SiliconANGLE theCUBESiliconANGLE theCUBE
  • info
  • Transcript
#4 The Future of AI: Moving Beyond the Browser and Into Our Hands in 2025
Clip Duration 03:48 / February 2, 2025
264 - Breaking Analysis - theCUBE Research Predictions 2025-.mp4
Video Duration: 46:34
search

Dave VellanteFrom theCUBE Studios in Palo Alto and Boston, bringing you data-driven insights from theCUBE and ETR. This is Breaking Analysis with Dave Vellante. Make no mistake, we are entering a technology cycle that is completely new, massively parallel computing in the GenAI awakening is creating an entirely different industry focus that has altered customer spending patterns. And moreover, this new computing paradigm has changed competitive dynamics almost overnight. Decisions whether to spend tens of billion dollars or hundreds of billions of dollars even on CapEx are being challenged by novel approaches to deploying AI. Geopolitical tensions are higher than at any time in the history of tech. Far surpassing the consternation over the Japanese ascendancy of the 1980s that some might remember. And while the pace of technology appears to be accelerating, causing a lot of concern and confusion and chaos, the reality is that broad technology adoption evolves over long periods of time.

This creates opportunities, it creates risks and tectonic shifts in industry structures. Hello and welcome to this week's theCUBE Research Insights powered by ETR. In this special Breaking Analysis, we're pleased to introduce new prediction episode featuring some of the top analysts at theCUBE Research. With us today are six of our industry analysts from theCUBE Research, Bob Laliberte, who covers networking. Scott Hebner on the AI front. Savannah Peterson who will be talking about the impact of consumer tech. Jackie McGuire, our newest cybersecurity analyst. Christophe Bertrand, who will be discussing his predictions on cyber resiliency. And Paul Nashawaty who leads our app dev practice. Rob Stretchay, we miss you. He's out of the office, so couldn't be here today. But thank you all for being here. We really appreciate the collaboration and we're very excited for our inaugural team predictions.

Now, before we get into it, I just want to share some survey data from ETR to show how much the industry has changed in the last two years. Each quarter, ETR does a spending intention surveys of more than 1800 IT decision makers, and we want to show you just how much of an impact the AI awakening has had on spending intentions. This graphic here shows spending by sector, so the vertical axis is net score or spending momentum within a sector. There's about 19 of them. And the horizontal axis is penetration or what they call pervasion in the data set for each sector. And this is based on account penetration, it's not revenue level spent. Now in this chart we're going back to 2023 and that red line at 40% indicates a highly elevated spending velocity.

And you can see we've highlighted in that box the ML/AI along with containers, cloud, and RPA were above or on that 40% dotted line. Now let's take a look at how fast that's changed over the last 24 months. It's not going to shock you, but look at this. Both the trajectory of machine learning AI over that time period and look what happens to other sectors. Ml AI shot to the top, other sectors became compressed. So this data underscores the transformation of the tech industry and specifically the spending priorities where ETR tells us that roughly 44% of customers that we talked to have been stealing from other budgets to fund their GenAI initiatives and that the ROI received is, let's say, tepid.

All right guys, let's get into the core of our episode today and turn our attention to the 2025 predictions. Here's a quick glance at all of our predictions. We're not going to stay on this graphic too long, but readers of Breaking Analysis will be able to scan this and then focus on the areas that they're most interested in. Scroll down. I'm going to kick things off. I'll give a quick overview then Bob Laliberte is going to follow with some predictions on networking. Scott is going to talk about the future of LLMs. Savannah is going to share her predictions about the impact of consumer tech on the broader industry. Jackie's going to get a little spicy and highlight some of the threats posed by recent policy changes from the Trump administration. And then Christophe is going to follow up with some predictions on data protection and AI. And then Paul is going to bring us home with a prediction about coding and developer impacts. All right, let me start things off.

The DeepSeek news is top of mind. Of course, my view has largely been misinterpreted. The DeepSeek innovations to the extent they are accurate, and I think they largely are, are only going to serve in my view, to expand the market for AI and value and volume are the two most important metrics here. In other words, the denominator of AI, of performing AI, doing AI projects, i.e. the cost that was just lowered, so the value goes up. Think benefit over cost equals ROI. You lower the denominator. We all know what happens. Adoption goes up, so volume increases, costs go down, volume goes up. This is what Satya tweeted out about Jevons Paradox, it costs less, people are going to consume more.

