Cloud and DevOps InfoQ Trends Report 2025

Transcript

Introductions [00:35]

Daniel Bryant: Hello and welcome to The InfoQ podcast. My name is Daniel Bryant and it’s that time of the year again where we examine the current trends in the cloud and DevOps space. I’m joined by three excellent panelists and we plan to cover everything from new technologies making an impact in the day-to-day to the big strategic issues that are affecting roadmaps in that 2, 3, 5, 10 timeline. Shweta, as you are guest today, would you kick off the introductions please?

Shweta Vohra: Absolutely. It’s my pleasure to be here. I’m Shweta Vohra. I am in technology industry over two decades now and worked across sectors from embedded domain to healthcare to banking and now serving the travel, which is Booking.com where I’m a lead architect. I’ve been fortunate doing this journey to really invent, write, speak, author of two books. Second one is coming next week, but the first one is Decoding platform patterns, as you know. That’s about me in a nutshell.

Steef-Jan Wiggers: Yes, I’m Steef-Jan Wiggers. I’m the InfoQ lead editor and I’ve been around the cloud space for quite some time. Also, author of books back in the day, not recent but more like BizTalk Server books and integration books. Also a little bit more cloud focused, but, yes, writer for InfoQ predominantly on cloud news. Working last couple of years in industry retail ERP systems and currently also pretty active in healthcare. Going forward, I will be more working in that area.

Matt Saunders: Hello, I’m Matt Saunders, one of the senior editors on the DevOps news queue for InfoQ. My day job is working for a company called Adaptivist. We work predominantly in the Atlassian space, so writing tools and running services around things like Jira and Confluence. Spreading our wings out into other areas, of course, we don’t just do that. Yes, I’m the head of DevOps working for the CTO in that team.

Daniel Bryant: Fantastic. Folks haven’t bumped into me before. I’m currently running the product marketing at Syntasso. Background in software architecture, built a bunch of platforms on cloud boundary, Mesos, Kubernetes, whole bunch of work in the cloud as well. Super excited to bring together all of our interesting opinions and thoughts today, and I would say they are opinions, so stand by. There’s going to be lots of good hot takes, I’m sure, as we cover some of the trends over the last 18 months or so since our last InfoQ trend report. Thanks so much for the introductions, everyone.

What cloud and DevOps trend has surprised you the most over the past 18 months? [02:57]

I’ll kickstart by offering to the room. And perhaps, Steef-Jan, I’ll look in your direction as well. Looking back at the last 18 months, what cloud and DevOps trend has surprised you the most?

Steef-Jan Wiggers: I think, of course, the infusion of AI in a lot of cloud services predominantly. Not sure they’re always useful, but you definitely see an uptick of AI. I find that interesting. Another, a little bit with geopolitics, upcoming new cloud platforms, giving the digital sovereignty. You see some movements in that area. Before that, it was like the big three, AWS, Google, and Microsoft Azure, and now it’s OCP cloud and some others popping up, providing hosting services, delivering more and more options. I’m pretty impressed by that trend as well, because before that in our spaces, it was just big three and now it’s more options. That’s definitely something that I saw coming the last couple of months or I would say the end of those 18 months, the shift from free to multiple options. And then the other thing is the uptick of AI going from what is the first wave, now all the way to agentic, where you also see agents popping up in some of the, I would say, cloud services as well.

Shweta, how about yourself? You’re working at Booking.com and other people you chat to on the conference circuit, what are you seeing is a popular trend over the last 18 months?

Shweta Vohra: Yes. Reflecting on my experience, I would say that cloud native adoption has deepened. Somewhere, I saw a report of CNCF saying that it’s much or nearly all companies using it. Vaguely, I remember it was some data around 80% or more, but I don’t fully go by that. However, in my own experience I’ve seen that definitely cloud native adoption has deepened more and more even for us. Those who are going cloud way have understood that only doing cloud will not help. Doing the cloud native way will help them.

Other thing which is coming to my mind, FinOps. FinOps has become mature. This is a space where we were first talking about the budgeting and forecasting. Nowadays, I can see the optimizations and more careful about and also AI bringing in magic to the game. FinOps space is maturing definitely. These two areas I can call out. Kubernetes is everywhere. We like it or not, but the Kubernetes is most dominated. CNCF top graduated one of the products they have and maturing with its own integrations and bit of complexity around. But almost all the companies I’ve worked on now fully rely on it definitely. These are the three things which are coming to my mind.

Daniel Bryant: Yes, no, love it. There’s three great topics there and we’ll definitely dive into a few of those later on as well. Matt?

Matt Saunders: Yes. Well, before I get to mine, I’ll just underline what Shweta mentioned about Kubernetes becoming the default for engineering and for deploying our applications, with a bit of a side hustle of functions as a service and people doing those things and also those things stabilizing. We see incremental Kubernetes releases once a quarter and it’s building on good stuff. There’s no revolutionary changes. So, yes, the visibility there. But the biggest surprise for me, actually this might be a little bit controversial and I’m picking up on the fact that last time we did this podcast… Well, I wasn’t on it, but you guys and girls were talking about is DevOps dead?

Since then, we’ve had things like the State Of DevOps Report coming out and saying that we seem to be getting worse in some areas doing DevOps. You look at the DORA metrics and the time to recover, the time to get code out, all those good things… Yes, I’m going to mention AI as maybe one of the reasons. In fact, I think that’s been cited as one of the reasons why we’re maybe not doing DevOps as well as we have been doing. We all talk about it being a mature thing, but maybe we’re just not talking about it and accidentally forgetting to do it properly anymore. I don’t know. Yes, that’s been the biggest surprise for me.

