NVIDIA, the GPU giant, is slowly turning into an acquisition machine. Most recently, it announced its intention to acquire Run:ai, a Kubernetes-based workload management and orchestration software provider. 

NVIDIA has entered into a definitive agreement to acquire Run:ai to help its customers make more efficient use of their AI computing resources. The Tel Aviv-based company simplifies the management and optimisation of AI hardware infrastructure for developers and operations teams. 

While the terms of the deal have not been publicly disclosed, sources close to the matter revealed that the acquisition was valued at $700 million.

“Run:ai has been collaborating closely with NVIDIA since 2020 and we share a passion for helping our customers make the most of their infrastructure,” said Omri Geller, Run:ai co-founder and CEO. “We’re thrilled to join NVIDIA and look forward to continuing our journey together.”

Run:ai was founded in 2018, and it launched its first product in 2020. The company’s founders, Omri Geller and Ronen Dar, met at Tel Aviv University while working on their master’s and PhD degrees, respectively, under the supervision of Professor Meir Feder.

Through their work, they identified a clear trend in the industry – a constant and growing demand for sufficient compute power to accelerate machine learning and deep learning, often surpassing available infrastructure. Recognising this challenge, they joined forces to find a solution, leading to the establishment of Run:ai.

Run:ai’s clientele includes some of the world’s largest enterprises across multiple industries such as Adobe, Sony, Zebra, and University of Oxford, utilising the platform to manage data-centre-scale GPU clusters. NVIDIA has stated that it will continue to offer Run:ai’s products “under the same business model” and will invest in Run:ai’s product roadmap as part of NVIDIA’s DGX Cloud AI platform. 

DGX for the win

NVIDIA’s Alexis Bjorlin said in the blog that as NVIDIA AI deployments for customers grow in complexity, with workloads spread across various infrastructures such as cloud, edge, and on-premises data centres, the need for effective management and orchestration becomes increasingly important.

The Run:ai platform provides AI developers and their teams with a centralised interface to manage shared compute infrastructure, ensuring easier and faster access for complex AI workloads. It offers functionality to add users, organise them into teams, grant access to cluster resources, control quotas, priorities, and pools, as well as monitor and generate reports on resource usage. 

Additionally, the platform enables the pooling of GPUs and sharing of computing power, from fractions of GPUs to multiple GPUs or multiple nodes of GPUs across different clusters, for separate tasks. This efficient utilisation of GPU cluster resources allows customers to maximise the return on their compute investments.

Run:ai specialises in enabling enterprise customers to efficiently manage and optimise their compute infrastructure, whether it is located on premises, in the cloud, or in hybrid environments.

The company has developed an open platform on Kubernetes, which serves as the orchestration layer for modern AI and cloud infrastructure. This platform is compatible with all popular Kubernetes variants and seamlessly integrates with third-party AI tools and frameworks.

Another AI investment for NVIDIA

This is one of the biggest acquisitions for NVIDIA after Mellanox in March 2019 for $6.9 billion. Apart from that, NVIDIA also acquired OmniML in March 2023, the company that helped NVIDIA move its ML models on edge.

Now, NVIDIA DGX server, workstation, and DGX Cloud customers will also have access to Run:ai’s capabilities for their AI workloads, particularly for generative AI deployments across multiple data centre locations. 

NVIDIA’s DGX Cloud is now hosted on Microsoft Azure, Google Vertex, and Oracle cloud infrastructure. Besides, NVIDIA is also going to self-host DGX on Blackwell systems this year. 

Although Run:ai faces limited direct competition, other companies are also exploring dynamic hardware allocation for AI workloads. One such example is Grid.ai, which provides software enabling data scientists to train AI models across GPUs, processors, and other hardware components simultaneously.

NVIDIA is vertically integrating their platforms, making it a single place for the AI infra needs. In 2023, NVIDIA made a total of 40 investments, and by this month in 2024, it has already reached its 12th investment. The idea is to simply fund and acquire companies that are using NVIDIA’s GPUs – which is virtually all companies at this point.