Google Cloud Next 2025 | Mark Cavage & Tushar Jain, Docker

Clips
More from Google Cloud Next 2025

Tushar Jain

EVP Product & Engineering

Docker

Mark Cavage

President & COO

Docker

play_circle_outline Boosting Developer Productivity and Security with AI: Docker, Kubernetes, and Cloud-Native Workloads Simplified

play_circle_outline Simplifying AI Applications with Docker Model Runner for Easy Use by Developers

play_circle_outline Docker evolving into a platform company, solving developer problems

Info
Transcript

Mark Cavage & Tushar Jain, Docker

Tushar Jain

EVP Product & Engineering Docker

Mark Cavage

President & COO Docker

**Video Title:**
Mark Cavage & Tushar Jain, Docker | Google Cloud Next 2025

**Video Description:**

At Google Cloud Next 2025, Mark Cavage, President and Chief Operating Officer of Docker, and Tushar Jain, Executive Vice President of Product and Engineering, discuss Docker's trajectory and innovations in the cloud-native landscape with hosts on theCUBE Research platform.

Cavage brings substantial leadership experience to Docker with their recent appointment, alongside Jain's extensive expertise in engineering. They explore Docker's ... Read more

explore Keep Exploring

What are some key areas of focus for Docker moving forward in terms of developer productivity and modernization of workloads? add

What is Docker Model Runner and how does it make it easier for developers to write applications against open-source models? add

What are the key components of being the container platform company according to Docker's perspective? add

bolt Powered by CUBE AI

Mark Cavage & Tushar Jain, Docker

search

Dave Vellante

>> Hi, everybody. Welcome back to Google Next 2025. We're here at the Mandalay Bay in Las Vegas, so my name is Dave Vellante. I'm here with Paul Nashawaty, and we're going to dig into Docker. Docker's a company that really got it all started in the whole DevOps movement and Kubernetes. And we were there from the beginning. Super excited to have Mark Cavage on. He's the newly minted president and COO at Docker. And Tushar Jain, who's the Executive Vice President of Product and Engineering at Docker. Gents, welcome.

Mark Cavage

>> Thanks.

Tushar Jain

>> Thank you.

Mark Cavage

>> Thanks so much for having us.

Dave Vellante

>> Okay, five weeks in. So what's the vision? What's the change? What's the 100 day plan? Ben?

Mark Cavage

>> We have the 100-day plan. Now you're going straight to brass tacks. All right. Okay.

Dave Vellante

>> What are the KPIs?

Mark Cavage

>> Yeah. You sound like my board. Great question. Look, we're super excited about everything. Really, Docker, as you said, you said it at the very beginning. Docker started DevOps. Docker started containerization. Docker's largely defined what it means to be cloud-native for a long time now. I think in the next generation and the years ahead, it's really, we're super focused on developer productivity, security, AI, and really helping developers simplify their workloads. And help some of the, Paul's mentioned this earlier, help bring some of the modernization of brownfield workloads into the container-native space, and really help that next generation of agentic apps.

Dave Vellante

>> What attracted you to Docker? I mean, it's a beloved brand and product, but what brought you there?

Mark Cavage

>> Great. Another good question. Well, there's a couple things, team and mission. So this character, Tushar and I have known each other for a very long time. But in terms of the asset and the company, I mean, we have 20 million registered developers. We're doing billions and billions of Docker Hub pulls a month. I mean, we're effectively the ecosystem. And so, in terms of a place that you could go drive a lot of impact and make a big dent in the industry, I think Docker's have a relatively unrivaled position. And really, we'll to talk some more in a bit with some really exciting innovation coming ahead.

Dave Vellante

>> Tim O'Reilly, we all know who Tim O'Reilly is. He's kind of inspired John and me when we started theCUBE, but he has a saying, "Deliver more value than you extract."

Mark Cavage

>> That's right.

Dave Vellante

>> And that's something that's been part of our ethos for a long time, and Docker sort of is in there. But that's got to be really satisfying from an engineering standpoint. The value, I mean, it's almost unquantifiable the value that you delivered. I wonder if you could comment on that, Tushar?

Tushar Jain

>> Absolutely. So actually, that phrase deliver more value than you extract is, I think, the definition of a successful platform. Like a platform is one where people on top of you make more value than you did. And Docker's been that for a generation, at least at this point, right? It's, what, 10 years old? And now as I look ahead, I think there's massive room for innovation here, and much need for the same thing to happen in the next phase of AI development. And making all the workloads that are running highly secure as we make that happen. So one, as a technologist, it's just a great position to be in. There are very few places, if you get out of the CSPs, there are very few places where you can actually go drive massive platform infrastructure impact across the board, and Docker, we have that opportunity, and I'm just really excited to be able to drive that forward. And we'll see this in a number of innovations we'll talk about.

