to think about the promotion of AI in the US and the role of the energy companies, how these people work together. Where the grids are burning out, we know that. So what is our growth policy as America or the United States? To grow the energy grid, whether it's going to be beyond the traditional sources of energy, whether it's going to be nuclear. I mean, Microsoft talked about investing in nuclear power stations for energy for the data centers. What's the role of solar? Then you got to worry about the new batteries that are coming out, the best batteries, BESS, it's an acronym. I forget what it stands for, like driving this massive amounts of energy buffers in case the grid were to fail. So my point is all of these things, then you got to think about the amount of heat this is going to generate these, what happens to the sustainability model. How do we recycle these energy sources or waste. I mean, it's a mind-blowing conversation that no one's talking about.
Dave Vellante
>> Nobody's talking about sustainability anymore. I mean, some people are people who are in the business.
Sid Nag
>> I mean, the whole ESG model, environmental ESG model, environmental, social and government.
Dave Vellante
>> That was the hottest topic five years ago, now nobody is talking about it.
Sid Nag
>> But with the advent of energy proliferation that's going to be needed for these AI data centers, I think it's time for us to talk about it as a society, as humanity. So what's the role of the government? I don't know. I mean, I only see the role of the government because there are monopolies that are running the energy companies. So there has to be some governmental innovation or some policy statement or some guidance because it's not really a free market. So I don't know. It's complicated, I guess. So I'm trying to lay out a few things that we ought to think about as-
Dave Vellante
>> One of my colleagues-...
Sid Nag
>> pioneers.
Dave Vellante
>> One of my colleagues posits that, look, without energy, obviously you're going to fail and you get the best people. You got guys like James Hamilton working on this problem. So they're going to solve it and put, whether it's LNG turbines next to the data center or nuclear, it's got to be resolved. So he's not worried about it, but it's not his problem to solve, I guess. But I want to come back to something we were talking about. We were using JPMC, Jamie Dimer as a metaphor for mainstream or enterprises with proprietary data. So we know that the hyperscalers are building out all this AI infrastructure. We know the neoclouds, they're taking the entire Nvidia kit, so they're getting some good allocation from Jensen Czech. What about mainstream enterprises? So all those, the first two building AI factories, will enterprises be building AI factories or will they be tapping into neoclouds and hyperscalers? Or because I guess colos like an Equinix or Digital Realty, I guess they could build an AI factory inside of their facilities. How do you see that playing out in the end? Will they bring ... I guess, because you talk to enterprises, many of them say, "Well, I'm using the cloud extensively, but I'm not going to put all my data into the cloud. I'm going to do some stuff on prem. It's going to be hybrid." Are they actually going to build AI factories and retool their data centers for liquid cooling and deal with the energy? How are they going to do that?
Sid Nag
>> I think, yeah, okay. So I think the only people that will build AI factories will be enterprises. And there will not be something massive. Like I said earlier, there'll be logical units that they will spin up and spin down on demand for a particular problem. I think that's a pedestrian use of AI that we're all waiting for. I mean, forget about all the esoteric stuff that people talk about in the context of AI and all these world models and high-flying silicon technology. That's all cool. But if I'm Jamie Diamond or I'm the CEO of Boeing, I'm telling myself and my people, how do I use these things in a pedestrian manner that gives me something to hang my hat on. And these are not going to be some massive pipelines of hardware and software and models and LLMs stitched together. They're going to be on demand things that'll get spun up, logical things get spun up and they're torn down as needed and solve a specific problem at a given time. But you raise a good point. It's a question is do all the enterprises have the skill to do that? That's going to be an interesting conversation. And I think one of the things that we're not talking about as an industry is a role in the cloud world, we talk about it in the IT world in general, but mostly in the cloud world, the role in the global system integrator. What is going to be the role of an Accenture or Deloitte to help the Jamie Diamonds as well? Maybe Jamie Diamond has PhDs working for him, so he probably doesn't need them. But the mid-tier guys.
Dave Vellante
>> Yeah, interesting.
Sid Nag
>> To help these enterprises to spin up these AI factories on demands? What are the methods and procedures? What are the best practices so that they can be replicated on demand and torn down as necessary. So that's a whole different conversation I think we got to have.
