The News
Mirantis announced it has joined the NVIDIA AI Cloud Ready Initiative as a founding ISV partner, bringing its k0rdent AI platform to NVIDIA Cloud Partners (NCPs). The platform delivers a validated “metal-to-model” stack designed to help neocloud providers and enterprises transform GPU infrastructure into scalable, multi-tenant AI cloud services.
Built on NVIDIA’s reference architectures and leveraging the NCX Infra Controller, k0rdent AI provides a unified control plane to orchestrate GPU infrastructure across bare metal, Kubernetes, virtual machines, and managed AI services. The goal is to move providers from manual operations and single-tenant deployments toward automated, multi-tenant AI cloud platforms with improved utilization and economics.
Analysis
AI Infrastructure Is Shifting From Capacity to Economics
Mirantis is entering a market that is rapidly moving beyond the initial phase of AI infrastructure buildout. While demand for GPUs and AI compute remains high, the challenge is no longer just deploying capacity; it is turning that capacity into profitable, repeatable services.
This is particularly relevant for neocloud providers, which are emerging as an alternative to hyperscalers by offering specialized AI infrastructure. However, many of these providers face operational challenges, including:
- Low GPU utilization due to single-tenant deployments
- Manual provisioning and fragmented tooling
- Difficulty monetizing infrastructure consistently
Mirantis’ positioning of k0rdent AI aims to target these issues. By enabling multi-tenancy, automation, and self-service consumption, the platform could shift AI infrastructure from a capital-intensive asset into a revenue-generating cloud service model.
This aligns with broader AppDev research trends. Internal data shows 61.8% of organizations operate in hybrid environments, and 60.7% prioritize cloud infrastructure investments, indicating strong demand for flexible, platform-based consumption models rather than fixed infrastructure deployments. Neoclouds are emerging to meet that demand, but they require operational platforms like k0rdent AI to scale effectively.
The Control Plane Becomes the Differentiation Layer
A key theme in this announcement is the importance of a unified control plane. Mirantis is not competing on hardware or raw GPU access; instead, it is focusing on the orchestration layer that sits above the infrastructure.
By integrating across NVIDIA’s full stack, from GPUs and networking (InfiniBand, Spectrum-X) to DGX/HGX systems and Kubernetes environments, k0rdent AI positions itself as the operational brain of the AI cloud.
This reflects a broader shift across AI infrastructure markets. As hardware becomes more standardized, differentiation is moving toward:
- Workload orchestration and scheduling
- Multi-tenant isolation and resource allocation
- Automation of provisioning and lifecycle management
- Integration with developer and MLOps workflows
For developers and platform teams, this is significant. The control plane determines how easily AI workloads can be deployed, scaled, and managed. In many cases, it becomes the primary interface between developers and infrastructure.
Market Challenges and Insights
The announcement highlights a fundamental challenge in the AI infrastructure market: utilization and efficiency. Despite strong demand, many GPU deployments are underutilized. This is due to a number of factors including over-provisioning for peak workloads, lack of workload scheduling optimization, siloed environments across teams or tenants, and limited automation in provisioning and scaling.
This creates a paradox similar to other parts of the AI ecosystem: high investment does not always translate into proportional returns. Industry projections of $2.5 trillion in AI spending by 2026 underscores the scale of investment, but also raises questions about how efficiently that capital is being used.
Our AppDev research reinforces this tension. While organizations are investing heavily in AI, only 53.4% report high confidence in scalability for peak loads, and 55.0% report full preparedness for resilience and failure recovery. This suggests that operational maturity is still catching up to infrastructure investment.
Mirantis’ focus on automation and multi-tenancy is a direct response to this gap. By improving utilization and reducing operational overhead, platforms like k0rdent AI aim to help organizations extract more value from existing infrastructure.
From Infrastructure to AI Cloud Platforms
Another important shift reflected in this announcement is the move from infrastructure to AI cloud platforms.
Mirantis’ “metal-to-model” framing emphasizes end-to-end lifecycle management, from bare metal provisioning to delivering AI services. This mirrors broader industry trends where infrastructure providers are evolving into platform providers that support:
- Model training and inference workloads
- Developer and data science workflows
- API-driven service delivery
- Multi-tenant consumption models
For neoclouds, this transition is critical. Competing with hyperscalers requires more than offering GPUs; it requires delivering a developer-friendly, operationally efficient platform that can support diverse workloads and use cases.
This also aligns with the rise of platform engineering in the AppDev landscape, where internal and external platforms abstract infrastructure complexity and provide standardized interfaces for developers.
Why This Matters for Developers and Platform Teams
For developers, the implications are clear: AI infrastructure is becoming more abstracted, automated, and consumable. Instead of managing infrastructure directly, developers will increasingly interact with self-service platforms for deploying AI workloads, APIs and orchestration layers that handle resource allocation, and integrated environments that combine infrastructure, data, and models.
For platform teams, the challenge is to build and operate these environments in a way that balances performance, cost efficiency, and governance.
Mirantis’ approach highlights several key priorities:
- Maximizing GPU utilization through intelligent scheduling
- Enabling multi-tenant environments without sacrificing isolation
- Automating lifecycle management to reduce operational overhead
- Providing a consistent control plane across hybrid environments
These priorities are becoming central to AI-native platform engineering.
Looking Ahead
The AI infrastructure market is entering a phase where economic efficiency and operational scalability will matter as much as raw compute capacity.
Mirantis’ participation in the NVIDIA AI Cloud Ready Initiative signals a growing recognition that the next wave of AI adoption will depend on platforms that can make infrastructure easier to deploy, more efficient to operate, and more profitable to run.
As neoclouds and enterprise AI factories continue to expand, the control plane (and the ability to orchestrate complex, multi-tenant environments) will likely become the primary battleground for differentiation.
For the broader market, the takeaway is clear: the future of AI infrastructure is not just about building clusters. It is about turning those clusters into scalable, monetizable platforms that developers can actually use.
How AI Teams Reclaim Time, Velocity, and Budget with Union.ai
SecOps Strain Grows as AI Adoption Outpaces Trust and Integration
Cloud Native Hits Critical Mass as Platform Engineering Reshapes Dev Workflows
MinIO Aligns With NVIDIA STX to Power AI Data Factories
AMD Scales AI Infrastructure With Open Ecosystem Strategy
