Swapnil Bhartiya: When it comes to AI, most of us think that scaling AI is just about buying more GPUs. But the real bottleneck is orchestration. Enterprises struggle to stitch together complex hardware and software into unified AI factories. It drains resources and starts innovation before it even begins to solve that. IREN is acquiring Mirantis to combine massive compute power they have with deep cloud native expertise that Mirantis has. And today we have with us once again, Alex Freedland, CEO of Mirantis, to unpack what this means for the future of AI infrastructure. Alex, first of all, it’s great to have you back on the show.

Alex Freedland: Great to be here.

Swapnil Bhartiya: It’s my pleasure. Of course, we have been covering Mirantis from very, very early. I even forgot how many, many decades ago it was. Talk a bit about what specific market dynamics make this the right time for you folks to join hands with and be acquired by them.

Alex Freedland: The market dynamics is an echo of what put Mirantis on the big stage in the first place. In 2011, when we focused on OpenStack and people learned about who Mirantis was, the market was kind of doing a similar dynamic. The cloud, the first generation of cloud was moving from hosting to cloud, right? And if you remember in the early days of cloud, the people like crack space and bunch of others, they were in a leasing business, they would go and put servers in cages and then would essentially lease those servers to people. And that’s what service providers did. And then of course AWS came in and they kind of pioneered the idea of a cloud which became self service orchestration and all of that. And the market moved from hosting to cloud when people understood that this is required for efficiency. So a very, very similar thing is happening right now where the world of AI started with massive training grounds, right? So to kind of get AI to do something useful, you first needed to train the models. And so for the first year or a couple of years, what was necessary was big foundational model companies would come in, they lease large bare metal clusters and they put their training runs on that to train chatgpts and all the other foundational models. And then as those models, be it foundational models or smaller models or open source models, have become useful, Inference became a thing and as inference is becoming a thing and you now need to run your inference workloads and they’re bursty by definition, some need to be adjusted this way. They need the right amount of hardware and then you need to do fine tuning and you have multiple generations of hardware which all work together. Suddenly the world woke up to the fact that hosting is no longer interesting. You actually need to have a self service cloud and that is kind of the time of the market that we are now in at. And like you said, orchestration is basically an essential element of what is required for an AI factory to be able to service real burst inference workloads.

Swapnil Bhartiya: I’m pretty sure that when the acquisition news came out and of course when you’re in discussion, you may also be aware of that a lot of your customers may also be worried that what does it mean for them? Because you folks have been betting a lot on neoclouds. Of course Mirantis has been very big player in the open source space as well. Can you talk about what kind of concerns that you have heard because of this acquisition may have towards kind of vendor lock in? How will you ensure that of course you will be running as an independent subsidiary, if I’m not wrong there, to ensure that you will continue to run the way you are running, whether it’s open source and also continue to sell to other New Clouds or other enterprise players as well.

