{ "name": "HPC providers, side by side", "url": "https://hpcproviders.com/", "license": "CC BY 4.0", "cite": "HPC Providers, https://hpcproviders.com/", "updated": "2026-11-11", "fields": [ { "key": "provider_type", "label": "Type", "type": "enum", "description": "What kind of provider it is: hyperscaler, managed HPC service, GPU cloud, sovereign or EU-hosted cloud, publicly funded supercomputer, or simulation-software cloud.", "filterable": true, "sortable": true, "enumValues": [ "hyperscaler", "managed-hpc", "gpu-cloud", "sovereign", "public-supercomputer", "simulation-cloud" ] }, { "key": "fabric", "label": "Fabric", "type": "enum", "description": "Primary node-to-node network the provider states for its HPC or GPU clusters. RDMA means the provider names RDMA cluster networking without naming the transport.", "filterable": true, "sortable": true, "enumValues": [ "InfiniBand", "EFA", "RoCE", "RDMA", "Ethernet", "Slingshot" ] }, { "key": "interconnect", "label": "Interconnect detail", "type": "string", "description": "Speeds and instance families as the provider states them.", "sortable": false }, { "key": "schedulers", "label": "Schedulers", "type": "string", "description": "Job schedulers and orchestrators the provider offers or documents.", "sortable": false }, { "key": "slurm", "label": "Slurm", "type": "bool", "description": "Offers or documents Slurm (managed, or as a supported deployment).", "filterable": true }, { "key": "kubernetes", "label": "Kubernetes", "type": "bool", "description": "Offers managed Kubernetes for its clusters.", "filterable": true }, { "key": "cpus", "label": "CPUs", "type": "string", "description": "CPU models the provider names for its HPC or GPU hosts.", "sortable": false }, { "key": "gpus", "label": "GPUs / accelerators", "type": "string", "description": "Accelerators the provider lists, as listed on its own pages.", "sortable": false }, { "key": "gpu_h100", "label": "H100", "type": "bool", "description": "Lists NVIDIA H100.", "filterable": true }, { "key": "gpu_blackwell", "label": "Blackwell", "type": "bool", "description": "Lists at least one NVIDIA Blackwell data-centre GPU (B200, B300, GB200, GB300).", "filterable": true }, { "key": "gpu_amd", "label": "AMD Instinct", "type": "bool", "description": "Lists at least one AMD Instinct GPU.", "filterable": true }, { "key": "bare_metal", "label": "Bare metal", "type": "bool", "description": "Sells non-virtualised bare-metal nodes.", "filterable": true }, { "key": "h100_hr", "label": "H100 $/GPU-hr", "type": "money", "unit": "USD/GPU-hr", "description": "Published per-GPU-hour price of an H100. Basis (on-demand, capacity block, derived from an 8-GPU price) is in the next column.", "filterable": true, "sortable": true }, { "key": "h100_basis", "label": "H100 price basis", "type": "string", "description": "What the H100 price is: purchase mode, instance size and any derivation.", "sortable": false }, { "key": "price_asof", "label": "Price checked", "type": "date", "description": "Date the price was read from the provider page.", "sortable": true }, { "key": "pricing_model", "label": "Pricing model", "type": "string", "description": "How the provider charges, as it states it.", "sortable": false }, { "key": "regions", "label": "Regions", "type": "string", "description": "Locations the provider names on its own pages.", "sortable": false }, { "key": "eu_hosted", "label": "EU-hosted", "type": "bool", "description": "Names data-centre capacity in an EU member state.", "filterable": true }, { "key": "sovereign", "label": "Sovereignty claim", "type": "string", "description": "The provider's own sovereignty or compliance statement, quoted or closely paraphrased.", "sortable": false }, { "key": "software", "label": "Software catalogue", "type": "string", "description": "Pre-configured application catalogue a platform provider advertises.", "sortable": false }, { "key": "underlying_clouds", "label": "Runs on", "type": "string", "description": "Clouds and hardware partners a platform provider lists.", "sortable": false } ], "records": [ { "id": "advania-gompute", "name": "Advania (Gompute)", "url": "https://www.advania.se/en/gompute-hpc-platform", "summary": "Managed HPC from Advania: dedicated bare-metal private clusters, pay-as-you-go bursting clusters and remote desktops on the Gompute platform, hosted in Nordic data centres.", "tags": [ "bare-metal", "managed-hpc" ], "fields": { "bare_metal": true, "regions": "Nordic data centres (cities not named); Iceland on-premises HPC deployments", "pricing_model": "Dedicated private clusters; pay-as-you-go bursting; credit-based on-demand; 48-hour trial. No list prices published", "sovereign": "Private cloud: full data sovereignty in accordance with European regulations; ISO 27001", "provider_type": "managed-hpc" }, "sources": [ { "url": "https://www.advania.se/en/gompute-hpc-platform", "title": "Gompute HPC platform", "accessed": "2026-11-11", "fields": [ "bare_metal", "regions", "pricing_model", "sovereign" ] }, { "url": "https://www.advania.co.uk/our-solutions/high-performance-compute/", "title": "Advania UK HPC", "accessed": "2026-11-11", "fields": [ "pricing_model" ] }, { "url": "https://advania.is/datacenters/", "title": "Advania Iceland data centres", "accessed": "2026-11-11", "fields": [ "regions" ] } ], "verifiedAt": "2026-11-11" }, { "id": "alibaba-cloud-ehpc", "name": "Alibaba Cloud E-HPC", "url": "https://www.alibabacloud.com/product/ehpc", "summary": "Elastic High Performance Computing: managed Slurm clusters on ECS with eRDMA-enabled GPU and bare-metal instance families, billed per core-hour.", "hq": "CN", "tags": [ "bare-metal", "hyperscaler", "rdma", "slurm" ], "fields": { "fabric": "RDMA", "interconnect": "Elastic RDMA Interface (eRDMA) on gn8is, gn8v, gn9g and ebmgn8* families", "schedulers": "E-HPC NEXT (Slurm cluster type)", "slurm": true, "gpus": "NVIDIA L20, A10, V100; Blackwell-based families (model not named on the page)", "bare_metal": true, "pricing_model": "E-HPC service fee $0.003 per core-hour above 200 vCPUs per region; ECS, storage and traffic billed separately", "provider_type": "hyperscaler" }, "pricing": { "model": "Per core-hour service fee", "from": 0.003, "unit": "core-hour (E-HPC service fee, ECS extra)", "currency": "USD" }, "sources": [ { "url": "https://www.alibabacloud.com/help/en/ecs/user-guide/on-the-gpu-instance-configuration-erdma", "title": "eRDMA GPU instance configuration", "accessed": "2026-11-11", "fields": [ "fabric", "interconnect" ] }, { "url": "https://www.alibabacloud.com/help/en/e-hpc/e-hpc-next/use-cases/deploying-elastic-high-performance-computing-clusters-based-on-erdma-e-hpc-next", "title": "Deploy E-HPC NEXT with eRDMA", "accessed": "2026-11-11", "fields": [ "schedulers", "slurm" ] }, { "url": "https://www.alibabacloud.com/help/en/egs/gpu-accelerated-compute-optimized-instance-families", "title": "GPU instance families", "accessed": "2026-11-11", "fields": [ "gpus", "bare_metal" ] }, { "url": "https://www.alibabacloud.com/product/ehpc/pricing", "title": "E-HPC pricing", "accessed": "2026-11-11", "fields": [ "pricing_model" ] } ], "verifiedAt": "2026-11-11" }, { "id": "aws-hpc", "name": "AWS HPC", "url": "https://aws.amazon.com/hpc/", "summary": "EC2 HPC instances (hpc7a, hpc7g, hpc6a), P5 H100 and P5en H200 GPU instances on EFA, the free ParallelCluster tool and AWS Parallel Computing Service, a managed Slurm controller.", "hq": "US", "tags": [ "efa", "hyperscaler", "slurm" ], "fields": { "fabric": "EFA", "interconnect": "EFA: 300 Gbps on hpc7a, 200 Gbps on hpc7g, 100 Gbps on hpc6a, up to 3,200 Gbps on P5/P5e/P5en", "schedulers": "AWS Parallel Computing Service (managed Slurm), AWS ParallelCluster (open source), AWS Batch", "slurm": true, "cpus": "hpc7a: 4th Gen AMD EPYC, up to 192 cores; hpc7g: AWS Graviton3E, 64 cores; hpc6a: 3rd Gen AMD EPYC, 96 cores", "gpus": "NVIDIA H100 (P5), NVIDIA H200 (P5en), 8 per instance", "gpu_h100": true, "h100_hr": 6.88, "h100_basis": "On-demand p5.4xlarge (1x H100) Linux in us-east-1 at $6.88 per instance-hour; the 8-GPU p5.48xlarge is $55.04 ($6.88 per GPU). EC2 Capacity Blocks run $5.97 per GPU-hour ($47.757 per p5.48xlarge hour).", "price_asof": "2026-11-11", "pricing_model": "On-demand and Capacity Blocks for EC2; PCS adds a controller fee and a per-node fee ($0.08 per instance-hour on standard types); ParallelCluster is free", "provider_type": "hyperscaler" }, "pricing": { "model": "On-demand, Capacity Blocks", "from": 6.88, "unit": "GPU-hour (H100, on-demand)", "currency": "USD", "note": "p5.4xlarge (1x H100) on-demand $6.88/hr and p5.48xlarge (8x H100) $55.04/hr in us-east-1. Capacity Blocks: $47.757 per p5.48xlarge hour ($5.97 per GPU-hour). p5en.48xlarge (8x H200) on-demand $63.296/hr." }, "sources": [ { "url": "https://aws.amazon.com/ec2/instance-types/hpc7a/", "title": "EC2 Hpc7a instances", "accessed": "2026-11-11", "fields": [ "fabric", "interconnect", "cpus" ] }, { "url": "https://aws.amazon.com/ec2/instance-types/hpc7g/", "title": "EC2 Hpc7g instances", "accessed": "2026-11-11", "fields": [ "interconnect", "cpus" ] }, { "url": "https://aws.amazon.com/ec2/instance-types/hpc6a/", "title": "EC2 Hpc6a instances", "accessed": "2026-11-11", "fields": [ "interconnect", "cpus" ] }, { "url": "https://aws.amazon.com/ec2/instance-types/p5/", "title": "EC2 P5 instances", "accessed": "2026-11-11", "fields": [ "interconnect", "gpus", "gpu_h100" ] }, { "url": "https://aws.amazon.com/pcs/", "title": "AWS Parallel Computing Service", "accessed": "2026-11-11", "fields": [ "schedulers", "slurm" ] }, { "url": "https://aws.amazon.com/hpc/parallelcluster/", "title": "AWS ParallelCluster", "accessed": "2026-11-11", "fields": [ "schedulers", "slurm" ] }, { "url": "https://aws.amazon.com/pcs/pricing/", "title": "AWS PCS pricing", "accessed": "2026-11-11", "fields": [ "pricing_model" ] }, { "url": "https://aws.amazon.com/ec2/pricing/on-demand/", "title": "Amazon EC2 On-Demand pricing (US East, N. Virginia, Linux)", "accessed": "2026-11-11", "fields": [ "h100_hr", "h100_basis", "price_asof" ] }, { "url": "https://aws.amazon.com/ec2/capacityblocks/pricing/", "title": "EC2 Capacity Blocks pricing", "accessed": "2026-11-11", "fields": [ "h100_basis" ] } ], "verifiedAt": "2026-11-11" }, { "id": "azure-hpc", "name": "Azure HPC", "url": "https://azure.microsoft.com/en-us/solutions/high-performance-computing/", "summary": "InfiniBand-connected HB-series CPU and ND H100 v5 GPU virtual machines, CycleCloud for Slurm, PBS Pro, LSF and HTCondor clusters, and Azure Batch scheduling as a service.", "hq": "US", "tags": [ "hyperscaler", "infiniband", "slurm" ], "fields": { "fabric": "InfiniBand", "interconnect": "ND H100 v5: 400 Gb/s NDR InfiniBand per GPU (3.2 Tb/s per VM); HBv5: 800 Gb/s per VM", "schedulers": "Azure CycleCloud (Slurm, PBS Pro, LSF, Grid Engine, HTCondor), Azure Batch", "slurm": true, "cpus": "HBv5: 4th Gen AMD EPYC, up to 368 cores; ND H100 v5 host: Intel Sapphire Rapids, 96 vCPUs", "gpus": "NVIDIA H100 80 GB, 8 per ND H100 v5 VM", "gpu_h100": true, "h100_hr": 12.29, "h100_basis": "Derived: ND96isr H100 v5 Linux pay-as-you-go, East US, $98.32 per VM-hour divided by 8. Azure publishes the VM price, not a per-GPU rate.", "price_asof": "2026-11-11", "pricing_model": "Pay-as-you-go, Spot and reserved; Azure Batch adds no extra charge", "provider_type": "hyperscaler" }, "pricing": { "model": "Pay-as-you-go, Spot", "from": 12.29, "unit": "GPU-hour (H100, derived from VM price)", "currency": "USD", "note": "$12.29 per GPU-hour is derived: ND96isr H100 v5 (8x H100) is $98.32 per VM-hour Linux pay-as-you-go in East US, divided by 8; Spot $18.17 per VM-hour. HBv5 (Standard_HB368rs_v5) is $19.80 per VM-hour; Spot $3.66." }, "sources": [ { "url": "https://learn.microsoft.com/en-us/azure/virtual-machines/sizes/gpu-accelerated/ndh100v5-series", "title": "ND H100 v5 series", "accessed": "2026-11-11", "fields": [ "fabric", "interconnect", "gpus", "gpu_h100", "cpus" ] }, { "url": "https://learn.microsoft.com/en-us/azure/virtual-machines/sizes/high-performance-compute/hbv5-series", "title": "HBv5 series", "accessed": "2026-11-11", "fields": [ "interconnect", "cpus" ] }, { "url": "https://learn.microsoft.com/en-us/azure/cyclecloud/overview", "title": "Azure CycleCloud overview", "accessed": "2026-11-11", "fields": [ "schedulers", "slurm" ] }, { "url": "https://learn.microsoft.com/en-us/azure/batch/batch-technical-overview", "title": "Azure Batch overview", "accessed": "2026-11-11", "fields": [ "pricing_model" ] }, { "url": "https://prices.azure.com/api/retail/prices?$filter=armSkuName eq 'Standard_ND96isr_H100_v5'", "title": "Azure Retail Prices API, ND96isr_H100_v5", "accessed": "2026-11-11", "fields": [ "h100_hr", "h100_basis", "price_asof" ] } ], "verifiedAt": "2026-11-11" }, { "id": "bull-jarvice", "name": "Bull HPC as a Service (JARVICE, formerly Nimbix)", "url": "https://www.bull.com/en/solutions/hpc/hpc-as-a-service", "summary": "Bull's HPC-as-a-service platform, built on the JARVICE software it took over with Nimbix, gives engineers and researchers point-and-click access to a catalogue of simulation, AI and quantum applications. It runs in the cloud, on premises or both, and can send jobs to existing Slurm clusters.", "tags": [ "simulation-cloud", "slurm" ], "fields": { "provider_type": "simulation-cloud", "software": "HyperHub catalogue of simulation, AI, machine learning, deep learning and quantum applications from partner ISVs, launched as point-and-click workflows (CFD, electromagnetics, EDA, structural, multiphysics, genomics, 3D rendering)", "schedulers": "JARVICE platform with unified HPC/AI workload scheduling; runs on Kubernetes and dispatches jobs to downstream Slurm clusters", "slurm": true, "underlying_clouds": "Cloud, on-premises or hybrid; JARVICE XE deploys on Amazon EKS, Google GKE, Microsoft AKS or self-managed Kubernetes" }, "sources": [ { "url": "https://www.bull.com/en/solutions/hpc/hpc-as-a-service", "title": "Bull HPC as a Service", "accessed": "2026-11-11", "fields": [ "provider_type", "software", "schedulers", "underlying_clouds" ] }, { "url": "https://github.com/nimbix/jarvice-helm/blob/master/SlurmScheduler.md", "title": "JARVICE Slurm Scheduler Overview (nimbix/jarvice-helm)", "accessed": "2026-11-11", "fields": [ "schedulers", "slurm" ] }, { "url": "https://github.com/nimbix/jarvice-helm", "title": "JARVICE XE Hybrid Cloud HPC platform Helm chart (nimbix/jarvice-helm)", "accessed": "2026-11-11", "fields": [ "schedulers", "underlying_clouds" ] } ], "verifiedAt": "2026-11-11" }, { "id": "cirrascale", "name": "Cirrascale Cloud Services", "url": "https://www.cirrascale.com", "summary": "US cloud provider of multi-accelerator AI training and inference infrastructure, offering NVIDIA, AMD Instinct, Qualcomm and Tenstorrent systems plus a managed inference platform. Sells hourly and in 3-, 6- and 12-month blocks.", "founded": 2017, "tags": [ "amd", "gpu-cloud" ], "fields": { "price_asof": "2026-11-09", "regions": "US (US-West, US-East, US-South per company about page)", "gpus": "GB300, B300, B200, RTX PRO 6000, MI355X", "gpu_h100": false, "gpu_blackwell": true, "gpu_amd": true, "interconnect": "Up to 3,200 Gb per server (fabric type not named)", "pricing_model": "hourly + term blocks", "provider_type": "gpu-cloud" }, "pricing": { "model": "hourly + term blocks", "from": 4.79, "unit": "GPU-hour", "currency": "USD", "note": "NVIDIA HGX B200 'as low as' $4.79/GPU-hr on cirrascale.com/pricing (2026-11-09); page says prices are examples and term blocks save up to 20%, so not confirmed as on-demand list price (list_price_b200_hr left null). No H100 listed. Others: B300 $5.69, GB300 $8.09, MI355X $4.29, MI350 $3.89. Founded = cloud division became a separate company in early 2017 (about page). ClusterMAX 3.0 lists it as 'Unavailable' (not rated)." }, "sources": [ { "url": "https://www.cirrascale.com/pricing", "title": "Cirrascale pricing", "accessed": "2026-11-09", "fields": [ "price_asof", "pricing_model" ] }, { "url": "https://www.cirrascale.com/", "title": "Cirrascale homepage", "accessed": "2026-11-09" }, { "url": "https://www.cirrascale.com/about", "title": "Cirrascale about", "accessed": "2026-11-09", "fields": [ "regions" ] }, { "url": "https://clustermax.semianalysis.com/cloudreview", "title": "SemiAnalysis