How AI Stacks are rewriting the Rules of Business

0:04Good morning, good evening, wherever you are. Welcome back to the Enterprise AI Show. I'm your host, Brian Grace Lee. And today we have a fun show. Longtime friend of the show, really good friend of the show, Dave Vellante is going to join us. Co-founder of The Cube. Many of you would know Dave if you've kept up with the tech industry for any period of time over the last, I don't know, five years, 10 years, 15 years. Dave runs one of the most popular industry shows that just keeps up with all things tech, all things AI, all things cloud. He's out at the events. He's speaking with executives. Just a guy who we have tremendous respect for, both as an analyst as well as somebody who really has his finger on the pulse of what's going on.

0:43We have a fun show today. We really dig into a bunch of topics. We get into some of the trends that are happening in AI from the perspective of how things are moving up the stack, where the value in the stack is going. His group has written a really, really good look at just how things are evolving. And then we dive into a whole bunch of topics, you know, just around big industry trends. So Dave does a great job of analyzing and of giving us the big picture of things and really kind of, you know, helping us frame a lot of the things that are happening really fast in our industry. So fun show. Hope you enjoy it.

1:14We'll get to it right after the break.

1:19Today's show is sponsored by Nasuni. There's a growing gap in AI right now between what's possible in theory and what successfully works at scale inside an enterprise. The difference comes down to unstructured file data. Many AI initiatives struggle because the file data they depend on is scattered, unstructured, and disconnected from where and how work actually happens. Nasuni changes that. It brings your unstructured file data into a single secure foundation so AI, both generative and agentic, can access it with the context, governance, and performance it needs in production. Bring AI to where your unstructured data lives.

1:54See what it takes to activate your data for AI and request a demo at nasuni.com slash AI.

2:03Today's show is sponsored by Sharegate. You're out of time. Copilot needs to be deployed ASAP, but your tenant really isn't ready for it. Years of data, permissions, and users lurk in its shadows, ready to be exposed by AI. Sharegate Protect sees it all, so you can find the exposure risks, fix them fast, and deploy AI with confidence. Microsoft 365 Governance, they've got this. Learn more at sharegate.com slash protect. And we're back. And, you know, folks, there's been a lot going on in the AI industry as we've, you know, constantly covering, but it feels like there's been maybe a lot of change.

2:41It felt like while there was a lot of announcements over the last six, nine, 12 months, a lot of them were sort of moving in the same direction, same sort of trajectory. And it feels like the last month, two months or so, there's been some stuff that's been making it bounce in unexpected ways. And so I'm really excited today to have one of my favorite people in the industry, somebody who I have had a chance to work for, I've looked up to for a long time. Very excited to have Dave Vellante, who is co-founder, CEO of The Cube. I know a lot of you that followed the cloudcast for a long time follow The Cube, but it's been a little while.

3:13Dave, great to have you back on, man. It's been way too long. Hey, Brian, mutual fan club here. You know how I feel about what you guys are doing, and I really appreciate the invitations. It has been a while. I know. I know. I follow your stuff. I listen to your stuff. You put out both a weekly really, really good kind of recap of you guys are always attending the show as kind of trends. And then obviously you and Dave or you and John have been doing, you know, Cube stuff and podcasts for a long time. I reached out to you not too long ago, just before vacation started and stuff. And you had written a great piece, another kind of long piece about something you called how AI stacks are rewriting the rules of business.

3:50And this felt like kind of the perfect thing that that Cube Research does that you are really good at doing, which is looking back, you know, looking at the big picture of things, especially with stuff changing. And I thought it would be good for us to really dig into it because it focuses a lot on how, you know, we've gone from talking, you know, very much about models and sort of a couple of dominant companies to, you know, this shift now that a lot of people are talking about, which is, you know, what are the software stacks going to look like on top of these models? Is there going to be a blurring of lines between an app and the data sources and all that stuff?

