Exclusive Coverage: theCUBE at IBM Think | Mark Beccue | Independent Analyst

theCUBE.net

Clips
More from Exclusive Coverage: theCUBE at IBM Think

Mark Beccue

Independent Market Research Analyst

play_circle_outline IBM's shift towards being a better partner and their strengths in AI

play_circle_outline The use of open source and proprietary models in AI

play_circle_outline Importance of understanding and leveraging AI in the enterprise

play_circle_outline The Role of AI Industry Analysts in Operationalized Enterprise: Integrating Data, Building Guardrails, and More

Info

Mark Beccue | Independent Analyst

Mark Beccue

Independent Market Research Analyst

search

>> Hello, welcome back to theCube's coverage of IBM Think in Boston, Massachusetts. I'm John Furrier, host of theCube. We are here on the ground getting all the action around what's going on in the enterprise, enterprise AI, and still the IBM and their comeback as they continue to rebuild the company and reinvent themselves. Mark Beccue is with me. He's an industry analysts in AI. Mark, great to see you. Thanks for coming on theCube.

>> Thanks, John. Good to be here.

>> This is our pop-up cube. We do whatever it takes. We're in a little corner here. Whatever it takes to get the story, we will get it. That's the way we operate, get that data. Speaking of data, Mark, this is the top number one conversation is that consumer AI is hot. The GPUs are selling like crazy, new custom silicon's coming out. You saw Google last week at their event talking about custom silicon, and they're ahead of the game there. You got Amazon figuring out what's going to go on with a new CEO, Matt Garman, a very product-oriented fella I've known very well. The world's completely changed, and it's clear that it's a platform shift. So you're starting to see some visibility in the market around where the AI is starting to drop a little bit. Certainly, there's a platform refresh at the bottom of the stack, middle layer and the app layer. Generative AI is changing everything 'cause it's a disruptive enabler. So there's enablement, there's value creation opportunities, but someone loses. Things will be disrupted.

>> Yeah. What's interesting, John, we talked a little bit before about, I think what happens is there's not a real good grip in the enterprise space about how much all this costs. So when you think about, one of the things that worries me is workloads. Where are you going to run these AI workloads? How much does it cost? There's training versus inference and all those things. I think what the problem is is so much is in flux we don't know... the models run big. Are they going to get smaller? What are those workloads going to look like? I don't really think there's a good idea of what that cost is.

>> The question everyone's asking is, okay, I have data, that's pretty clear to most folks that data is a valuable asset here, "How do I configure the data, number one?" That's a platform architecture decision. Then two, "How do I run it in the end-to-end workload that was predefined before generative AI?" So a lot of these companies have these workloads. They know what they look like. They're somewhere in the cloud, some are on hybrid, but now you have to change it with the new data model and coming in. So the first question is, what do I do with my data? How do I figure that out?

>> Well, that's an age-old question, isn't it? It's like big data was our big... how long have we been talking about big data?

>> Since 14 years ago? I do.

>> So-

>> Data's the new oil I heard back then too.

>> Data is the new oil and we've heard that a lot, right? I guess, again, a big flux really, right? When you've got silos, some of this stuff is, some of it's structured, some of it's unstructured. Do they want all of it? Do they have privacy restrictions about financial data that they can use, whatever it is? I think it's less of a technical issue and more of a management issue still. So it's really outside of AI. I think what's different, and maybe you touched on is that GenAI now makes it affordable to think that you can get at it.

>> What do you see as the number one challenge in the enterprise? Is it compliance, security, skill set, what to do? What's the first three steps to take? In your analysis, what are you seeing in your research?

>> Yeah, that's a great question. I think it's a lot about lifecycle management. You really don't have a whole lot of companies that understand AI lifecycle management. If you look at the market at the ones that have succeeded, and let's call them that actually have products out, I'll pick on some people like Adobe, Salesforce, Zoom, they spend a lot of time in AI so they understand all the risk management pieces, all of that stuff that goes with it. So it's really the structures, the soft stuff that goes around the tech and how to run it.