Dave Vellante
>> Interesting, Sid, because you remember the days of SAP, you'd have to hire a global system integrator to implement it. Then it got too expensive. There was the SAP backlash. Oh, system integrators, they're in big trouble. And then what happened? I mean, and the cloud came along, they said the same thing. Now AI comes along, oh, the system integrated services companies are in big trouble. But it's a complicated matter for a lot of organizations. They've got to figure out their data. They've got to harmonize that data. They've got to figure out how to inject process into that whole equation, build a digital representation of the enterprise. A lot of organizations, to your point, just don't have the skill sets to do that. So they're going to need the GSIs to do that. And I suppose vendors could have solutions. Right now I can buy a full Nvidia stack if I want, but not everybody wants to do that. So that's why they've introduced capabilities to integrate in. It's a heterogeneous world, as you know.
Sid Nag
>> Yeah. And then there's the elephant in the room, agentic.
Dave Vellante
>> Oh, yeah. I was going to ask you about that.
Sid Nag
>> And everyone's waving their hand and saying, "Oh, agentic will do that." I mean, yeah, it'll do that, but right now it's a lot of smoke and mirrors that people are preaching. So people say, "Oh, I don't need a GSI, agentic is going to do that for me." Is that really going to be the case? I mean, I don't think so. GSIs will use agentic as a tool, but they will still have to build the process, the methodology that is bespoke for that particular environment.
Dave Vellante
>> I got to make you laugh. So George Gilbert built the graphic for me. I'll show it to you. For those who can't see it's a camel And it's got fleas on the camel's back and it's surrounded by logos of basically agentic company with a lightweight interface to a backend LLM. And there's more of those than there are fleas and the camel's back, and they're probably in big trouble. The fundamental value proposition is shifting or it's going to get subsumed by the LLMs or maybe it's not going to deliver that much value. But that's really what's required here is you can't just wave a wand and expect agentic is going to solve all your problems. So how do you see these agents? Are they worker bees that are highly governed, but they're not deterministic? They are AGI or GenAI. Is it RPA 2.0 or is it more than that?
Sid Nag
>> Oh no, I don't know. I don't know if I want to go there.
Dave Vellante
>> You hope not, right?
Sid Nag
>> I don't want to go to RPA 2.0 please.
Dave Vellante
>> That would really suck.
Sid Nag
>> That scares me. Shut the interview down now. No, no. I'm thinking of these agentic agents or whatever as a helping hand. Now, I think there's one step between what people need today and AGI. I don't think we ever going to see AGI in my lifetime. Other people are talking about world models, blah, blah, whatever. Let's park that conversation. That's a whole different show. But I do think that agentic is being a little bit overblown in terms of how it's going to be used. I mean, they are going to be using front office operations, back office operations, and a whole host of other typical enterprise tasks, but I still think that it's going to be hybrid where some things will be automated and some things will be working in hand in hand with the humans and as an assistant. I think that's what's going to be evolving. And I think that's necessary because that's going to give the buyer and the enterprises or the CIOs the confidence that this is not all smoke and mirrors. They want to see it in action. I mean, everybody I think wants to see everything in action tomorrow morning, but they're damn scared that something is going to go wrong. So I think that's the step in the right direction where you have certain things you use agentic for, certain things you use agentic for assisting humans, and then the role of the GSI who can help stitch all these things together for those who don't have the wherewithal to do it themselves.
Dave Vellante
>> Do you worry about jobs? I mean, you know a lot of young people, I do too, they struggle coming out of college to find jobs. You have a deep, deep technical background. You went to IIT, you got, I think, multiple PhDs. Would you recommend a kid today taking a similar path? Would you recommend him?
Sid Nag
>> Hey, listen, listen, listen, I heard Peter DeSantis and Dave Brown talk about EC2 solving the jitter problem and virtualization in the nitro cards, and that was based on a book by John Hennessy and Patterson that I used 30, 40 years ago to write my-
Dave Vellante
>> David Patterson....
Sid Nag
>> master's thesis and I wrote quote to build a virtual memory for an operating system. Guess what? Some things are timeless, alright?
Dave Vellante
>> Yeah. Okay. So computer science got a ways to go.
Sid Nag
>> A ways to go.
Dave Vellante
>> Alright, Sid. Good. Great to see you, man. Thanks so much for coming in. We really appreciate it.
Sid Nag
>> Thank you for having me. It was very fun to talk to you.
Dave Vellante
>> Alright. And thank you for watching the AI Factories, NYSE Wired + theCUBE. I'm Dave Vellante for John Furrier and the entire CUBE team. Right back right after this short break.
Dave Vellante