4:28And so I thought it'd be good to dive into it. As you were thinking about this, what was the big framing you were trying to get across? Well, I have to give credit to George Gilbert and David Floyer, whom we just recently lost, sadly. But those guys have been working with me for many years now. I mean, George and I We go back to what we called, I think we called it the fifth or the sixth data platform. We were trying to figure out, okay, what's next beyond Snowflake and Databricks? And then it evolved into the whole Gen AI thing and then agents and everything else. But the observation and the framing that we were using for that post is, we observed that when we went from on-prem software to SaaS, everything changed, right?

5:13The pricing model, the delivery model, the whole way in which we deployed software completely changed. Right. We went from CapEx to OpEx, et cetera. But one thing that really didn't change was the operating model of the business. It affected IT. It obviously created huge wealth for software companies, SaaS companies, but nobody else really noticed. Maybe the software was a little bit better, but the operating model of the company didn't change. We feel like the same thing is happening here, but different. In other words, the pricing model is going toward maybe from consumption going to outcome pricing.

5:55Certainly the technology model is changing. And we think it's going to affect the business side of organizations, not just the IT department. And that had implications in our research in terms of software-like marginal economics moving throughout the business to the extent that you can deploy AI quickly. And it may portend a sort of winner take most, not for the technology vendors, but for the buyers, those that can apply AI most effectively. And so we've been kind of looking at the software stack. You know, shout out to Alex Williams. It's a great name, right? The new stack. Well, there's definitely a new AI software stack emerging, and we wanted to define that more precisely.

6:40And it continues to evolve. I mean, daily, it's unbelievable how fast things are moving. Right. Yeah. You know, as I look through it, I think you guys have always done a really good job. I know, I mean, I remember many years ago when I was working with you guys, you know, David Floyer was really introducing this idea of systems of engagement, systems of record, and how those were evolving. And it was, you know, some of it was for sort of slow-moving data, slow-moving tasks in your business. You had more that was more dynamic. I think now you guys are kind of introducing this idea of systems of intelligence.

7:12And that feels like, you know, it feels different. It feels like, okay, what is that going to mean? And I think you guys are kind of unlocking the beginning of not just how does this technology work, but what can I do with it? And then, like you said, as you're thinking about what you can do with it, you have to be really cognizant of what is the cost of that? What's the marginal cost of that? And then how are we going to measure intelligence, measure outcomes of intelligence? It feels like a whole new area that has some aspects of the old stuff because everything builds on it. But what are these new things that you guys are really kind of starting to think about, bigger picture things you're thinking about as you're thinking about these systems of intelligence?

7:57Yeah, I mean, I'll start with sort of the buyer perspective. And anytime we see these waves, people get confused. They see the stock market attacks SaaS, the so-called SaaSpocalypse. And so if you're an IT buyer or CIO, you're like, well, maybe there's a smoking gun there. and I might want to rethink. I had a CIO tell me the other day, he goes, I can't promise this, but I think I just signed my last ERP, you know, license. We'll see, you know, and so it gets confusing, right? Do I use the agents that SAP and Salesforce and Workday and everybody else ServiceNow are bringing to me? How are they going to interact with my other agents?

8:39So that's, I think, confusing, obviously, a lot of people. And the flip side of it is the vendors. You know, Ray Wong wrote a book one time, Everybody wants to rule the world. And so, you know, it's will the LLM vendors take over? How will the traditional SaaS companies and the deterministic software defend its turf? And I guess the point to answer your question is we see this emergence of what we are calling the system of intelligence. I mean, we've been talking about ontology and using Palantir as an example for many, many years now. So it's an extension of that sort of Palantir ontology. Imagine that for everybody.

9:16And as you think about the stack, we've got this$50 billion BI business, which is largely building dashboards and metrics and dimensions. And it's now sort of becoming the way in which you interact with software is changing. It's not necessarily through the application anymore. It's through either an LLM or some kind of harness or a coding agent or sort of a business agent or a co-pilot. And we see sort of the front end, that user surface and that back end system of intelligence as creating this powerful learning loop where the tacit knowledge of the enterprise starts to get incorporated. And I'd say it this way, Brian.