>> I've been cutting my notes here. I pulled it out. Platform shift is a new architecture. The AI stack, developers, the role of applications, cloud-native infrastructure, role of open source, all of these things are hot buttons-

>> Yes....

>> in the enterprise and certainly they're buzzwords, for sure. IBM's playing in all of those. We're here at IBM Think. You got Red Hat, lucky to buy those guys when you got OpenShift now in all of the enterprises, so check that and platform shift in good position. HashiCorp, there's a debate there if that's truly going to be cloud-native AI enablement or will Aegis take over what Terraform does? It's a whole nother question. Open source, certainly get that with Red Hat and then this whole new stack of AI stacking it. So what does the AI stack look like? This is going to be the big question. So assume that compute gets better. IBM's got a good storage story. Okay, check, check. Assume the infrastructure's fast and can perform it. What happens at the software layer above Red Hat in the IT, in the IT ops, in the AI ops? Is it agent-driven? We hear Rob Thomas on stage in the Partner Summit here talking about observability. He's checking all the boxes, but I'm not hearing anything.

>> Yeah.

>> Does IBM have a good story there?

>> Yeah.

>> What's your take on that?

>> Yeah, again, a really interesting question. Where IBM sits, they can play a lot of different ways if you think about it. But I think that the prevailing wind right now is less about, think about it, the top layer would be applications. We're less about that, and we're more about development tools and platforms to build your own. So if you look at what IBM's done that I think is really, really smart is they've built a series of these language models, and they've done it very responsibly, but they're almost narrowly focused. They have different ones that are focused on finance and different pieces. All the different granite models are really cool, but think of it as they've built a way to have a trusted partner that can help you build your own stuff. That's where I think it is right now and where they're headed and they have some legs to do that.

>> Yeah, the old expression certainly in theCUBE has been our mantra is that developers drive the standards. Remember the old days back, not to date ourselves 'cause we're kind of old, well, I'm older, older than most in the industry, we had standards bodies, IEEE, all these groups that would decide on standards. Standards drive a lot of benefits, especially open standards. The developers have been that proxy open standard de facto body because-

>> Like crowdsource....

>> whatever they adopt tends to be implemented.

>> That's right.

>> You've seen that bottoms up of success work, certainly with Linux Foundation and Apache. Is there an open source opportunity with AI to say, "Hey, the tail can wag the dog here?" Obviously, the proprietary models like OpenAI and Anthropic and others that are proprietary, even though they're technically called open, might actually get replicated in open source. This could swing the app development side to be the dictator of what happens in the middle layer. What's your take on that? It's a radical concept.

>> Oh, my gosh, yeah. Yeah, there's so much going on. You and I talked a little bit about what's a good model, what's a bad model? Large, small? You and I both decided it's all good. You need all of them. You're probably going to need open source and proprietary models as well because they bring different things to the table, right?

>> Yeah.

>> When you look at the pros and cons of open source, the pro is yes, cheaper, maybe collectively known, but the problem is you don't have the support. There's not like somebody to call and say, "Hey, this is broken. Can I fix it?" Those kinds of things and security, which is another big issue when you're looking at those things. On the other side, the private stuff, you don't know where it's coming from, and one of the biggest challenges I think to language models right now is they're not accurate. So accuracy issues have to do with what's the training data that's going on? If you don't know what the training data is... Now what IBM and some of the open source folks are focusing on is naming the metadata, tagging the metadata, telling you where it's coming from, and that's a big step.

>> One of the things Arvind said today on the partner keynote, this is pre-THINK. We're here Monday, we're in Boston, and tomorrow's the big day for the main event, it's partner day today. So Arvind's up there with the partners. He said the pie's big enough for everybody. You know, we really haven't heard that from IBM. They haven't been a pie sharer in the past. They've been, "Give me the services I want and all," so they're not really a pie sharer.

>> Yeah.

>> But now Arvind's saying, "Hey, we're pie sharers. We want to be more open," and so they're betting the ranch on partnership and the ecosystem. Not their best move in the past.

>> Right.

