SUSE today revealed it has extended an alliance with Switch to enable organizations to build digital twins using an artificial intelligence (AI) platform that integrates software infrastructure from both SUSE and NVIDIA.
Announced at the SUSECON 2026 conference, the expansion of the alliance allows Switch to use the AI Factory platform to run NVIDIA Omniverse libraries and blueprints to deploy digital twins in data centers operated by Switch.
The SUSE AI platform is based on SUSE Rancher Prime container management and orchestration software and SUSE Linux Enterprise Server that has been optimized for graphical processor units (GPUs). The Omniverse libraries make it possible to simulate physical environments using NVIDIA DGX systems that can process AI models and machine learning algorithms alongside each other.
That approach also makes it possible to ensure the NVIDIA DGX systems can be deployed in air-gapped IT environments. That capability is critical because in many cases organizations are being deployed to create a simulation of a business that organizations cannot afford to lose control over, notes Frank Feldmann, chief strategy officer for SUSE. “Organizations will need to be able to control it,” he says.
In fact, Switch, in addition to making these platforms available to its customers, is also using the integrated platform to run its own internal AI models.
While digital twins as a concept have been around now for several years, recent AI advances are making it more feasible to build and deploy them.
In general, digital twins and AI are converging to create environments that accelerate the training of robotic systems. That approach provides access to a simulation that, in addition to accelerating training, also reduces the total cost of building a robot.
It’s not clear at what rate AI models embedded into robots will become pervasively deployed, but every large enterprise is by now at least experimenting with one or more use cases, spanning everything from stacking shelves in warehouses to performing complex tasks for various branches of the military. The challenge, as always, is infusing the ability to consistently correlate space and distance to prevent a robot from miscalculating, for example, the amount of physical force required to perform a task.
Hopefully, as the cost of building and deploying robots continues to decline the number of use cases where they can be practically applied will continue to increase. In 2025, global shipments of humanoid robots jumped to roughly 18,000 units, more than five times the previous year. China accounted for over 85% of those deployments while the U.S. accounted for about 13%.
One Shanghai startup, AGI Bot, shipped more than 5,000 units by itself — nearly 40% of the global market. Another Chinese company, Unitree, expects to ship between 10,000 and 20,000 humanoids in 2026. Their starting price is about $13,500.