9:56The observation is, again, we always talked about data as an asset. Now process is becoming an asset. and a first-class citizen, that tacit knowledge of the enterprise is now becoming incorporated into that software stack. And that is novel and it's going to be a high-value piece of real estate. And we call it the system of intelligence and we borrowed this, I have to give credit to Jeffrey Moore, George and he are working together very closely. George is working on a feature article for The Economist. And so this research has sort of grown over the last several years now. Yeah, no, and it's interesting.

10:36I think when you talk about that sort of like process being valuable, I think that is something that companies have focused on for a long time around, you know, how do we improve process and so forth. But I don't think we've necessarily tried to connect the dots between, you know, what are the things that happen in process or the technology behind it or the steps or the, you know, and what does that mean for the business? Because if you can quantify that, if you can codify it, now, like you said, that becomes the foundation of agents, right? You don't have to think about it entirely as like that becomes the thing that displaces a person, but it does become something that if you have great understanding and great control of your processes, that's just as valuable as having an incredible data set.

11:24Because in a lot of cases, companies now can buy data. They have similar data sources, similar data sets, but it's what they do with it, whether it's that process or that manipulation that oftentimes becomes those differentiators. So, yeah, it'll be really interesting to see how that starts to evolve in terms of can companies do that. I think the other thing that's interesting is we hear these stories of people who are saying, like you said, did I write my last check for an ERP system? And then there's talk of, hey, I'm going to build my own ERP system. And what's your current thinking around that?

12:02Because it's obviously on one hand, it sounds sort of interesting because obviously some of these systems get big and old and clunky and slow and complex because thousands of companies have been asking for features. But at the same time, I don't know that there's this newfound hunger by companies to say, I really want to build an ERP system. It feels like we're sort of somewhere in between there right now. Like the, we haven't seen the example of like the great homegrown ERP after 20 years, but it does feel like there's some, you know, interest because of what LLMs and agents and stuff are able to build.

12:39Well, I mean, you saw this, right? The corner office, the CEO said, okay, thou shalt do AI. And we said, okay, so what do you do? You start playing around with AI. And so you had this sort of top-down edict. and and now you see all this bottoms up sort of activity going right kind of like the original pc revolution where it was all kind of personal productivity and everybody's everybody's doing agents and you know like the joke as you're walking around with your laptop open you don't want to close it to stop the job right and uh and and so and you realize okay wow these llms are really really powerful but they certainly are not deterministic and and a lot of talk about bringing determinism together with stochastic software.

13:22And we are really far from having that system perfected. And so I'd say a couple of things. One is, I think in many respects, determinism is sort of a myth because the determinism lives within departments. Your CRM, there's a single source of truth maybe for the sales department, another one for the finance department, and maybe that's the one in theory for the company, but logistics and HR, et cetera, they all have their own systems. And the promise of AI is to bring all those together. And I think that, you know, when you think about the LLMs, they, you know, don't necessarily have even close to that capability.

13:59So how's that going to play out? Presumably what they're going to do, back to Ray Wang's theory about everybody wants to rule the world, they're going to partner with the SaaS companies who have that process knowledge and that sort of underlying application and business logic, you know, embedded. And of course, Alex Karp says, well, they're just going to steal your alpha. Right. And you're screwed if that happens. So we're the answer. You come to Palantir, which, you know, we all know the story about forward deployed engineers and, you know, a million dollar, you know, a year that you're paying those guys.

14:31And so it's too complicated there. The promise of the yellow limbs as they simplified, I guess I would say on the one hand, And I'll tell you, George has laid out this amazing framework, but it's complicated. It feels like the Hadoop stack in some regards. The flip side of that is, OK, well, the LLM vendors, the agents don't do a great job yet, certainly across the enterprise mess, the messy middle. But I do think that ultimately it's got to be simpler in order for AI to live up to its promises. And here's what I would say to the, I mean, I definitely understand Alex Karp and his rant and how the LLMs have the potential to steal your alpha.