Daniel Bryant: Yes, love it, love it. For my starter, I think it’s all really good points. I’m more in the platform engineering space and I’m surprised at how quickly the internal developer platform topic has made its way into the boardroom. I’m chatting to C-level folks now and they’re like, “Where’s my platform? Where’s my portal?” I think AI is also infusing into some of that now as they want an AI-powered platform.

I will just plus one what you said, Steef-Jan. My work shadows the UK and the US and a bit of Europe, too, and the digital sovereignty thing is becoming a big deal. My US colleagues, my US friends are a bit like, “Huh?” And I’m like, “No, no. In the UK and the EU, it’s really becoming a thing now about who owns their data, who has access”. We don’t get into the politics too much in this podcast, but just the changing nature of the world is really driving the uptick of digital sovereignty. We’re covering that quite a bit on InfoQ and I’m quite keen to see how that develops over the next few years.

Fantastic lay of the land we’ve all talked about there. I think I’ll go straight on now to the AI elephant in the room, because we’ve all mentioned it, right? We got to do it. We hear a lot about AI and LMs. Now, are they genuinely transforming workflows or are they mostly tooling augmentation? Shweta, I’m going to start with you on this one.

Shweta Vohra: Yes. I’m one of those person who is loving the AI wave. I have seen it a lot of times when we started with JVM and there was a noise. “How can code run in the browser? How can this happen?” We rely on browsers now so much. So many waves have come and gone, but I’m super excited about AI and I see the lots and lots of opportunities it’s going to give us. However, it is certainly a step closer but not there yet. I mean, we cannot say that I can pick up my existing applications and I can bring AI ingrained into that seamlessly. This is still a lot of work needs to be done.

Soon, we will see that the current cloud things, people will start calling it legacy, but we will have to weave in that legacy. I mean, whether that’s mainframe or on-brands or cloud, we will have to weave those things if we have to make it reality. Enterprise, we are not there yet, but a lot of good things which AI is doing for us, making things easier… As an architect if I think a lot more I can do, which was really difficult oddly. For example, if I want to enforce a checklist somewhere to my engineers, I can automate it now. Earlier, it was an act of explaining everyone, making sure that they have buy-in, they follow it, but now I have opportunity to include it somewhere in my pipelines, introduce a bit of AI there, and get it checked that whether they can go ahead or not.

There are so many opportunities. I’m really excited about that, but maybe one more thing I can call out where it has disturbed me. Because of this increasing focus partners, our leaders, business, everybody started focusing on AI, whereas our engineers still struggle with those cloud services which are in their own products. One team I was recently working with, I’ll give an example on OpenSearch. OpenSearch is a complex service to really have those clusters designed, indexing, querying, fetching data, making sure your search goes through, you have that latency targets met. That’s where, when partners are focusing on the other side of things, engineers get less of support. That is the disturbance I would like to call out, but apart from that, yes, this elephant is welcome.

Daniel Bryant: I love it, Shweta. Thank you very much. Steef-Jan, I see you nodding a few times there as well.

Steef-Jan Wiggers: Well, I definitely agree with what was said previously. It’s not enterprise ready. Especially in healthcare, there are initiatives, right? What could you do first is your customers, the insurees, and there’s also healthcare in the medical center or in hospitals, discharge letters being automated but still requires some doctor to review it. From that perspective, it also puts more load on enterprise in general because you still face legacy, maybe even mainframes, then you’ve got your compute services, your cloud native that starts maturing or where you need to get value from it. Then all of a you can also AI. Yes, your context is getting bigger, so you need the context switch. More and more is being brought in. It definitely has potential, but, yes, I also am a bit worried about, because it adds more to what we’re already doing and that also leads to, governance-wise, what can you do with data? Can we completely trust AI or things like human in the loop? Some feel maybe you should automate everything. I’m not so sure. On one way, I’m still a little bit on the fence with AI as well. What can it actually bring today?

Daniel Bryant: Yes, good comments. I’ll just quickly check in, from the platform engineering perspective, are not seeing AI writing full Terraform scripts yet or CDK or… Clue me, pick your poison there. There is a lot of great stuff going on around providing insight, as Shweta mentioned, being able do document, explore, and share concepts. I use ChatGPT a lot for that. I know we do internally for explaining our code bases, explaining our infrastructure as code. But when it comes to generation still, at least from what I’m seeing in Syntasso, there is a whole lot of humans in the loop.

Even if you are auto-generating your infrastructure as code, you’re being very careful about all you’re deploying, because, A, it costs you money. That’s exactly what you mentioned, Steef-Jan, compliance and legal regulations. There’s a lot of promise there, but I do think, as Shweta said, the vendors have gone a bit heavy into this space and are now pulling back a little bit going, “Actually, security has got to be top of mind for these things”. Matt, I like the way you pulled out the DevOps connection you’re on. You are steeped in that world. Is AI contributing to DevOps, disconnecting DevOps? What do you think?

Matt Saunders: Much like most other environments, it is starting to get a foothold. It’s not an elephant in the room. We’re actually on the savanna in Africa and there are elephants everywhere and there’s no exception to that. Yes. I mean, really, now September 2025, we’re now seeing tools coming out in this space that are not just chatbots, because that’s where we started when ChatGPT came out a couple of years ago and it’s just revolutionized everything. So much of what I’ve seen in the DevOps and cloud space since then has basically just been putting a chatbot that knows a bit more than ChatGPT does alongside things.