>> They haven't been that good at it, let's be honest. IBM's ecosystem has been very blue. If you're a blue washer or you've been with part of IBM's indoctrinated into their blue world, yeah, you're great. But now, the Amazons of the world, Google showing major moves on how to build an ecosystem. You got to get distribution of software, you got to get money making going on. You got to have some synergies. IBM's now trying hard. So they're looking at, they allowed a deal with AWS, okay, with SageMaker-

>> Yep....

>> and Watsonx governance. Now all of that's in this marketplace, which by the way, billions and billions of dollars being sold through Amazon's marketplace. Adobe, collaboration experience platform. Again, worker, good for applications. Meta Llama 3 indemnification, huge deal. Mistral again, that's just another model to throw on top of there and integrate that and indemnify it, run Watsonx on Azure, checking all the boxes. Salesforce Granite model available in Salesforce Einstein platform. SAP granite, and then the Saudi Data Artificial Intelligence Authority, check that box. Are they checking the boxes? This sign that IBM's actually really doing it, what's your take on this? What's your analysis? If they do want to go the distance, I do believe that they have authentic strategy. What have they got to do to be successful?

>> Right. Well, I want counter and ask you a question. I think that there was a major shift with IBM when they re-orged and they put in the software and they started talking hybrid cloud and AI, when they really did that it's 2019. When they did that, it brought a lot of focus. So I felt like maybe we're not being fair that they weren't that good a partner before. I think they're better because of that shift part of it, but-

>> Well, when Arvind came on, he brought a product-led focus, not a-

>> Yes. Yes....

>> services-led focus-

>> Right, which helps....

>> which totally helps because now you have product leadership that's going to attract-

>> Yes....

>> partners. Nobody wants-

>> Right, and it's more you can see the vision. Whereas before it's like, "I had no idea," right?

>> Yeah. Partners want to add services too. Why compete on services? Again-

>> Right....

>> a little nuance point inside baseball, but-

>> Exactly....

>> it's important.

>> Yeah, and well, here's what I think is interesting about IBM and those partners is think of it this way. I'll throw you another one. They're probably the closest thing to a Accenture where they can go in and really help somebody. So they've got the whole thing. They're an end-to-end help. So it's interesting that maybe they're more attuned now to being a better partner because they were able to think that way.

>> Well, you've got a great point. The whole SI, GSI global integrators. IBM used to be an integrator. They spun out, they got cloud Kindle, they outsourced and all that stuff. But Dave just wrote a post on SiliconANGLE breaking analysis, really highlighting that IBM is undervalued mainly because they have so much in their portfolio that isn't popping right now with Gen.AI. But if you look at their main core pillars, Red Hat, Watson and their core products that drive customer value, those are going to have a direct injection of nutrients, if you will-

>> Yes....

>> and with Gen.AI because they have scale, they got data. Dave and I are both very bullish on the fact that IBM's got the pieces in place.

>> Yes.

>> Other stuff might come on later. That's that rising tide will float that boat. But I think IBM's got so much it's hard to squint through that and say, well, that'll be phase two.

>> Well, that's another good point. I would say this, in my opinion, they are one of the most experienced and responsible AI leaders in the world, and there's some things that they do, they're doing it right. So I think they're going to have good guidance to the customers they have. You think about, I don't know if you looked at governance to some of the Watson tools they have, but what I told you before about Granite, those Granite models, building that with open and saying, "Here's the data," all of this is the way to do this from what I call an adult in the AI rumpus room. So I think that ABM's one of them.

>> Well, first of all, I think IBM, not to be completely giving them a complete A+ here-

>> Right....

>> IBM's definitely done well. I'll Give them an A-.

>> They are checking some boxes, but what they're doing is some innovative things. For instance, their InstructLab that they announced at Red Hat Summit is very impressive, and that is actually something completely different. It's an open source initiative that levels the playing field because it allows models to integrate with each other, specifically, collectively contribute code and changes to other LLMs. That's a huge deal. That's the first time I've seen someone come out and saying and admit publicly via a product announcement that we will have impact. We'll say that this is going to be an integration between two models-

>> Right....

>> where there'll be parameters being passed, things changing and integrating together. That's fusion. That's alchemy.

>> And multimodal.

>> That's multimodal.

>> Multimodal.