Dave Vellante

>> Yeah. I mean, Docker Hub is really doing that, right? I mean, we see in our research that 75% of our respondents in our research indicated that they're using 6 to 15 different tools across their platforms. That's a lot. With the desire, actually in 2025 alone, 54% of those respondents indicate that they want to move towards a more unified approach. But to Dave's point, delivering more value, that's what Docker Hub was doing. That's what your intentions are, right?

Tushar Jain

>> Yeah. So Docker Hub is, it's sort of the backbone of open source container development, right? We get 11 billion pulls a month. If you want to get open source images, you come to Hub, it's the source of everything. That's where everyone's pulling this from. And it's all hosted for free and we give it to you, right? And now on top of that, we can start capitalizing and solving new problems for customers, which we're doing as we're hearing from them.

Dave Vellante

>> Yeah, and it was a paradigm shift over the last year, right? I mean, big changes in the platform, more adoption. It sounds like you have a significant amount of uptick from what I'm seeing in the market.

Mark Cavage

>> It's been growing for a long time, and it keeps growing. Really, to double down on the point Tushar said, we effectively are the upstream for the Internet's open source distribution now. And so you come to one place, you're coming to Docker to get it. And so that's really, to your point on all the numbers and the consolidation and what you're seeing inside of the enterprise totally tracks with what we see too.

Tushar Jain

>> Absolutely.

Dave Vellante

>> Well, and we are seeing a massive shift. I mean, we always sort of bromide, but these waves only come around every so often. This one, I think when you listen to people like Jensen and Thomas talk about this wave, you realize the entire stack is changing. And I can go through whether it's compute, or storage, or the data platform, I mean, all the way up through the applications. So my question is, how is... So two-part question. One for the president and COO, interested in how you see the vision of the company changing as a result of this AI wave. And from a technical standpoint, what does it mean to be AI optimized in your world? Maybe start with the vision.

Mark Cavage

>> Great question. So today we launched Docker Model Run as an example. So I'll come back to this in just a second. But to your question of where are we going with the company and how are we thinking about AI and how much changes? Two things. Number one, we actually think containers are the right packaging for most open source app, or most AI applications, excuse me. And when you think about the workload that an agent has to go off and do, independently be versioned, independently be shipped. Talk to another agent, talk to tools. Like all of that still feels like containers. And so, we deeply believe that Docker is going to be the home run place, or the center of still in the AI world. Packaging up agentic apps, deploying agentic apps, securing agentic apps. And the Docker Model Runner announcement I just mentioned a second ago, we're making a lot of investments. They were really excited about this. We've got tremendous... Like the booth, a lot of people are lined up over there. We're making it incredibly easy at Docker, simple, to use that kind of word, to download and run any open source model that you want from Docker Hub. So you get GPU acceleration on your laptops, and ultimately that will translate to flexibility across cost, performance, latency, and product, so.

Dave Vellante

>> Tushar, based on what Mark just said, it's almost like the market is just coming to you. It doesn't require rethinking the architecture. Like for instance, storage. It's got to be disaggregated. Or compute. Just think of cluster management. I mean, those are massive changes. The engineering task for you is more extending as opposed to rewriting, is that right?

Tushar Jain

>> It's a mixture of things, actually.

Dave Vellante

>> Yeah, please.

Tushar Jain

>> As I think about it, to follow on what Mark said, there are a few aspects to think about of, like what we have to deliver, right? What does it mean for AI applications for us to do a good job with them with containers? One, you need really good developer experience. All the way from inner loop all the way through to outer loop. Second, you need performance across all of it. And then third, you need security across all of this. This is what containers did for microservices when that started. So now as I take that and I think about what's happening in the AI space, I take Docker Model Run, it's done a few things. One, we made the developer experience really easy in the inner loop, and we follow this all the way through. Doing this did require interesting technical challenges and changes, because how you expose the GPU inside a container is non-trivial.

Dave Vellante

>> How you exploit the capabilities, yeah.

Tushar Jain

>> Exploit the capability, right? And so there's deep work we've done there, and then as you bring this out. As you go to production, there are interesting challenges here that we'll face. If you look at it, two fundamental physics have changed. The size of containers with models have gone bigger. And second, now, instead of CPUs we are willing to throw away in the old world, we now have GPUs we care a ton about and want to optimize and get every microsecond out of them. But the infrastructure still hasn't caught up. Everything from how we are actually deploying packaging and how we are running in Kubernetes and stuff. And those are all areas and opportunities for us to go drive. And the last point here, even as you think about tools like MCP coming out, we've worked with Anthropics since the beginning. That's another place, I think, where there is a lot of room across both developer productivity, developer experience, security, and sort of discovery and management of all of that.

