The Tool Desk
Outbyte Driver Updater FREEScan for outdated or missing drivers - takes under a minuteDriver Scan →Outbyte PC Repair FREERepair Windows errors before they cause bigger problemsFix Now →Yes. Nvidia completed its acquisition of Israeli AI-infrastructure startup Run:ai on December 30, 2024. The deal was widely reported at approximately $700 million, although Nvidia did not disclose an official purchase price. Run:ai develops software that helps organizations schedule, share and govern expensive GPUs across AI workloads—it does not make chips, operate a cloud or build AI models.
For investors, enterprise technology buyers and competitors, the important point is strategic: Nvidia acquired a control layer above the GPU. The company can now offer software that helps customers decide which teams receive scarce computing capacity, how GPUs are shared and how AI workloads run across data centers and clouds.
As an Amazon Associate I earn from qualifying purchases.
The short version
- Nvidia announced the acquisition on April 24, 2024, and completed it on December 30, 2024.
- The purchase price was reported at about $700 million by TechCrunch; Nvidia’s announcement did not confirm a price.
- The European Commission cleared the transaction unconditionally on December 20, 2024.
- Run:ai’s technology is now marketed as NVIDIA Run:ai, an enterprise platform for AI workload and GPU orchestration.
- Nvidia also promotes KAI Scheduler, an open-source Kubernetes scheduler based on Run:ai technology.
What Run:ai actually does
Run:ai manages shared GPU infrastructure. A company might have one cluster serving researchers, model-training teams and production inference services. Without specialized orchestration, some GPUs may sit idle while other teams wait, or small amounts of unused capacity may be impossible to allocate efficiently.
Run:ai’s software provides a management and scheduling layer for that environment. Its capabilities include:
#1 Best Overall
- NVIDIA Volta GV100 Architecture — 4,608 CUDA Cores, 640 1st-Gen Tensor Cores delivering 14 TFLOPS FP32 and 112 TFLOPS deep learning performance for AI training, inference, HPC, and scientific computing workloads
- 32GB HBM2 ECC Memory — 900 GB/s Bandwidth — High-bandwidth memory on a 4096-bit bus with ECC error correction provides the memory capacity and throughput required for the largest AI models, simulations, and datasets
- PCIe 3.0 x16 Interface — 250W TDP — Standard PCIe Gen3 connectivity with passive cooling designed for enterprise rack server deployment in HPE ProLiant, Dell PowerEdge, and Supermicro platforms with adequate chassis airflow
- NVLink — Scale to 96GB Unified Memory — Connect two V100 GPUs via NVLink at 300 GB/s bi-directional bandwidth to scale GPU memory from 32GB to 96GB for larger AI training and HPC workloads
- Multi-Precision Computing — Supports FP64 (7 TFLOPS), FP32 (14 TFLOPS), FP16 (112 TFLOPS) and INT8 precision modes for flexible deployment across training, inference, and scientific simulation workloads
- Assigning users and teams quotas, priorities and resource pools.
- Pooling GPU capacity across clusters.
- Allocating fractional GPUs where workloads do not need an entire device.
- Scheduling training, development and inference jobs.
- Monitoring resource consumption and reporting on usage.
- Managing workloads through Kubernetes across on-premises, cloud and hybrid infrastructure.
In practical terms, the platform helps administrators answer questions such as: Which team gets capacity first? How many GPUs can a project use? Can a short inference workload run alongside a larger training job? Can unused capacity in one cluster be made available elsewhere?
That makes Run:ai infrastructure software rather than a conventional semiconductor acquisition. It does not replace Nvidia GPUs; it is designed to help organizations operate them more effectively.
Why Nvidia wanted Run:ai
Nvidia’s business has traditionally centered on accelerated computing hardware, networking, systems and related software. Run:ai adds a deeper role in the day-to-day allocation and governance of that computing capacity.
Free tools Windows power users keep installed
One-click scans. No signup required.
This is an analytical interpretation of the companies’ stated capabilities, not a disclosed internal explanation of Nvidia’s motives. But the strategic logic is clear: as GPUs become more expensive and difficult to obtain, customers need more than hardware. They need software that controls how the hardware is consumed.
Run:ai can help Nvidia become more embedded in enterprise operating workflows by managing:
- GPU access between departments and customers.
- Priority rules for urgent or high-value jobs.
- Utilization of fragmented capacity.
- Workloads spanning multiple clusters or clouds.
- Policy, monitoring and reporting for enterprise AI operations.
That matters commercially because software can create recurring relationships and switching costs around infrastructure that customers have already purchased. It also gives Nvidia a stronger position in the orchestration layer, not just the hardware layer.