15:14The flip side of that is if you can create a new operating model that allows you to sort of dominate your field, maybe you don't care. Maybe you say that new operating model is my moat because I can move faster because I'm AI native and everybody else is trying to sort of bolt on AI to their business. And so I think these are the types of things that organizations are struggling with right now, and therein lies the confusion. Yeah. No, I completely agree, and I think it's something we've been wrestling with a little bit, which is we all sort of lived through the cloud era where it was like, okay, you had this bottleneck, which was IT.

15:55AWS comes along and sort of says, hey, what if that, you know, department of no could kind of go away, but you still were using things that you're familiar with. And, you know, you could kind of unlock this potential of building it at whatever speed you want. You can do scale, whatever you want. And what we've talked about a number of times is like, as soon as the AWS thing took off, and maybe it was a year after it took off and it was the first reinvent or whatever, There were just a number of companies that kind of came out of, you know, they were startups or it was a Netflix or it was whatever.

16:27And there were sort of lighthouse customers that you could point to and say, that's what cloud native looks like in terms of how they operate. It's what cloud native looks like in terms of how they're applying it to their business model. And I don't know, maybe you have better insight to this or you've had side conversations. Like, I feel like for all the talk of companies saying they're going to be AI native and they're throwing a lot of AI at stuff, I don't feel like there's the Netflix of AI yet, right? And I don't know that that doesn't mean they won't emerge, you know, or, you know, the Instagram of AI or whatever it is.

17:02But I can't quite figure out if like that company hasn't emerged yet because people really haven't figured out how to build that new thing, that new moat, or if they're out there and they're just keeping it really, really quiet because they're like, you know, the more information is out there, the more somebody else is going to be able to sort of replicate it because they have the power of LLM. Do you get any sense yet of, you know, is there a Netflix, a lighthouse, you know, a light tower kind of company out there that people that we're looking at the same way we did with Netflix, with the AWS world?

17:35It's a really interesting question, Brian, and I kind of get it a lot. And if I had to bet, I'll tell you, three months ago, if I had to bet, I would have said it's, you know, Anthropic and OpenAI. Yeah. And I still kind of lean that way, although you see what, you know, Thinking Machines is doing. You see what the Chinese open weight models are doing. You see how Jensen is going after this. I still think that the leading frontier model vendors, either they figure out that they, and I believe they understand this, have to go hard after the enterprise stack that we can talk about consumer as well.

18:15I don't see consumer as a negative because of volume and learnings. But I do think the real prize is in the enterprise. Unlike, like you're saying, you know, Instagram and the social media wave was sort of a consumer play. Right. I really believe that the frontier model vendors understand that they have to own a major chunk of that high value real estate or they're screwed. All this money that they've raised and the valuations aren't going to hold up. And to do that, they either have to partner with and or acquire, you know, these companies and they have to invent. they've got to do it through organic development.

18:52They've got to do it through partnerships and they've got to develop it through ecosystems. And so I would bet on, you know, those companies or Google, I think is within its sort of narrow world is going to do great, but I don't see it as the sort of horizontal play. And I come back to, and this is kind of counter to the narrative right now. I think the leading frontier model vendors, their advantage ultimately is going to be simplicity because this stuff is way too hard. And my hope is that the LLMs can do more than they can do today and do it more reliably. But in order to do that, they're going to have to bring that sort of determinism into focus.

19:32And that's going to happen through partnerships and M &A. And the partners are going to have to be very careful. But on the other hand, for the SaaS vendors, their survival may depend on it. Because if they try to... And you see what Salesforce is doing, going headless, and some other SaaS vendors are saying, no, you've got to come through our application. Remember the whole Benioff and Satya thing? Benioff called them Clipper, and Satya said that it's just going to be agents talking to CRUD databases. Oh, wow, it turns out that Satya's personal productivity software was the most exposed because the file formats are already open.