But now we’re seeing things like you mentioned, Pulumi, earlier that just launch a thing called Neo, which is basically an agent for cloud engineering. Adam Jacob and the system initiative team are working in those areas as well. We’re only now people coming to market with products that are a step ahead here. I think that’s a good thing that it’s taken a while, because harking back to what Shweta and Steef-Jan have said about adoption in enterprises, I think we are not yet there with understanding the social and the ethical implications of working with these AIs and how that works in an enterprise context.

Daniel, again, you mentioned security. Also, when you think about things like the human in the loop, I think we’re probably coming to a phase now where we’re starting to make value judgments on whether a bunch of humans can do something well or can an AI plus the same bunch of humans or a smaller bunch of humans with the right oversight do that better, faster, cheaper, more effectively, safer for those sort of things. I think those are universal across organizations, not just in the DevOps world. But yes, I don’t think we, as a society or even in the microcosms of our organizations, enterprise or SMEs or whatever, have really got a handle on that yet. Until we do, we’re treating tools like Pulumi’s Neo…

I was talking to a vendor of observability products yesterday and the AI bots. They’ve got to go and do deep dives for root cause analysis on tools on how much you trust those things to give you the right answer versus how much exposure you give them to your logs, to your sacred corporate data. There’s a lot to unpick there.

Daniel Bryant: Yes. Agreed, man. Great points but all concerned. I think enterprise adoption, there’s definitely progress being made, but there is some caution to be applied there. I think legacy, as we’ve all mentioned, the socio-technical side is always quite a challenge as well, so big plus one to that.

Shweta Vohra: Can we put it this way that doing the newer products and the newer ideas with AI is easier, has become easier? There are toolings and bunch of a lot of things are happening. Of course compliance is always something to look into, but enterprise, we are not there yet. Yes, not at all.

Are companies adopting platform engineering and building platforms as a product? [15:54]

Daniel Bryant: That’s one, Shweta. Plus one. I want to switch up a little bit. I think this is related to this, but I wanted to look at architecture and platform engineering because the InfoQ audience definitely looks through a lot of things through an architecture lens. That’s one of our core premises in some ways. I was going to ask what are the most significant friction points in adopting things like platform engineering, arguably an evolution of DevOps as well to keep it in that space? But even things like internal platforms, we see people building platforms as a product and maybe not so much from an AI theme. But are people generally seeing companies adopting platform engineering, building platforms as a product? Shweta, I’ve seen lots of your good writing in this space as well. Are you generally positive about platforms as a product and internal developer platforms?

Shweta Vohra: Maybe you’ll hear of anti-pattern from me, which you may or may not expect, Daniel. I have not seen yet a successful platform as a product. A fully formed product is yet something which I know a lot of things are in favor of that build. Let’s build it as platform as a product. The reason being I tell you the ground realities, which I have seen, still things are getting built in isolation and always this tool fragmentation because of that. Of course, different teams owning different set of tools and they’re building in isolation, and then when they build in isolation, the glue is always a problem. When you are just putting the puzzle pieces one on top of other and there is no proper fixture to that, of course none of the puzzles can fix right way.

Platform engineering as a practice, I’ve seen it getting adoption to a little maturity in terms of everybody is talking about that developer experience. Whether you take one step or many steps based on the culture to culture of the company what many teams are taking. Platform engineering is definitely a buzzword on a keyboard. Some places, it’s really doing its job in terms of bringing that focus what we need to be bringing and making this management of being manageable or having that glue where this isolation can make sense and we bring a bit more gluing to that. Does that answer your question, Daniel?

Daniel Bryant: Yes, that’s a great insight. I mean, definitely in my day-to-day work, I’m seeing a lot of people aspiring to build platforms as a product and making reasonable success. But I do take your point of, like we’ve all said it in an enterprise context, there’s just so many departments, so many competing goals, so many competing budgets. It’s quite a challenge. I think that’s a great thinking point, Shweta. I might dive into that more in the report with Steef-Jan if we can get some data and see what we can find around that. I think that’s a great comment. And Matt, I’ll throw it to your way. Any thoughts in relation to that too?

Matt Saunders: Yes, Shweta’s absolutely right. It’s all very tied into the developer experience. But equally, I’m a little bit blase about it and about the adoption of platform engineering platforms as products, things like self-service portals. You kind of need to understand how value flows through your organisation and how you actually do things well for your customers, be they externally or internally. I think quite a lot of people haven’t got the value stream mapping memo or value stream management memo. I think all of us here are familiar with what that looks like and how there’s a switch in your brain when you start to understand how value actually gets delivered.

I think that’s a bit of a prerequisite for platform engineering being done well, because if people can’t actually see the value of an engineer being able to click some buttons and get the things that they want, be that infrastructure or monitoring or GRQs or CICD pipelines rather than going to an external team and filling a spreadsheet or a big complex form and we’ll get back to a week next Tuesday, it’s a bit of a harder sell. But having said that, it’s not all doom and gloom, because a lot of the principles do seem to be landing, both internally in my day job and I speak to a lot of the customers of my day job as well. We’re getting there. We’re not getting these big exemplars of 100% self-service and developers able to click their fingers and, before you know it, the code that they thought of is already in production, but we’re getting there. If this whole movement didn’t exist, we’d be in a much worse place, so yes, it’s not all bad. In fact, it’s very good.

