Build your AI factory with open source
The fastest, most trusted path to value for AI infrastructure
Enterprises are racing to embrace AI while balancing infrastructure costs. If you’re procuring new hardware, you want to make the most of your investment. Canonical helps you get value from your AI hardware with state-of-the-art silicon optimization for Ubuntu. With our portfolio, you get everything you need to take AI into production, operate it securely at scale, and meet your compliance requirements.
Ubuntu turns AI hardware into AI value
The largest AI companies in the world trust Ubuntu to run their infrastructure. Here’s why.
- Ubuntu unlocks the capabilities of the latest GPUs, accelerators, networking systems and servers
- Fastest time to value with optimized kernels
- Latest features straight from the largest silicon vendors
- Built in AI libraries like CUDA, ROCm, DOCA-OFED (simplified installation saves you days or hours in GPU set up)
- Long-term security maintenance and support to ensure you get the “bang for your buck” and reliability
Optimized for leading silicons
Tested and certified with the world’s largest hardware vendors
Everything you need to get agentic AI into production, at scale
Enterprise model serving infrastructure
Serve frontier models with maximum performance & efficiency and the latest open source innovations. Enrich your models’ knowledge without retraining.
Leverage disaggregated serving for stable token throughput under load with KServe, LLM-d and vLLM, and implement LLM governance with MLFlow.
Stand up a production LLMOps platform quickly
Deploy the software you need to train & finetune models at scale on hybrid cloud infrastructure of your choice.
Scale with industry standard Slurm or collaborate with Canonical’s Kubeflow, MLFlow and Feast stack on Kubernetes.
From the public cloud to behind the firewall
Work with the right AI infrastructure for your use cases. For instance, prototype on public clouds, then migrate your workload to your private cloud when you’re ready for production.
With Canonical’s solutions, you can run your workloads anywhere, including self-governed, hybrid, and multi-cloud environments.
Get the most out of AI with Kubernetes for service orchestration
Canonical Kubernetes is optimized to enhance GenAI, incorporating features developed and integrated in tight collaboration with NVIDIA.
Download the solution brief: Kubernetes by Canonical delivered on NVIDIA DGX systems ›
Full-stack AI infrastructure
Applications
Get everything you need to bring your models from experimentation to production. Build, train, and track models with Charmed Kubeflow. Serve LLMs with vLLM, and scale GPU accelerated model serving with LLM-d.
There’s no AI without data. Canonical also delivers a comprehensive portfolio of integrated open source solutions for database management, data processing, search, and caching.
Application lifecycle management
Our software operators enable you to use a single codebase to automate the lifecycle management of your data applications across public clouds, Kubernetes, virtual (VM) and bare metal machines.
This is possible thanks to Juju and charms. Juju is an open source orchestration engine for software operators that enables the deployment, integration, and lifecycle management of applications at any scale, on any infrastructure. A charm is an operator – business logic encapsulated in reusable software packages that automate every aspect of an application's life.
Infrastructure
Canonical’s stack provides a comprehensive infrastructure foundation for your AI workloads, from the operating system to virtualization, and even hardware automation.
Ubuntu is the operating system of choice for AI. It gives you access to the broadest ecosystem of machine learning tools and libraries, and enables you to diversify your silicon supply chain for resiliency with an OS optimized for all leading architectures. And with Landscape, you can automate security patching, auditing, access management, and compliance tasks across your Ubuntu estate.
Using MAAS, you can automate the provisioning and management of your data center hardware, including the GPUs, SmartNICs, and DPUs you need to drive your AI workloads.
Certified hardware
With Ubuntu Certified hardware, you don’t need to spend months testing whether a new piece of silicon will work in your fleet. You know that the hardware is validated for long-term support and reliable AI operations. Rapidly deploy inference fleets and training clusters without having to perform custom OS engineering for every new piece of hardware.
AI infrastructure in the real world, powered by Canonical
Lowering the cost and complexity of critical space missions
“We wanted one partner for the whole on-premise cloud because we’re not just supporting Kubernetes but also our Ceph clusters, managed Postgres, Kafka, and AI tools such as Kubeflow and Spark. These were all the services that were needed and with this we could have one nice, easy joined-up approach.”
Michael Hawkshaw, IT Service Manager, European Space Agency
Fortune 500 fintech company goes from constant troubleshooting to scalable machine learning
“We like the whole integration umbrella that Canonical offers with Canonical Kubeflow. We have easy access to popular ML frameworks like TensorFlow, PyTorch, and XGBoost, as well as the native Kubernetes tools for monitoring and logging. It helps our data scientists quickly build and experiment with their models in an efficient manner.”
Senior data scientist, Fortune 500 financial services company
Building a sustainable cloud for AI workloads
“The level of engagement from the Canonical team was remarkable. Even before we entered into a commercial agreement, Canonical offered valuable advice around OEMs and hardware choices. They were dedicated to our project from the get-go.”
Tim Rosenfield, CEO and Co-Founder, Firmus
Machine learning drives down operational costs
“Partnering with Canonical lets us concentrate on our core business. Our data scientists can focus on data manipulation and model training rather than managing infrastructure.”
Machine learning engineer, entertainment technology company
Cost-effective, security-maintained, and supported
Enterprise support for our open source AI infrastructure stack is available with a transparent subscription: Ubuntu Pro. Pay for the support you need with simple pricing per node.
Fast-track compliance and run AI projects securely. Let us handle the CVEs.