20:07And so Anthropa can just go in and get them. And so it's just so interesting right now, the dynamics. And that's why I do come back to ultimately somebody has to make this a lot simpler than it is to be adopted because we just not everybody can afford to go higher a million dollar a year forward deployed engineers. Right, right. Yeah, it's it is really interesting. I mean, you know, again, you and I have been doing this for a while, so I think we tend to draw on on historical stuff. I mean, you know, there was a long period of time where we would watch every AWS reinvent and you would just go, okay, AWS just talked about, you know, six new startup partners.

20:46And then five minutes later, Andy Jassy would announce six new features that you're like, well, he just, you know, they learned enough about the customer usage patterns and they just built those into features. And so, you know, we're all watching Anthropic or OpenAI or whatever roll out, you know, Claude for marketing, Claude for work, Claude for business. And it's interesting because on one hand, you're like, okay, we've seen this movie before. We're seeing, you know, kind of look at their data, seeing patterns. They're going to make that capable. And that plays to your idea of like, they got to kind of make it simpler.

21:18They got to make it built in. The interesting part of it is, you know, the software business has always been built on, you know, in a lot of cases, volume, per seat licensing, number of cores that it's running on. And it's going to be interesting to see how they price this thing because, again, you've got massive costs that are constantly ongoing. And you get this idea of like, well, maybe they're moving towards a pricing model that's around outcomes, right? Like a value of intelligence or outcomes or whatever. And that's a hard thing to do, right? It's sort of like trying to build an ROI model.

21:54We always talk about it, and then you really don't necessarily achieve it, but it made some progress and all. I think that's going to be a really interesting thing to watch as well because they're piling on capabilities. I mean, at a faster rate than even AWS piled things into their console, which eventually got overwhelming. And they're trying to change the pricing model at the same time, which is like – it's always amazing to me how many variables they're playing with and throwing at the market at the same time before anything's stabilized. which it's hard to wrap your head around, having a point of focus, I guess.

22:31I think Phil Hurst coined the term service as software. He uses services as software. We talk about service as software. And I guess the way these FDE companies, forward deploy engineer companies, popping up all over the place. and you're seeing them, everybody's announcing it. Alex Karp says, everybody wants an ontology. Amazon's announced it. Google's announced it. Databricks has announced it. And you're seeing these sort of startups also. The way they price is they have their FDE go in and say, with their strategy or the salesperson, if you will, say, okay, this is the outcome you want. And by the way, these projects are very narrow.

23:14They're not dealing with the whole enterprise mess. It's really, okay, this is the objective and it's very narrow. But okay, that's the outcome. These are the steps that we're going to have to take to get there, but very much like a services company would. And we're going to accrue points. And this is what it's going to cost you when we hit these milestones. And that's how they get paid. And the incentive system for the FDE is just very tightly tied to that. And so the client pretty much knows what the outcome is going to cost. And they can track it. I mean, you know this. This is a software development sort of agile life cycle.

23:50And they're sort of copying that. You know this better than I do. And bringing that in. And so to your point is, okay, well, how do the LLM vendors play in all that? And what you're seeing, of course, everybody's now token maxing is out of favor. Everybody's freaking out about token costs. I do feel like if you look at the macro, about four or five percent globally of company revenues, organizational revenues or budgets, if they're nonprofit, are spent on information technology. Yep. I think that number is going to explode. And if it doesn't, frankly, then this whole AI thing is just a bunch of BS.

24:27And I think that could double or triple by the middle of the next decade as organizations essentially scale with a lot less labor. I mean, if you can get two to three X your revenue with the same or 10 % more labor, that's a big win. And I think you're going to see a shift toward tokens is going to be that sort of bigger expense away from potentially labor. That's how you used to scale. You just hire more people. Right. And I think you're already seeing it. People want to be token generators. They're building, you know, bringing whether it's a Mac mini or, you know, mini AI factories on prem to be their own token generators.

25:09And I think there's going to be two things. One is it'll be a combination. It's going to be a highly distributed system all the way out to the edge. And I think token price, token costs are going to keep coming down. I mean, it's going to be, you're seeing 90 % a year, but the token usage is exploding. But I think that trend continues. Right, right. In a favorable way. Right. Well, and I know you and I talked, you were just chatting before the show. you know, we've talked a bunch. The interesting challenge about the economics right now is like you said, yeah, you know, inference costs are getting better.