>> That is the future, I think, and I think that's where the AI can get smarter, and so that got my attention. I think that's a very interesting play. Not sure the world's ready for it prime time, but directionally it feels like that's where it's going to go.If you're talking about picking on them a little bit, let's say, what I don't understand about that is that OpenShift and Red Hat, they talk about open source-

>> Right....

>> but you also have to be part of that platform to use it. So it's not hugging face. You know what I'm saying?

>> Exactly.

>> It's not like you can go get these models at GitHub git-reflogging, you can't. So it's like, "Okay, well, you know."

>> Look at this, Mark. You go back to IBM five years ago, Watson failed, One.O failed. They overplayed their hand on that. A lot of NLP in there. But now with generative AI, they missed the cloud. That soft layer open blue shift didn't work. So here they are, 2024 partner with Amazon Azure, they don't need to have a cloud.

>> No.

>> They have a cloud partnerships. So this idea of distributed computing. My conspiracy theory all along, this goes back to when I interviewed Arvind before he became CEO, before they bought Red House out in San Francisco, he squinted, I could see in his eyes, I stared him in the eyes. I said, "You're building an operating system for the internet," and he smiled 'cause he's a distributed computing guy.

>> Yeah, yeah, yeah.

>> So IBM actually could be the player that becomes not an integrator, but a runtime assembly of assets and run an operating system for AI. Meaning there could be a Linux-like moment coming in the world where something different runs everything.

>> Right, well, that too, but to your point, what David said the other day on something he's writing on was what happens when you try to navigate the cloud and the on-prem? So is IBM positioned to help companies that are doing both or can't decide? That's another thing.

>> I think IBM is, their integration chops-

>> Where are you at?...

>> actually, this is where your point about their services background actually helps them because they can bring an integrator-like vibe with real product knowledge, technology knowledge and the ability to execute that project. So it's going to be a little bit of everything. If you look at how APIs are connecting everything, you're seeing that 85% of all internet traffic is through APIs. So they're betting on that with their relationships by saying, "We're not going to be in the public cloud-

>> Right....

>> we'll be the cloud operator-

>> Right....

>> cloud operation." So what is DevOps anyway? It's cloud ops.

>> Right.

>> So I think they're in a position, Arvind's smart.

>> Yep.

>> You can see his thinking. If you look at his moves, distributed computing and AI falls on their lap.

>> Yeah, but-

>> It just lifts everything up.

>> To your point, for me, I always look at this AI first and I do say even though they gave away a lot of IP or sold it, they understand AI very well. So are they going to leverage it? That's a different story, but they do understand it very well.

>> Well, the good news is the customers who have the data are going to need a partner to cross the bridge to the future, but first you got to build the bridge.

>> Right.

>> Right? I think IBM could be that bridge builder for companies so then they cross together, and that's a very lucrative partnership. That's an old IBM kind of mindset, customer-centric. I think they can pull it together. We'll see. Well, final question for you, what are you working on? Obviously you've been doing AI industry analyst for a while. A lot changes every week. It's a whole new thing.

>> That's why I have a great beer.

>> You got to hit the home runs. You're on the consumer.

>> That's right.

>> The enterprise side's developing fast. What's your research tell you?

>> So we're still in this very, very early stage for GenAI, we really are. The more operationalized enterprise AI is just starting to roll out. I think it goes back to something you said, this is the year where the pragmatists start to talk about, "How do I integrate my data? How do I protect all of this? How do I build guardrails?" They need help doing that, they're trying to do it themselves.

>> It's the year we start to see the line where it's like old way, new way. I love these disruptive enablement points. That happen very rarely in our industry. I think I've only seen two in my generations over the past 30 years. But once you see it, that's old, that's new-

>> Right....

>> so when start to line up, and by the way, disruptive means people lose.

>> Yes.

>> Technology's transitioning.

>> Chaotic.

>> Yeah, well, you could be incompetent and move over, just everyone doesn't want to be on the wrong side of the street.

>> No, you don't, and it's difficult to experiment without taking a risk.

