Hugo Huang
on 28 May 2026
This new release brings the stability, security, and expansive ecosystem of Ubuntu to Cloud TPU virtual machines on Google Cloud.
May 28, 2026 – Today, Canonical and Google Cloud announced the availability of certified Ubuntu images for Google’s Cloud TPU Virtual Machines, which are included by default when setting up the VMs.
By combining Google’s custom AI accelerators with the reliability and familiarity of Ubuntu, we are providing a streamlined path for developers and enterprises to adopt the newest AI infrastructure on Google Cloud.
As AI is entering the Age of Inference, the focus has shifted from training massive models to powering real-time interactions. Agentic workflows require seamless coordination between general-purpose compute and machine learning acceleration, which is why Google Cloud and Canonical are working together to make it easier for organizations to adopt advanced AI infrastructure.
With this launch, Cloud TPU VMs offer a simplified user experience, with Ubuntu preinstalled, that mirrors launching any other standard Google Cloud Compute Engine instance. Developers benefit from the stable foundation and broad ecosystem offered by Ubuntu, paired with direct access to Google’s seventh-generation custom silicon, Ironwood (TPU 7x). To ensure continuity for existing production environments, Cloud TPU v5 and Trillium (v6) instances are paired with Ubuntu 22.04 LTS, allowing teams to maintain their current workflows without disruption.
A broad ecosystem and operational efficiency from Day One
Different AI workloads demand different infrastructure configurations, but they all require a reliable software foundation. Canonical has collaborated closely with Google Cloud to ensure that the Ubuntu-based TPU VM environment provides AI innovators with out-of-the-box readiness and comprehensive support for the machine learning lifecycle. This deep integration includes validation and optimization of:
- Open-source AI frameworks: Seamless access and optimized performance for industry-standard frameworks like JAX, PyTorch, and TensorFlow, alongside tools like Ray for scaling ML workloads.
- Operational consistency: Platform and MLOps teams can use familiar tooling for monitoring, automation, and configuration management (from Kubernetes to Snap packages), backed by up to five years of standard security maintenance through Ubuntu LTS.
- Sustainability and efficiency: Ubuntu’s lightweight footprint minimizes system overhead, directing compute resources to the models themselves. Combined with the performance-per-watt efficiency of Cloud TPUs and Google’s commitment to 24/7 carbon-free energy, these workloads represent one of the most sustainable ways to run large-scale AI.
- Enterprise-grade AI security (coming Q3 2026): Customers will soon have the choice to use Ubuntu Pro for Cloud TPUs, introducing advanced capabilities for mission-critical environments. This will include live kernel patching for security updates on long-running training jobs and Expanded Security Maintenance (ESM) for over 30,000 open-source packages for up to 15 years with the Legacy add-on for Ubuntu Pro.
This streamlined foundation ensures that organizations can move from experimentation to production faster than ever before.
Getting started
To get started, simply select your desired Cloud TPU machine type—including TPU 7x, v6e, v5p, or v5e—when creating a VM in Compute Engine, and your optimized Ubuntu OS will be automatically provisioned and ready to use out of the box.
gcloud compute instances create –image-project=ubuntu-os-accelerator-images –machine-type=[SELECT TPU Machine TYPE] –zone=[SELECT AVAILABLE ZONE] [INSTANCE NAME]
About Canonical
Canonical, the publisher of Ubuntu, provides open source security, support and services. Our portfolio covers critical systems, from the smallest devices to the largest clouds, from the kernel to containers, from databases to AI. With customers that include top tech brands, emerging startups, governments and home users, Canonical delivers trusted open source for everyone. Learn more at https://canonical.com/