25:40Token prices are coming down, but usage is sort of going through the roof. And the challenge of it is up until now, everything has been NVIDIA's ball game, right? NVIDIA is the collector of sort of all of that. You recently were at, you know, at the NVIDIA show, you were engaged around what NVIDIA just dropped out there. We've been waiting for a number of years for some sort of, I don't want to say cracks in NVIDIA's monopoly, but like some other alternative player to come along? Where are you seeing NVIDIA start to make some moves that are interesting to you? Well, I think NVIDIA still has a substantial lead.

26:15Jensen put out something today. Of course, you're seeing his open source mantra. But he puts out something the other day about the challenges of keeping up with that one year cadence, energy and other supply challenges, data center build outs, et cetera. Because Moore's Law, we all know Moore's Law and Jensen's Law is kind of every 12 months, buy more, make more kind of thing. AMD last week kind of laid down the gauntlet. I put out a post prior to AMD Advancing AI, said, look, AMD doesn't have to take out NVIDIA. It just has to be an indispensable possible second AI platform. I feel like Lisa Su kind of rejected that premise and basically stood up and claimed leadership across the portfolio, CPUs, GPUs, RAC, software.

27:08They did the Xilinx acquisition, so they got out to the edge. And she sort of basically said, we are the leader in all of these. Now, after the AMD Advancing AI conference, the sort of narrative, which we were buying into as well, was you have this idea of extreme co-design. Extreme co-design, for people that don't know, it sounds like you're going to co-design with all these other entities, but it's really extreme co-design inside of NVIDIA. And that's how they've been able to do the one-year cadence. Well, AMD is a separate or different philosophy, which is open co-innovation across an ecosystem.

27:45They've trotted out Meta and Anthropic and OpenAI, and they've invested in TensorFlow, which is an alternative NeoCloud, which is an AMD exclusive, et cetera, et cetera. And they are co-innovating in an open, everybody's open, but they call it open, with the ecosystem. Okay. So I'm not saying one is better than the other. Clearly, NVIDIA is working. The question is, it now gets down to, so Semi-Analysis put out a really detailed technical analysis this weekend, talking about some of the challenges that AMD is having. And it's sort of a light bulb went off in my head, which is the real competitive moat is going to be engineering velocity.

28:30In other words, not the engineering talent, but the pace at which you can actually iterate, incorporate feedback, whether it's from the ecosystem, if in the case of AMD or your sort of internal flywheel, if you're NVIDIA, the pace at which you can keep innovation going. And that's kind of the new challenge. The whole ecosystem, Brian, as you know, wants an alternative to NVIDIA just because they want a second source. They want negotiating leverage. You know, they don't, they feel like sometimes they feel like they're getting bullied, but they're afraid to say so. So everybody's cheering on AMD.

29:08If I had a bet, I would say NVIDIA is still going to maintain, you know, 75, 80 % of the market. AMD will get, you know, it's mid single digits of a huge market. So their revenue can easily double if they do that. And then Broadcom comes through, you know, the backend with Google and TPUs and their other relationships with, you know, the co-development that they're doing with ASICs. And so those are kind of the big three. But I don't see cracks in the NVIDIA armor. I do see a little bit more competition. But I think that with Jensen at the helm and the founder-led CEO, that continues for quite some time.

29:47But I love the competitive posture that Lisa Su put forth. Yeah. No, I think it's, you know, we've said for a while, like, it's good for our industry, especially when we're going through these new inflections, to be a lot of competition. I know it makes things complicated to analyze and kind of get the big picture of it, but it's kind of the only way for this stuff to scale at a price that becomes reasonable for people, that is not just only available to the biggest companies or people with the deepest pockets. So whether we're talking about the internet or mobile or whatever, it never was built upon only the most expensive, you know, computing hardware.

