By Enterprise AI projects don’t stall at the GPU procurement stage. They stall at the operational layer — the moment a platform team has to figure out how to deploy, orchestrate, monitor, and recover AI workloads at production scale, across heterogeneous environments, with hard multi-tenancy requirements and no standardized tooling to lean on. GPU clusters get provisioned and then sit underutilized. Infrastructure teams work around gaps in scheduling, observability, and lifecycle management that the cloud-native community hasn’t fully closed yet. The distance between “we have the compute” and “we have AI in production” is measured not in hardware but in operational debt.
This is the problem that IREN Limited (NASDAQ: IREN), a vertically integrated AI Cloud provider spanning power, land, data centers, and large-scale GPU infrastructure, is moving to solve through software. On May 5, 2026, IREN announced a definitive agreement to acquire Mirantis, Inc. in an all-stock transaction valued at approximately $625 million. The deal is straightforward in its logic: IREN brings raw GPU infrastructure at scale; Mirantis brings the software, orchestration, and enterprise operations expertise needed to actually run production AI on top of it.
Mirantis is not a newcomer to this challenge. The company has spent more than a decade building open infrastructure — first on OpenStack, then on Kubernetes — and counts over 1,500 enterprise customers globally. Its k0rdent AI platform manages AI infrastructure across bare metal, virtual machines, and Kubernetes environments, and Mirantis is a founding Independent Software Vendor partner of the NVIDIA AI Cloud-Ready ISV Validation Initiative, one of only three inaugural ISV partners in that program. The combination of IREN’s infrastructure scale and Mirantis’s operational software layer is designed to collapse the time it takes to bring AI workloads from provisioning to reliable production.
Mirantis will operate as an independent subsidiary with its own governance board — including both company representatives and independent board members — preserving the open source commitments and customer continuity the company has built its reputation on. The open source community commitments remain intact. The existing customer contracts, roadmaps, and day-to-day staff relationships remain in place. The acquisition is structured to accelerate, not absorb.
The strategic context is broader than any single deal. As AI infrastructure emerges as the next major technology wave — on a par with the mobile revolution, the Linux-led compute era, and the Kubernetes moment in cloud — the industry is splitting along its historical fault line: proprietary stacks versus open standards. IREN and Mirantis are explicitly betting on the open standard side of that split, with k0rdent AI positioned as the candidate platform for industry-wide adoption.
The Guest: Dominic Wilde, SVP and General Manager, Core Business at Mirantis
Key Takeaways
- IREN’s $625M all-stock acquisition of Mirantis targets the AI operations gap — not the GPU supply problem — by adding a software and orchestration layer to large-scale GPU infrastructure.
- Mirantis will operate as a standalone subsidiary with independent governance, preserving open source commitments and existing customer contracts without disruption.
- k0rdent AI is designed from the ground up for composability — opinionated reference architectures with full freedom to swap components, purpose-built for GPU virtualization, multi-tenancy, and sovereign AI requirements.
- As a founding ISV partner of the NVIDIA AI Cloud Ready Initiative, Mirantis occupies the middle tier of a three-layer ecosystem (NVIDIA foundational software → ISV orchestration → cloud provider delivery) that NVIDIA formalized to accelerate industry-wide AI deployment.
- Day-two operations — observability, FinOps, failure recovery, lifecycle management — are the next major frontier for AI infrastructure, and the IREN-Mirantis combination is explicitly focused on closing those gaps for distributed inference, edge workloads, and large-scale training environments.
***
In this exclusive interview with Swapnil Bhartiya at TFiR, Dominic Wilde, SVP and General Manager of Core Business at Mirantis, discusses the strategic rationale behind IREN’s $625 million acquisition of Mirantis, the mechanics of operating as an independent subsidiary, the technical challenges Kubernetes must still solve for AI-scale GPU workloads, how k0rdent AI was architected for composability and open standards, what the NVIDIA AI Cloud Ready Initiative means in practice, and how observability, FinOps, and day-two operations become the critical differentiator for enterprise AI infrastructure at scale.
The Core Problem: AI Operations, Not AI Hardware
The framing of enterprise AI as a compute problem misdiagnoses the actual bottleneck. Platform engineering teams face a different challenge: the operational and software layer required to deploy, orchestrate, monitor, and recover AI workloads at scale remains immature, fragmented, and poorly standardized. The IREN-Mirantis deal is a direct response to this gap — combining infrastructure at scale with software and operational expertise.
Q: Mirantis has built its reputation on open infrastructure, open stack, Kubernetes, and no vendor lock-in. Now you’re being acquired by IREN, which is very much an AI cloud provider with its own stack. How does Mirantis stay neutral, and what happens to customers who want to run k0rdent or Mirantis software outside of IREN’s cloud?
