According to foreign media SiliconANGLE,NVIDIAIntendedAcquisition AI Infrastructure Virtualization Startup Run:ai, the transaction amount can be up to US$1 billion (currently approximately RMB 7.2 billion).
Run:ai's workload management platform of the same name was recently certified by NVIDIA DGX SuperPOD. Its AI orchestration technology can help users easily run AI and machine learning projects to meet the growing requirements for generative AI and large models.
Run:ai was founded in 2018 by CEO Omri Geller and CTO Ronen Dar, who met while studying at Tel Aviv University’s School of Electrical Engineering. Dar was previously a researcher at Bell Labs and an algorithm engineer at Anobit Technologies, which is now acquired by Apple.
Run:ai combines traditional OS virtualization with GPU-based AI workloads.Its virtualization layer can effectively pool the computing resources of large GPU clusters and enable high-efficiency sharing across multiple AI workloads..
The company also introduced a computational abstraction layer,Leverage graph-based parallel algorithms to analyze deep learning models and decompose them into smaller parts, these small parts can then be trained in parallel, thus reducing the time required to complete the entire training process.
This computational abstraction layer also helps overcome the limitations of GPU memory capacity, allowing larger and more complex AI models to be trained on smaller GPU clusters.
If the deal is completed, it will be Nvidia's first acquisition of an Israeli company since it announced in March 2019 that it would buy Mellanox for approximately $7 billion.