Alex Freedland: Yes, we did hear a few concerns from our prospects and Neoclouds who were wondering whether or not this would change anything to them. But you know what I responded to them is how I’m going to respond here. The whole thesis of us coming together with IREN came as a result of our discussions with the founders and of course IREN has become a customer before they decided that they’re willing to acquire us. And what happened with us was we focused as a company on New clouds and we also focused with Nvidia on becoming the early ISV in the Nvidia AI Ready cloud program. So of course Nvidia is today still remains a dominant provider of GPUs to the market and they also understand that the future of workloads is in inferencing. And ultimately, you know, the AI native startups and companies and all that, they need to be able to consume AI on their own terms, right? And of course there is a huge amount of real estate that people are building those AI factories on, which is energy and data center space. And a lot of the companies that are building energy and data center space, unless they have easy way to orchestrate from bare metal all the way up to a workload, they will not be able to bring AI directly to the customers, right? But they will have to lease their bare metal to a hyperscaler or to, you know, Facebook and all that. And the market is going to be very, very narrow and of course the whole idea of bringing the AI ready cloud ISV program was for Nvidia to trying to democratize this ability to enable all the new clouds and enterprises and telcos and everybody to be able to orchestrate all the way from bare metal to a workload and enable a much more democratic way for customers to adopt AI. And that was the thesis, and that was exactly the thesis that brought us even closer together. And you might remember that at GTC we were announced as one of the three inaugural partners and we worked very closely with Nvidia because of our experience with intel and what we did before in the OpenStack world. And we wore that ISV. So we brought that expertise to Nvidia and IREN was our launch customer of this large scale program that later constituted in this contract a $3.4 billion deal where Nvidia is actually adopting Nvidia as a customer, adopting our stack, our offtake stack from IREN as a provider. So we announced the three way partnership there at the show in March. So the thesis hasn’t changed. And the way IREN looks at the market is they are first and foremost an energy provider and a GPU provider. And in the market today and the market for the next five to 10 years is going to be a supply constrained market. So nobody is worried about being able to find customers. The one thing you worry about is whether or not you can remove friction for all these customers to actually consume AI efficiently, to be able to get all the way to the token and to be able to make sure that the token economics is such that you can actually be efficient in the way you scale. And the number one challenge that we have to solve for is to be able to go to the end customer, give them the AI supply with orchestration and the good economics and you expect that all these customers will be looking for availability wherever it exists. So they can consume it in the hyperscaler, they can consume it from IREN or they consume it from anybody else, because we’re all going to be supply constrained. And they will be, in addition to supply constraint, they will be, there’ll also be regulation. You’ll have to be in the country, you will have to be of a certain security levels and all that. So from that perspective, IREN never looked at other Neoclouds as competitors. They looked at other Neoclouds as partners with whom you can create an environment where a customer can truly adopt AI at scale. And that was the thesis that the founders from IREN and myself kind of met each Other and we understood that the opportunity is very large. The pipeline is more than we can possibly digest. And my conversation with IREN was, how about you buy us? My conversation was, hey, we’re looking to raise money because we, you know, our investors have been in this company for 10 years. They’re not in the position to give us a lot more money. So we need to refinance the company. We actually started this conversation with Nvidia who was supporting us in that. And IREN came in and said, hey, we have a lot of capital. Our thesis is one on one. How about we essentially refinance and become your investor and we will accelerate you and everything you do with us will be available to the world and all the other use cases we’ll be able to adopt and we’ll just remove friction from the adoption of compute all over the world. And that was this thesis of our merger. And if we would have gone and done this with trying to raise money from capital, which we had a lot of interest and Nvidia would have helped us, it would have taken us six months to refinance the company. Instead, it was such a good opportunity that the whole deal came together in two weeks. So to me it’s an opportunity to get more capital, more use cases, access to infrastructure at scale where it can get the use cases. Take that learning to the rest of the market and scale, scale, scale and build, build, build. So that was the thesis and that’s what we’re doing. And the deal wouldn’t have happened if it had been any other way.

Swapnil Bhartiya: First of all, thank you for explaining in such a detail. And also a lot of things that people are not aware of, you know, talking those it gives and brings more clarity. And second thing is the market is changing so fast, just doing one bit is not going to help. And you can’t do everything either way. So I think you use the term merger, so it’s more or less like, you know, and there are a lot of companies that operate in the same fashion, like Red Hat and IBM is a great example, right? Red Hat runs like an independent company within IBM. They actually work with each other as their clients and partners versus ownership.

Alex Freedland: So by the way, VMware, which is the most successful infrastructure story, was originally acquired by amc, right? And then became part of Dell and went public and then, you know, became, you know, it’s the single most successful infrastructure play, unfortunately taking awry by the Broadcom acquisition. But before that it served the industry. It removed the friction, it created the control plane which for virtualization for two generations, right? For two decades. So I think similar play exists today, except, you know, we’re not going to be closed source, we’re going to be open source. So it’s more like red hat from that perspective. But it is important to have a large infrastructure player backing a software player. It’s been done before and in fact it’s been the most successful play infrastructure companies have seen, even sometimes.

Swapnil Bhartiya: I have also asked this question in past that, hey, how do we look at Mirantis? Because I’ve been covering for so long and you keep evolving with where the market is. Many people still associate Mirantis strictly with Kubernetes and OpenStack. What does this deal signal about your future directions and where the broader AI infrast heading? Because these days everything is about AI VR. I mean the. We have moved past all the infrastructure. The most important thing is to put that in people’s hands. Make it cheaper, make it affordable, make it more powerful, and of course make sure it doesn’t hallucinate. But let’s talk about how should we look at Mirantis now?