I totally agree with that. We've seen the history of the computer industry. If you think about it, we put humans on the moon with a computer that was less powerful than a calculator. And since then we've been consuming compute like crazy and I don't think that's going to change. So winners, NVIDIA, I think Broadcom, I think AMD especially because of the edge opportunities, I think infrastructure players like Dell and HPE, the networking folks that Bob Laliberte is going to talk about. I think the hyperscalers, if Satya Nadella is going to spend 80 billion and the hyperscalers is going to spend north of $200 billion on AI this year, again, they just lowered their denominator, so they're going to get more for less. And software companies are going to benefit big time from this because we're of course injecting AI and agents into all software if it's less expensive for them.

If they can use smaller language models that are more efficient, that's good for everybody. Neutral, we saw Constellation Energy, the stock drop dramatically on Monday of this past week. I think it's neutral because I think they can dial up and dial down as needed. So we'll see about the energy trade. I think it's negative for closed source LLMs, especially Anthropic. IBM Granite got a little validation with their open source, smaller language model. OpenAI is TBD in my view. At first I felt like this is a bit of a negative for them." They're concerned. There was an article in the Journal this week about, ironically, how they stole IP, how DeepSeek stole IP from OpenAI.

Of course, OpenAI is trained on internet data and it's been accused of similar things, although they cleaned some of that up. But look, they're still the leader in innovation and firms, like SoftBank, are willing to spend a lot of money and invest in them. So we'll see how that plays out. Now how are we going to measure the accuracy of this prediction in 2026? First of all, watch NVIDIA. Watch the stock price. Watch their volume, watch their profitability. Let's see if pricing trashes as a lot of people had expected, but I think they're going to continue to thrive. Edge computing revenue takes off. That was one of my predictions from last week. AI projects get less expensive. Enterprise ROI becomes more attainable and hyperscalers get more bang from their buck on their CapEx spend. Savannah Peterson is here with me in studio. Savannah, thank you so much for coming in today. It's pretty heavy.

Savannah PetersonSuch a joy to be here, Dave, first time in the Boston studio.

Dave VellanteWell it's awesome to have you. I hope you can come back more often.

Savannah PetersonDefinitely. Maybe not in January next time.

Dave VellanteWell, what do you think?

Savannah PetersonIt's little brisk for this calibrating.

Dave VellanteWhat do you think about that prediction, DeepSeek trend?

Savannah PetersonI agree with you. I think that the market really whiplashed with the DeepSeek conversation in a way that was very knee-jerk. We've been talking to Jonathan Ross and groq for a year and a half now. Jonathan being the inventor of the LPU. He's also the brand behind the TPU at Google. What I found interesting about this is the conversation was always going to be about inference. It was always going to be... What was going to make AI real was always going to be inference. People have been focused exclusively on specifically, basically NVIDIA GPUs. The reality is the experience that makes AI real and makes it magical and saves lives or changes lives in our future is going to come from that on the edge or that in device. It can be on-prem, it can be anywhere, but that very real-time data-driven utility and value, whatever that might be. What I like about this DeepSeek situation, I realize a lot of people are paranoid, tons of privacy concerns. You talked about how Italy just banned DeepSeek earlier today. They're accelerating the conversation around competition.

Competition is a good thing in AI, in my opinion. Now, do necessarily want to risk privacy to the geopolitical mention you made earlier with China? Maybe not. Not necessarily where I would personally like a lot of my data process, but on the flip side, it's not just going to be an OpenAI world and there's going to be more in the compute world than NVIDIA. And I think that this conversation has really blown that open and it was not surprising at all to me that this happened. But I do think that it is important since we are referencing DeepSeek specifically as an example. DeepSeek very gently qualified the fact that they were able to train this model for less than $6 million, which is 1/100th of what some of the competition has been able to do that for compelling number, except that doesn't include any of the costs going up front.

It also doesn't reveal how much they have spent on hardware. Early estimate I just saw one today from another report, estimates that they actually spent $500 million on what they've done just on the hardware side. So I think that you're absolutely right in the sense that it's going to be multiple players, multiple Silicon parties happening here, and that there are going to be... There's going to be more players. I don't think that the conversation will be as brand-based as it's been at the end of this year as it is at the beginning of this year. I'm hoping that we're looking at solution and value-based discourse or industry specific discourse than we are just saying ChatGPT is going solve our problems or DeepSeek is going to be the thing that disrupts ChatGPT.