Acquisition timeline
| Date | Event |
|---|---|
| 2020 | Run:ai and Nvidia had already been collaborating, according to Run:ai CEO Omri Geller’s statement in Nvidia’s acquisition announcement. |
| April 24, 2024 | Nvidia announced a definitive agreement to acquire Run:ai. |
| September 30, 2024 | Italy referred the transaction to the European Commission under Article 22 of the EU Merger Regulation. |
| October 31, 2024 | The European Commission accepted the referral. |
| November 15, 2024 | The transaction was formally notified to the Commission. |
| December 20, 2024 | The Commission approved the acquisition without conditions. |
| December 30, 2024 | Nvidia completed the acquisition. |
| 2025–2026 | Nvidia repositioned the technology as NVIDIA Run:ai and promoted KAI Scheduler as an open-source component. |
Why a relatively small startup drew antitrust scrutiny
The transaction received EU attention even though Run:ai’s revenue was negligible. Italy used its national merger-referral powers to send the deal to the European Commission, which accepted the referral under Article 22 of the EU Merger Regulation.
Rank #2
- GPU-Modell: Gefoce RTX 3080
- Memory Type: GDDR6X Memory Capacity: 20GB Memory Bus Width: 320bit Output Interfaces: 3*DP + HDMI Core Clock: 1710MHz Memory Clock: 19Gbps Power Interface: 8+8pin Recommended Power Supply: 850W or higher
The Commission examined two relevant areas: discrete data-center GPUs and GPU-orchestration software. Nvidia and Run:ai did not directly compete in the same product market, but the products are connected. Orchestration software must work with the GPUs that customers use.
The Commission therefore considered whether Nvidia might:
- Make its GPUs less compatible with rival orchestration software.
- Make Run:ai less compatible with AMD, Intel or other competing accelerators.
- Use its position in GPUs to favor its own orchestration product.
In its December 20, 2024 decision, the Commission concluded that the acquisition raised no competition concerns and cleared it unconditionally. It said Nvidia lacked both the technical ability and incentive to impair compatibility, while customers had credible alternatives and could develop software internally. The Commission also noted that Run:ai was not yet a significant orchestration-software provider.
Unconditional EU clearance does not mean regulators found that Nvidia could never disadvantage competitors. It means the Commission reached that conclusion for this transaction based on the evidence and competitive conditions examined in its review.
What’s actually slowing this PC down?
Pick the symptom - the matching free tool is one click away.
What about the United States?
Contemporary reporting said the U.S. Department of Justice also examined the transaction. The available sources establish reported scrutiny, but not a definitive public outcome equivalent to the European Commission’s formal clearance.
Accordingly, it is inaccurate to say that the DOJ approved the deal, blocked it or definitively ended an investigation without a primary DOJ statement supporting that claim.
What changed after the acquisition?
Nvidia now markets the product as NVIDIA Run:ai. Its current positioning extends beyond basic GPU scheduling to include AI-native workload orchestration, dynamic allocation, policy-based governance, hybrid and multi-cloud support, inference scaling and lifecycle management from development through deployment.
Rank #3
- No Processor Installed; Supports 2x AMD EPYC 9004 Series Processors
- No Memory Installed; Supports 24x DDR5 4400/4800 Regsitered Memory Modules
- 8x 3.5" Trays; (Bring Your Own SATA/NVMe Drives)
- 4x H200 NVL Tensor Core 141GB HBM3e PCI Express 5.0 x16 GPU Accelerator Card
- In Original Packaging; Includes Rails and ASUS GPU Cables
Nvidia also says Run:ai is included in NVIDIA AI Enterprise. The company promotes features including fractional GPU use and GPU memory swapping for inference workloads, and presents the platform as part of its broader enterprise AI-factory and DGX software stack.
These are current Nvidia product claims. Nvidia’s product page also cites figures such as “10x GPU availability,” “20x workloads running” and “5x GPU utilization.” Those figures are vendor marketing claims, not independent benchmarks, and should not be treated as results every customer will achieve.
Open source does not mean the entire platform is free
At the time the acquisition closed, Run:ai said it planned to open-source its software. The current result is more specific: Nvidia identifies KAI Scheduler as an open-source scheduler based on Run:ai technology, while NVIDIA Run:ai remains the enterprise offering.
That distinction matters. Open source does not automatically make every enterprise feature, user interface, governance control, integration, support service or commercial entitlement free. It also does not prove that the platform runs equally well on every accelerator.
Version and deployment caveat
Nvidia’s main product page currently advertises NVIDIA Run:ai version 2.25, while the reviewed self-hosted support matrix lists version 2.24 as its latest supported release. This may reflect different release timing or deployment tracks. Buyers should verify the exact version-specific support matrix and release notes before installation.
Who should consider NVIDIA Run:ai?
Run:ai is most relevant to organizations that:
- Operate shared Nvidia GPU clusters used by multiple teams.
- Have fragmented or unpredictable GPU demand.
- Need quotas, queues, priorities, resource pools and policy controls.
- Run Kubernetes across on-premises and cloud environments.
- Want commercial support and integration with Nvidia’s enterprise stack.
- Need governance and monitoring rather than only a basic scheduler.
The product may be a poor fit for a small team with only a few GPUs, an organization that does not use Kubernetes, a primarily non-Nvidia accelerator fleet, or a buyer seeking a fully self-service open-source stack. It may also be unsuitable for Jetson, DGX Spark and workstation configurations listed as unsupported in the cited documentation.
