Ubuntu Server: a platform made for enterprise scale

A platform is an environment that allows software to run smoothly across the infrastructure, runtime, and application layers. The key word there is “smoothly”: a good platform connects those layers so well that you don’t notice it. That’s what Ubuntu Server has become: the essential layer between bare metal and the apps running on top, continuously optimized across resource management, networking, and security.

Ubuntu 26.04 LTS represents over 12 years of that work coming together. In this blog, we’ll highlight the features that show just how far it’s come, and how you can use Ubuntu Server in combination with other Canonical products to deliver solutions for enterprise scale.

Bare-metal automation with MAAS

Enterprise environments rarely run on uniform hardware, and making the most of what you have can be a large manual undertaking. Ubuntu Server goes beyond running standard workloads by enabling full bare metal automation through its integration with MAAS (Metal as a Service).

MAAS automates the full lifecycle of your machines, from discovery, OS deployment, and configuration, to secure decommissioning . Before a server even enters production, integrated diagnostics stress-test the hardware to flag connectivity or component issues, which is vital for maintaining large-scale fleets. You can also tag which servers in your estate are best suited to deploy AI workloads. This enables you to deploy resources efficiently, as GPU workloads become a bigger part of enterprise infrastructure planning. 

By integrating with Infrastructure as Code (IaC) tools like Terraform, MAAS enables you to manage physical hardware using the same automated workflows typically reserved for the cloud. This eliminates the manual overhead that makes bare-metal management difficult at scale, by allowing you to orchestrate a fleet of physical servers as easily as if they were virtual machines. 

As with any platform, the work is always ongoing to make the integration tighter. A particular highlight is instant AI readiness. Enhanced NPU drivers allow MAAS to automatically detect and pre-configure the kernel modules and runtimes for AI accelerators, turning raw silicon into an active AI node instantly.

Virtualization support

Software-defined infrastructure (SDI) relies on the ability to abstract hardware into flexible, digital resources. While virtual machines (VMs) have long been the standard, containers are rising in adoption. Any enterprise server platform needs to support mixed environments.

Ubuntu Server addresses this need through LXD, its native virtualization layer. Traditional approaches to virtualization often force you to choose between a container manager or a hypervisor. LXD allows you to host both system containers and VMs together on a single infrastructure. For smaller-scale private cloud deployments, LXD clustering lets you extend this unified environment across multiple nodes, giving you a lightweight, flexible option without additional orchestration overhead. For a deeper technical rundown of how this works, I recommend exploring our detailed documentation.

For private cloud infrastructure, OpenStack tends to be the dominant choice. Ubuntu is already the primary OS for around half of all OpenStack deployments, and Ubuntu Server is the target platform for KVM, the hypervisor most commonly used with OpenStack. Canonical OpenStack builds on this as an opinionated distribution running on Ubuntu Server, adding structure to what is otherwise a complicated process of deployment, patching, and ongoing maintenance. This combination provides a robust enterprise engine when you need to enforce strict isolation boundaries through virtual machines at large scales using Ubuntu Server infrastructure.

All of this matters as you can select the model that best works for you. System containers deliver near bare-metal performance for workloads where density and efficiency are the priority. VMs provide a stronger hardware-level isolation boundary for environments where that separation is required. Whichever option you select, Ubuntu Server helps you to scale it.

Ubuntu 26.04 LTS also brings updates that strengthen virtualization capabilities of Ubuntu Server, through support for confidential computing. Ubuntu 26.04 LTS adds hardware-level encryption through AMD SEV-SNP and Intel TDX. This means workloads running on Ubuntu Server can remain encrypted and inaccessible even to the host system, strengthening isolation for sensitive enterprise workloads. GPU-passthrough for AI containers: as with prior releases, optimizations in Ubuntu 26.04 LTS now allow you to share a single physical GPU across multiple system containers. This enables you to set up AI factories,  with independent workloads drawing upon the same powerful hardware simultaneously.

Native orchestration capability

Orchestration illustrates what platform cohesion actually means: the value isn’t in any single component, but in how reliably the pieces connect.

Kubernetes is the standard for container orchestration, but it doesn’t operate in a vacuum. It needs a solid, well-integrated platform beneath it. Ubuntu Server is that platform, whether you’re running on the public or private cloud. Optimized Ubuntu images bring this capability to the public cloud, while bare metal automation through MAAS allows you to do the same on-premises.

This makes orchestration simpler, as the OS experience is the same across platforms. But what about Kubernetes itself? Because the underlying platform is consistent, you have a single baseline regardless of which Kubernetes flavor you run. Canonical Kubernetes takes this further by integrating with Ubuntu Server. Unlike upstream Kubernetes, where new releases come out every 4 months, Canonical Kubernetes is supported for up to 15 years, as is the rest of the Ubuntu Server stack.  

So what’s new in terms of your own container orchestration? Well, Ubuntu 26.04 LTS brings dual-track containers. Administrators can now choose the update velocity that matches their risk profile by selecting between two distinct paths for the container runtime. Opt for the “Stable” track to prioritize production reliability and a consistent 15-year baseline, or the “App” track to access the latest upstream features for development and testing.

Integration with leading AI toolkits and libraries

If AI is to become a truly enterprise-standard tool, then setup needs to become intuitive. The days of it being acceptable for models to require manual compilation and deep hardware expertise are coming to an end. Silicon vendors work directly with Canonical to streamline how Ubuntu interacts with libraries and frameworks like NVIDIA CUDA and AMD ROCm. In fact, for Ubuntu 26.04 LTS, these packages have been integrated directly into the Ubuntu archive, making hardware enablement a simple, standard package installation.

When it comes to selecting a model, inference snaps (pre-packaged models and inference runtimes that run on Ubuntu) automatically detect the underlying hardware and deploy a  silicon-optimized inference engine to get the best performance possible on that hardware.

From layers to interconnection

Hopefully this blog has given you a sense of what Ubuntu offers as a cohesive enterprise server platform. Rather than a layered set of tools, Ubuntu functions as an interwoven, interconnected platform. For a closer look at the specific features we released in Ubuntu 26.04 LTS, Christian’s blog is the place to go.

What that release represents, though, is something broader. The individual components, such as bare-metal management, virtualization, container orchestration, AI readiness, have always been strong. What Ubuntu Server 26.04 LTS marks is the point at which they fully come together. Years of platform work have brought us to a place where the conversation shifts from capability to optimization. That’s a meaningful milestone.