The reality is ChatGPT still uses an excessive energy. A lot of companies have not found ROI or return on their investment when it comes to generative AI specifically. DeepSeek just reminds us that there's a big broad conversation to have and a lot of players and a lot of money still at stake.

Dave VellanteValue and choice. It's interesting to see how many organizations, AWS, Microsoft, I think IBM as well are putting DeepSeek, whether it's in Bedrock or Azure, et cetera. So thank you.

Savannah PetersonThank you.

Dave VellanteBob Laliberte is up next. Bob, we're showing your prediction that networking for AI gets a big boost in 2025. What do you think, Bob? Which companies are going to be impacted positively or negatively? State your prediction and then defend it, please.

Bob LaliberteAbsolutely. And I think there's a couple of predictions within my prediction. The overall one is that networking will finally move out of that basic plumbing and be recognized for the value it's delivering today's modern, highly distributed and certainly supporting all of these GenAI initiatives that are going on. And there's multiple ways it's going to happen. So obviously everyone's talking about AI data centers needing to be built, and so there's a whole new network architecture that needs to go in and support that environment, all the GPU clusters and so forth. So you're seeing winners like Arista and NVIDIA, Cisco's playing in there, Juniper's playing in there, Dell is playing in there. So they all have offerings. The one prediction within the prediction I'll do is that for these AI data centers, we're really going to see, I think, Ethernet take hold and probably force out that InfiniBand.

Today InfiniBand is perceived as the performance lead, but Ethernet's quickly closing that gap. And with the skills differences, unless you already have an HPC environment, I think we're going to see more and more organization shift and do Ethernet. So that's the first piece. That's an AI data center. The second piece is obviously you need to get data to those places where you're doing the training and so forth. So suddenly that WAN is going to be a lot more important. And so I'm starting to see organizations starting to come out with solutions for that, whether it be private WAN solutions, highly performant ones. You're seeing Juniper routers, Cisco and so forth playing in that space or the telcos themselves recognizing the opportunity and creating AI connect services to enable organizations, large enterprises, global enterprises.

To be able to get the data where it needs to be for either training or inferencing or fine-tuning, et cetera. So we're definitely going to see a lot more. And in those cases, certainly for the US, AT&T, Verizon, T-Mobile, et cetera, are really going to have a big play there as well as all of the routing vendors for organizations looking to move to private WAN environments. And then the last piece of that would be the edge component of it. And so from the edge it's all about how do you collect the data that's going to be used. So you're going to see a tremendous uptick, I believe, and not only wifi, but private 5G and more importantly, the convergence of those technologies. Studies that we've done in the past have indicated that organizations really don't want to have separate management planes and technologies for both of those. They want to have it integrated. We're starting to see the first signs of that where organizations are starting to put wifi and 5G in the same access point, and I think we're going to see more of that.

So companies like Meter who are doing that, Ericsson who's leading away with a lot of that, you've got HP Athonet, Solana, Federated. There's a lot of smaller companies driving that highway nine. So keep looking for that. Look at the opportunities to converge wifi and private 5G for that consistent management and data collection. So ultimately, like I said, 2025, lot more focus on networking, especially as it relates to these AI environments, all the way through from those front end networks, all the way back into the back end AI data centers as well.

Dave VellanteThank you, Bob. So plumbing becoming increasingly important. Interesting what you're saying about InfiniBand and Ethernet. A lot of people think that's a real negative for NVIDIA. On the other hand, NVIDIA is working on Ethernet as well, so we'll see how that plays out. I know, Jackie, that you had some thoughts on this. Can you chime in here and give us your opinion on Bob's prediction?

Jackie McGuireSo I wholeheartedly agree with the Ethernet component of this. I've been trying to draw attention to the whole skilled trade side of AI and how... One of the things I think is good is that this will be the rise of hardware again. So we've been a pretty software focused technology economy, but we are going to start to see physical limitations. So I think on the skilled trade side, Ethernet is certainly a much easier to ramp thing. And then I also couldn't agree more on both WLAN as well as the kind of convergence of wifi and mobile because we don't have enough fiber in as many places as we're going to need it to service businesses. And it's also been a long time since most average businesses had to worry about the speed of their internet, the capacity, who they were sharing capacity with around them, and the absolutely monumental amounts of data we're talking about moving back and forth.

We're going to start having to be creative with balancing the load on various different networks. And then lastly, I think obviously from the security standpoint, creating some of the closed networks Bob was talking about will be an imperative. So you'll start to see businesses that was historically never have considered air gapping things. Air gap them because it's going to be necessary. So I wish I could wholeheartedly disagree, but I think Bob's pretty spot on.