>> Mark, thanks for coming on theCube. Mark Beccue's here, industry analyst bringing down the AI action IBM THINK CUBE cover. It's our pop-up CUBE, small CUBE. We got a small spot here. Whatever it takes, we will get the story. We'll set up a pop-up cube was with the big CUBE. Dave Vellante's in Dell Tech World right now at the big stage. I'm here getting the job done here in Boston. I'm John Furrier. Thanks for watching.

Mark Beccue | Independent Analyst

search

>> Hello, welcome back to theCube's coverage of IBM Think in Boston, Massachusetts. I'm John Furrier, host of theCube. We are here on the ground getting all the action around what's going on in the enterprise, enterprise AI, and still the IBM and their comeback as they continue to rebuild the company and reinvent themselves. Mark Beccue is with me. He's an industry analysts in AI. Mark, great to see you. Thanks for coming on theCube.

>> Thanks, John. Good to be here.

>> This is our pop-up cube. We do whatever it takes. We're in a little corner here. Whatever it takes to get the story, we will get it. That's the way we operate, get that data. Speaking of data, Mark, this is the top number one conversation is that consumer AI is hot. The GPUs are selling like crazy, new custom silicon's coming out. You saw Google last week at their event talking about custom silicon, and they're ahead of the game there. You got Amazon figuring out what's going to go on with a new CEO, Matt Garman, a very product-oriented fella I've known very well. The world's completely changed, and it's clear that it's a platform shift. So you're starting to see some visibility in the market around where the AI is starting to drop a little bit. Certainly, there's a platform refresh at the bottom of the stack, middle layer and the app layer. Generative AI is changing everything 'cause it's a disruptive enabler. So there's enablement, there's value creation opportunities, but someone loses. Things will be disrupted.

>> Yeah. What's interesting, John, we talked a little bit before about, I think what happens is there's not a real good grip in the enterprise space about how much all this costs. So when you think about, one of the things that worries me is workloads. Where are you going to run these AI workloads? How much does it cost? There's training versus inference and all those things. I think what the problem is is so much is in flux we don't know... the models run big. Are they going to get smaller? What are those workloads going to look like? I don't really think there's a good idea of what that cost is.

>> The question everyone's asking is, okay, I have data, that's pretty clear to most folks that data is a valuable asset here, "How do I configure the data, number one?" That's a platform architecture decision. Then two, "How do I run it in the end-to-end workload that was predefined before generative AI?" So a lot of these companies have these workloads. They know what they look like. They're somewhere in the cloud, some are on hybrid, but now you have to change it with the new data model and coming in. So the first question is, what do I do with my data? How do I figure that out?

>> Well, that's an age-old question, isn't it? It's like big data was our big... how long have we been talking about big data?

>> Since 14 years ago? I do.

>> So-

>> Data's the new oil I heard back then too.

>> Data is the new oil and we've heard that a lot, right? I guess, again, a big flux really, right? When you've got silos, some of this stuff is, some of it's structured, some of it's unstructured. Do they want all of it? Do they have privacy restrictions about financial data that they can use, whatever it is? I think it's less of a technical issue and more of a management issue still. So it's really outside of AI. I think what's different, and maybe you touched on is that GenAI now makes it affordable to think that you can get at it.

>> What do you see as the number one challenge in the enterprise? Is it compliance, security, skill set, what to do? What's the first three steps to take? In your analysis, what are you seeing in your research?

>> Yeah, that's a great question. I think it's a lot about lifecycle management. You really don't have a whole lot of companies that understand AI lifecycle management. If you look at the market at the ones that have succeeded, and let's call them that actually have products out, I'll pick on some people like Adobe, Salesforce, Zoom, they spend a lot of time in AI so they understand all the risk management pieces, all of that stuff that goes with it. So it's really the structures, the soft stuff that goes around the tech and how to run it.

>> I've been cutting my notes here. I pulled it out. Platform shift is a new architecture. The AI stack, developers, the role of applications, cloud-native infrastructure, role of open source, all of these things are hot buttons-

>> Yes....