ClusterMAX 3.0 (Sep 2026) rankings image", "accessed": "2026-11-09" }, { "url": "https://cirrascale.com/pricing", "title": "Cirrascale pricing", "accessed": "2026-11-11", "fields": [ "interconnect", "gpus" ] } ], "verifiedAt": "2026-11-09" }, { "id": "core42", "name": "Core42", "url": "https://www.core42.ai", "summary": "G42 company selling sovereign AI cloud (GPU-as-a-service on NVIDIA, AMD, Cerebras and Qualcomm accelerators), a sovereign Azure-based public cloud and private cloud, mainly to governments and enterprises in the UAE, with deployments in the US and Europe.", "hq": "AE", "tags": [ "amd", "bare-metal", "eu", "gpu-cloud", "infiniband", "slurm" ], "fields": { "h100_hr": 2.5, "price_asof": "2026-11-09", "interconnect": "InfiniBand, Ethernet, NVLink", "fabric": "InfiniBand", "bare_metal": true, "slurm": true, "kubernetes": true, "regions": "UAE; US (New York deployment); Europe (Italy, France; European HQ in Dublin). Company cites 9 global data centers.", "gpus": "H100, H200, B200, GB300, MI300X, MI355X", "gpu_h100": true, "gpu_blackwell": true, "gpu_amd": true, "schedulers": "Slurm, Kubernetes", "h100_basis": "\"From\" price for H100", "pricing_model": "on-demand pay-as-you-go + large clusters", "eu_hosted": true, "provider_type": "gpu-cloud" }, "pricing": { "model": "on-demand pay-as-you-go + large clusters", "from": 2.5, "unit": "GPU-hour", "currency": "USD", "note": "'From' price for NVIDIA H100 (SXM/PCIe not stated); H200 from $4.00, B200 from $5.00, MI300X from $3.50; GB300, MI355X, Cerebras on request. ClusterMAX 3.0 (Sep 2026) lists Core42 under 'Participation', not a medal tier." }, "sources": [ { "url": "https://core42.ai", "title": "Core42 homepage", "accessed": "2026-11-09" }, { "url": "https://www.core42.ai/products/ai-cloud", "title": "Core42 AI Cloud", "accessed": "2026-11-09" }, { "url": "https://www.core42.ai/resources/news/core42-raises-usd-550-million-from-hsbc-to-scale-global-ai-infrastructure?hsLang=en", "title": "Core42 Raises USD 550 million from HSBC to Scale Global AI Infrastructure", "accessed": "2026-11-09" }, { "url": "https://clustermax.semianalysis.com/img/neocloud-ranking-v3.png", "title": "SemiAnalysis ClusterMAX 3.0 ranking (Sep 2026)", "accessed": "2026-11-09" } ], "verifiedAt": "2026-11-09" }, { "id": "coreweave", "name": "CoreWeave", "url": "https://www.coreweave.com", "summary": "Publicly listed GPU cloud selling large NVIDIA GPU clusters, mostly under multi-year contracts, to AI labs and enterprises. Runs Kubernetes-native infrastructure (CKS, SUNK Slurm-on-Kubernetes) across 50+ data centers in the US and Europe.", "hq": "US", "founded": 2017, "tags": [ "bare-metal", "gpu-cloud", "infiniband", "slurm" ], "fields": { "h100_hr": 6.16, "price_asof": "2026-11-09", "interconnect": "InfiniBand", "fabric": "InfiniBand", "bare_metal": true, "slurm": true, "kubernetes": true, "regions": "US and Europe; 50+ active data centers (company site, Nov 2026)", "gpus": "GB300, GB200, B300, B200, RTX PRO 6000, GH200, H200, H100, L40S, L40, A100", "gpu_h100": true, "gpu_blackwell": true, "gpu_amd": false, "schedulers": "Slurm, Kubernetes", "h100_basis": "On-demand HGX H100 8-GPU instance, $49.24 per hour divided by 8", "pricing_model": "on-demand + reserved", "provider_type": "gpu-cloud" }, "pricing": { "model": "on-demand + reserved", "from": 6.16, "unit": "GPU-hour", "currency": "USD", "note": "HGX H100 (SXM) on-demand, $49.24/hr per 8-GPU instance (~$6.16/GPU-hr); HGX B200 $68.80/hr per 8 GPUs ($8.60/GPU-hr). No single-GPU H100 instance listed." }, "sources": [ { "url": "https://www.coreweave.com/pricing", "title": "CoreWeave GPU pricing", "accessed": "2026-11-09", "fields": [ "h100_hr", "price_asof", "pricing_model" ] }, { "url": "https://investors.coreweave.com/news/news-details/2026/CoreWeave-Reports-Strong-Second-Quarter-2026-Results/default.aspx", "title": "CoreWeave Reports Strong Second Quarter 2026 Results", "accessed": "2026-11-09" }, { "url": "https://www.coreweave.com/about-us", "title": "About CoreWeave", "accessed": "2026-11-09" }, { "url": "https://docs.coreweave.com/", "title": "CoreWeave docs (CKS, SUNK)", "accessed": "2026-11-09" }, { "url": "https://www.coreweave.com/products/networking", "title": "CoreWeave networking", "accessed": "2026-11-09" }, { "url": "https://newsletter.semianalysis.com/p/clustermax-30-the-industry-standard", "title": "ClusterMAX 3.0 (SemiAnalysis)", "accessed": "2026-11-09" } ], "verifiedAt": "2026-11-09" }, { "id": "crusoe", "name": "Crusoe", "url": "https://www.crusoe.ai", "summary": "Denver-based, vertically integrated AI infrastructure company that builds data center campuses and runs Crusoe Cloud, selling NVIDIA and AMD GPU instances, managed Kubernetes, Slurm and inference. Built the Abilene, Texas campus used by OpenAI.", "hq": "US", "tags": [ "amd", "gpu-cloud", "infiniband", "slurm" ], "fields": { "h100_hr": 3.9, "price_asof": "2026-11-09", "interconnect": "InfiniBand, Ethernet", "fabric": "InfiniBand", "slurm": true, "kubernetes": true, "regions": "Data center sites named in company releases: Abilene and Armstrong County (Texas), Wyoming, Norway (with Polar); Paraguay deployment announced Oct 2026", "gpus": "Vera Rubin NVL72, GB300, B300, GB200, B200, H200, H100, A100, L40S, MI355X, MI300X", "gpu_h100": true, "gpu_blackwell": true, "gpu_amd": true, "schedulers": "Slurm, Kubernetes", "h100_basis": "On-demand HGX H100 80 GB; rises to $4.50 on 1 Dec 2026", "pricing_model": "on-demand + spot + reserved", "provider_type": "gpu-cloud" }, "pricing": { "model": "on-demand + spot + reserved", "from": 3.9, "unit": "GPU-hour", "currency": "USD", "note": "HGX H100 80GB on-demand per GPU-hour; rises to $4.50 from 1 Dec 2026. B200 on-demand is contact sales until 1 Dec 2026, then $8.50. Row-to-price mapping read from page text." }, "sources": [ { "url": "https://www.crusoe.ai/cloud/pricing", "title": "Crusoe Cloud pricing", "accessed": "2026-11-09", "fields": [ "h100_hr", "price_asof", "pricing_model" ] }, { "url": "https://www.crusoe.ai/cloud", "title": "Crusoe Cloud", "accessed": "2026-11-09" }, { "url": "https://docs.crusoe.ai/", "title": "Crusoe Cloud docs", "accessed": "2026-11-09" }, { "url": "https://www.crusoe.ai/resources/newsroom/crusoe-announces-series-f-funding", "title": "Crusoe Raises $3.9 Billion Series F", "accessed": "2026-11-09" }, { "url": "https://www.crusoe.ai/newsroom", "title": "Crusoe newsroom", "accessed": "2026-11-09" }, { "url": "https://newsletter.semianalysis.com/p/clustermax-30-the-industry-standard", "title": "ClusterMAX 3.0 (SemiAnalysis)", "accessed": "2026-11-09" } ], "verifiedAt": "2026-11-09" }, { "id": "cudo-compute", "name": "CUDO Compute", "url": "https://www.cudocompute.com", "summary": "Builds and operates dedicated NVIDIA GPU clusters and AI factories (land, power, colocation, managed services) for enterprise and regulated workloads. Pricing is by quote; no public on-demand rates.", "tags": [ "bare-metal", "gpu-cloud", "infiniband" ], "fields": { "interconnect": "InfiniBand, Ethernet", "fabric": "InfiniBand", "bare_metal": true, "regions": "North America, Europe, UK, Middle East/MENA (company site; cities not named)", "gpus": "H100, H200, B100, B200, B300, GB200, GB300, VR200", "gpu_h100": true, "gpu_blackwell": true, "gpu_amd": false, "pricing_model": "reserved / dedicated clusters (quote)", "provider_type": "gpu-cloud" }, "pricing": { "model": "reserved / dedicated clusters (quote)", "unit": "GPU-hour", "currency": "USD", "note": "All SKUs (H100 PCIe/SXM, H200, HGX B200, GB200 NVL72, B300, GB300, VR200) are 'Quote on request' as of 2026-11-09. Listed as 'Not Recommended - Unavailable' in ClusterMAX 3.0 (no environment provided for testing), not a rating." }, "sources": [ { "url": "https://www.cudocompute.com/pricing", "title": "CUDO Compute pricing", "accessed": "2026-11-09", "fields": [ "pricing_model" ] }, { "url": "https://www.cudocompute.com/", "title": "CUDO Compute homepage", "accessed": "2026-11-09" }, { "url": "https://www.cudocompute.com/about-us", "title": "CUDO Compute about