CoreWeave AMD Genoa (9274F) - High [...], $ 2,455.20/mo. on Cloud

AMD Genoa (9274F) - High Performance has been added on Jul 5, 2026

💪 CPU/Cores :	AMD Genoa 9274F; 96 vCPUs
🔋 RAM :	768000 MB
🔌 Hosted domains :	1
🆓 free domains :	0
📌 Dedicated IPs :	0
💳 Payment Methods :	Credit / Debit / Prepaid Cards
🔨 Control Panel :	[In-house]
✍️ Support Options :	EmailHelp DeskAvailable 24/7
🌏 Server Locations :	Canada Denmark Norway Spain Sweden United Kingdom United States
⚑ Targeting :	US
🚀 Uptime :	99.9 %

See also initial AMD Genoa (9274F) - High Performance plan location on their website!

---

*📜 Plan description

📄 Editorial Review

CoreWeave is a U.S.-based cloud infrastructure company founded in 2017 by Michael Intrator, Brian Venturo, and Brannin McBee. The business initially concentrated on GPU-intensive cryptocurrency workloads before redirecting its computing capacity and engineering expertise toward artificial intelligence and high-performance computing. This transition turned CoreWeave into one of the most prominent specialized AI cloud providers. The company became publicly traded on Nasdaq under the ticker CRWV in March 2025 and has grown to approximately 2,500 employees.

CoreWeave is substantially different from a conventional web hosting provider. It does not concentrate on shared hosting, domain registrations, email hosting, website builders, cPanel accounts, or inexpensive virtual servers. Its platform is designed for AI laboratories, software companies, enterprise engineering departments, research organizations, and customers requiring large amounts of accelerated computing capacity.

The central service is GPU Compute, with infrastructure based on NVIDIA (nvidia.com) Vera Rubin, Blackwell, Hopper, Ada Lovelace, and earlier accelerator generations. Available systems include GB300 NVL72, GB200 NVL72, B300, B200, RTX PRO 6000 Blackwell Server Edition, H200, H100, GH200, L40S, L40, and A100 hardware.

Many GPU configurations run on complete bare-metal servers rather than virtualized instances. Customers receive direct access to the GPUs, processors, local storage, and high-performance networking without an additional hypervisor layer. This architecture is particularly suitable for distributed model training, large-scale inference, reinforcement learning, scientific computing, rendering, and other workloads where hardware consistency and predictable performance are important.

CPU Compute complements the GPU infrastructure with AMD Genoa, AMD Turin, Intel Emerald Rapids, Intel Ice Lake, and NVIDIA Grace processors. These systems can handle preprocessing, data preparation, orchestration, application services, model serving, and the supporting workloads that accompany large GPU clusters.

CoreWeave Kubernetes Service, commonly known as CKS, provides managed Kubernetes directly on bare-metal infrastructure. Each cluster runs inside a private VPC, with network separation enforced through NVIDIA BlueField data processing units. CoreWeave manages the Kubernetes control plane, drivers, storage integration, networking, and physical infrastructure, while customers retain control over their containers, services, namespaces, and applications.

CKS works with Kubernetes-native automation, REST APIs, Terraform, and kubectl. It is designed for engineering teams that want the flexibility of Kubernetes without having to build and maintain the underlying accelerated infrastructure themselves.

SUNK combines Slurm scheduling with Kubernetes operations. Researchers can continue using familiar Slurm workflows, while infrastructure teams receive Kubernetes-based lifecycle management, observability, identity provisioning, topology-aware scheduling, and automated hardware-health controls. SUNK Self-Service simplifies cluster deployment, while SUNK Anywhere can extend the same management approach to infrastructure located outside the CoreWeave cloud.

Inference services are available through several deployment models. Serverless Inference provides API-based access to managed models with token-based billing. Dedicated Inference runs custom or open-source models on selected GPU configurations, with CoreWeave managing routing, scaling, cluster operations, and infrastructure maintenance. Inference on CKS provides more direct control over Kubernetes, runtimes, scheduling, topology, and autoscaling.

CoreWeave Sandboxes provide isolated environments for reinforcement learning, software agents, code execution, model evaluation, and automated tool use. Sandboxes can operate on existing CKS capacity or through a fully managed serverless environment. They include namespace isolation, network policies, configurable resource limits, Python tooling, and Grafana-based monitoring.

Storage services include AI Object Storage, Distributed File Storage, local NVMe or SSD storage, and dedicated single-tenant storage clusters. Object storage includes hot, warm, cold, and archive tiers for datasets with different access patterns. Distributed File Storage provides a scalable POSIX-compatible environment suited to training datasets, shared files, checkpoints, and high-throughput AI workloads.

Local ephemeral storage can reach approximately 60 TB on selected systems and is included with the corresponding compute configuration. This provides fast scratch space for temporary datasets, caches, model checkpoints, and intermediate processing.

Dedicated storage environments can use technology from VAST Data (vastdata.com), WEKA (weka.io), DDN (ddn.com), IBM (ibm.com) Spectrum Scale, or Pure Storage (purestorage.com). Dedicated VAST Storage operates in a single-tenant environment with support for NFS, S3, NVMe over Fabrics, snapshots, replication, audit controls, SAML and SSO integration, encryption, and multi-petabyte datasets.

Networking services include private VPCs, public IPv4 addresses, bring-your-own-IP support, ingress DNS, high-performance InfiniBand, GPUDirect RDMA, and Direct Connect circuits reaching 400 Gbps. Internet ingress, internet egress, internal VPC transfer, inter-region transfer, NAT gateway traffic, and VPC usage do not carry separate data-transfer charges.

Mission Control combines infrastructure telemetry, automated node management, hardware lifecycle controls, operational visibility, and direct engineering assistance. CoreWeave Observe provides cluster metrics and Grafana dashboards, while Telemetry Relay can forward audit and access information into customer SIEM platforms, Prometheus Remote Write endpoints, HTTPS services, or S3-compatible storage.

GPU Straggler Detection identifies underperforming GPUs and ranks inside distributed training jobs. The Mission Control Agent can deliver infrastructure information and operational assistance through Slack, allowing engineers to diagnose problems without continuously switching between several management systems.

CoreWeave Omni brings the CoreWeave operating model into customer-owned facilities. The customer retains control of the data center and hardware, while CoreWeave installs and manages a dedicated cloud region containing Kubernetes orchestration, storage, networking, observability, lifecycle management, and the wider CoreWeave software stack.

The company has expanded beyond infrastructure through several AI software acquisitions. CoreWeave completed its acquisition of Weights & Biases (wandb.ai) on May 5, 2025, in a transaction valued at approximately $1.7 billion, although the official financial terms remained undisclosed.

OpenPipe (openpipe.ai), a reinforcement-learning platform, joined CoreWeave in September 2025. Marimo (marimo.io), developer of an AI-oriented Python notebook, was acquired in October 2025. CoreWeave also entered into an agreement to acquire Monolith AI (monolithai.com) on October 6, 2025, extending its reach into automotive, aerospace, manufacturing, and engineering applications.

The company operates a modern and responsive English-language website. The documentation, customer console, legal information, product details, and support resources are primarily available in English. Public prices are expressed in U.S. dollars.