>> in the enterprise and certainly they're buzzwords, for sure. IBM's playing in all of those. We're here at IBM Think. You got Red Hat, lucky to buy those guys when you got OpenShift now in all of the enterprises, so check that and platform shift in good position. HashiCorp, there's a debate there if that's truly going to be cloud-native AI enablement or will Aegis take over what Terraform does? It's a whole nother question. Open source, certainly get that with Red Hat and then this whole new stack of AI stacking it. So what does the AI stack look like? This is going to be the big question. So assume that compute gets better. IBM's got a good storage story. Okay, check, check. Assume the infrastructure's fast and can perform it. What happens at the software layer above Red Hat in the IT, in the IT ops, in the AI ops? Is it agent-driven? We hear Rob Thomas on stage in the Partner Summit here talking about observability. He's checking all the boxes, but I'm not hearing anything.

>> Yeah.

>> Does IBM have a good story there?

>> Yeah.

>> What's your take on that?

>> Yeah, again, a really interesting question. Where IBM sits, they can play a lot of different ways if you think about it. But I think that the prevailing wind right now is less about, think about it, the top layer would be applications. We're less about that, and we're more about development tools and platforms to build your own. So if you look at what IBM's done that I think is really, really smart is they've built a series of these language models, and they've done it very responsibly, but they're almost narrowly focused. They have different ones that are focused on finance and different pieces. All the different granite models are really cool, but think of it as they've built a way to have a trusted partner that can help you build your own stuff. That's where I think it is right now and where they're headed and they have some legs to do that.

>> Yeah, the old expression certainly in theCUBE has been our mantra is that developers drive the standards. Remember the old days back, not to date ourselves 'cause we're kind of old, well, I'm older, older than most in the industry, we had standards bodies, IEEE, all these groups that would decide on standards. Standards drive a lot of benefits, especially open standards. The developers have been that proxy open standard de facto body because-

>> Like crowdsource....

>> whatever they adopt tends to be implemented.

>> That's right.

>> You've seen that bottoms up of success work, certainly with Linux Foundation and Apache. Is there an open source opportunity with AI to say, "Hey, the tail can wag the dog here?" Obviously, the proprietary models like OpenAI and Anthropic and others that are proprietary, even though they're technically called open, might actually get replicated in open source. This could swing the app development side to be the dictator of what happens in the middle layer. What's your take on that? It's a radical concept.

>> Oh, my gosh, yeah. Yeah, there's so much going on. You and I talked a little bit about what's a good model, what's a bad model? Large, small? You and I both decided it's all good. You need all of them. You're probably going to need open source and proprietary models as well because they bring different things to the table, right?

>> Yeah.

>> When you look at the pros and cons of open source, the pro is yes, cheaper, maybe collectively known, but the problem is you don't have the support. There's not like somebody to call and say, "Hey, this is broken. Can I fix it?" Those kinds of things and security, which is another big issue when you're looking at those things. On the other side, the private stuff, you don't know where it's coming from, and one of the biggest challenges I think to language models right now is they're not accurate. So accuracy issues have to do with what's the training data that's going on? If you don't know what the training data is... Now what IBM and some of the open source folks are focusing on is naming the metadata, tagging the metadata, telling you where it's coming from, and that's a big step.

>> One of the things Arvind said today on the partner keynote, this is pre-THINK. We're here Monday, we're in Boston, and tomorrow's the big day for the main event, it's partner day today. So Arvind's up there with the partners. He said the pie's big enough for everybody. You know, we really haven't heard that from IBM. They haven't been a pie sharer in the past. They've been, "Give me the services I want and all," so they're not really a pie sharer.

>> Yeah.

>> But now Arvind's saying, "Hey, we're pie sharers. We want to be more open," and so they're betting the ranch on partnership and the ecosystem. Not their best move in the past.

>> Right.

>> They haven't been that good at it, let's be honest. IBM's ecosystem has been very blue. If you're a blue washer or you've been with part of IBM's indoctrinated into their blue world, yeah, you're great. But now, the Amazons of the world, Google showing major moves on how to build an ecosystem. You got to get distribution of software, you got to get money making going on. You got to have some synergies. IBM's now trying hard. So they're looking at, they allowed a deal with AWS, okay, with SageMaker-

>> Yep....