30:29You know, that was a little bit of like Cisco back in the day. And then obviously lots of open stuff came along. Yeah, it's, I think it's really interesting to sort of watch how, you know, the players who would like a NVIDIA alternative are, it'll be interesting to watch and see if they act like a community, like sort of open source communities do. And granted, this isn't exactly open source, But at least it's a number of people with a lot of smart people kind of working together on stuff where it's almost like they have to realize that collectively, whether it's OpenAI and Anthropic and Meta and whoever else, maybe it's 10 other companies, that they collectively have to build a very good substrate, that hardware or accelerator substrate, so that they can then go differentiate themselves at the harness layer or at the application layer.

31:21But if they sit there and fight over their contributions to, you know, sort of an open-ish AMD ecosystem, they're just going to watch NVIDIA, you know, continue to sort of race ahead of them, no matter what they do with their harnesses because they don't have any control over the hardware. And it'll be interesting to see if these guys who inherently are so competitive with each other can figure out some of that sort of open source way of working together. Because that feels like the thing they have to figure out if they want to, again, like you said, have supplier flexibility and bring cost down and move the value chain up a little bit higher and stuff.

31:56And I think they will. I mean, I think AMD will sell every system it can make. But they've been remarkable. I mean, they've gone from a company that was attacking x86 and gaining share, you know, with the Zen microarchitecture and all that with a wounded Intel, granted, NVIDIA is not wounded as Intel was. But they compressed a 15-year NVIDIA lead into like five years through, you know, core R &D. They continue with their, you know, Epic line and Instinct as their AI accelerator. And they bought Xilinx. They'd done like$60 billion of acquisitions. About$49 billion of that was Xilinx. But then they built out Rockum, which is their CUDA alternative.

32:41And they made a big investment ecosystem. They compressed a 15-year lead into five years with organic R &D, M &A, and an ecosystem build out. I mean, truly remarkable second act or third act even for AMD. And to your point about open source, I mean, open source has been a big contributor to this. And I think that's how they've done it. And this is what is so interesting is, you know, open source is going to be always this sort of interesting lever that people can pull. It's just that, and by the way, NVIDIA heavily, you know, investing in open source and using open source and open sourcing selectively.

33:22So they're, you know, they have the marketing wars of who's more, I'm more open. No, I'm open, but that's fine. Let them do that. But I think that this is really interesting setup between extreme co-design and open co-innovation. And I think it comes down to that engineering velocity and the speed at which your engineering factory can put out new innovation. Yeah. Yeah. I'll wrap it up with one last thing. And you mentioned it a little bit. Alex Karp, CEO of Palantir, had his sort of famous CNBC interview. I don't know. It's probably been a month now. But it's interesting to watch. I mean, I think I may have done a show about it.

33:58I'm sure people have seen it and read about it. I mean, essentially what he said was, you know, the point we were sort of talking about is the more companies are using some of these frontier models, again, whether it's Anthropic or OpenAI or whatever it might be, the more you're providing them your data, the more they're learning about your business, the more they're going to be coming after your business. And we've seen that, you know, whether they're talking about, you know, legal services or marketing services or, you know, things with hedge funds or whatever it might be. It's very interesting to watch all of these folks position themselves sort of on both sides of the fence, right?

34:34It's, you know, we want to be open and our part of the stack is the most important piece. The other piece is a commodity. Do you get a sense that like all – like I don't remember a time in the past where companies had to be so much on both sides of the fence sort of all the time because the dynamics were sort of changing. I feel like that in and of itself sort of confuses people is, you know, if you're an end customer, you're like, I want as much flexibility as I can get. but things are confusing and I'm not sure if like, is this Palantir sort of harness centric way of doing it? Is that the right way of doing it?

35:10Should I be looking for an open harness? I mean, do you get the sense that people are more confused than they were six months ago, a year ago, just trying to understand where the vendors are going or where the technology is going? I do. I mean, I think people, you know, people watch the all in, you know, listen all in podcasts. They say, oh, so this is so easy. Download, you know, open claw and go, go, go. And then they do it and they go, oh, Maybe it's not so simple, number one. I guess the other thing I would say is Alex Karp, the counter to Alex's rant is, well, if Anthropoc is going to exit the year at$100 billion run rate, which is about 10x what Palantir is, somebody's buying this stuff.