Dominic Wilde: “The companies have a shared vision on this. We have a long track record of building on open source, delivering solutions that serve customers, giving them both an on-ramp and an off-ramp if they should require it. If you look at history, every time there’s a major technology wave, platforms split into proprietary and open. We’ve seen this in the mobile space with Android, we’ve seen this in compute with Linux, and we’ve seen this in cloud with Kubernetes, where the open standard tends to win. Now, as AI infrastructure is really starting to emerge as the next big technology wave, now is the time for an open standard to emerge. We had already built k0rdent as an open source-first platform. IREN and ourselves were aligned to this idea that it is good for the industry overall to deliver an open standard platform — it kind of floats all boats. The more you remove friction and accelerate things with open standards, the more people will consume. This deal helps us accelerate the cause.”
Independent Governance: What Standalone Subsidiary Actually Means
For enterprise customers evaluating the risk of a supplier acquisition, the governance model of the post-close entity matters more than the deal announcement itself. Mirantis’s structure as a standalone subsidiary is designed to provide concrete reassurance — not just a press release commitment.
Q: Mirantis is expected to operate as a standalone subsidiary. What does that independence really look like in practice, and how does it protect the existing customer base?
Dominic Wilde: “We will have our own governance model, our own governance board, which will include people from both Mirantis and IREN, but also independent board members as well. There will be independent governance, and I think that can give confidence to other customers in the industry — including our existing customers — that we will continue to deliver on our roadmaps, our promises, all of the good things that we’ve been doing. We will continue to do that. But we will also build forward and accelerate for the industry, to remove friction and accelerate the adoption of AI infrastructure.”
Kubernetes at AI Scale: What Still Needs to Be Solved
Kubernetes has effectively won the cloud-native orchestration battle. The question for AI infrastructure is not whether Kubernetes is the right foundation — it is — but how the community evolves it to meet the specific demands of GPU workloads, hard multi-tenancy, and geopolitically-driven data sovereignty requirements.
Q: Where does upstream Kubernetes still fall short specifically around GPU scheduling, multi-tenancy, and lifecycle management? What is Mirantis actually solving there today?
Dominic Wilde: “I would never characterize it as Kubernetes is failing. This industry is moving fast. In many ways, Kubernetes has already won the cloud battle, and is now evolving to match the needs of the AI cloud and AI services. Like any good community-driven technology, the evolution path is part of the attraction — we’re not ripping this up and starting again, we’re evolving the technology forward, and that is a community effort. Specifically around technology challenges: utilization, virtualization of GPUs — particularly with respect to hard multi-tenancy requirements that get driven by sovereign issues. Those are emerging, and in some cases are geopolitically driven. In other cases there is just the need for hard segregation of data for privacy reasons. Those are challenges that with GPU architectures have to be solved and optimized. Most of what we are doing is optimization. The industry has solved problems like these before, and now it’s taking prior art and asking how we move forward and optimize for the business needs that are out there. Inference is a whole new world — this is the big coming wave and revolution. There are a lot of challenges around how you deliver and broker inference sessions. Some of these things are using technology to solve business problems, and some are optimizing technology for better performance.”
k0rdent AI Architecture: Composability by Design
The k0rdent AI platform was designed from its inception around a principle of composability — delivering opinionated reference architectures that get teams started quickly, while preserving the freedom to swap components without being locked into a proprietary dependency chain. This architectural philosophy predates and survives the IREN acquisition.
Q: IREN brings the raw GPU horsepower. Mirantis brings the orchestration brain. How tightly will these two layers get integrated, and will operators still have the freedom to swap in their own schedulers, runtimes, or networking components?
Dominic Wilde: “The industry needs a full suite and a full stack of capability, and all of the layers of that stack are related to and dependent on each other. One of the tenets we started with from the very beginning with k0rdent — and the fact that k0rdent is open source — was the idea of composability. We will come with an opinion. Opinionated stacks are good. They get people started. Reference architectures are good. But you shouldn’t have an agenda. It shouldn’t force people to stay and be tightly coupled into an opinionated approach. If somebody has different elements they want to swap out or they want to use alternative things, the option is there. We’ve designed and architected that from the ground up. As far as the approach moving forward with the broader industry, we are building for the broader industry, and we are driving k0rdent with the intention that there is an opportunity to deliver an open standard for the industry. We believe k0rdent AI has tremendous opportunity and potential to be that standard. Now with the backing of IREN — they are bringing together infrastructure at scale and proven delivery capability — we’re adding to that software operational expertise, and we will strengthen the ability for customers to deploy this infrastructure. Together, we can drive AI infrastructure online a lot faster, but we’ll continue to support our existing customers as we do that. We’ll continue to advance k0rdent as the AI platform of choice — the AI open standard, really.”