>> and Watsonx governance. Now all of that's in this marketplace, which by the way, billions and billions of dollars being sold through Amazon's marketplace. Adobe, collaboration experience platform. Again, worker, good for applications. Meta Llama 3 indemnification, huge deal. Mistral again, that's just another model to throw on top of there and integrate that and indemnify it, run Watsonx on Azure, checking all the boxes. Salesforce Granite model available in Salesforce Einstein platform. SAP granite, and then the Saudi Data Artificial Intelligence Authority, check that box. Are they checking the boxes? This sign that IBM's actually really doing it, what's your take on this? What's your analysis? If they do want to go the distance, I do believe that they have authentic strategy. What have they got to do to be successful?

>> Right. Well, I want counter and ask you a question. I think that there was a major shift with IBM when they re-orged and they put in the software and they started talking hybrid cloud and AI, when they really did that it's 2019. When they did that, it brought a lot of focus. So I felt like maybe we're not being fair that they weren't that good a partner before. I think they're better because of that shift part of it, but-

>> Well, when Arvind came on, he brought a product-led focus, not a-

>> Yes. Yes....

>> services-led focus-

>> Right, which helps....

>> which totally helps because now you have product leadership that's going to attract-

>> Yes....

>> partners. Nobody wants-

>> Right, and it's more you can see the vision. Whereas before it's like, "I had no idea," right?

>> Yeah. Partners want to add services too. Why compete on services? Again-

>> Right....

>> a little nuance point inside baseball, but-

>> Exactly....

>> it's important.

>> Yeah, and well, here's what I think is interesting about IBM and those partners is think of it this way. I'll throw you another one. They're probably the closest thing to a Accenture where they can go in and really help somebody. So they've got the whole thing. They're an end-to-end help. So it's interesting that maybe they're more attuned now to being a better partner because they were able to think that way.

>> Well, you've got a great point. The whole SI, GSI global integrators. IBM used to be an integrator. They spun out, they got cloud Kindle, they outsourced and all that stuff. But Dave just wrote a post on SiliconANGLE breaking analysis, really highlighting that IBM is undervalued mainly because they have so much in their portfolio that isn't popping right now with Gen.AI. But if you look at their main core pillars, Red Hat, Watson and their core products that drive customer value, those are going to have a direct injection of nutrients, if you will-

>> Yes....

>> and with Gen.AI because they have scale, they got data. Dave and I are both very bullish on the fact that IBM's got the pieces in place.

>> Yes.

>> Other stuff might come on later. That's that rising tide will float that boat. But I think IBM's got so much it's hard to squint through that and say, well, that'll be phase two.

>> Well, that's another good point. I would say this, in my opinion, they are one of the most experienced and responsible AI leaders in the world, and there's some things that they do, they're doing it right. So I think they're going to have good guidance to the customers they have. You think about, I don't know if you looked at governance to some of the Watson tools they have, but what I told you before about Granite, those Granite models, building that with open and saying, "Here's the data," all of this is the way to do this from what I call an adult in the AI rumpus room. So I think that ABM's one of them.

>> Well, first of all, I think IBM, not to be completely giving them a complete A+ here-

>> Right....

>> IBM's definitely done well. I'll Give them an A-.

>> They are checking some boxes, but what they're doing is some innovative things. For instance, their InstructLab that they announced at Red Hat Summit is very impressive, and that is actually something completely different. It's an open source initiative that levels the playing field because it allows models to integrate with each other, specifically, collectively contribute code and changes to other LLMs. That's a huge deal. That's the first time I've seen someone come out and saying and admit publicly via a product announcement that we will have impact. We'll say that this is going to be an integration between two models-

>> Right....

>> where there'll be parameters being passed, things changing and integrating together. That's fusion. That's alchemy.

>> And multimodal.

>> That's multimodal.

>> Multimodal.

>> That is the future, I think, and I think that's where the AI can get smarter, and so that got my attention. I think that's a very interesting play. Not sure the world's ready for it prime time, but directionally it feels like that's where it's going to go.If you're talking about picking on them a little bit, let's say, what I don't understand about that is that OpenShift and Red Hat, they talk about open source-

>> Right....

