If you want to apply AI to sensitive data, there's a constraint you can't negotiate around: the data, the models, and the infrastructure they run on all have to stay inside your private data center and under your control. Which means running the whole thing yourself, and that's a much bigger job than calling an endpoint.
You need a model and something to serve it with, GPUs divided up sensibly, an answer for what happens when a fine-tuning job wants capacity that inference is already using, and retrieval that reaches your private data sources without making them reachable from anywhere else. None of it is machine learning work, and it all has to be right before the model earns its keep.
Today we're announcing two products for that: Talos Linux for AI and Talos Cupar. Between them, they bring what Talos Linux already does for Kubernetes and VMs to AI workloads.
What Talos Linux for AI does
Talos Linux for AI is an open-source, immutable operating system for AI workloads. It supports NVIDIA and AMD GPUs, model serving and inference, supervised fine-tuning, and RAG connectivity to private data sources.
The design principles remain the same. Like the rest of Talos Linux, it's immutable, minimal, and managed entirely through declarative configuration, so your AI infrastructure cannot drift from its intended state. AI workloads bring more moving parts, so being able to rebuild a machine from its config rather than repairing it matters more, not less.
One console for AI on private infrastructure
Talos Cupar is a commercial platform for deploying and managing AI models, GPU capacity, private documents, and data on infrastructure you own. Cupar is to AI what Talos Director is to the data center.
From one interface, teams select models and inference engines, allocate GPU capacity through sharding and slicing, configure fine-tuning including LoRA, set up RAG pipelines and MCP connectors, and deliver chat and agentic AI applications to their users.
GPU allocation is the item on that list I'd pay attention to. Allocating by slice rather than by whole card decides how much of the hardware you've already paid for you can actually put to work.
Yardi is already running on Cupar. What Rob Teel, CEO of Yardi, says about it:
"Virtuoso Enterprise, Yardi's AI platform, is built on Talos Cupar and is used today by hundreds of Yardi clients. Yardi provides real estate software for managing 15 billion square feet of commercial space and 15 million apartment units, and we are very pleased with the Talos Cupar architecture and functionality."
Talos Cupar is in Limited Availability, which means full production support and direct engagement with our engineering team during onboarding. General availability is planned for March 2027.
FAQ
What is Talos Linux for AI?
Talos Linux for AI is an open-source, immutable operating system for AI workloads from Sidero Labs. It supports NVIDIA and AMD GPUs, model serving and inference, supervised fine-tuning, and RAG connectivity to private data sources. Like all of Talos Linux, it is immutable, minimal, and managed entirely through declarative configuration, meaning AI infrastructure cannot drift from its intended state.
What is Talos Cupar?
Talos Cupar is a commercial platform from Sidero Labs for deploying and managing AI models, GPU capacity, private documents, and data on infrastructure an organization owns. From one interface, teams select models and inference engines, allocate GPU capacity through sharding and slicing, configure fine-tuning including LoRA, RAG pipelines and MCP connectors and deliver chat and agentic AI applications to their users. Cupar is to AI what Talos Director is to the data center.
What is the difference between Talos Linux for AI and Talos Cupar?
Talos Linux for AI is the open-source operating system that AI workloads run on. Talos Cupar is the commercial management layer above it, providing one interface for models, GPU allocation, fine-tuning, retrieval pipelines, and the chat and agentic applications built on top.
How do Talos Linux for AI and Talos Cupar support sovereign AI?
Both are built so the data, the models, and the infrastructure they run on stay inside an organization's own data center and under its control. Talos Linux for AI is the open-source, immutable operating system the AI workloads run on. Talos Cupar is the commercial platform above it for deploying and managing AI models, GPU capacity, private documents, and data on infrastructure the organization owns.
Is Talos Linux for AI open source?
Yes. Talos Linux for AI is an open-source, immutable operating system for AI workloads, and part of Talos Linux, the open-source operating system for Kubernetes, virtual machines, and AI workloads.
If you're working through a sovereign AI requirement and want to know whether this covers it, email sales@siderolabs.com.