Dave VellanteGood points. Thank you for bringing that up. Maybe DeepSeek can help us with that little network bandwidth challenge. All right, Scott Hebner is up next. Scott, awesome to have you here today. Let's take a look at your prediction around LLMs. You say they alone won't get the job done in 2025, they need some help. What do you mean by that? Enlighten us on what are the limitations that you see that LLMs are going to run into this year and how are we going to fill those gaps?

Scott HebnerSo a little bit of context before I jumped into that. As I was thinking through the predictions, I think there's definitely a consensus view that this would be the year that marks the rise of agentic AI. And I didn't want to repeat that. So I published six predictions today for AI and I'm really focusing on the underlying changes that I think are going to happen that are going to address the limitations in today's AI and therefore unleash ultimately the value of agentic AI. So I put six of them out, you can go read about those. And as you said, my first prediction is that LLMs run out of gas to fuel enterprise ROI. And what I mean about that is we kind of think about what's happening out there, certainly in 2024, is some 70% of enterprises have deployed or rely on LLMs as part of their generative AI solutions, but only 24% of them have rolled out LLMs in production, at scale.

And some half of them the projects that they have underway are simply just not getting deployed. When you look at ROI, the challenge is it's about productivity and analyzing information and automating repetitive tasks. It's hard to measure, it's more implicit versus explicit. But those that have tried to measure it, like Harvard Business School for example, they say only 18% have reported a high ROI impact. Yet 67% are planning to move beyond LLMs and to derive these higher value AI use cases, which kind of gets us into the challenges with the LLMs. Based on their correlative designs, they're prone to inaccuracies and bias and the concealment of influential factors.

They can confuse correlation with causation, which could be a big problem. I think we all have to recognize that a prediction is not a judgment and you need to make judgments to make decisions, and you have to understand consequences and understand what influences what in an outcome. And most importantly, I think, absolutely most important is the models have to be able to explain themselves so you can have trust in the outcomes. Otherwise, when you start using AI for decision intelligence and problem-solving or even autonomous action, unless you can trust it, it is just simply not going to happen. So my prediction when it's all said and done here is that there will be a doubling of the use of more advanced AI technologies that will form an ecosystem, an architected ecosystem of models, and the reliance on an LLM alone is going to fade into the past.

And just to give you some numbers there so at this time next year we can measure whether we had some accuracy in this prediction. Today, 75% of enterprises using GenAI, 55% are using predictive AI models, and those will start getting coupled to the GenAI in a greater extent. Only 35% are using domain-specific LLMs or small language models. The real action is going to be in the reasoning and the explainability. And today there's basically four levels. There's chain of thought or self-taught reasoning, which is used by about 20% of enterprises. Then there's semantic reasoning, about 20% are using that, and that's knowledge graphs and neuro-symbolic AI, things of that nature.

Then there's the causal AI that understands the causal mechanisms and therefore consequence. 12% of enterprises are using causal AI today. And then ultimately there's the notion, the swarm intelligence, which is multi-agent reasoning, and it's about 5% today. So those are the baseline numbers. My projection is that those double in 2025.

Dave VellanteAwesome. Scott, thank you. There's a lot to unpack there, Savannah. What do you think about Scott's prediction?

Savannah PetersonI agree with Scott in the sense that there's this inflated hot hair balloon around LLMs and considering that it's been the first point of interaction for a lot of people with generative AI and with LLMs, there's a little more emphasis on it than I think there is utility from what we're seeing there. So on that front, I think that's right. I would say though right now everyone is in a mass experimentation phase with AI and how they're going to look at it at scale and how that's going to work out. So when I look at those numbers of 24% are at prod and scale, like Scott mentioned, that's actually a pretty significant number of MVPs. When we talk to other companies, we're seeing more like 10% of their experiments or projects are getting to scale at this point or getting through production up and out.

So I think that the hype will definitely balance out. We're obviously in an era of agentic right now. I think that 12 months from now, we are going to be at a position where we're talking about something completely different than that. So I think the LLM has just got the lion's share of attention when we had the GPT craze come, but I genuinely... I don't know that they were ever going to be the fuel for enterprise ROI. I think they were the fuel for enterprise budgeting. So I'm curious to see. I think we're going to learn that it's going to take a lot of different models, specific models and industry solutions for these larger companies across their offerings to actually achieve a greater ROI with AI.