>> but you also have to be part of that platform to use it. So it's not hugging face. You know what I'm saying?

>> Exactly.

>> It's not like you can go get these models at GitHub git-reflogging, you can't. So it's like, "Okay, well, you know."

>> Look at this, Mark. You go back to IBM five years ago, Watson failed, One.O failed. They overplayed their hand on that. A lot of NLP in there. But now with generative AI, they missed the cloud. That soft layer open blue shift didn't work. So here they are, 2024 partner with Amazon Azure, they don't need to have a cloud.

>> No.

>> They have a cloud partnerships. So this idea of distributed computing. My conspiracy theory all along, this goes back to when I interviewed Arvind before he became CEO, before they bought Red House out in San Francisco, he squinted, I could see in his eyes, I stared him in the eyes. I said, "You're building an operating system for the internet," and he smiled 'cause he's a distributed computing guy.

>> Yeah, yeah, yeah.

>> So IBM actually could be the player that becomes not an integrator, but a runtime assembly of assets and run an operating system for AI. Meaning there could be a Linux-like moment coming in the world where something different runs everything.

>> Right, well, that too, but to your point, what David said the other day on something he's writing on was what happens when you try to navigate the cloud and the on-prem? So is IBM positioned to help companies that are doing both or can't decide? That's another thing.

>> I think IBM is, their integration chops-

>> Where are you at?...

>> actually, this is where your point about their services background actually helps them because they can bring an integrator-like vibe with real product knowledge, technology knowledge and the ability to execute that project. So it's going to be a little bit of everything. If you look at how APIs are connecting everything, you're seeing that 85% of all internet traffic is through APIs. So they're betting on that with their relationships by saying, "We're not going to be in the public cloud-

>> Right....

>> we'll be the cloud operator-

>> Right....

>> cloud operation." So what is DevOps anyway? It's cloud ops.

>> Right.

>> So I think they're in a position, Arvind's smart.

>> Yep.

>> You can see his thinking. If you look at his moves, distributed computing and AI falls on their lap.

>> Yeah, but-

>> It just lifts everything up.

>> To your point, for me, I always look at this AI first and I do say even though they gave away a lot of IP or sold it, they understand AI very well. So are they going to leverage it? That's a different story, but they do understand it very well.

>> Well, the good news is the customers who have the data are going to need a partner to cross the bridge to the future, but first you got to build the bridge.

>> Right.

>> Right? I think IBM could be that bridge builder for companies so then they cross together, and that's a very lucrative partnership. That's an old IBM kind of mindset, customer-centric. I think they can pull it together. We'll see. Well, final question for you, what are you working on? Obviously you've been doing AI industry analyst for a while. A lot changes every week. It's a whole new thing.

>> That's why I have a great beer.

>> You got to hit the home runs. You're on the consumer.

>> That's right.

>> The enterprise side's developing fast. What's your research tell you?

>> So we're still in this very, very early stage for GenAI, we really are. The more operationalized enterprise AI is just starting to roll out. I think it goes back to something you said, this is the year where the pragmatists start to talk about, "How do I integrate my data? How do I protect all of this? How do I build guardrails?" They need help doing that, they're trying to do it themselves.

>> It's the year we start to see the line where it's like old way, new way. I love these disruptive enablement points. That happen very rarely in our industry. I think I've only seen two in my generations over the past 30 years. But once you see it, that's old, that's new-

>> Right....

>> so when start to line up, and by the way, disruptive means people lose.

>> Yes.

>> Technology's transitioning.

>> Chaotic.

>> Yeah, well, you could be incompetent and move over, just everyone doesn't want to be on the wrong side of the street.

>> No, you don't, and it's difficult to experiment without taking a risk.

>> Mark, thanks for coming on theCube. Mark Beccue's here, industry analyst bringing down the AI action IBM THINK CUBE cover. It's our pop-up CUBE, small CUBE. We got a small spot here. Whatever it takes, we will get the story. We'll set up a pop-up cube was with the big CUBE. Dave Vellante's in Dell Tech World right now at the big stage. I'm here getting the job done here in Boston. I'm John Furrier. Thanks for watching.