Alex Freedland: Well, I mean, Mirantis has always tried to focus on where the market is going, like you said. Right. And we pivoted into OpenStack because we understood that, you know, the compute now is becoming a cloud. And we went where the demand was and we came up with OpenStack as a vehicle and we became the winner of that game. Right? Then compute evolved from virtual machines to containers. And you know, we did the Docker acquisition and we became relevant in the container world while being true to our IIS roots. Right. As that happened and as that evolved, AI became first and foremost a workload. Right? Now because of the models, you can create a workload that has a lot more power to change the world. Right? And so as a result, the demand for compute has gone through the roof and it pushes with it, you know, storage and networking as well. And so Mirantis of course is there to serve that overwhelming demand. And we are positioned probably at the best just in the industry today with this acquisition to become a major player in that. But when you think about this, when you have a workload that requires a lot of compute, this workload is not just the model. And this workload has a lot that’s going on that there are agentic things that will be running and calling that model. There’ll be databases, there’ll be regular traditional compute things. I mean, you’ve seen intel and AMD becoming going to all time high because now that we’re seeing agentic workloads being adopted using GPU infrastructure, the traditional computers being pooled in spades. Right. Intel is an all time high. In fact, there was a funny article saying that 51% of all the Nvidia investments is now intel because they invested $5 billion to kind of help the company. Now that $5 billion is worth more than 50, more than $25 billion. Right. Why? Because the accelerated computing demand is pooling all the other compute. Right. And now that we’re moving from bare metal to full stack, suddenly IIS is so important. Why is Nvidia suddenly got so close to Mirant? Is it because Mirantis is, you know, a company that can pull Nvidia? No, because Mirantis is the only company out there that actually kept our IIS business. Because when private cloud kind of stopped being a thing, everybody went to hyperscalers and all that. Mirantis was there, still kept our OpenStack business and had a very significant understanding how to build the iis, which nobody else did. Now Red Hat is kind of slowly coming back to that. They reinstated their OpenStack business and they’re doing this. But. But we’re 50 times more nimble than $6 billion red hat. Right. So we were able to kind of out execute because of that layer of the stack. And then all the infrastructure orchestration is still done by kubernetes. Right. So Kubernetes as a service is equally as important and we have that now. There are many companies that are doing kubernetes and that’s still relevant right now you can run models on top of that and we need to invest at the top of the stack. But in the end, AI doesn’t change the fundamentals of the full compute stack. You need all of the iis, you need all of the kubernetes and we’re doing that in space. Now another thing that happened to us is we do have 1500 enterprise customers, big and small. And that was one of the things that again Ayden is interested in because they need to go back. These are the end customers they’re going to be selling to. So today many of those customers are enterprises. They’re experimenting with AI or on, you know, on top of ChatGPT and all those other things. But as they scale, they will need to consume their own infrastructure. They’ll lease it from everywhere it’s going to be available and they’re modernizing their infrastructure to be AI ready because that’s where the world is going. So if you look at our traditional business which we call core, right? We’re actually everything we’re learning from our k0rdent experience, k0rdent AI is, is flowing into our MTE platform and that is slowly being implemented into our customers. And then our customers get everything we learn from service providers. So that’s a natural progression of how we do things. It’s just, you know, everything is going in the same direction. The only difference is for some future is today, for some future is tomorrow and some for future is day after tomorrow. And if you look at how we’re building our business, we, we’re going to our customers exactly where they are. For those who are moving slower and they need traditional compute, we’re helping them to modernize so they can run AI in two years. For those who need AI experimentation, we already run those experimentations and Cordon is working together with MKE and everything else. And for those who need it yesterday, which are Neoclouds and service providers, we’re leading the charge and leading the market. So it’s not a stretch for us. It’s actually we are evolving with the industry and we’re very much where our customers need us to be.

Swapnil Bhartiya: Excellent. Once again, thank you so much for explaining that in detail earlier. You did of course mentioned Nvidia and you also when talking IREN, you mentioned that you’ll continue to work with other. Now of course you have been leaning heavily into Nvidia aligned infrastructure and of course the whole AI factory concept. Are you also open to work with other chip makers and integrate with the wider range of AI software or you are going to focus more on Nvidia?

Alex Freedland: Well, we focus on Nvidia because Nvidia has been 90 plus percent of the market. Right. That’s where all the action is and that’s where most of the innovation has been. Right. So we focused on that but we haven’t been. Ultimately we focus on the industry and our customers because Nvidia doesn’t pay our bills, our customers do. Right. So we need to make our customers successful. As we’re seeing our customers expand their GPU reach and all that. We’ve been already partnering with AMD and you have seen some of the announcements, you know, if you’ll go to AMD website, they see that, you know, we’re supporting wide range of their chipsets with k0rdent. There have been an announcement with Super Micro that we’ve been validated on their AMD chips and we’re, you know, we’re working on some, on some go to market arrangements that way. But beyond that I think there is a bigger motion that’s happening and this is through the open source community. You might have seen Nvidia open sourcing a large portion of their stack, starting from NICO infrastructure all the way up to Lepton and Run AI and all that that happened around March of this year. We’ve been very closely aligned with this whole open source motion and we’ve been talking to the Linux foundation where it’s very interesting that the folks who built OpenStack foundation, with whom we’ve been on boards together on all that now have merged into Linux Foundation. You know, Jonathan is the chairman of CNCF now and all that. So it’s kind of, you know, everything is coming back, you know, into, into one motion. So we have been working very closely with Nvidia and the Linux foundation to make sure that open sourcing is actually done in the way that it supports broader industry momentum. And in that you will see more of this work being, being, you know, happening. And this will also invite other players like AMD and other chip providers and networking providers and everything to be able to support that. Because again, in the end we are in a supply constrained market. And what that means is every chip that Nvidia makes will get sold, every networking piece that Nvidia makes will get sold. Right? The question is same is true for amd, by the way. There is no capacity that AMD has produced that is not being sold. And anytime soon that’s not going to change. What’s going to happen is we need to create enough solutions and hopefully it’s going to be open source first so that a customer, when they’re able to get their arms and hands to whatever infrastructure is available, can have the software support that just works with everything. And we’ve seen how it’s been done with OpenStack and how we did the hardware compatibility lists and how we worked with all the vendors. Similar motions are now being set up to enable the rest of the industry because ultimately it’s about the customer, it’s not about the chip that they produced. And Nvidia understands that, AMD understands that. And that’s the whole motion that we’re pioneering because of our open source pedigree and conviction.

