Nvidia opens up Israeli-founded Run:ai software two years after $800 million acquisition
The chip giant is making the AI workload management platform open source, giving developers and cloud providers access to tools designed to improve the use of costly GPUs.
Nvidia is making the full software platform of Israeli-founded Run:ai available as open source, giving developers, cloud providers and other technology companies access to tools designed to make more efficient use of the expensive graphics processing units powering the artificial intelligence boom.
The move opens up technology Nvidia acquired in late 2024 for an estimated $800 million, following an eight-month regulatory review in the United States and Europe. Run:ai develops software that helps organizations allocate computing resources across AI workloads, from training models to running them in production. As demand for AI computing has surged, the ability to extract more work from existing hardware has become an increasingly important part of the industry’s infrastructure.
The newly open-sourced platform allows organizations to examine and adapt the code, use the complete system or adopt individual components. Nvidia is also making the software available through its developer infrastructure, while a number of cloud and infrastructure providers are integrating it into their own offerings.
The announcement follows a commitment made when Nvidia acquired Run:ai. At the time, the companies’ founders, Omri Geller and Dr. Ronen Dar, said they planned to open-source the software to help developers build AI systems faster. They argued that making the platform available as open source could extend its reach beyond Nvidia’s own GPUs and into the wider AI ecosystem.
Founded in 2018, Run:ai developed a software layer that manages how AI workloads use GPUs, the processors that have become central to training and operating modern AI models. Rather than allowing different teams and applications to compete for computing resources without coordination, the platform can allocate capacity dynamically, including sharing GPU resources among workloads that do not require an entire chip.
The open-source release includes the complete Run:ai platform as well as four components that can be adopted separately: KAI Scheduler, GPU Fractioning Operator, KAI Resource Management and Workload Map. Together, they cover different parts of the process, from deciding which workloads run on which resources to dividing GPU capacity and allocating it among teams.
The release does not mean that every use of Run:ai is now free of charge or includes enterprise support. The community product does not require a paid Run subscription or Nvidia support entitlement, but enterprise support is not included, and the software remains subject to its applicable open-source licenses and other terms. Nvidia will continue to offer commercial versions through its enterprise AI products and supported cloud offerings.