Daniel Bryant: Great. Steef-Jan, are you nodding away there as well?

Steef-Jan Wiggers: Yes, I’m still a little bit more relatively new in this platform engineering. I’ve learned it also a couple of months ago, but I do see it at the retailer I worked for but also at the healthcare company. Still early days. I’ve heard Matt and also Shweta put on a good points that we’re currently struggling a little bit too. We’re trying to sell it but it doesn’t land yet. It also has to do with the value streams. It also has to do with some of the discussions. It doesn’t magically just generate everything for you, but what does the platform team do versus what does the business team do? Yes, there’s some discussions there.

I do see an uptick in my area of work and it does get the attention. It even reaches the boardroom, but, yes, it’s still like, “Okay, what’s the added value of having a platform on our end?” It does give you some of the separation of concern, because the platform being this complete platform, including maybe deployments and stuff and the monitoring, and then the business team just focus on getting the logic out to get it somehow deployed. But yes, there’s a thin line whether you have the cut, what does one team do versus what does the platform team do. At least that’s what I see the struggle currently.

Daniel Bryant: I see the same thing. That’s really good point, Steef-Jan. This is kind of interesting. Friends of ours, Matthew Skelton and Manuel Pais, have just released edition two of Team Topologies. And I’m sure, again, many of us here, I see lots of nods, are like, “It’s an obvious book to read. Everyone’s read that one”. Not so much. A great talk by Stefania Chaplin at a meetup earlier in the week in London, she was talking a lot about Team Topologies. Again, it’s the socio and the technical. How do you offer everything as a service? How do you collaborate? how do you facilitate? I encourage listeners, if they haven’t bumped into Team Topologies, do check it out, because I think exactly what we’re all saying, really, flow of value, APIs, abstractions, separation of concerns, all the good classic architecture things. You’ve got to think about them from a socio-technical context. Yes, that’s great thoughts by everyone there.

Are organisations doing “more with less” in the cloud and DevOps space? [22:04]

I’ll move on if that’s all right and go a bit more the do more with best thing. The global economy is up and down a little bit at the moment. Are we seeing some push back into consolidation? Is AI taking all the budgets? Is best of breed a thing? Are people buying whole stacks? I’d love to hear your thoughts on what you’re seeing out in the real world.

Matt Saunders: Well, Shweta mentioned FinOps as well earlier. I think a lot of people have struggled with maybe not optimizing their cloud costs. We hear these horror stories of cloud costs sparring out of control, which I don’t think is a reality. But the reality is that we’ve just talked about putting teams together and platforms so that engineers can click and get exactly what they want. In some ways, the lack of friction in organizations that are doing DevOps and doing platform engineering well is almost causing like profligate spending to some degree.

Again without wishing to get into politics or economics, but there is pressure to do more with less. That’s a global macroeconomic thing going on. If you put that together with the fact that we do our experiments or we spin up our CloudWatch metrics and all of a sudden we’re spending what may have been just like “oh, that’s a bit much but okay” problem and now it isn’t. Doing more with less is key. I can’t really apply this directly to DevOps or clouds, but if we look at where IT budgets are going, and particularly in the big vendors who are running GPU-based infrastructure and, by extension, where smaller companies… I mean not just like the massive NVIDIAs and OpenAIs of this world, but the medium to large enterprises buying up compute and using a lot of budget to run LMS, that’s squeezing everything out.

I mean, it’s a sea change. Whether that’ll continue, I don’t think people are getting value for money with these massive numbers that we’re seeing spent on AI, but it’s all part of the hype cycle. But yes, TLDR, DevOps infrastructure cloud do more with less in the precious few moments that people have working on improvements that are not AI-related. A lot of them do come down to cutting costs and getting that done to their minimum budgets.

Daniel Bryant: Yes, plus one. Shweta, what are you seeing in the world?

Shweta Vohra: I think all the points are covered by Matt, but one of the CTOs I met in conference and… We were discussing about the same topic that how much you’re investing on it, how much you’re focusing on AI, and how much you’re focusing on the existing tech person. Made a statement saying that, “But Shweta, I cannot afford to not do it, so I have to invest my money and IT budget on the AI thing as well”, which is fair because they have that kind of pressure all over the place that they cannot afford to not do it.

My point, which I have been saying and will again repeat that, before you look into your platform or your other technology things, first, look into your technology strategy. You don’t need to do a hundred things, for sure. Even if you are invested on the AI, even if you are invested on your cloud, if you have to balance both, first look into that. “What space in AI really meaningful for me and I really need to have a center of excellence developed around it and invest there?” Of course you cannot afford to not do it, that space we are operating from. But on the other hand, that same thing applies to cloud and DevOps. You don’t need a hundred tools to manage your estate. Cut down on that. Look at your technology strategy first, then think about other things. That can help bringing that balance to, with less, do more, but there’s no ideal answer, there.

Daniel Bryant: No, it’s a good point, Shweta. It’s almost from first principles there. Don’t get too caught up on the AI or do more less. Actually look at your overall strategy. I think that’s really good advice and I definitely bumped into companies that are not doing that and could do with that advice. I think that’s a really good call out. I mean, Steef-Jan, what are you seeing in healthcare? I’m thinking people are getting squeezed budgets.