Dave VellanteWell, thank you. I mean, history in technology adoption shows that while things come out of the factory very quickly and there's tons of innovation and the speed of innovation feels like it's accelerating. The speed of adoption is oftentimes challenged by some of these other issues that you need to fill, whether it's security or governance or just being able to keep the lights on. But let's stay with Savannah who is fresh off of CES and great to have you in studio today.

Savannah PetersonI know it's so nice to be here, Dave. Thanks for having me to Boston.

Dave Vellante<inaudible> comes east. So here, Savannah, we're showing your very intriguing prediction that we're finally moving past the browser as our path to knowledge. We've often said consumer innovations, that's what drives the tech industry and it seeps into the enterprise and becomes ubiquitous. Consumer is the starting point. What can we look forward to in 2025 and what does it all mean?

Savannah PetersonWell, as you know, I'm all about putting the human in tech and I'm very passionate about making AI real, not just talking about AI. And I think what happened here in the beginning is what happens at the beginning of every technological revolution. When we got the internet, we did what we'd always done. We sent letters and we called it email. When we brought LLMs to the consumer, we put it in a Google search browser and we called it ChatGPT. It's equivocally the same thing, but that's not going to bring the real value and utility to humanity until people are able to interact with their devices in a way that creates things that we can't even think about. AI is going to give us a Z access in terms of creation and innovation. And I find myself very soured about the democratization of AI conversation often because right now it's not, it's centralized power in a lot of places.

Yes, there's open source models that a lot of different people are using, but the democratization of anything actually comes when you put it physically in people's hands. And that's why I'm very excited about 2025 being the year that takes AI from a browser based experience for most people beyond a chatbot, beyond something that we're using in customer service for greater productivity. And two, an actual creation vessel or curing cancer with data at the edge or being able to predict tornadoes with more efficacy. The things that are actually going to change our lives and make them better is going to happen when we finally have AI hardware in our hands, which is why I really think that 2025 is going to be the year of the AI PC. Bringing AI to the edge and the ability to process workloads of varying types of AI at the edge means that we get to interact with 50% more data.

50% of the data is at the edge, got 1.5 billion laptops due for refresh over the next five years. So the AI PC is coming to our hands, and I think that is... I just can't wait to see what people create in unexpected ways and the solutions that we're going to make. It's going to be awesome.

Dave VellanteWell, Savannah, you millennials are kind of really a force in the marketplace now, and you're going to demand the humanization of all this technology. Jackie, speaking of which, what are your thoughts on Savannah's prediction?

Jackie McGuireWell, I'm a security guru. So I think one interesting application of AI at the edge is privacy because one of the things that... I got into blockchain really early before it was taken over by Crypto Bros. And one of the things I thought would be great was what if I could keep all of my data at the edge and encrypt it and you could just push the advertising model to me and process it at the edge so I maintain all my privacy. But we've never really had devices powerful enough to process the types of algorithms that social media is using because there are these crazy algorithms. So I think one of the interesting cases... I do agree with Savannah, I think really smart people get really annoyed with bad user experience and build things out of nothing more than frustration all the time. I know I've done that myself. And I think if we start to see AI phones and AI PCs, it'd be really interesting to see if we can start pushing algorithms to the edge instead of pulling data into the middle to run algorithms on them. I hope people become more privacy conscious in 2025. I keep my fingers crossed, but that's one place I think that we could really see a lot of improvement for people's lives in terms of privacy.

Dave VellanteReal-time inferencing at the edge is something that I predicted last week that I think it's really going to become a meaningful source of revenue this year. Jackie, let's stay with you. You're not pulling any punches here. You're not afraid to go right after the Trump administration and point out some potential unintended consequences of so-called Doge. The quest for efficiency could spell trouble, and Jackie predicts that gutting CISA, CSRB and shifting certain authority to DHS is really ill-advised and risks critical infrastructure and exposes other vulnerabilities. Jackie, let's hear it. Bring the analysis here.

Jackie McGuireAll right. So I want to be somewhat fair and balanced here because I think there are a few pieces of what this administration's doing with DHS that makes sense and some things that are introducing unnecessary risk. I think there's a conversation to be had about whether these types of agencies should be subject to the whims of an administration. So Jenny <inaudible> shout out, badass who's been running CISA for the last several years. CISA was established in 2018 by the Trump administration, which is kind of interesting that it's now in their crosshairs. So one of the things they did was disband the CSRB, which is the group that was investigating salt typhoon. And when we looked at... So my prediction was essentially that we already saw attacks and the cost of attacks increased 48% over the first half of 2024. We're still compiling the numbers for the second half.