Day-Two Operations: Where Enterprise AI Actually Lives or Dies
Provisioning AI infrastructure is the solved part of the problem. Keeping it running reliably — with full observability, cost accountability, and fast recovery — at the scale and diversity of enterprise AI workloads is where the operational debt accumulates. The IREN acquisition gives Mirantis the resources to move faster on exactly these day-two challenges.
Q: Provisioning infrastructure is the easy part. Keeping it running reliably at AI scale is where teams lose sleep. What should platform engineers realistically expect to improve in observability, failure recovery, and upgrade management as a direct result of this acquisition?
Dominic Wilde: “It’s about having the backing to accelerate and move forward faster. Day-two operations are key here. Things like observability, FinOps — being able to build correctly the observability, the recovery factors, all of those usual day-two things that in the Kubernetes community we take for granted — all of those have to be brought in and optimized for AI infrastructure, whether that’s distributed inference or inference jobs at the edge, large training models, all of the different kinds of workloads. We have to optimize for all of those. This is the journey we’re on. Our intent is to move a lot faster, remove any friction, remove any uncertainty for the industry. That’s the path we’re on.”
NVIDIA AI Cloud Ready Initiative: The Three-Layer Ecosystem
Mirantis’s status as one of only three founding ISV partners of the NVIDIA AI Cloud Ready Initiative places it at the center of a deliberately structured three-layer ecosystem: NVIDIA providing and open-sourcing foundational components, ISV vendors integrating those components into enterprise-grade platforms, and cloud providers delivering vertically integrated end-to-end services. The IREN acquisition completes this stack for Mirantis.
Q: Mirantis is a founding ISV partner in the NVIDIA AI Cloud Ready Initiative. How does IREN’s scale change what Mirantis can offer within that ecosystem?
Dominic Wilde: “When NVIDIA announced the AI Cloud Ready Initiative, they did it because they are very ecosystem-driven — they want to advance the industry as a whole. They recognized they had a lot of pieces of technology they had built, and they started to open source a lot of it. The bare metal NCX controller was a big announcement at GTC, and we’ve integrated that into k0rdent AI. As an ISV vendor, we have the expertise to take those pieces, bring them into a usable format, align to that the expertise and system-level capabilities. Then cloud service companies like IREN come in and deliver vertically integrated cloud platforms — adding their delivery capability, their scale. There’s always been a natural affinity and combination of these three constituents. What NVIDIA did was formalize that into an initiative and a program. We were very proud to be one of the inaugural ISV partners — one of only three. The proof point is that this model works. You bring these three pieces together and you can drive faster and solve a lot of problems.”
Open Source and Sovereignty: Nothing Changes
The acquisition triggers legitimate questions from the open source community and from customers operating in regulated, sovereignty-sensitive environments — particularly in Europe. Dominic Wilde’s answer is unambiguous.
Q: What does the acquisition mean for Mirantis’s involvement in open source and in sovereign cloud strategies — particularly in Europe?
Dominic Wilde: “Nothing changes. It’s a very simple answer. Nothing changes. We’ve reached out to our customer base and let them know that everything they have relied on Mirantis for remains in place. All of the promises, the roadmap, the services, the contracts they have, the staff that they work with on a day-to-day basis — all of it remains in place and goes forward. No change at all, as is also true of our commitment to the open source community. Our CTO, Sean O’Mara, has written a blog on our website specifically to the open source community to explain our position. I encourage people to go read that. We are absolutely committed to our existing customer base, our emerging customer base, and to the community. Nothing changes. We will be governed as an independent entity and we will move the industry forward. We will continue to make those contributions to the community.”
The IBM-Red Hat Parallel: A Framework for Understanding the Relationship
The acquisition’s closest structural analog in enterprise tech is IBM’s acquisition of Red Hat — a deal that preserved Red Hat’s open source identity, market neutrality, and brand while giving it access to IBM’s distribution scale. The parallel is imperfect but instructive for understanding how IREN intends to relate to Mirantis post-close.
Q: If we draw a parallel between Red Hat and IBM versus IREN and Mirantis — what kind of chemistry is going to be there? Will Mirantis continue to operate as a peer to IREN in terms of customer and partner relationships?
Dominic Wilde: “There are regulatory things we have to go through, but from our perspective, we are driving forward as we were. We will continue to deliver as we have before. We don’t see any major changes from what has gone before or even our approach. What we have now is a partnership and a capability that wasn’t there before — one that enables us to move faster and accelerate.”