Steef-Jan Wiggers: Well, one particular end, more or less, is also done a little bit with constraints on people and capacity. The business, whether it’s retail or whether it’s healthcare, they want more, but you don’t have the capacity. That’s also one of the drivers. I think you have to do more with less, because you don’t have the people. I’m not sure if AI could or the elephant in the room might participate and then it could help, but there’s just what I see a lot of the cognitive strain on people as well, wearing more and more hats. That’s more given by, like I said, constraints on resources. Then you have to do more with less people in this instance today, so your cognitive load becomes bigger and bigger as well. At least that’s what I see as well. Other than what I’ve heard before, having less services, what do you need to do? Sometimes you get a zillion tools or services that are not necessary. Really think about what is necessary that’s good old requirement engineering, et cetera, and this is what you need instead of we just build a bunch of stuff and you can use it but we don’t need it. This is back in the service oriented days as well, so yes.

Are leaders more aware of factors such as developer experience and cognitive load now? [27:18]

Daniel Bryant: Guys, really interesting. All of us have talked about the cognitive load of juggling all these tools and I think AI was promised to help with that. And we actually touched on that 18 months ago when we did the previous trend report. We talked about DevEx shifting more widespread adoption and people being more conscious of cognitive load. I’ll throw this question out there to anyone who wants to answer it. Are people, particularly leaders, managers, are they more aware of things like developer experience and cognitive load these days?

Matt Saunders: Yes, I think so. I think just seeing some people doing good Devex and the emergence of some standards around what that looks like. I mean, whether that equates to people that are actually getting more budget for this or not, I don’t really know. Does it mean that the cognitive load goes down? No, probably not. I would also say referring to our good friend the elephant, again, I think cognitive load has gone up because managers and directors are now looking at cases where they don’t want to take people’s jobs away and replace them with AIs necessarily, but people are having those thoughts. Coming back to what Shweta said about it’s an arms race, you can’t afford not to do this stuff, because at some point someone’s going to use AI in a way and have been using it for a number of months or years in a way that starts to blow away competition in all sorts of markets, all sorts of verticals, healthcare, finance, et cetera. I don’t know what that looks like, but it’s probably going to happen. I think that thinking of making sure that you’re not left behind is probably shoehorning out some of the thinking about making sure the developers are happy.

Daniel Bryant: Yes, the FOMO, Matt, as the kids say, right?

Matt Saunders: FOMO, exactly. Yes.

Daniel Bryant: Fear of missing out.

Matt Saunders: Yes, it’s real. FOMO is real.

Daniel Bryant: Yes, I think basically Shweta you said the same thing, as in people are like, “I’ve got to have an AI strategy because my competitors have an AI strategy”. Whether that strategy actually adds value… We’re seeing the MIT study, I think, that came out not long ago saying actually AI isn’t delivering on the value but… There we go.

Matt Saunders: I think that’s only going to be a temporary thing. I think we’ll get the hang of it. I mean, we all know it’s revolutionary. It’s massive. We’re going to get it wrong before we get it right, and whoever gets it right is going to make a killing. So yes, we want to be that.

Shweta Vohra: Of course we are focusing on developers and cognitive load. I think what cognitive load, it’s increasing all numbers. We should start looking at it and address, because I don’t think developers take that much load as the leadership has or the architects or the other roles. For example, if you tell any program manager now to get AI ready, that’s more of load than our developers load if you tell me. We should do something about it, and the thing which we can do is consolidation. Consolidation is driven by some organizations. Maybe we should keep reminding them. Maybe we should tell CNCF to don’t add more projects please. You have your existing graduated projects really well integrated. Please don’t allow anyone to add more project. Some of these steps can help, but hopefully this message will go long and somebody will hear it.

Daniel Bryant: Well, Shweta, we’re making you pitch that to CNCF. I do hear you. It is hard to stop innovation, but you can provide some guardrails around it. Yes. While I always put the joking slide up in my presentations by showing the eye chart that is the CNCF landscape and it always gets a laugh. We love the CNCF and it’s not an easy job being part of that. I know many of my friends and colleagues are involved heavily with the CNCF, but big plus one to what you’re saying there, Shweta, about, yes, not adding to the cognitive load.

I want to switch it up a little bit now and look at hybrid cloud, because that came up in our last trend reports. I think you said, Steef-Jan, in the last report cloud is switching more into the evolution now rather than revolution. As Matt said, I think the revolution’s going on in the AI world. I think that was nicely phrased, Matt, but I do see hybrid cloud getting more uptick in the conversations I’m involved with day in, day out with customers. Steef-Jan, I’ll put it your way. What are you seeing around legacy systems, hybrid, multi-cloud? I know you’re doing a lot in the Azure space, a lot in the .NET world being an MVP there.

Steef-Jan Wiggers: Well, that’s a good one, because recently I saw a presentation with my own company that we’re talking about AWS and Microsoft in the hybrid cloud version. At QCon London, I saw a talk about hybrid. In retail, I see it because your warehouse system is not something you want to have connected online because you always want to do your order picking. From that perspective, you will see things running on-prem but also in the cloud when it comes to IoT or Edge. You still have devices accumulating telemetry and push that out for your data platform. Those are the kinds that I still see.