Independent reader supportYour contribution helps us test, update, and keep practical guides available for everyone.Technical checks before deployment
NVIDIA Run:ai is not an isolated software installation. The support matrix and system requirements should be checked against the proposed environment.
Rank #4
- 【Brilliant AI Performance for production】 on-device processing with up to 100 TOPS AI performance with low power and low latency, Due to the high thermal demands of Super mode, only the J30 Series supports upgrading to Super mode via the JetPack 6.2 update
- 【Hand-size edge AI device】 compact size at 130mm x120mm x 58.5mm, includes NVIDIA Jetson Orin NX 16GB production module, a cooling fan with a heatsink, enclosure, and a power adapter. Support desktop, wall mount, fit in anywhere
- 【Expandable with rich I/Os】4x USB 3.2, HDMI 2.1, 2xCSI, 1xRJ45 for GbE, M.2 Key E, M.2 Key M, CAN, and GPIO
- 【Accelerate solution to market】pre-installed Jetpack with NVIDIA JetPack 5.1 on the included 128GB NVMe SSD, Linux OS BSP, 128GB SSD, support Jetson software and leading AI frameworks and software platforms
- 【Comprehensive certificates】FCC, CE, RoHS, UKCA
- Kubernetes: The reviewed v2.24 matrix lists upstream Kubernetes 1.33–1.35 and Red Hat OpenShift 4.17–4.20, with support for major managed Kubernetes services based on upstream compatibility.
- GPU Operator: The deployment requires Nvidia GPU Operator, with the cited matrix listing versions 25.3 through 25.10.
- Worker nodes: Documented minimums include two CPU cores and 4 GB of memory, although production sizing depends on cluster size and monitoring needs.
- Shared storage: Workloads need consistent access to training data, code, checkpoints, weights and other artifacts across worker nodes.
- Networking: An ingress controller is required, and certain inference endpoints need suitable DNS configuration.
- Version compatibility: Run:ai, Kubernetes, OpenShift, CUDA, GPU Operator, frameworks and Nvidia inference components may all affect supportability.
- Multi-node NVLink: The documentation says Kubernetes 1.32 or newer is required for scenarios such as GB200 multi-node NVLink deployments.
Fractional allocation and aggressive scheduling can improve utilization, but poorly tuned policies can also create contention, unpredictable inference latency, memory pressure or interference between training and production workloads.
Alternatives to compare
| Option | Best suited to | Main trade-off |
|---|---|---|
| KAI Scheduler | Developers and smaller teams wanting AI-aware Kubernetes scheduling. | It is a scheduler component, not necessarily the full supported enterprise control plane. |
| Volcano | Open-source Kubernetes batch, queueing, gang-scheduling and HPC workloads. | More end-to-end AI operations may need to be assembled and operated separately. |
| Slurm | Established HPC and large-scale batch environments. | It is not a drop-in replacement for a Kubernetes-native ML platform. |
| Native Kubernetes plus the Nvidia GPU Operator | Small or relatively simple GPU clusters. | Advanced quotas, sharing, governance and observability may require additional engineering. |
What the deal means for Nvidia and its competitors
The acquisition increases Nvidia’s presence in the control plane of AI infrastructure. That could strengthen customer dependence on Nvidia’s software and hardware ecosystem, particularly where orchestration policies, monitoring and inference operations become deeply integrated into production systems.
PC Slower Than It Used to Be?
A free scan shows the junk files, broken settings and background clutter dragging Windows down - then fixes them in one click.Free scan · Windows 10 & 11Crashes, No Sound, or Screen Glitches?
Random freezes, missing sound and display glitches usually trace back to one bad driver. Find and replace yours safely.Free scan · under a minuteAt the same time, open-source scheduling technology could make adoption easier for organizations that do not want to purchase the full enterprise platform immediately. It may also give other ecosystem participants an opportunity to adapt parts of the technology. But open-source components do not eliminate the broader questions of hardware compatibility, support, enterprise features or commercial control.
For competitors such as AMD and Intel, interoperability is the central issue. The European Commission examined whether Nvidia could make Run:ai incompatible with rival GPUs and concluded that the transaction did not warrant intervention. Future product changes, market conditions and regulatory questions are separate matters.
What investors and buyers should take away
For investors, Run:ai is not a major revenue event on the scale of Nvidia’s chip business. Its importance is strategic: it supports Nvidia’s effort to capture more value from the software, networking and operational layers surrounding AI computing.
For buyers, the decision should not be based on the acquisition headline alone. Compare the cost of Nvidia’s sales-led enterprise product with the engineering and support costs of open-source or native tooling. Nvidia does not publish a standard list price on the reviewed product page; it directs prospects to sales or authorized partners. The relevant questions are therefore deployment fit, licensing, support, hardware coverage and operational complexity—not an invented per-GPU price.
Recommended Free Tools
Quick Recap
Product prices and availability are accurate as of the date/time indicated and are subject to change. Any price and availability information displayed on Amazon at the time of purchase will apply.