I'm sure everybody who's ever worked in the insurance industry or dealt with insurers knows it's this really crazy tangled web of insurance and reinsurance. And at the end of the day, there's maybe a dozen very large companies that hold the vast majority of insurance. And so if we think about something like a systemic cyber attack coming from China, taking out our critical infrastructure and taking out our financial sector, that would have billions of dollars worth of damage. I'm a veteran of the financial crisis, and so I always think about systemic risk. If all of that risk ends up on the heads of these handfuls of insurers, which is kind of what happened during the financial crisis, Congress is going to have to step in and save them. We could be looking at another bailout, but for the insurance sector at this point. And so then there's a question of if we're trying to be more fiscally conservative, are they going to do that?

Will they bail out the insurance companies? And if they don't, that doesn't just affect cyber risk insurance. That's things like excess FDIC insurance, SIPC, which is what covers your brokerage accounts. And so we're really playing with fire here, and when we talk about the fact that the number of attacks went down last year from ransomware, but the cost of those attacks went up enough to raise the cost by more than 40%. And so we've disbanded the group that was investigating these attacks, and that probably means they're not going to get a lot better. So hopefully the private sector steps up and starts collaborating more and creating more of a private sector approach. But I really think this could have significant systemic impacts significantly beyond cybersecurity.

Dave VellanteWow. You're scaring me here. But you're right. I mean, this is a risk that not a lot of people are talking about. If insurance companies become insolvent, that could be a significant risk to the economy. We're seeing with climate change. There's risks of that all over the country and the world. So Christophe, great to have you here today. Thank you. Bring you into the conversation, any thoughts on what Jackie just predicted?

Christophe BertrandMultiple thoughts on the topic. So clearly, first of all, these were just advisory committees. There are still a bunch of other authorities in place to keep things reasonably protected. But I think though the real issue that you brought up is very clear, Jackie. We have massive exposures anyway, whether in the private sector or with public infrastructure. And we just recently ran a cybersecurity, cyber resiliency summit, and what we heard from vendors and experts was very interesting, their lack of preparedness. So I do agree with you that there is a private component to fixing these exposures. At the same time, everybody's also intertwined from an economic standpoint. What would be really good for maybe a rogue adversary like North Korea I would think may not be good for a large business partner like China.

So I think it's probably partially advised to disband a smart regulatory and advisory type of committee, but everybody's kind of connected. So I'd like to see what the final risk really is when it's all said and done, but it's definitely infrastructure. That's a weakness and that's a strategic weakness for the U.S.

Dave VellanteAll right, thank you. Well up next we're going to stay with Christophe. Christophe Bertrand, as you said just off the Cyber Resiliency Summit, which was held in our Palo Alto offices. Christophe, you must have learned a lot from that summit. And here you're predicting that AI will be the next battleground in data protection. And you think, it's interesting, subcomponent of prediction that the regulations that are going to create more havoc for enterprises that are trying to keep up and that in itself creates challenges and risks. You've got different regulations, whether it's in California or in different parts of Europe, different countries, which is clearly onerous and difficult for organizations to respond to. But Christophe, please share the details of your prediction and then Paul is going to follow up.

Christophe BertrandYes, absolutely. So first of all, I think everybody's talking about AI and great, so how do you protect all that stuff? Is it an afterthought? Like we've seen for SaaS applications and well, cloud applications before. I'm going to say for my colleague Paul, what about originally protecting the various container environments that were out there. So any new infrastructure play, anything that's important to the business typically goes out, doesn't get protected well, so my prediction is simple. The next battleground is, in terms of workloads, and I'm going to say workload as a term here to depict AI infrastructure. That's the next battleground for backup recovery vendors, storage vendors, et cetera, right? So I expect to see a lot of announcements. We've already seen some of those in the Cyber Resiliency Summit and some initiatives that are in existence.

I see a lot more going on around that, and I think the AI ecosystem should start thinking really hard about how to better play with the vendors that protect the actual infrastructure they're delivering against. That's the first statement. The other is, well, AI is a great technology for automation. We talked about agentic AI, about certain processes getting smarter and better and maybe some ability to have agents make some smart decisions to protect the environment, whether AI or not in general terms. Well, I think that's also going to be a very important area. I predict we are going to see dozens of AI-related feature announcements from both the storage, the backup recovery community, cyber resiliency community. It's probably going to become pretty boring after 2025 because everybody's going to be talking about it.