35:51Somebody's leaning in and adopting it. And of course, OpenAI is accelerating. And again, I come back to simplicity. I generally, by the way, I agree with you. I think people are more confused than ever and concerned. I think sovereign is a huge topic that we haven't touched on. And, you know, what is a sovereign AI stack? I think everybody should think about what that means and define what that means and figure out where you want to be on that spectrum. I think to answer your question, I mean, the closest I'm trying to think of examples. I mean, back in the PC era, people, you know, had to get close to Intel.

36:25Remember all the risk players, you had digital and data general and Sun and HP and IBM. Okay. But Intel was the main player and you wanted to have a good relationship with them. I guess similar things with Microsoft. And I think there's a little bit of that with the AWS ecosystem. You know, thou shalt not say multi-cloud at our reinvent. Right, right. You know, and so there was some of that. I think this is, you know, much bigger. We've never seen anything like this in terms of valuation and money raised and commitments that are being raised. You know, I do say, I mean, we've heard people talk.

37:06This is not my novel idea, but people have said, and it's true, Sam Altman and Dario Modi, they're not the greatest spokespeople for AI. They haven't helped with AI positivity. And you look at what's happening in China. Everybody's like super excited about AI. So there is that backlash. I mean, it's really a bifurcated, fragmented mindset. You know, AI is bad versus AI is good. And, you know, it begs the question. And you have China is now the big wild card. I mean, what China is doing is remarkable. Right, right. Yeah, that's probably a whole other conversation we should have about just how, you know, like, I don't remember.

37:47I mean, yes, there's always been a certain amount of politics and geopolitics that is around certain parts of technology. But it is now deeply ingrained in what feels like every announcement, every decision, where funding is coming from, whether technology is allowed. That's probably a whole other conversation we should have at some point just about people have to think as much about what are the political ramifications, What are the regulatory ramifications as like, am I just getting the right technology? Is this good technology? Can I afford what this stuff is? And we've never had to deal with that on like a day-to-day kind of core buying decision basis, which is a whole new element to this whole thing.

38:32Yeah. I mean, I guess I would say the promise of AI, of course, is this massive productivity boom, the ability to scale with a lot less labor. where the North Star is this full stack digital twin, this digital representation of my business in real time. Yeah. And that is the promise. John and I were in a debate at Advancing AI, but I was saying, well, of course it's a bubble. I feel like we're in a bubble. No, there's no bubble. I think we're in a bubble. The question is, when will that bubble pop? Right. And that's what I think as everybody, maybe not near-term concerned, but with all the circular reference and all the debt now, you know, people taking on, companies taking on debt, you know, people are kind of sleeping with one eye open there.

39:16I don't think we're anywhere close given all the constraints, but eventually that's something that people I think are a little concerned about, but you know how it is, Brian, when the surf's up, you got to go. You can't not participate in these markets. So it's worse to miss them than it is to, you know, get hit. And so I think you got to build fortress balance sheets. You You got to think about what happens when, but you got to put the pedal to the metal and get on the AI curve and go for it. Yeah, yeah. David, I'm going to let you go. We could go on for a long time, talk about a lot of things, but I want to make sure we get you back on more frequently.

39:52So hopefully we can make that happen. Folks, if you're not listening to Dave's stuff over on the Cube or on his breaking analysis, I'll put a link in the show notes. Make sure that's on your reading list every week. It is very, very well worth all your time. So Dave, great to see you again. Great to have you back on the show. And hopefully we can do this again real soon. Thanks, Brian. Thanks, everybody. All right. Thanks, everybody. Thanks for listening. And we will talk to you all next week. Thanks for listening. Check us out at theenterpriseaishow.com for past shows, newsletters, and all things Enterprise AI.