us", "accessed": "2026-11-09" }, { "url": "https://newsletter.semianalysis.com/p/clustermax-30-the-industry-standard", "title": "SemiAnalysis ClusterMAX 3.0", "accessed": "2026-11-09" } ], "verifiedAt": "2026-11-09" }, { "id": "denvr-dataworks", "name": "Denvr Dataworks", "url": "https://www.denvr.com", "summary": "Canadian GPU cloud offering on-demand and reserved NVIDIA and Intel Gaudi instances in bare-metal and virtualized configurations, plus on-prem AI infrastructure. Runs liquid-immersion-cooled Modular SuperClusters and is an NVIDIA Cloud Partner.", "founded": 2017, "tags": [ "bare-metal", "gpu-cloud", "infiniband" ], "fields": { "h100_hr": 2.45, "price_asof": "2026-11-09", "interconnect": "InfiniBand, RoCE", "fabric": "InfiniBand", "bare_metal": true, "regions": "Canada (Calgary)", "gpus": "GB200, B200, H200, GH200, H100, A100", "gpu_h100": true, "gpu_blackwell": true, "gpu_amd": false, "h100_basis": "On-demand H100 SXM, 8x with 3200G InfiniBand", "pricing_model": "on-demand (per minute) + reserved (up to 20% off)", "provider_type": "gpu-cloud" }, "pricing": { "model": "on-demand (per minute) + reserved (up to 20% off)", "from": 0.58, "unit": "GPU-hour", "currency": "USD", "note": "From = A100 MIG 20 GB slice. H100 SXM (8x, IB 3200G) $2.45; GH200 $3.87; A100 SXM 80 GB $1.45; Gaudi 2 $1.10. B200, GB200, H200 and Gaudi 3 are reserved only, no public price. ClusterMAX 3.0 (Sep 2026) lists Denvr as 'Unavailable' (not rated)." }, "sources": [ { "url": "https://www.denvr.com/pricing", "title": "Denvr Dataworks Pricing", "accessed": "2026-11-09", "fields": [ "h100_hr", "price_asof", "pricing_model" ] }, { "url": "https://www.denvr.com/", "title": "Denvr Dataworks", "accessed": "2026-11-09" }, { "url": "https://www.denvr.com/about-us", "title": "About Denvr Dataworks", "accessed": "2026-11-09" }, { "url": "https://clustermax.semianalysis.com/img/neocloud-ranking-v3.png", "title": "SemiAnalysis ClusterMAX 3.0 (Sep 2026) rankings image", "accessed": "2026-11-09" } ], "verifiedAt": "2026-11-09" }, { "id": "eurohpc-ju", "name": "EuroHPC JU access", "url": "https://eurohpc-ju.europa.eu/access-our-supercomputers/access-policy-and-faq_en", "summary": "Free allocations on Europe's EuroHPC supercomputers (LUMI, Leonardo, MareNostrum 5, JUPITER and more) through Regular, Extreme Scale, Benchmark and AI Factories calls, with an industry track.", "tags": [ "eu", "public-supercomputer" ], "fields": { "pricing_model": "Free of charge through competitive access calls", "regions": "LUMI, Leonardo, MareNostrum 5, JUPITER, MeluXina, Discoverer, Deucalion, Arrhenius", "eu_hosted": true, "sovereign": "Publicly funded European infrastructure; Regular Access has Scientific, Industry and Public Administration tracks", "provider_type": "public-supercomputer" }, "sources": [ { "url": "https://eurohpc-ju.europa.eu/access-our-supercomputers/access-policy-and-faq_en", "title": "EuroHPC access policy and FAQ", "accessed": "2026-11-11", "fields": [ "pricing_model" ] }, { "url": "https://eurohpc-ju.europa.eu/eurohpc-ju-call-proposals-regular-access-mode_en", "title": "EuroHPC Regular Access call", "accessed": "2026-11-11", "fields": [ "regions", "eu_hosted", "sovereign" ] }, { "url": "https://www.eurohpc-ju.europa.eu/playground-access-ai-factories_en", "title": "AI Factories Playground access", "accessed": "2026-11-11", "fields": [ "sovereign" ] } ], "verifiedAt": "2026-11-11" }, { "id": "gmi-cloud", "name": "GMI Cloud", "url": "https://www.gmicloud.ai", "summary": "GPU cloud selling dedicated NVIDIA GPU clusters, bare metal and serverless inference to AI companies. Headquartered in Mountain View with roots in Taiwan; building a Taiwan AI factory and a planned 1 GW site in Japan.", "hq": "US", "founded": 2021, "tags": [ "bare-metal", "gpu-cloud", "infiniband", "slurm" ], "fields": { "h100_hr": 2.6, "price_asof": "2026-11-09", "interconnect": "InfiniBand", "fabric": "InfiniBand", "bare_metal": true, "slurm": true, "kubernetes": true, "regions": "North America, Europe, Asia-Pacific (company site; specific sites not named); Taiwan AI factory (announced Nov 2025); Kagoshima, Japan site planned (announced Mar 2026)", "gpus": "H100, H200, B200, GB200, GB300", "gpu_h100": true, "gpu_blackwell": true, "gpu_amd": false, "schedulers": "Slurm, Kubernetes", "h100_basis": "\"From\" price for H100", "pricing_model": "on-demand + reserved", "provider_type": "gpu-cloud" }, "pricing": { "model": "on-demand + reserved", "from": 2.6, "unit": "GPU-hour", "currency": "USD", "note": "H100 from $2.60/GPU-hr (SXM/PCIe not stated). H200 from $3.20, B200 from $5.00 (limited availability), GB200 from $8.00; GB300 pre-order. Reserved pricing via sales. Page shows '$' without a currency code." }, "sources": [ { "url": "https://www.gmicloud.ai/pricing", "title": "GMI Cloud pricing", "accessed": "2026-11-09", "fields": [ "h100_hr", "price_asof", "pricing_model" ] }, { "url": "https://www.gmicloud.ai/", "title": "GMI Cloud homepage", "accessed": "2026-11-09" }, { "url": "https://www.gmicloud.ai/en/company/about", "title": "GMI Cloud about", "accessed": "2026-11-09" }, { "url": "https://www.gmicloud.ai/en/blog/gmi-cloud-brings-taiwans-first-ai-factory", "title": "GMI Cloud brings Taiwan's first AI factory", "accessed": "2026-11-09" }, { "url": "https://www.gmicloud.ai/en/blog/gmi-cloud-announces-1gw-sovereign-ai-infrastructure-in-japan-accelerated-by-nvidia-vera-rubin-nvl72-tm", "title": "GMI Cloud announces 1GW sovereign AI infrastructure in Japan", "accessed": "2026-11-09" }, { "url": "https://newsletter.semianalysis.com/p/clustermax-30-the-industry-standard", "title": "SemiAnalysis ClusterMAX 3.0", "accessed": "2026-11-09" }, { "url": "https://pulse2.com/gmi-cloud-raises-668-million-to-expand-global-ai-infrastructure/amp/", "title": "GMI Cloud raises $668 million (Pulse 2.0, Sep 2026)", "accessed": "2026-11-09" } ], "verifiedAt": "2026-11-09" }, { "id": "google-cloud-hpc", "name": "Google Cloud HPC", "url": "https://cloud.google.com/solutions/hpc", "summary": "H3 compute-optimised VMs, A3, A4 and A4X GPU machines and Cluster Toolkit, an open-source tool that deploys Slurm-based HPC clusters.", "hq": "US", "tags": [ "ethernet", "hyperscaler", "slurm" ], "fields": { "fabric": "Ethernet", "interconnect": "H3: default network up to 200 Gbps", "schedulers": "Cluster Toolkit (open source) with Slurm", "slurm": true, "cpus": "H3: 4th Gen Intel Xeon Scalable (Titanium)", "gpus": "A3 Mega and High: H100; A3 Ultra: H200; A4: B200; A4X: GB200 and GB300", "gpu_h100": true, "gpu_blackwell": true, "provider_type": "hyperscaler" }, "sources": [ { "url": "https://docs.cloud.google.com/compute/docs/compute-optimized-machines", "title": "Compute-optimized machine family", "accessed": "2026-11-11", "fields": [ "fabric", "interconnect", "cpus" ] }, { "url": "https://docs.cloud.google.com/cluster-toolkit/docs/overview", "title": "Cluster Toolkit overview", "accessed": "2026-11-11", "fields": [ "schedulers", "slurm" ] }, { "url": "https://docs.cloud.google.com/compute/docs/gpus", "title": "GPU machine types", "accessed": "2026-11-11", "fields": [ "gpus", "gpu_h100", "gpu_blackwell" ] } ], "verifiedAt": "2026-11-11" }, { "id": "hot-aisle", "name": "Hot Aisle", "url": "https://hotaisle.xyz", "summary": "AMD-only GPU cloud selling on-demand MI300X virtual machines (1x, 2x, 4x) and 8x bare-metal nodes billed by the minute, launched through a terminal UI, API or CLI. Also lists MI355X.", "founded": 2023, "tags": [ "amd", "bare-metal", "gpu-cloud" ], "fields": { "price_asof": "2026-11-09", "bare_metal": true, "regions": "US (Michigan)", "gpus": "MI300X, MI355X", "gpu_h100": false, "gpu_blackwell": false, "gpu_amd": true, "pricing_model": "on-demand, billed per minute", "provider_type": "gpu-cloud" }, "pricing": { "model": "on-demand, billed per minute", "from": 2.99, "unit": "GPU-hour", "currency": "USD", "note": "MI300X VM (1x, 2x, 4x) $2.99/GPU-hr; 8x MI300X bare metal $3.39/GPU-hr. No NVIDIA GPUs; MI355X listed without a price. Company says 700+ customers served; founders Jon Stevens and Clint Armstrong; datacenter described as 100% green." }, "sources": [ { "url": "https://hotaisle.xyz/pricing/", "title": "Hot Aisle