Back in the day when cloud came, it was like, “Yes, we’re going all the way into the cloud”. But that never came to a fruition, because in my day-to-day job, let’s say 80% was always hybrid, maybe 10, 20% when I was dealing with completely SaaS solution exchanging data. Sure, CRM, Salesforce with another SaaS cloud that are using platform services. Yes, completely Cloud, I would say, where you do cloud native solution at a insurance company of what I’ve done. But most of the time, it’s hybrid connectivity with on-prem systems.

When you go to industries, you get your purchase order or your production order that’s sent to a factory, basically a machine online that’s not connected. So it comes in, it’s being validated, and fed into the machine that does its prioritization and does the job and then reports back. Maybe I’ve used this amount of ink, let’s say you’re printing ink labels or anything and then reports that back to the EAP. That’s a completely hybrid solution and will stay that way because the machine will not go in the cloud. You still need to feed that with certain orders and then it can prioritize based on what you do on-prem and then it does it. But yes, hybrid is still there. Yes, it’s not going away at all. When there’s, for instance, low latency, then you’re completely on-prem. You’re not using the cloud at all.

Daniel Bryant: Yes, plus one for that. Now, we’re seeing literally all the customers I’m working with, it’s always hybrid. There’s always some element of multi-cloud. It’s quite common, again, for digital sovereignty reasons or compliance or disaster recovery business.

Steef-Jan Wiggers: What I’ve heard, too, is I have workload A that I’m doing on one cloud platform while workload B I’m going to push to another cloud platform. I’ve seen it too.

What trends are you seeing around hybrid cloud adoption? [33:33]

Daniel Bryant: 100%. We see this all the time, yes. And Shweta, I’ll throw over to you now, but something that is wriggling around in my brain and this is somewhat… I’m part of a self-selecting audience, I think, but I do see Kubernetes as the API for a lot of this cloud, even from folks that are doing Edge stuff, but they’re still using Kubernetes on the Edge. I remember Chick-fil-A were famous for running a Kubernetes cluster in every one of their restaurants in the States and that was when I was my previous job. I was working with them. We’re seeing more stories now of people doing hybrid stuff, but the common substrate is the Kubernetes API. It’s not always Kubernetes behind the scenes. It might be virtualized or whatever in the cloud, but on-prem it’s actually they’re running their own Kubernetes cluster. But are you seeing a mix of hybrid cloud and Kubernetes being in the mix too?

Shweta Vohra: Yes, I think hybrid cloud is reality. You rightly said it, Steef. I mean, it’s not going anywhere. I dabbled with a lot of hybrid cloud customers working for IBM, Accenture, and those companies where this helps so many others. I still saw that we have mainframes which are much stronger than any other cloud offerings they have. Hybrid cloud is a reality. Multi-cloud, I’ve seen that people… There is no much fuss nowadays. I’ve heard about whether you should be vendor locking or whether you should be going with multi-cloud. That disturbance, I’m not seeing. The trend wise, if I may call that, people are going with one cloud provider as the base and then topping it up with various services from others.

For example, if I’m going with AWS, then I’m using Gemini or other things from Google as well or other services which I may need. Multi-cloud wise, at least in my experience, I’ve not seen that, again, those articles or those trends popping up that whether we should be still going. On the strategy-wise, one thing which I’ve seen working well as compared to the others who are trying to go all cloud is that if you pick up your major workload, it’s like 80/20. Pick those 20% which have 80% of impact. If you can move them, your organization is there. I think leadership can try and focus on those things. Once that needle moves, rest becomes easier for your business. Those things are what I have noticed.

Daniel Bryant: Ah, it’s good commentary, Shweta

Matt Saunders: People are using Kubernetes as that common substrate. For a very long time, I’d have conversations with people saying, “Oh, we’re going to launch our new products on cloud. Let’s choose two of them so we have resilience. Let’s use AWS and let’s use Google as well”. And I always say, “No, don’t do that. You’ll spend more time and effort doing that than you would just deciding you’re going to move away from one because you don’t want to lock in or whatever and just blitzing that over a couple of months to move across”. That’s now even easier. I think different cloud providers are being commoditized to some extent and people are making choices of where to put their workloads based on bigger contracts, discounts that you get across the board.

That’s the different cloud providers, but also picking up on data center stuff that people are still running, seeing big pushes away from people running their own data centers. The cloud providers are now wise to the reasons why the last whatever percentage of people are still running their own data centers. Workloads like Steef-Jan, you’ve been talking about, which can’t ever actually move because she’s had to physically be somewhere else. I see in my conversations with cloud providers a big push to… It’s like the late majority adoption of cloud and they seem to be having a lot of success on that.

Other things I want to mention is things like cloud repatriation, which occasionally you see David Heinemeier Hansson from the 37signals come up and say how we’ve moved a whole load of stuff off of the cloud and we’re saving an awful lot of money. That is still happening. Less with more does cause that to be a thing, but, again, I think only a small number of people are doing that. Small number of people are committing to have engineers again who know how to reboot servers, that whole thing. For me, I’d be delighted to do that, because I used to do that 30 years ago and I loved it. I quite like tinkering with that sort of stuff and there is money to be saved doing that, but I don’t think we’ve really had a slam dunk case that caused these people to really sit up and take notice and go, “Let’s go off and do that”.

How should leaders measure the success of their cloud, DevOps, and platform engineering investments? [37:42]

Daniel Bryant: We’re coming into the end, folks. I want to do metrics, if that’s all right, and then we’ll do a little wrap up around the room. We have a few minutes left. How should leaders measure the success of their cloud DevOps platform engineering investments? What really matters? Shweta, I’m going to throw it to you first, because I think you’ve said it several times in this podcast in terms of the strategy and the business impact. But I’d love to get your final thoughts wrapping up as in how leaders should measure the success.