So that's the second part of the conversation here. And then yes, compliance is actually an interesting topic, compliance and governance, because we're talking about AI, we're talking about protecting those environments. But look, do you even know what data you have and where it is? Have you classified it? Now you want to go do all of this great AI stuff and sorry, Scott and others and all this great agentic stuff. Well, hang on. What data are you using? Are you authorized to use it? Are you augmenting it in the right way so that it is compliant? Because we can talk about deregulating AI all we want. If you don't comply with GDPR, CPRA, CCPA, something's going to happen that's not going to be good. So I think that's going to become a very interesting, again, talking back to this interconnected world we live in.

This is going to be a regulation sort of factor here. It doesn't matter if it's in the US or somewhere else. If somebody has a regulation and you're talking about a global business, it kind of applies to everybody. And I think that's going to affect data and AI very, very much in 2025. I expect it's going to be become probably a major, major headache for many organizations. Expect compliance governance to become a hot topic.

Dave VellanteIt's interesting. I mean over the last decades we've seen compliance and things like that oftentimes be a checkoff item. We have a policy, rubber stamp, but those companies that have actually tried to ingrain compliance into their workflow and make it a value producing entity are now in a better position to take advantage of AI. Many have not. So Paul, what do you think about Christophe's prediction? Anything you would add?Absolutely. First of all, I completely respect where Christophe's coming from. This is definitely an area of interest. It's an exciting time. We know that there's growth. We know that there's acceleration and modernization efforts and data is, dare I say, the new oil of the business. It's the lifeline of business and where we have access to this information doesn't just apply across the net new things that are happening. Whenever I'm talking about applications and such, I talk about past, present, and future and bridging the gap across all three of those areas. Dave, you talked about a checklist item of over the last 10 years and such, but you got to remember that those applications and those data sets are still relevant today. And I agree with Christophe talking about that. The AI workloads will be the next battleground to help with this, but I do truly believe that there will be a human in the loop.

I don't believe that there won't... We're at the point where we're just going to let it go and let systems just kind of do their thing. Automation is key, but when we look at the modernization efforts and we look at that data set, it does matter where that data comes from. The second part of his prediction, Christophe, your prediction, I was really intrigued by because it dovetails nicely into where I'm going with my predictions as well. And when I think about the number of data or the number of applications that are using data, it not just comes from the application itself and it's not net new data. This is existing data that has policies, regulations, and governance around it. But when we look at the creation of these applications moving forward, these applications are going to be accelerating at exponential growth. And when we look at this exponential growth, that data set that's in place, everything that Christophe was talking about around DOOR compliance and the EMEA and Europe, GDPR issues, governance and data sovereignty issues, all that applies. And that protection of that information and data is incredibly important with these new applications.Excellent. Well thank you for that. And let's stick with Paul, and thanks for hanging with us here. You're predicting, Paul, that companies are going to try and replace portions of their developer workforce with AI, but it might not be so straightforward. And you say that at least one organization is going to try this shoot for half of its devs gone and replace them with AI systems, but it might not be so successful. Can you please explain?Absolutely, Dave. This is an area that's kind of near and dear to my heart. When I look at the research that we're talking about and we talk about application modernization and development, it really does talk about this exponential growth and scaling of these applications. So when we see this, it's a huge competitive advantage to get your application out the door quickly. So as an example, in theCUBE Research that we recently ran a study on, we found that 24% of organizations are looking to release code on an hourly basis, yet only 8% can do so, right? And so part of this is that competitive force that says these organizations need to scale and modernize very quickly. So they're looking to replace or try to replace... This is a prediction now. They're looking to replace or try to replace development with AI-driven systems.

And my prediction is they will fail. While I see AI tools and power tools really increasing the assistance of developers with the focus around automation and repetitive tasks by code generation, debugging, testing, these systems really need to catch up in order to handle those complex problem-solving. So developers can focus on innovation, collaboration and really work with their peers on the next set, which also kind of ties into my next set of predictions, which is really looking at the tool set and developer productivity is going to come into factor here. We see that the tools that organizations are using today in order to accelerate this competitive growth of driving these applications forward, what we're feeling and what we're seeing in data is 50% of organizations are going to abandon the bespoke or best of breed development tools in favor of using a unified platform.