Dave Vellante

>> Yeah, I think that makes a lot of sense. And when I think about the announcements that I saw at the keynote today, there's a lot of synergies between Docker Model Runner and the announcements, right? I mean, when you think about the hybrid cloud and multi-cloud approaches, when you think about the AI powered, like you were talking about the size of the containers. But the other thing I also want to understand a little bit more, and I know we've talked about this in a lot of briefings that we've had, but is the bridging the gap between the old and new, right? That's a big factor, because obviously we're here looking at the new opportunities, but when we look at bridging the gap between old and new people need a path to get to the new. What would you recommend for them?

Mark Cavage

>> Well, great question. I just talked to two customers today about this very problem. You know, some interesting challenges that come up. For example, many of them are on Windows environments versus Linux environments, and it's just not so easy to port. So what we see a lot of is across both in solving the developer productivity problem of the, what's called the brownfield app, giving them a container-native workflow on the laptop so they can actually just have better velocity. We see this with people displacing things like virtual box and moving to container environments, again with that old-school app. And then the other is, the classic problem of, I've got a monolith and I need to decompose it. Just eating away at the edges, and starting to change the packaging model first and foremost so you get velocity. And then eating away at the edges with ancillary services that kind of live in the orbit of the big monolith. And ultimately that takes a long time. Many people are on a many year, or many decade journey, I suppose, in some cases to transform these things. But, yeah.

Dave Vellante

>> Can we get into Docker Model Runner? What does it do? What problems are you solving?

Mark Cavage

>> Do you want to take it?

Tushar Jain

>> Docker Model Runner really simply is, we have open-source models in catalog. You can just start running them locally to develop apps against them locally. You do Docker Model Run with the name of the open-source model and you get it. That's it. That's simple. Runs locally on your Mac today, tomorrow it'll be on all the platforms as we roll this out across Linux. So for developers, what it's really doing is making it really easy to write applications against open-source models. We'll also be modifying compose. So you can write an application, write a compose stack that spins up your app server and an open-source model and get all that working. And so for developers, this just makes it where they are. We meet them where they are and fit into their workflow and let them run against open-source models. And the bet here really is that developers will use open-source models and embedded models and small language models in a number of ways. We'll also let them publish their own into our catalog.

Dave Vellante

>> But again, you're coming back to unification too, right? So if a developer can have a bespoke deployment and use the tools that they're comfortable with, but still use Docker Model Runner as the way they run.

Mark Cavage

>> That's absolutely right. I mean, to answer the question a little bit more on like why do this in the first place? We're believers that the world is going to be multi-LM. We're going to see a guard, many people use this word, the large language model garden. People are going to pick the right thing for the right job and the right agent at the right time, and might use one for OCR, and one for text processing, and one for customer support, and so on. And so developers need flexibility. We think local LLMs are super important, because again, whether it's for cost latency, privacy, data sovereignty, any number of reasons, it just makes sense in some cases. In other cases, people use the big hosted cloud models. But ultimately, to your point of unification, we want it to be easy. And you want to take all the muck and the crap away from a developer of what it means to go build an app that works with LLM, whether it's locally or on a big one in the cloud. And so just make that easy for them to focus on the job they want to do, which is, I just want to build an app. I don't want my app to be an agent. And I just want to make that easy and painful. And hopefully we have developers from Docker continue to hug us the way they do now.

Dave Vellante

>> Do you think it allows for organizations to hire more application generalists and specialists? Because what we're seeing in our research is, 67% of respondents in our research are looking to hire generalists over specialists, because they can't find the skill to hire for the specialists. Does this help?

Tushar Jain

>> Go ahead.

Mark Cavage

>> Go ahead.

Tushar Jain

>> I was saying, I'm thinking absolutely, right? In large part by making... Like what's happening in this space is, instead of having MLEs and specialists like machine learning engineers for everything, what's happening is, software development is having to build generic general gen AI applications. Anything we can do to plug into their workflows makes it easier to have generalists within. There's another example of what Docker Model Run. In addition to running, we have created, we've standardized how models are packaged up into open source OCI models to those specs. So now they can also take the models they build and put them into any OCI registry, including their own on-prem ones. This again, fits into where they are, and fits in to the normal workflow, which allows generalists to keep doing what they're doing and use this.

Dave Vellante

>> So Docker, I mean, platforms all the rage, but they've been the rage for a while. Platforms beat products. So as the newly minted head, how do you think about Docker evolving from what was once a beloved dev tool to a platform? What does that mean? What is the Nikesh Arora from Palo Alto uses the term platformization of the industry. How do you platformize the company? How do you think about platforms as the future?

Mark Cavage

>> Great question. Well, one, I think Docker's actually always been a platform when you think about it.

