Mirantis Backs 'Neoclouds' for AI
Cloud-native infrastructure company Mirantis announced it is actively partnering with "neoclouds," a new class of AI infrastructure providers. The collaboration aims to build and operate the specialized "AI factories" needed to power the next wave of artificial intelligence development.
The move by Mirantis is a strategic pivot led by co-founder Alex Freedland, who returned as CEO in January 2024 to spearhead the company's focus on AI infrastructure. An open-source pioneer, Freedland was previously named one of the most important people in cloud computing by Business Insider. "Neoclouds" are a new generation of cloud providers focused specifically on AI-as-a-service, offering GPU-centric architecture optimized for AI workloads. Unlike traditional hyperscalers such as AWS or Google Cloud, neoclouds provide bare-metal access to the latest GPUs and are designed to handle the high-performance computing required for training and inference. This specialization allows neoclouds to offer more cost-effective and powerful solutions for AI development compared to the broader, premium-priced services of hyperscalers. Their infrastructure is purpose-built to minimize I/O bottlenecks and accelerate training for the largest AI models, often leveraging high-bandwidth interconnects like NVIDIA's NVLink and InfiniBand fabrics. The collaboration centers on Mirantis' k0rdent AI, a Kubernetes-native platform for engineering and operating the entire AI stack, from hardware to models. Recently, Mirantis partnered with VAST Data, integrating its AI Operating System with k0rdent to reduce data bottlenecks that can leave expensive GPU capacity underutilized. This helps neoclouds maximize GPU utilization and accelerate the delivery of high-performance inference and AI services. The market for AI infrastructure is experiencing explosive growth, with some analysts forecasting spending to surpass $200 billion by 2028. This demand is driven by the compute-intensive nature of AI, from training massive models to running inference at scale. Mirantis is also collaborating with hardware providers like Supermicro to create verified, Kubernetes-native stacks for deploying and operating GPU-accelerated infrastructure. This approach aims to provide neoclouds and enterprises building their own AI factories with standardized, repeatable models for building out their infrastructure, reducing the need for costly and time-consuming custom integration work.