Shweta Vohra: If you ask me, I want to do a PhD in that topic because I want to spend some time, but there’s no ideal answer that is ready with me. I think I heard Nicki saying one some space, Nicki Watt, that tension metrics is what’s working best for us. If I’m saying my house is clean, I have to say that these floors are clean, my kitchen is clean, my fridge is in order. You need a couple of things to say that, yes, my house is in order and my house is clean.

Developer experience tells you about your platform engineering, whether it’s working for you or it’s not working for you. Then your maturity assessment of your tooling and platforms will tell you about how your independent components are working. The percentage of things you have really started and finished, that will tell about whether it is the migration or modernization or those things, whether those are reaching the end line or not. One thing to add here is that you would also like to say that if I’m developing a service and integrating something, whether that is really completed in all sense, is it ready to be adopted by anyone in the system, design-wise it’s complete? As I said earlier, it’s gluable, it’s integrated well, then, yes, that is working.

Maybe what I’m trying to say is that you will have to start from the overall start to finish one thing that things which you started, you’re finished. Second is component’s health. You need to monitor each of those, because even if one component will fail, it’ll impact the developer experience. Those component’s health is what you need to major. Of course ultimately when you’re putting all the blocks together, those are integrated and developer experience. These are a couple of metrics putting together. I think we should be able to say that, “Okay, what works, what doesn’t work?” And of course DORA adoption, I’ve seen that a lot of people have taken in developer experience surveys. A lot of companies are adopting it, a lot of tooling around it get DX, et cetera. Those are indications that people are adopting to those things differently.

Daniel Bryant: Yes. Plus one, Shweta. Actually as we’re recording the podcast today, it was just announced DX has been acquired by Atlassian for a billion dollars. I think I’m loving the work that Abi Noda and Nicole Forsgren and Laura Tacho made doing amazing work in that space. Big plus one to all that. Matt, what do you think about-

Matt Saunders: I was about to mention that DX thing. Well, to answer the question on the surface, I always used to just say, “Well, DORA metrics”. How quickly can you get an idea from thinking about it into production? There’s a whole load more complexity there, as Shweta has gone into some excellent detail, on percentages of projects that actually managed to make it to the big time. I would caution that somewhat with my culture of experimentation hat on. Try and drift away a bit or try and pull away from initiatives where you have to be right the first time every time. I think we’re hopefully a little bit past that in many ways. But yes, I wanted to give a bit of a meta answer to that question by referring to that Atlassian acquisition.

I mean, when I first saw it I was like, “Oh, that’s quite interesting”. And then when I saw the number, the price, I’m like, “Oh, my God. Right, okay”. Other people in the industry are also betting big on things like this and taking in tools which… I’m being a bit reductive here, but DX have the core form measurement framework and that seems to be a big part of what they do. Echoing your sentiments, Daniel, on the people involved who I’m sure you may know these people. It’s good stuff and I think we’ll probably see over the next couple of years some different ways of measuring how people are actually being effective.

We all know that numbers of lines of code written is not a good way of working out if you developers are productive or all those stereotypical ways of badly measuring developer productivity. But simply through the sheer intent of what these big number acquisitions, these big value acquisitions convey suggests that we’re going to see quite a lot on this in the future. Whether it’s an AI play or not, I don’t know. It’s too early. I work at the Atlassian ecosystem, but I’ve only got a day’s notice of this as well. As these things come together, you can envisage things like deep integration with wikis, with ticketing systems to work out what people are actually doing. It’s going to be quite interesting to see what products and add-ons we get.

Daniel Bryant: Love it. Yes, great stuff, man. Yes, fantastic. Steef-Jan, just go to you in terms of the metrics.

Steef-Jan Wiggers: Well, it’s funny you’re saying metrics, because I was talking today with a customer. “How could you measure the success of application being built?” Then one of the things they say, “Yes, we do measure how much people are using the application if it’s third party or internally”. That’s one thing. When you build a service that’s ultimately subscription-based, it’s going to be sold, then it’s more like what are the costs versus what your net of your service.

A great example is Figma. We’ve written about at InfoQ that there was like 300,000 US costs to running what they were offering. But on the other end, if you see the numbers of how much they net on their service, and I think Corey Quinn fleshed that out pretty interestingly… You do know there’s a lot of costs on one way, but then on the other end, if you see what it nets and what you’re getting from the service, then that’s another metric, right?

Daniel Bryant: Yes.

Steef-Jan Wiggers: I wouldn’t say lines of codes and that kind of stuff as a metric. It’s more like what are you building, what time, and what does it ultimately get, where it’s user experience and uptake using. You’ve built a lot because that’s one thing you offer. It’s something that you build as a platform, as SaaS or solution or something like that or subscription-based. You can derive the success from that too on one end, I would say. I’ve seen it in the past too, where you build a service, then there’s a lot of costs involved and people say, “Oh, yes, using all these platform services and they’re expensive”. But if you look at a business case and what it ultimately leads to, the costs are not that significant.

