The Forever AI Factory: Mirantis and NVIDIA End the Re-Platforming Cycle

CAMPBELL, CA – July 08, 2026 – In the relentless race to build out the world's AI capacity, the biggest cost isn't always the silicon itself, but the constant, disruptive cycle of re-engineering infrastructure to accommodate it. Today, Mirantis announced a move that strikes at the heart of this problem, aligning its k0rdent AI platform to offer 'day-zero' support for NVIDIA's entire accelerated computing roadmap—from the current Grace Blackwell architecture to the forthcoming Vera Rubin and beyond. This isn't merely a software update; it's a structural shift aimed at creating the 'forever AI factory,' an infrastructure that evolves seamlessly with the hardware that powers it.

For enterprises and the burgeoning class of 'neocloud' providers, this development signals a potential end to the punishing cycle of re-platforming. By integrating deeply with NVIDIA's foundational software, Mirantis is building a bridge to the future, allowing organizations to treat AI infrastructure not as a series of disposable projects, but as a single, continuous, multi-generational investment.

The End of the Upgrade Tax

NVIDIA's annual hardware cadence, while a marvel of engineering, has created a significant operational tax for its customers. Deploying a new generation of GPUs, like moving from Hopper to Blackwell, has traditionally forced organizations into complex and costly redesigns of their entire infrastructure stack, from bare-metal provisioning to networking and orchestration. This constant churn slows down innovation and inflates the total cost of ownership.

Mirantis aims to eliminate this tax. The company's promise of “continuous architecture readiness” means that an AI factory built today on its platform for NVIDIA's Grace Blackwell chips will be able to incorporate next-generation Vera Rubin hardware without a fundamental operational overhaul. The underlying workflows, management planes, and provisioning models remain consistent, allowing operators to simply integrate new hardware as it becomes available.

“AI infrastructure is now a multi-generation investment strategy with AI factories representing a fundamental shift in how infrastructure is designed and operated,” said Kevin Kamel, vice president of product management at Mirantis. “The combination of k0rdent AI and NVIDIA delivers lifecycle automation for a fully integrated, enterprise-ready platform that is future-proofed to deliver AI factories so that users can take advantage of NVIDIA’s latest hardware technologies without disruption.”

This is particularly critical for neocloud providers and large enterprises making long-horizon purchasing decisions. By abstracting away the hardware-specific complexities, the platform allows them to invest with confidence, knowing their operational foundation won't become obsolete in twelve months.

Automating the Engine Room

The magic behind this seamless evolution lies in a deep, multi-layered integration with NVIDIA's own open-source infrastructure tools. Mirantis hasn't just built on top of NVIDIA's ecosystem; it has woven it into the core of its k0rdent AI platform, transforming a collection of powerful but complex components into a validated, production-ready system.

At the foundation is the NVIDIA Infra Controller (NICo), a suite of microservices for zero-trust, bare-metal lifecycle management. NICo automates the gritty, essential tasks that are often handled manually: discovering new hardware in a rack, validating firmware, provisioning NVIDIA's powerful BlueField DPUs for network control, and securely sanitizing servers between tenants. While NICo is open source, operationalizing it at scale is a significant engineering challenge. Mirantis has done this heavy lifting, integrating NICo as the primary provisioning layer for k0rdent AI.

Layered on top are components of the NVIDIA DSX OS, a modular platform for building and operating AI factories. This provides a standardized framework for everything from runtime consistency and health automation to multi-tenant operations. By incorporating these elements, k0rdent AI ensures that the factories it deploys adhere to a proven, high-performance blueprint, drastically reducing the complexity and time required to go from racked hardware to a consumable AI service.

The Economics of Next-Generation AI

This deep integration is engineered to drive two critical metrics in the AI economy: performance per watt and cost per token. As AI models grow, the energy consumption and computational cost of running them are becoming primary business concerns. Mirantis's alignment with NVIDIA's roadmap directly targets these efficiencies.

Research indicates that NVIDIA's upcoming Vera Rubin platform is projected to deliver a staggering 10x improvement in inference throughput per watt and a corresponding 10x reduction in inference token costs compared to its predecessor. By providing day-zero readiness for Vera Rubin, Mirantis ensures its customers can immediately capitalize on these economic gains without a painful and protracted migration period.

Furthermore, the platform tackles one of the most persistent inefficiencies in AI infrastructure: low GPU utilization. Expensive, powerful GPUs often sit idle or are underutilized in single-tenant architectures. k0rdent AI addresses this head-on by enabling robust, hardware-enforced multi-tenancy and virtualization. This allows multiple users and workloads to securely share GPU resources, dramatically increasing utilization rates and, for neocloud providers, profitability. By automating the entire stack, the platform also slashes operational overhead, freeing up engineering talent to focus on higher-value AI/ML work instead of infrastructure maintenance.

A Strategic Alliance Shaping the Ecosystem

This announcement is more than a product integration; it's a testament to a deep strategic alliance that is helping shape the future of the AI datacenter. Mirantis is not merely a downstream consumer of NVIDIA's technology. The company is a primary open-source infrastructure contributor to the NICo project, actively collaborating with NVIDIA to define how AI infrastructure provisioning evolves.

This is a crucial distinction. It means Mirantis has a seat at the table, influencing the direction of foundational technologies and gaining early access to innovations. This co-development model is further solidified by the company's status as an inaugural partner in the NVIDIA AI Cloud Ready initiative, a validation program that certifies software platforms for scalable AI cloud infrastructure. This certification assures customers that the Mirantis platform is a pre-validated, pre-integrated solution ready for immediate deployment on NVIDIA-certified hardware.

This collaborative approach is becoming a defining feature of the AI landscape, as seen with competitors like Red Hat also announcing day-zero support for NVIDIA's roadmap. The sheer complexity of building and operating AI factories at scale necessitates this level of deep partnership between hardware and software players. By working to build the foundational plumbing of the AI world together, companies like Mirantis and NVIDIA are creating the stable, scalable, and economically viable systems that will power the next fifty years of progress.