Pricing", "accessed": "2026-11-09", "fields": [ "price_asof", "pricing_model" ] }, { "url": "https://hotaisle.xyz/", "title": "Hot Aisle", "accessed": "2026-11-09" } ], "verifiedAt": "2026-11-09" }, { "id": "hyperstack", "name": "Hyperstack", "url": "https://www.hyperstack.cloud", "summary": "Self-serve GPU cloud from London-based NexGen Cloud, launched in 2023, selling on-demand, spot and reserved NVIDIA GPUs from A4000 to B300, plus managed Kubernetes and private clusters, with data centers in Europe and North America.", "hq": "GB", "tags": [ "gpu-cloud" ], "fields": { "h100_hr": 3.2, "price_asof": "2026-11-09", "kubernetes": true, "regions": "Europe and North America (sites not named)", "gpus": "B300, B200, GB300, H200, H100, A100, RTX Pro 6000, L40", "gpu_h100": true, "gpu_blackwell": true, "gpu_amd": false, "schedulers": "Kubernetes", "h100_basis": "On-demand H100 SXM (PCIe $2.50, NVLink $2.60)", "pricing_model": "on-demand + spot + reserved", "provider_type": "gpu-cloud" }, "pricing": { "model": "on-demand + spot + reserved", "from": 0.15, "unit": "GPU-hour", "currency": "USD", "note": "From = RTX A4000 on-demand. H100 SXM $3.20, H100 NVLink $2.60, H100 PCIe $2.50, H200 SXM $3.99, B200 $6.00, B300 $7.40 on-demand; reserved from $2.72 (H100 SXM) and $5.10 (B200)." }, "sources": [ { "url": "https://www.hyperstack.cloud/gpu-pricing", "title": "Hyperstack GPU pricing", "accessed": "2026-11-09", "fields": [ "h100_hr", "price_asof", "pricing_model" ] }, { "url": "https://www.hyperstack.cloud/about-us", "title": "Hyperstack about us", "accessed": "2026-11-09" }, { "url": "https://www.hyperstack.cloud/", "title": "Hyperstack homepage", "accessed": "2026-11-09" }, { "url": "https://clustermax.semianalysis.com/img/neocloud-ranking-v3.png", "title": "SemiAnalysis ClusterMAX 3.0 ranking (Sep 2026)", "accessed": "2026-11-09" } ], "verifiedAt": "2026-11-09" }, { "id": "ibm-cloud-hpc", "name": "IBM Cloud", "url": "https://www.ibm.com/high-performance-computing", "summary": "IBM's public cloud for HPC and AI: hourly H100, H200 and Gaudi 3 GPU servers, bare metal, and IBM Spectrum LSF and Symphony clusters for enterprise simulation, risk and AI workloads.", "hq": "US", "tags": [ "hyperscaler", "roce", "bare-metal", "eu", "amd" ], "fields": { "provider_type": "hyperscaler", "fabric": "RoCE", "interconnect": "gx3d-160x1792x8h100: cluster network of 8 x 400 Gbps NICs (3.2 Tbps dedicated) carrying RoCEv2 for RDMA; H100 and H200 cluster networking offered in us-east (wdc06, wdc07) and eu-de (fra04)", "schedulers": "IBM Spectrum LSF, IBM Spectrum Symphony", "cpus": "Intel Xeon 8474C (4th Gen) on gx3/gx3d H100, H200, A100, L40S and L4 profiles; Intel Xeon 8568Y+ (5th Gen) on Gaudi 3 profile; 4th Gen Intel Xeon or AMD CPUs on bare metal", "gpus": "NVIDIA B300, H200, H100 SXM5, A100 PCIe, L40S, L4; AMD Instinct MI300X; Intel Gaudi 3", "gpu_h100": true, "gpu_blackwell": true, "gpu_amd": true, "bare_metal": true, "h100_hr": 10.73, "h100_basis": "On-demand gx3d-160x1792x8h100 virtual server (8x H100 SXM5) at $85.85/instance-hour in us-east and us-south, divided by 8; USD, US-account list price effective 1 Nov 2026. Same profile is $99.60/hr in eu-de; spot is $21.46/hr in us-east", "price_asof": "2026-11-11", "pricing_model": "GPU virtual servers billed hourly on-demand or as spot; bare metal billed hourly, monthly or on reserved 1-year terms", "regions": "H100 profile: Dallas, Washington DC, Toronto, Sao Paulo, Frankfurt, Madrid, Paris, London, Sydney, Tokyo; H200 adds Chennai", "eu_hosted": true }, "pricing": { "model": "on-demand + spot", "from": 10.73, "unit": "GPU-hour", "currency": "USD", "note": "gx3d-160x1792x8h100 (8x H100 SXM5) on-demand $85.85/instance-hour in us-east and us-south, $99.60 in eu-de, divided by 8 for the per-GPU figure. Spot $21.46/instance-hour in us-east and us-south (about $2.68/GPU-hour). US-account list prices from IBM's Global Catalog pricing API, effective 1 Nov 2026; OS licences extra." }, "sources": [ { "url": "https://cloud.ibm.com/docs/vpc?topic=vpc-accelerated-profile-family", "title": "IBM Cloud Docs: Accelerated profile family (GPU profiles)", "accessed": "2026-11-11", "fields": [ "fabric", "interconnect", "cpus", "gpus", "gpu_h100", "regions", "eu_hosted" ] }, { "url": "https://globalcatalog.cloud.ibm.com/api/v1/gx3d-160x1792x8h100?complete=true", "title": "IBM Global Catalog API: gx3d-160x1792x8h100 profile", "accessed": "2026-11-11", "fields": [ "regions", "eu_hosted", "interconnect", "pricing_model" ] }, { "url": "https://globalcatalog.cloud.ibm.com/api/v1/99548819-fcbc-486c-8eeb-0355a56b035e:us-east443a2d16/pricing", "title": "IBM Global Catalog pricing API: gx3d-160x1792x8h100, us-east", "accessed": "2026-11-11", "fields": [ "h100_hr", "h100_basis", "price_asof", "pricing_model" ] }, { "url": "https://globalcatalog.cloud.ibm.com/api/v1/99548819-fcbc-486c-8eeb-0355a56b035e:eu-de76e789ca/pricing", "title": "IBM Global Catalog pricing API: gx3d-160x1792x8h100, eu-de", "accessed": "2026-11-11", "fields": [ "h100_basis" ] }, { "url": "https://www.ibm.com/high-performance-computing", "title": "IBM high-performance computing", "accessed": "2026-11-11", "fields": [ "schedulers", "bare_metal" ] }, { "url": "https://github.com/terraform-ibm-modules/terraform-ibm-hpc", "title": "IBM Cloud HPC deployable architecture (IBM Spectrum schedulers), terraform-ibm-modules", "accessed": "2026-11-11", "fields": [ "schedulers" ] }, { "url": "https://www.ibm.com/products/gpu-ai-accelerator", "title": "IBM Cloud GPUs and AI accelerators", "accessed": "2026-11-11", "fields": [ "gpus", "gpu_blackwell", "gpu_amd" ] }, { "url": "https://www.ibm.com/products/bare-metal-servers", "title": "IBM Cloud Bare Metal Servers", "accessed": "2026-11-11", "fields": [ "bare_metal", "cpus", "pricing_model" ] } ], "verifiedAt": "2026-11-11" }, { "id": "lambda", "name": "Lambda", "url": "https://lambda.com", "summary": "San Francisco GPU cloud (formerly lambda.ai) selling self-serve on-demand NVIDIA instances, reservable InfiniBand clusters (1-Click Clusters) and large private clusters to AI developers and labs. Building new data centers in Oklahoma and Alabama.", "hq": "US", "founded": 2012, "tags": [ "eu", "gpu-cloud", "infiniband", "slurm" ], "fields": { "h100_hr": 4.89, "price_asof": "2026-11-09", "interconnect": "InfiniBand", "fabric": "InfiniBand", "slurm": true, "kubernetes": true, "regions": "On-demand regions in the US (Virginia, Washington DC, Illinois, Texas x3, California, Arizona, Utah), Japan (Tokyo, Osaka), India, Germany, Israel (company docs, Nov 2026); new data centers announced in Mayes County, Oklahoma (Sep 2026) and Lowndes County, Alabama (Nov 2026)", "gpus": "B200, H100, GH200, A100", "gpu_h100": true, "gpu_blackwell": true, "gpu_amd": false, "schedulers": "Slurm, Kubernetes", "h100_basis": "On-demand 1x H100 SXM, price list effective 16 Nov 2026", "pricing_model": "on-demand + reserved", "eu_hosted": true, "provider_type": "gpu-cloud" }, "pricing": { "model": "on-demand + reserved", "from": 4.89, "unit": "GPU-hour", "currency": "USD", "note": "1x H100 SXM on-demand per published price list effective 16 Nov 2026 (H100 SXM $4.59-$4.89 by instance size; 1x row mapped from specs). B200 SXM6 $7.69-$7.99 by instance size, 1x size not clearly identified. Prices before 16 Nov shown only to logged-in users. 