What cloud and DevOps topics will be trending over the next 12-18 months? [44:23]

Daniel Bryant: So wrapping up this great conversation, we’ve covered everything from the details to the strategy. I think this has been fantastic. Thank you so much to all of you for that great insights. I always like to do a wrap up. What do you think is going to be the trend up and coming over the next 12, 18 months? When we look back on this podcast in that timeframe, hopefully we’ll all have a chat at conferences or wherever, maybe sharing a coffee, sharing a beer, for example. Will we go, “Oh, I wish I hadn’t said that”, or where we’re going, “Oh, no, I was on point. I’m a genius”? Where do we all think that that’s going to be the most interesting discussion points? Any trends are going to be overrated or things that should be cautious about? As you were guessing, Shweta, I’m going to throw it your way. Let’s open up. What’s your wrapping up phrase of the trends that are up and coming?

Shweta Vohra: I’ll use three terms here. Software engineering, platform engineering, and AI engineering, because AI engineering is thrown into the mix. I would say that software engineering has been working for us. All of us may or may not need platform engineering. Now, platform engineering I would like to see, in a year’s time or the trend which I believe will happen, is that people will be mindful about the platform engineering where they really need it. Same thing for the AI engineering, people will… As of now, this is quite chaotic, but in a year’s time, of course, we may not be able to read the enterprise reality, but we should be mindfully using AI engineering. That for my company, for my role, for my learning, what is that makes more sense to me and going deeper into that solution. Maturity trend is what I think I’m saying, maturing more and more in software engineering, continue doing that. In platform engineering space, be more mindful. AI engineering, start bringing this in.

Daniel Bryant: Classic. That was very nice, Shweta. That’s very good. Steef-Jan, how about to follow? What are you thinking?

Steef-Jan Wiggers: Well, I think this digital cloud sovereignty and more and more cloud providers coming into play will be interesting. Also, when you default on Kubernetes, then you can host your workload in any kind of cloud except for those free. I think that’s going to be interesting. Are those cloud providers still going to be big if you can shift around or if you are forced to stay within your, let’s say, boundaries? Let’s say the EU. You see a couple of big cloud providers coming up there, French cloud providers predominantly. If you can land your workloads there when they’re more like leveraging also Kubernetes, containerization and such, then I’m a bit worried about servers or platform services. And then on the other end, you can also host your AI solutions on virtual machines and they can run in any kind of cloud provider, too, at least also on the up and coming. I’m going to say this I’m not so sure about platform servers and serverless in that area.

Daniel Bryant: Very interesting, Steef-Jan. Very interesting. I’ll double down, actually, and just say if you’re not looking at digital sovereignty, you really should be. I definitely see a bit of a mismatch like we’re seeing on our readership numbers on in InfoQ. Sometimes the pieces about digital sovereignty get a load of views a lot of times from Europe and not so much the States and other parts of the world, but other times they don’t. I think people are not realizing the impact it’s going to have to be in the EU with things like GDPR and a bunch of other requirements. We’re seeing it again all over the world at being implemented too.

I definitely think the other bit of advice I would get people thinking about is the build versus buy when it comes to the platforms. I’m seeing a lot of people that are wanting to build everything. I get it. I was an engineer one time and I still am a bit. I do building and some things get easier now, but I do think you’ve got to look at the ROI. Sometimes actually buying stuff and plugging stuff together or glue, I think a few of us mentioned… I like your analogy, Shweta, the gluing. I think my role as an architect feels to have shifted a little bit more towards those discussions, like the trade-offs and the economics and the politics sometimes as we all danced around it. But I think that’s becoming a reality with my architect hat on that that is a thing that I’m going to buy. If you are not familiar with these things as an architect now, I think you’ve pretty much got to be aware of some of this stuff because it’s actually going to impact your day job whether you like it or not.

Matt, do you want to round us up?

Matt Saunders: Yes. Yes. I’ll just riff on the build versus buy thing a little bit. I think we’ll see some evolution there. I’m still seeing lots of people building empires and a lot of this seems to be because of a lack of pragmatism over how much it costs to actually run things and not help by some pricing models being, let’s just say, a little bit aggressive with some of the SaaS things you can get. But yes, just wrapping up, I mean just looking at the year ahead, I’m really looking forward to solving real problems and we’re starting to see this.

I’ve talked about this already. Solving real problems in the DevOps space using AI in ways that are not highly dependent on the particular model that’s being used and they’re not just able to be thrown out because we think an AI model is hallucinating about something. But seeing people put out products and hopefully some open source stuff that we can use and implement to help us build stuff faster and better and more reliable. I mentioned observability as well. And using those AI bots for good to help fix things and make them better.

I’m anticipating an MCP apocalypse in the next year or so. Where we are, everyone’s adopting MCP and the standard settles down and we resolve all of these issues over how do you give granular access to an MCP server. I see us coming in, with our DevOps and cloud knowledge of how to automate things and how to do infrastructure as code, as being quite a significant contributor to solving that so that we can be giving AIs access to things in the right ways. I think there’s a lot of DevOps learnings that we could be used in there. Yes, sorry to end it on the elephant in the room again, but it is just so ubiquitous and I’m not really sorry.

Daniel Bryant: It’s a positive spin, Matt.

Shweta Vohra: Very good point, frankly.

Daniel Bryant: Yes. Plus one to that, Shweta. Yes, plus one. Well, super. It’s been a fantastic conversation. We’ve covered all the things. I’m super happy to be chatting to you all. I really appreciate your time. Thank you very much, everyone, for joining. Appreciate it.

Matt Saunders: Thank you.

Shweta Vohra: Bye.

Steef-Jan Wiggers: Thanks, everyone.