Swapnil Bhartiya: Excellent. Once again, thank you for explaining that so beautifully. Of course, we talked about this earlier. You mentioned it briefly. I also want to talk a bit about Open Source because you folks have a long and active history in the open source source space. How does the IREN acquisition influence your open source working strategy? And will you continue to engage with the Community or something might change there.

Alex Freedland: We will engage in the community a lot more. You will see us contribute a lot more. Again, our understanding from our partnership with IREN is, is exactly this, right? It’s about removing friction and creating standards. And we have full support and full conviction from the, from the IREN organization and founders directly. Right? The challenge for a company like Mirantis and one of the reasons we’re, we’re not able to maybe contribute as much as we have in the, in the, in the past is because we, we’re a much smaller organization compared to Red Hat for example, right? There are 6 billion and we are a little over 100 billion kind of growing quickly. So to be able to kind of punch way above our weight, we had to have partners who had this belief and in the first iteration of this thing, we had big support of intel who gave us a lot of money so we could actually contribute things under our name into the communities. So now we’re having similar conversations with Nvidia who is looking to back us into those things. And then of course it does help to have a $20 billion business with tens of billions of dollars in the balance sheet behind us. So we can actually allocate some of that and into creating an open source motion and again remove friction for the industry. And IREN will greatly benefit from that. Because open source is good for company like IREN, because ultimately open source removes lock in and as a result, if you have a lock in in the stack, and we’ve seen this with VMware who just decided to raise prices 10x. So what happens when this happens is, you know, if you run a workload and somebody in the middle with a proprietary lock in puts in, you know, takes all the margin, whatever is left is being essentially the money has been taken out from anything underneath. So lock in is good for people who build, is bad for people who build infrastructure. And IREN understands that they’re putting tens of billions of dollars into building the most efficient infrastructure. They don’t want to have some proprietary actor just put a chokehold on the customer and say you can only have a percentage of that. That’s not good for IREN, that’s not good for the industry. So the conviction of open source stays and infrastructure has always been the case where, where open source is actually a predominant player in infrastructure software. I don’t think it’s going to change.

Swapnil Bhartiya: You have addressed this question, but I just want to ask it separately, is that once the acquisition is finalized, what impact, what immediate impact will have on the day to day operations of your customers and roadmap or it will have zero impact. They will continue to work with Mirantis as they were doing. It is just happening at Mirantis Financial and it will not have any impact on customers.

Alex Freedland: Well, it will have an impact on our customers and it’s going to be a very positive impact because the first thing that is happening is we have access to tens and hundreds of thousands of GPUs and many different use cases. So everything we’re building will be tested at scale. And that means to whoever is trying to build an AI factory at scale, Mirantis will be by far the best company with the best experience to do things at scale. And you understand that in the world of GPUs getting access to GPUs at scale is the main problem of any software provider. So we will have solved for that as number one. Number two is we will be able to greatly accelerate our roadmap because we have, you know, a company with tens of billions of dollars of, you know, the balance sheet available to us, maybe not all of the tens of billions, but hundreds of millions of dollars will be available to us to accelerate R and D and to accelerate everything that we do. And of course our relationship with Nvidia will continue which means that they are very interested in making sure that all the partners are able to do offtake and bring AI all the way to the end consumer. And that is again going to continue at length. So what’s going to happen is we will continue to operate independently. I will remain a CEO. There is I’ve given the commitment to stay for five more years and it’s a great opportunity that I’m so excited about. Mirantis will have an independent, well, a board today, you know, it’s discussed. It’s going to be board of five where IREN will have two seats and will be two. I will be on the board, one of our independent directors will be on the board board and we will bring another independent director and ultimately we will just go and create a control plane for AI cloud industry.