Sidero ships Talos Linux for AI and Talos Cupar for private AI stacks

Sidero Labs said Thursday it introduced Talos Linux for AI, an open-source immutable OS for AI workloads, and Talos Cupar, a commercial platform to deploy and manage models, GPUs, and private data on customer-owned infrastructure.

Talk of keeping AI under your own control is cheap until you control the operating system under the graphics chips. Sidero is packaging an open-source Linux that does not drift from the setup you declared, plus a console, so a company can keep models, documents, and those chips in its own data center instead of renting someone else’s.

On Thursday, 1 October 2026, Sidero Labs said it introduced Talos Linux for AI and Talos Cupar. The release is datelined Santa Barbara, California. Sidero says the two products are for organizations that want to use AI on sensitive data while keeping the data, the models, and the machines inside a private data center they control.

What each product is, as the release states it. Talos Linux for AI is an open-source operating system for AI workloads, and it is immutable. Immutable means the system is not edited in place. It stays on the setup you declared, so a machine cannot quietly drift away from that setup. Talos Cupar is the commercial platform for deploying and managing models, graphics-chip capacity, and private documents on infrastructure an organization owns. A graphics chip, often called a GPU, is the accelerator that runs the model. The release says the pair extends the security, manageability, and reliability of Talos Linux to AI.

What the operating system includes, as the release states it. Talos Linux for AI allows graphics-chip support for NVIDIA and AMD, model serving and inference, supervised fine-tuning, and a connection from retrieval-augmented generation to private data. Inference is running a trained model so it answers. Serving is keeping that model up so people and apps can call it. Supervised fine-tuning is further training on labeled examples from your own work. Retrieval-augmented generation, or RAG, pulls passages from private documents into an answer, instead of relying only on what the model already stored. The release says that, like the rest of Talos Linux, this system is immutable, minimal, and managed entirely through declarative configuration. Declarative means you state the setup you want and the system applies it, rather than logging in and changing files by hand. The release’s line is that AI infrastructure cannot drift from its intended state.

What the console is for. Talos Cupar is the management layer for AI on private infrastructure. From one screen, the release says, a team selects models and inference engines, allocates graphics-chip capacity through sharding and slicing, configures fine-tuning including LoRA, sets up RAG pipelines and MCP connectors, and delivers chat and agent applications to its users. Sharding and slicing, here, mean dividing one chip’s capacity so more than one job can use it. LoRA is a lighter way to fine-tune: it trains a small add-on instead of rewriting the whole model. MCP is the Model Context Protocol, a connector an assistant uses to reach tools and data. An agent application, in this release, is software that can take a next step on a task, not only reply in chat. The release’s comparison is that Cupar is to AI what Talos Director is to the data center.

Steve Francis, president of Sidero Labs, is quoted on the release. “If you don’t control the operating system and servers your models run on, you don’t control your data,” he said. He said Talos Linux for AI gives organizations an open-source, immutable foundation for running AI on their own infrastructure, and that Cupar makes operating that stack as simple as Talos Director makes operating the data center.

Who already runs it, as the release states it. Yardi, which acquired Sidero Labs in July 2026, already runs its AI platform on Talos Cupar. Rob Teel, chief executive of Yardi, said Virtuoso Enterprise, Yardi’s AI platform, is built on Talos Cupar and is used today by hundreds of Yardi clients. He said Yardi provides real estate software for managing 15 billion square feet of commercial space and 15 million apartment units, and that Yardi is very pleased with the Talos Cupar architecture and functionality. The client count and those two property figures are his, in the release.

When a customer can get the console. Talos Cupar is in Limited Availability now. On the release, Limited Availability customers receive full production support and direct contact with Sidero Labs engineering during onboarding. General availability, the broader release, is planned for March 2027.

Who the about box says the companies are. Sidero Labs, a Yardi company, builds Talos Linux, which the release calls the secure, immutable, open-source operating system for Kubernetes, virtual machines, and AI workloads, plus the commercial platforms that manage it at scale. Those platforms are Omni, for Kubernetes fleet management; Talos Director, for virtual machines, storage, and networking across the data center; and Talos Cupar, for the AI lifecycle. Kubernetes is the system many companies use to run packaged apps across a cluster of machines. The release says organizations choose Sidero for security, manageability at scale, and reliability, with enterprise support and a strong compliance posture. Yardi, the release says, develops software for real estate companies of all types and sizes, and has more than 10,000 employees.

The picture is Sidero’s official release graphic. A black field carries the Talos by Sidero mark, an orange-red burst beside the word Talos. The headline reads “Introducing Talos Linux for AI & Talos Cupar,” with a thin red line under it. The line along the bottom reads “Your models. Your data. Your infrastructure.” On the right, a stacked server chassis glows orange inside a wireframe cube, and the top layer is a grid of red lights. The graphic does not print a calendar date.

In plain terms, Sidero said on Thursday that Talos Linux for AI is an open-source operating system for running models on NVIDIA and AMD chips, and that Talos Cupar is the commercial console for choosing models, splitting chip capacity, fine-tuning, connecting private documents, and handing users a chat or an agent. Cupar is in limited availability, with a broader release planned for March 2027. Yardi, which bought Sidero in July, says its Virtuoso Enterprise platform already runs on Cupar for hundreds of its clients.