And the reason around that is really it's focused on reducing the operational complexity, streamlining workflows, and really improve integration allowing for organizations to really accelerate delivery cycles going towards that releasing code on a very rapid cadence. We're also looking at how developer productivity will change. We see that... Early research is showing that developers continue to report that they're only spending 24% of their time writing code and they really need to spend that time writing that code in order to do what they really need to get those applications out the door. The time that they're spending on ancillary tasks like designing and testing and debugging and shareholder meetings. This is where I think AI is going to come in and really help out a lot. AI is going to come in and take those tedious tasks and move away from those and allow the developers to focus on that innovation, as I said.

And I do want to touch on one other piece here on my predictions. I do believe that with this explosion of growth of application and information, we're finding that low-code and no-code expansion will really transform roles, right? And we're seeing that the organizations that this adoption here of low-code and no-code will grow over 30% and empowering the citizen developer to really take on those routine applications to build those applications, which goes back to what Jackie was talking about, what Bob was talking about and what Christophe was talking about, and Scott, as well as Savannah. We were all talking about the governance, compliance, regulation and security that is incredibly important when we look at releasing code development for the citizen developer and the business lines to take on their own rules.

So a lot's happening in the space, a lot's happening in the modernization space and application development space. I'm really excited to see what really matures and we didn't even touch on the future state of applications for things like serverless and Wasm and where that's going. So I think there's a lot happening in 2025.I'll say your point about best of breed versus integrated suites, it's an age-old debate and pendulum swing in the tech industry and obviously hitting your world pretty hard and that notion of low-code, you're right. So many of our other CUBE analysts have talked about agents and if agents are going to work and AI is going to be democratized, they're going to have to learn from the reasoning traces of business people who know what the workflows are. And the more low-code and no-code that is out there, the more democratized AI becomes, the smarter those agents can be and the more productive and useful. Savannah, I mean you know a lot about this space. You're sort of leading with Rob, our KubeCon efforts and our CNCF efforts. Along with Paul as well, what do you think about Paul's prediction? Any thoughts?

Savannah PetersonI agree entirely with Paul in terms of the developer experience and that 24, 27% stat in terms of the time being spent doing the things hypothetically they actually enjoy doing. I've always found quite shocking and quite alarming. I think what Paul's essentially saying is less of an emphasis on DevOps, more of a focus on platform engineering. It's a conversation that we've been hearing at the last couple of KubeCons. I'm not a hundred percent sure that we'll be abandoning some of the bespoke solutions or whether or not there's going to become an orchestration layer or something else that comes in across the top of platform that allows folks to continue to have that autonomy and benefit of choice within certain working suites and just pulling that all up into something that creates a cohesive observability and a utility for the folks on the architecture side of that or the developers themselves.

I hundred percent agree with Paul also on the AI is not going to be generating all of our code. I think that it's going to make code more efficient to test in certain cases. I think there's going to be ways to be giving better health in real time about what's going on with different systems and stuff that's happening, but there's absolutely no way. I do think that AI helps bring certain tools together in form factors that allow us to create quicker, but that's not a replacement for the person who is creating them. I think we'll just see the rate of innovation and creation accelerate with a... The unity is going to be a really big key here. There's going to be a lot of platform plays in 2025 and tool sets that have to work across teams and actually pull a lot of these teams together so that you don't have these silos reinventing the wheel every time they're building something new or spinning up a new cluster. So very much largely agree with Paul, developers are going to hopefully be developing more and while AI is not going to take your job, it's going to make your job suck less.

Dave VellanteAll right, thank you for those additional points and thank you, CUBE Research analysts, awesome insights. Really appreciate your time and your efforts here. All right. I want to also thank Alex Meyerson who's not here today, but he's usually on production and manages the podcast. Ken Schiffman is on production today. Thank you Ken. Great job today. Kristin Martin and Cheryl Knight help get the word out on social media and in our newsletters. And Rob Hoef is our editor in chief at siliconangle.com. Remember all these episodes are available as podcasts wherever you listen. All you got to do is search Breaking Analysis Podcast. I publish each week on siliconangle.com and TheCubeResearch.com. You got some ideas, email me, david.vellante@siliconangle.com or DM me at dvellante or comment on our LinkedIn post and check out etr.ai. They've got great survey data, opinions without data are just that, they're just opinions. This is Dave Vellante for theCUBE Research Team and theCUBE Research Insights powered by ETR. Thanks for watching and we'll see you next time on Breaking Analysis.