Dave Vellante

>> Fair enough.

Mark Cavage

>> I'm going to call you on that one a bit.

Dave Vellante

>> No, it's fair enough.

Mark Cavage

>> Okay. But in terms of what does that mean to be the container platform company? I think we have to solve all the problems developers have, everything from the IDE all the way through to prod and that entire life cycle. And again, today, Docker is already present every step along the way. But what we hear from our customers and developers all the time is, "I really need help securing it. I really need help with CI. I really need help with bringing down the cost and time to my deployments. I really need help building AI applications, and I really need help modernizing all the things in my stack." And so for us, being a platform means we're going to have first class offerings across every step of that journey from your laptop to prod. And as Tushar said, and you said, having an ecosystem, people that get more value out of it than we do on top of us to keep working with Docker, the platform as it is today.

Dave Vellante

>> Tushar, what's your engineering team's superpower?

Tushar Jain

>> I think really good developer experience, but really go really deep technically, and combine those two. And it's rare to find both, where you have people with strong, deep, but can also pop back up and build a really good experience. And that's sort of Docker's superpower, I think, over time, right?

Mark Cavage

>> I'm also going to jump in for you actually. Which one is like passion, actually.

Tushar Jain

>> Passion.

Mark Cavage

>> Like the developers inside Docker, inside of Tushar's org, they've all come from outside of Docker and they're fervent, fervent believers in making Docker better because they live the pain.

Dave Vellante

>> As is the community.

Mark Cavage

>> Yes, it's community.

Tushar Jain

>> That's because they've all been active Docker users themselves, whether they're at Docker or not, right? It's tech that's everywhere.

Dave Vellante

>> And you keep investing too. Like the test suite is another area where you're invested. That engineering team that came in was a solid team. So yeah, all this investment that you're putting in the company is really building your bench strength.

Mark Cavage

>> Yep. And yeah, again, what you're going to see over the months ahead is just these things coming together in a tighter and tighter integration loop to make it faster to deploy, make it faster to develop, make it cheaper to develop, let you build AI apps with hiring less generalists, and ultimately secure the supply chain. Which is, oh I need it again and again.

Dave Vellante

>> I have to ask the question because you led it here. All right. Over the next few months, what's coming?

Mark Cavage

>> All right. Well, I see I walked into that one.

Dave Vellante

>> You did, you did.

Mark Cavage

>> No, great question. Actually, next week I think you'll see some very exciting stuff coming out of us around the MCP protocol. Tushar said this earlier, we've been working with Anthropic since the very beginning. We've got an investment in Docker Model Run, we just talked about. You've told us all the numbers on Docker Hub that you look at. It makes a lot of sense. There's a natural place to go find, manage, discover, package, and install MCP tooling. So-

Dave Vellante

>> Pretty cool.

Mark Cavage

>> Stay tuned for more.

Dave Vellante

>> Well, Savannah took a break, so I'm going to ask the Savannah question. I'll start with Tushar, and then Mark, you'll bring us home. What do you want to be able to say next year at Google Cloud Next 2026 that you cannot say today?

Tushar Jain

>> Everything that Mark said we will do has happened, and we've got lots of customers on it. So meaning, we're a continuing platform that solves. We have customers using us across to solve problems across productivity all the way through production. They're securing the end-to-end SCLC. It's easier to get compliance by running on us all the way through, no matter where they run. AI workloads are running on us because it is much better developer experience. It's highly secure. Also, we haven't touched on this too much here, but the publishing network for AI has also grown rapidly on us. Satellite hub is also a massive marketplace where everyone publishes images on us. And last, yeah, I think those three things. We've got security. And yeah, security, develop productivity and AI, all that is running, and we've got customers running on us. That's it.

Dave Vellante

>> And Mark, the board may have you on the 100-day shot clock, but what do you want to say a year from now that you can't say today?

Mark Cavage

>> I would just underline everything Tushar just said. And actually I'd flip it back to you. What you should be asking me in a year is, what do you want to be saying in five years? And so my hope is a year from now all the things we're talking about exist, and you'll have a whole lot more press releases to point at. And really you're going to be asking us how much bigger we get.

Dave Vellante

>> Well guys, thanks for coming in on a pinch. I know it was a tight timeframe, but-

Mark Cavage

>> Thanks for having us....

Dave Vellante

>> it really, it was a pleasure having you on and best of luck.

Tushar Jain

>> Okay.

Mark Cavage

>> Yeah, likewise.

Tushar Jain

>> Thank you.

Dave Vellante

>> Thank you. All right. Dave Vellante for Paul Nashawaty, Savannah Peterson and John Furrier also. You're watching theCUBE's coverage of Google Cloud Next 2025. We'll be right back right after this short break.