The Pulse
Nutanix Buys Ryax to Bring GPU Scheduling to Its AI Platform
Nutanix has acquired French AI infrastructure company Ryax Technologies to add GPU utilization and workload scheduling capabilities to Nutanix Kubernetes Platform and Nutanix Enterprise AI. The companies say the planned integration will hel
AI.info Team ·
Nutanix is buying French AI infrastructure company Ryax Technologies as it moves to add GPU scheduling and resource optimization to its enterprise AI products. The acquisition, announced September 22, will bring Ryax’s technology into future releases of Nutanix Kubernetes Platform and Nutanix Enterprise AI.
Ryax develops software for orchestrating AI compute across private data centers, public clouds, neoclouds and high-performance computing environments. Nutanix says the platform can improve GPU and CPU utilization, automate workload placement and reduce the manual configuration that often slows the move from AI experiments to production systems.
Ryax Adds Placement Logic to NKP and NAI
The deal gives Nutanix two main capabilities to fold into its AI stack: intelligent resource optimization and AI-aware scheduling. The first is designed to match workloads with available computing resources more precisely. The second is intended to place training, inference and batch jobs on hardware that meets performance and cost requirements.
Nutanix says the combined platform will be able to work across heterogeneous infrastructure rather than assuming that every job runs on the same cluster or accelerator. That matters for enterprises splitting AI workloads among owned servers, cloud capacity and specialized providers because of GPU availability, data location or cost.
"Agentic AI requires a flexible architecture, simple operations and the ability to use infrastructure far more efficiently," said Rajiv Ramaswami, CEO of Nutanix.
Rajiv Ramaswami, CEO, Nutanix
Ramaswami said the Ryax acquisition addresses infrastructure-management problems that enterprises face as they deploy agentic AI. Nutanix did not disclose a purchase price and said the financial impact of the transaction is not material to the company.
Ryax Targets Idle and Oversized GPU Allocations
A Nutanix technical post published alongside the acquisition describes several planned capabilities. Ryax can use historical workload telemetry to estimate resource needs, adjust allocations and retry jobs after out-of-memory failures. The system is also designed to divide GPU capacity among multiple containerized workloads and release accelerators when active computation ends.
Those features target a familiar problem in AI infrastructure: teams often reserve more capacity than a job needs, then leave expensive accelerators idle between runs. Fractional GPU allocation can allow several workloads to share one machine, while serverless GPU allocation can limit billing and reservation time to periods when a job is actually running.
Nutanix’s blog says Ryax testing on a 30-run deep-learning burst reduced node-hours by 62% while finishing 5.7% faster through per-execution sizing. The same post says a document-intelligence pipeline reduced GPU hold times from hours to minutes per run, and that NVIDIA MIG-based fractional GPUs allowed four concurrent executions on one H100, reducing cost per execution by 52%.
Scheduling Will Span Clouds, Kubernetes and HPC
The planned scheduling layer is broader than a Kubernetes-only queue. Nutanix says Ryax can evaluate workloads against performance, cost and energy targets before selecting a destination across multiple clusters, public clouds and Slurm-based HPC systems.
The approach also includes support for different accelerator fleets, including NVIDIA and AMD systems. Nutanix says history-based autoscaling could set CPU, memory and fractional GPU allocations from earlier runs, while energy-aware placement could steer workloads toward clusters with lower predicted power use.
Those capabilities are planned rather than generally available product features. Nutanix says it intends to integrate Ryax into future versions of NKP and NAI, and its release states that the integrations remain under development and may change before release.
Ryax Team Will Join Nutanix in France
Ryax’s team will join Nutanix in France and continue working on the technology, according to the announcement. Andry Razafinjatovo, CEO of Ryax Technologies, said the company’s platform was built to handle the backend complexity of running AI workflows across different infrastructure environments.
The acquisition extends a series of AI infrastructure moves by Nutanix, which has been expanding its Kubernetes, inference and hybrid-cloud offerings. In this deal, the immediate focus is not a new model or accelerator, but the software layer that decides where AI workloads run and how much hardware they receive.
Nutanix’s announcement sets no general-availability date for the Ryax capabilities. For now, the transaction adds a French orchestration team and a scheduling technology platform to Nutanix’s planned AI stack, with the product integration still ahead.