1-Click Clusters: H100 from $5.54, B200 from $8.87 per GPU-hour." }, "sources": [ { "url": "https://lambda.com/pricing", "title": "Lambda pricing", "accessed": "2026-11-09", "fields": [ "h100_hr", "price_asof", "pricing_model" ] }, { "url": "https://lambda.com/1-click-clusters", "title": "Lambda 1-Click Clusters", "accessed": "2026-11-09" }, { "url": "https://docs.lambda.com/public-cloud/on-demand/", "title": "Lambda on-demand instances docs", "accessed": "2026-11-09", "fields": [ "regions" ] }, { "url": "https://lambda.com/about", "title": "About Lambda", "accessed": "2026-11-09" }, { "url": "https://lambda.com/blog/lambda-closes-1-billion-senior-secured-fixed-rate-financing", "title": "Lambda closes $1 billion senior secured fixed rate financing", "accessed": "2026-11-09" }, { "url": "https://lambda.com/blog/lambda-closes-926-million-senior-secured-term-loan-b-facility", "title": "Lambda closes $926 million senior secured term loan B facility", "accessed": "2026-11-09" }, { "url": "https://lambda.com/blog/lambda-to-invest-33-billion-in-alabama-with-new-data-center", "title": "Lambda to build new data center in Lowndes County, Alabama", "accessed": "2026-11-09" }, { "url": "https://lambda.com/blog/new-lambda-data-center-in-oklahoma", "title": "Lambda to build new data center in Mayes County, Oklahoma", "accessed": "2026-11-09" }, { "url": "https://newsletter.semianalysis.com/p/clustermax-30-the-industry-standard", "title": "ClusterMAX 3.0 (SemiAnalysis)", "accessed": "2026-11-09" } ], "verifiedAt": "2026-11-09" }, { "id": "lumi", "name": "LUMI (CSC)", "url": "https://lumi-supercomputer.eu/", "summary": "Finland's EuroHPC pre-exascale system: LUMI-G with AMD Instinct MI250X GPUs and LUMI-C on AMD EPYC Milan, linked by HPE Slingshot 11. Companies buy time through CSC at EUR 0.535 per GPU-hour.", "hq": "FI", "tags": [ "amd", "eu", "public-supercomputer", "slingshot", "slurm" ], "fields": { "fabric": "Slingshot", "interconnect": "HPE Slingshot 11", "schedulers": "Slurm (partitions and quickstart in the LUMI docs)", "slurm": true, "cpus": "LUMI-C: AMD EPYC Milan", "gpus": "LUMI-G: AMD Instinct MI250X (a GPU-hour is a full MI250X module, 2 GCDs)", "gpu_amd": true, "gpu_h100": false, "pricing_model": "CSC company price list excl. VAT: EUR 0.535 per GPU-hour, EUR 0.57 per CPU node-hour, EUR 1,000 one-time project setup. Free AI SME access through AI Factories calls", "eu_hosted": true, "provider_type": "public-supercomputer" }, "pricing": { "model": "Pay per use (CSC price list)", "from": 0.535, "unit": "GPU-hour (LUMI-G, excl. VAT)", "currency": "EUR", "note": "Company route; Finnish companies get 20% of Finland's national capacity. Non-Finnish company eligibility not stated on the fetched pages." }, "sources": [ { "url": "https://docs.lumi-supercomputer.eu/hardware/", "title": "LUMI hardware", "accessed": "2026-11-11", "fields": [ "fabric", "interconnect", "schedulers", "slurm", "cpus", "gpus", "gpu_amd" ] }, { "url": "https://csc.fi/en/customers/high-performance-computing-for-companies/", "title": "CSC HPC for companies", "accessed": "2026-11-11", "fields": [ "pricing_model" ] }, { "url": "https://docs.lumi-supercomputer.eu/runjobs/lumi_env/billing/", "title": "LUMI billing", "accessed": "2026-11-11", "fields": [ "gpus" ] }, { "url": "https://lumi-supercomputer.eu/access-calls-for-ai-factories/", "title": "LUMI AI Factories access calls", "accessed": "2026-11-11", "fields": [ "pricing_model" ] } ], "verifiedAt": "2026-11-11" }, { "id": "nebius", "name": "Nebius", "url": "https://nebius.com", "summary": "Amsterdam-based, Nasdaq-listed AI cloud selling NVIDIA GPU capacity, managed Kubernetes and inference services to AI companies and enterprises. Publishes per-GPU on-demand prices and live spot prices, with regions in Europe, the US, the UK and Israel.", "hq": "NL", "tags": [ "eu", "gpu-cloud", "infiniband" ], "fields": { "h100_hr": 4.5, "price_asof": "2026-11-09", "interconnect": "InfiniBand", "fabric": "InfiniBand", "kubernetes": true, "regions": "Finland, France (2), Spain (Madrid), UK (2), Israel, US (Kansas City MO; Woodbury MN); private region in Iceland (company docs, Nov 2026)", "gpus": "GB300, B300, GB200, B200, H200, H100, RTX PRO 6000, L40S", "gpu_h100": true, "gpu_blackwell": true, "gpu_amd": false, "schedulers": "Kubernetes", "h100_basis": "On-demand HGX H100", "pricing_model": "on-demand + reserved + spot", "eu_hosted": true, "provider_type": "gpu-cloud" }, "pricing": { "model": "on-demand + reserved + spot", "from": 4.5, "unit": "GPU-hour", "currency": "USD", "note": "NVIDIA HGX H100 on-demand per GPU-hour; HGX B200 $8.50, HGX B300 $9.50; GB200/GB300 NVL72 by quote" }, "sources": [ { "url": "https://nebius.com/prices", "title": "Nebius AI Cloud prices", "accessed": "2026-11-09", "fields": [ "h100_hr", "price_asof" ] }, { "url": "https://docs.nebius.com/overview/regions", "title": "Nebius AI Cloud regions", "accessed": "2026-11-09", "fields": [ "regions" ] }, { "url": "https://nebius.com/about", "title": "About Nebius", "accessed": "2026-11-09" }, { "url": "https://nebius.com/ai-cloud", "title": "Nebius AI Cloud", "accessed": "2026-11-09" }, { "url": "https://nebius.com/services/managed-kubernetes", "title": "Nebius Managed Kubernetes", "accessed": "2026-11-09" }, { "url": "https://nebius.com/newsroom/nebius-secures-3-5-billion-financing-facility-with-citi", "title": "Nebius secures $3.5 billion financing facility with Citi", "accessed": "2026-11-09" }, { "url": "https://nebius.com/newsroom/nebius-brings-compute-pricing-into-the-open-as-europe-s-ai-builders-scale-on-its-cloud", "title": "Nebius brings compute pricing into the open", "accessed": "2026-11-09", "fields": [ "h100_hr", "price_asof", "pricing_model" ] }, { "url": "https://nebius.com/newsroom", "title": "Nebius newsroom (convertible notes Aug 2026)", "accessed": "2026-11-09" }, { "url": "https://newsletter.semianalysis.com/p/clustermax-30-the-industry-standard", "title": "ClusterMAX 3.0 (SemiAnalysis)", "accessed": "2026-11-09" } ], "verifiedAt": "2026-11-09" }, { "id": "nscale", "name": "Nscale", "url": "https://www.nscale.com", "summary": "UK-based AI cloud combining GPU compute, its own data centers and power, selling large reserved clusters to hyperscalers and AI labs plus managed Kubernetes, Slurm, bare metal and serverless inference. Filed for a NYSE IPO (proposed ticker NSCL) in Sep 2026 and acquired Anyscale in Jul 2026.", "hq": "GB", "tags": [ "bare-metal", "eu", "gpu-cloud", "slurm" ], "fields": { "bare_metal": true, "slurm": true, "kubernetes": true, "regions": "Norway (Glomfjord, Narvik, Stavanger, Oslo), UK (Loughton, Hayes), Portugal (Sines), Iceland (Keflavik, Blonduos), US (Texas, North Carolina, West Virginia); mix of Nscale-operated and partner sites", "gpus": "GB300", "gpu_h100": false, "gpu_blackwell": true, "gpu_amd": false, "schedulers": "Slurm, Kubernetes", "eu_hosted": true, "provider_type": "gpu-cloud" }, "sources": [ { "url": "https://www.nscale.com/", "title": "Nscale homepage", "accessed": "2026-11-09" }, { "url": "https://www.nscale.com/about", "title": "About Nscale", "accessed": "2026-11-09" }, { "url": "https://www.nscale.com/services/platform-services", "title": "Nscale platform services", "accessed": "2026-11-09" }, { "url": "https://docs.nscale.com/docs/getting-started/overview", "title": "Nscale docs overview", "accessed": "2026-11-09" }, { "url": "https://www.nscale.com/press-releases/nscale-files-initial-public-offering", "title": "Nscale files for initial public offering", "accessed": "2026-11-09" }, { "url": "https://www.nscale.com/press-releases/pre-ipo-convertible-financing", "title": "Nscale $3.36bn pre-IPO convertible financing", "accessed": "2026-11-09" }, { "url": "https://www.nscale.com/press-releases/nscale-series-c", "title": "Nscale Series C", "accessed": "2026-11-09" }, { "url": "https://www.nscale.com/press-releases/execution-at-scale", "title": "Nscale execution at scale update", "accessed": "2026-11-09" }, { "url": "https://www.nscale.com/press-releases/nscale-microsoft-2025", "title": "Nscale and Microsoft GB300 agreement", "accessed": "2026-11-09" }, { "url": "https://www.nscale.com/press-releases/nscale-acquires-american-intelligence-power-corporation", "title": "Nscale acquires American Intelligence & Power Corporation", "accessed": "2026-11-09" }, { "url": "https://www.nscale.com/newsroom", "title": "Nscale newsroom", "accessed": "2026-11-09" }, { "url": "https://clustermax.semianalysis.com/v3", "title": "SemiAnalysis ClusterMAX 3.0 (Sep 2026)", "accessed": "2026-11-09" } ], "verifiedAt": "2026-11-09" }, { "id": "nsf-access", "name": "NSF ACCESS", "url": "https://allocations.access-ci.org/", "summary": "Free US allocations on academic supercomputers such as NCSA Delta (Slingshot, A100, A40, H200 and MI100 GPUs) for researchers and educators; no NSF award required.", "hq": "US", "tags": [ "amd", "public-supercomputer", "slingshot" ], "fields": { "fabric": "Slingshot", "interconnect": "Delta: 200 Gb/s HPE Slingshot", "gpus": "Delta: NVIDIA A100 (40 GB), A40, H200 (141 GB); AMD MI100", "gpu_amd": true, "pricing_model": "Free for researchers and educators; entry-level allocation needs only a project description", "regions": "Delta at NCSA, Illinois", "provider_type": "public-supercomputer" }, "sources": [ { "url": "https://delta.ncsa.illinois.edu/", "title": "NCSA Delta", "accessed": "2026-11-11", "fields": [ "fabric", "interconnect", "regions" ] }, { "url": "https://www.ncsa.illinois.edu/research/project-support/delta/", "title": "NCSA Delta hardware", "accessed": "2026-11-11", "fields": [ "gpus", "gpu_amd" ] }, { "url": "https://allocations.access-ci.org/", "title": "ACCESS allocations", "accessed": "2026-11-11", "fields": [ "pricing_model" ] }, { "url": "https://delta.ncsa.illinois.edu/delta-allocations/", "title": "Delta allocations", "accessed": "2026-11-11", "fields": [ "pricing_model" ] } ], "verifiedAt": "2026-11-11" }, { "id": "oracle-cloud-hpc", "name": "Oracle Cloud HPC", "url": "https://www.oracle.com/cloud/compute/hpc/", "summary": "OCI bare-metal HPC and GPU shapes with RDMA cluster networking: BM.Optimized3.36 for CPU codes and BM.GPU.H100.8 with eight H100 GPUs.", "hq": "US", "tags": [ "bare-metal", "hyperscaler", "rdma" ], "fields": { "fabric": "RDMA", "interconnect": "Cluster networking: 8 x 2 x 200 Gbps RDMA on BM.GPU.H100.8; 1 x 100 Gbps RDMA on BM.Optimized3.36", "cpus": "BM.Optimized3.36: Intel Xeon 6354, 36 cores; BM.GPU.H100.8: 2x Intel Sapphire Rapids 8480+, 112 cores", "gpus": "8x NVIDIA H100 80 GB (BM.GPU.H100.8)", "gpu_h100": true, "bare_metal": true, "provider_type": "hyperscaler", "h100_hr": 10, "h100_basis": "Pay-as-you-go list price for OCI Compute GPU H100 (part B98415), per GPU-hour, sold as the 8-GPU BM.GPU.H100.8 bare-metal shape. H200 is also $10.00 per GPU-hour.", "price_asof": "2026-11-11", "pricing_model": "Pay-as-you-go per GPU-hour or per OCPU-hour" }, "pricing": { "model": "Pay-as-you-go", "from": 10, "unit": "GPU-hour (H100)", "currency": "USD", "note": "OCI publishes H100 and H200 at $10.00 per GPU-hour pay-as-you-go; NVIDIA AI Enterprise adds $2.50 per GPU-hour if you need it." }, "sources": [ { "url": "https://docs.oracle.com/en-us/iaas/Content/Compute/References/computeshapes.htm", "title": "OCI compute shapes", "accessed": "2026-11-11", "fields": [ "fabric", "interconnect", "cpus", "gpus", "gpu_h100", "bare_metal" ] }, { "url": "https://www.oracle.com/cloud/compute/pricing/", "title": "OCI Compute pricing", "accessed": "2026-11-11", "fields": [ "pricing_model" ] }, { "url": "https://apexapps.oracle.com/pls/apex/cetools/api/v1/products/?currencyCode=USD", "title": "OCI price list API (USD)", "accessed": "2026-11-11", "fields": [ "h100_hr", "h100_basis", "price_asof" ] } ], "verifiedAt": "2026-11-11" }, { "id": "ovhcloud", "name": "OVHcloud", "url": "https://www.ovhcloud.com", "summary": "Listed French cloud and hosting provider whose Public Cloud sells hourly NVIDIA GPU instances (H200, H100, A100, L40S, L4 and older) alongside servers, storage and managed Kubernetes, with a European data sovereignty focus.", "hq": "FR", "founded": 1999, "tags": [ "eu", "sovereign" ], "fields": { "h100_hr": 2.99, "price_asof": "2026-11-09", "kubernetes": true, "regions": "GPU instances in Gravelines (France) and Beauharnois/Montreal (Canada) per regions page; 46 data centers overall", "gpus": "H200, H100, A100, L40S, L4", "gpu_h100": true, "gpu_blackwell": false, "gpu_amd": false, "schedulers": "Kubernetes", "h100_basis": "On-demand 1x H100 80 GB, excl. VAT", "pricing_model": "on-demand hourly + monthly / savings plans", "eu_hosted": true, "sovereign": "SecNumCloud qualification for the SNC Cloud Platform (Sept 2026); public-cloud GPU SecNumCloud status not confirmed", "provider_type": "sovereign" }, "pricing": { "model": "on-demand hourly + monthly / savings plans", "from": 1, "unit": "GPU-hour", "currency": "USD", "note": "From = L4 (l4-90) or A10 at $1.00/hr. H100 price is h100-380, 1x H100 80GB, $2.99/hr excl. VAT (form factor not stated on the page); 8x H200 (h200-1920) $49.56/hr = $6.20/GPU-hr; L40S $1.80/hr; A100 $3.07/hr. USD prices from the international English price page." }, "sources": [ { "url": "https://www.ovhcloud.com/en/public-cloud/prices/", "title": "OVHcloud Public Cloud prices", "accessed": "2026-11-09", "fields": [ "h100_hr", "price_asof" ] }, { "url": "https://www.ovhcloud.com/en/public-cloud/regions-availability/", "title": "OVHcloud Public Cloud regions", "accessed": "2026-11-09", "fields": [ "regions" ] }, { "url": "https://corporate.ovhcloud.com/en/investor-relations/", "title": "OVHcloud investor relations", "accessed": "2026-11-09" }, { "url": "https://corporate.ovhcloud.com/en/company/", "title": "OVHcloud company", "accessed": "2026-11-09" }, { "url": "https://clustermax.semianalysis.com/img/neocloud-ranking-v3.png", "title": "SemiAnalysis ClusterMAX 3.0 ranking (Sep 2026)", "accessed": "2026-11-09" }, { "url": "https://www.ovhcloud.com/en/compliance/secnumcloud/", "title": "OVHcloud SecNumCloud", "accessed": "2026-11-11", "fields": [ "sovereign" ] }, { "url": "https://corporate.ovhcloud.com/en/company/history/", "title": "OVHcloud history", "accessed": "2026-11-11" } ], "verifiedAt": "2026-11-09" }, { "id": "penguin-on-demand", "name": "Penguin On Demand (POD)", "url": "https://www.penguinsolutions.com/pod", "summary": "Penguin Solutions' single-tenant HPC cloud: non-virtualised bare-metal nodes in a US Tier III data centre, run with its own ClusterWare cluster management software.", "hq": "US", "tags": [ "bare-metal", "managed-hpc" ], "fields": { "schedulers": "ClusterWare cluster management software (no job scheduler named on the page)", "bare_metal": true, "regions": "One US Tier III data centre (city not named)", "interconnect": "\"Low-latency networks\"; InfiniBand named only in the DGX-Ready managed services section", "provider_type": "managed-hpc" }, "sources": [ { "url": "https://www.penguinsolutions.com/pod", "title": "Penguin On Demand", "accessed": "2026-11-11", "fields": [ "schedulers", "bare_metal", "regions", "interconnect" ] }, { "url": "https://www.penguinsolutions.com/?p=8182", "title": "Penguin DGX-Ready managed services", "accessed": "2026-11-11", "fields": [ "interconnect" ] } ], "verifiedAt": "2026-11-11" }, { "id": "rescale", "name": "Rescale", "url": "https://rescale.com", "summary": "Cloud HPC platform for engineering simulation and AI: a catalogue of 1,250+ software packages scheduled across AWS, Azure, Google Cloud, Oracle and CoreWeave capacity, sold by quote.", "hq": "US", "tags": [ "simulation-cloud" ], "fields": { "software": "1,250+ HPC and AI software packages (platform page)", "underlying_clouds": "AWS, Microsoft Azure, Google Cloud, Oracle, CoreWeave, AMD, Intel and NVIDIA (platform logos and partner pages)", "pricing_model": "Quote only: Enterprise and Government editions; On-Demand Priority, On-Demand Economy and Commitment Plans; no rates published", "provider_type": "simulation-cloud" }, "sources": [ { "url": "https://rescale.com/platform/", "title": "Rescale platform", "accessed": "2026-11-11", "fields": [ "underlying_clouds", "software" ] }, { "url": "https://rescale.com/pricing/", "title": "Rescale pricing", "accessed": "2026-11-11", "fields": [ "pricing_model" ] } ], "verifiedAt": "2026-11-11" }, { "id": "sabalcore", "name": "Sabalcore", "url": "https://www.sabalcore.com", "summary": "Orlando-based bare-metal HPC cloud for engineers and scientists running tightly coupled MPI jobs up to 768 cores, on InfiniBand-connected Intel Xeon nodes. Pay per core-hour with HPC engineers onboarding every customer.", "hq": "US", "tags": [ "bare-metal", "infiniband", "managed-hpc" ], "fields": { "provider_type": "managed-hpc", "fabric": "InfiniBand", "interconnect": "NVIDIA/Mellanox InfiniBand up to 200 Gbps (HDR) between compute nodes; parallel file system on a 100 Gbps InfiniBand network", "cpus": "Intel Xeon through Xeon Scalable, up to 3.3 GHz, bare metal (Xeon 6 announced as coming soon); up to 16 GB RAM per core, 384 GB per node", "bare_metal": true, "pricing_model": "Pay-as-you-go at $0.14 per bare-metal core-hour, billed to the second with no minimums; prepaid volume rates from $0.10 per core-hour; academic discount; storage $0.10/GB-month after 100 GB free; free data transfer and login nodes; 1,000 free core-hours to start" }, "pricing": { "model": "pay-as-you-go + prepaid volume", "from": 0.14, "unit": "core-hour", "currency": "USD", "note": "$0.14 per bare-metal CPU core-hour pay-as-you-go; prepaid volume rates as low as $0.10/core-hour. The billing FAQ on the same page works an example at $0.12/core-hour; the headline rate is $0.14. No GPU nodes listed." }, "sources": [ { "url": "https://www.sabalcore.com/", "title": "Sabalcore: HPC High Performance Computing in the Cloud", "accessed": "2026-11-11", "fields": [ "fabric", "interconnect", "cpus", "bare_metal", "pricing_model" ] }, { "url": "https://www.sabalcore.com/hpc-services/pricing/", "title": "Sabalcore HPC cloud pricing", "accessed": "2026-11-11", "fields": [ "interconnect", "cpus", "bare_metal", "pricing_model" ] } ], "verifiedAt": "2026-11-11" }, { "id": "scaleway", "name": "Scaleway", "url": "https://www.scaleway.com", "summary": "French cloud provider selling per-minute GPU instances (L4, L40S, H100 PCIe/SXM, B300), GPU bare metal and dedicated inference alongside general cloud services, aimed at European developers and AI companies.", "tags": [ "bare-metal", "eu", "sovereign" ], "fields": { "price_asof": "2026-11-09", "bare_metal": true, "regions": "EU (Paris fr-par-1/fr-par-2, Warsaw pl-waw-2)", "gpus": "H100, B300, L40S, L4", "gpu_h100": true, "gpu_blackwell": true, "gpu_amd": false, "pricing_model": "On-demand, billed per minute; H100 PCIe EUR 2.87 per hour, H100 SXM 8x EUR 25.33 per hour (EUR 3.17 per GPU-hour)", "eu_hosted": true, "provider_type": "sovereign" }, "pricing": { "model": "on-demand (billed per minute)", "from": 0.79, "unit": "GPU-hour", "currency": "EUR", "note": "From = L4-1-24G at EUR 0.013125/min. H100 PCIe (H100-1-80G) EUR 2.87/hr; H100 SXM (H100-SXM-8-80G) EUR 25.33/hr for 8 GPUs = EUR 3.17/GPU-hr; L40S EUR 1.47/GPU-hr; B300 (B300-SXM-8-288G, Paris) EUR 60/hr for 8 GPUs = EUR 7.50/GPU-hr. Prices from Scaleway's public product-catalog API, excl. VAT; the pricing web page returned 403. No USD prices. ClusterMAX 3.0 (Sep 2026) lists Scaleway as 'Unavailable'." }, "sources": [ { "url": "https://api.scaleway.com/product-catalog/v2alpha1/public-catalog/products", "title": "Scaleway public product catalog API", "accessed": "2026-11-09", "fields": [ "price_asof" ] }, { "url": "https://www.scaleway.com/en/pricing/gpu/", "title": "Scaleway GPU pricing (HTTP 403 to automated fetch)", "accessed": "2026-11-09", "fields": [ "price_asof", "pricing_model" ] }, { "url": "https://clustermax.semianalysis.com/img/neocloud-ranking-v3.png", "title": "SemiAnalysis ClusterMAX 3.0 ranking (Sep 2026)", "accessed": "2026-11-09" } ], "verifiedAt": "2026-11-09" }, { "id": "stackit", "name": "STACKIT", "url": "https://www.stackit.com/en/products/compute-engine/stackit-compute-engine-gpu", "summary": "Schwarz Group's European sovereign cloud: H100 HGX, H100 NVL, A100 and L40S GPU servers billed pay-as-you-go in EUR.", "hq": "DE", "tags": [ "eu", "sovereign" ], "fields": { "gpus": "NVIDIA H100 HGX 80 GB (up to 8 per VM), H100 NVL 94 GB, A100 PCIe 80 GB, L40S 48 GB", "gpu_h100": true, "pricing_model": "Pay-as-you-go, EUR excl. VAT. 8x H100 HGX server EUR 64.33 per hour (EUR 8.04 per GPU-hour); L40S from EUR 1.74 per GPU-hour", "regions": "European data centres (EU01 server naming)", "eu_hosted": true, "sovereign": "European ownership; the company states it is not subject to the US CLOUD Act; BSI C5 Type 2", "provider_type": "sovereign" }, "pricing": { "model": "Pay-as-you-go", "from": 1.74, "unit": "GPU-hour (L40S, excl. VAT)", "currency": "EUR", "note": "H100 HGX 8-GPU server EUR 64.33 per hour = EUR 8.04 per GPU-hour." }, "sources": [ { "url": "https://www.stackit.com/en/products/compute-engine/stackit-compute-engine-gpu", "title": "STACKIT Compute Engine GPU", "accessed": "2026-11-11", "fields": [ "gpus", "gpu_h100", "pricing_model" ] }, { "url": "https://www.stackit.com/en/products/compute-engine/stackit-compute-engine-gpu/gpu-models-in-comparison", "title": "GPU models in comparison", "accessed": "2026-11-11", "fields": [ "gpus" ] }, { "url": "https://www.stackit.com/en/why-stackit/benefits/hyperscaler", "title": "Why STACKIT", "accessed": "2026-11-11", "fields": [ "regions", "eu_hosted", "sovereign" ] } ], "verifiedAt": "2026-11-11" }, { "id": "t-cloud-public", "name": "T Cloud Public (Open Telekom Cloud)", "url": "https://www.t-cloud-public.com/en/solutions/high-performance-computing", "summary": "Deutsche Telekom / T-Systems sovereign cloud with HPC cluster options, bare-metal nodes and InfiniBand-backed Lustre storage, billed per core-hour. Open Telekom Cloud until January 2026.", "hq": "DE", "tags": [ "bare-metal", "eu", "infiniband", "slurm", "sovereign" ], "fields": { "fabric": "InfiniBand", "interconnect": "InfiniBand for parallel file systems (Lustre); compute fabric not stated", "schedulers": "Slurm, Moab/NODUS cloud bursting, UNIVA, SGE", "slurm": true, "gpus": "NVIDIA P100, V100, T4 (older generations; no H100 listed)", "bare_metal": true, "pricing_model": "Pay-per-use; HPC compute from EUR 0.015 per core-hour", "regions": "Germany (Biere, Magdeburg), Netherlands, Switzerland", "eu_hosted": true, "sovereign": "Sovereign cloud from Germany; BSI C5, ISO 27001, TISAX, TCDP 1.0", "provider_type": "sovereign" }, "pricing": { "model": "Pay-per-use", "from": 0.015, "unit": "core-hour", "currency": "EUR", "note": "\"From\" price, operating costs included, per the HPC solution page." }, "sources": [ { "url": "https://www.t-cloud-public.com/en/solutions/high-performance-computing", "title": "T Cloud Public HPC", "accessed": "2026-11-11", "fields": [ "fabric", "interconnect", "schedulers", "slurm", "gpus", "bare_metal", "pricing_model" ] }, { "url": "https://www.t-cloud-public.com/en/", "title": "T Cloud Public", "accessed": "2026-11-11", "fields": [ "regions", "eu_hosted", "sovereign" ] } ], "verifiedAt": "2026-11-11" }, { "id": "taiga-cloud", "name": "Taiga Cloud (Northern Data)", "url": "https://northerndata.de/taiga-cloud-by-northern-data-group", "summary": "European sovereign GPU cloud (H100/H200 with InfiniBand, managed Kubernetes and Slurm) run by Frankfurt-based Northern Data Group, which was acquired by Rumble via an exchange offer in 2026 and now operates as part of Rumble's Quake AI (formerly Rumble Cloud).", "hq": "DE", "tags": [ "bare-metal", "eu", "gpu-cloud", "infiniband", "slurm" ], "fields": { "interconnect": "400 Gb InfiniBand at the Boden (Sweden) and Lefdal (Norway) sites; H100 SXM and H200 SXM; managed Kubernetes and SLURM", "fabric": "InfiniBand", "bare_metal": true, "slurm": true, "kubernetes": true, "regions": "UK (Chester, London), Portugal (Sines), Sweden (Boden), Norway (Kristiansand, Lefdal)", "gpus": "NVIDIA H100 SXM and H200 SXM", "gpu_h100": true, "gpu_blackwell": false, "gpu_amd": false, "schedulers": "Slurm, Kubernetes", "eu_hosted": true, "provider_type": "gpu-cloud" }, "sources": [ { "url": "https://northerndata.de/taiga-cloud-by-northern-data-group", "title": "Taiga Cloud by Northern Data Group", "accessed": "2026-11-09" }, { "url": "https://northerndata.de/home", "title": "Northern Data Group homepage", "accessed": "2026-11-09" }, { "url": "https://northerndata.de/en/investor-relations/news/delisting-of-northern-data-ag-shares-munich-stock-exchange-decides-on-the-timeline?hsLang=en", "title": "Delisting of Northern Data AG shares - Munich Stock Exchange decides on the timeline (25 Jun 2026)", "accessed": "2026-11-09" }, { "url": "https://quake.ai/", "title": "Quake AI (formerly Rumble Cloud)", "accessed": "2026-11-09" }, { "url": "https://northerndata.de/taiga-cloud-by-northern-data-group", "title": "Taiga Cloud by Northern Data Group", "accessed": "2026-11-11", "fields": [ "interconnect", "slurm", "kubernetes", "bare_metal", "regions" ] } ], "verifiedAt": "2026-11-09" }, { "id": "tensorwave", "name": "TensorWave", "url": "https://tensorwave.com", "summary": "AMD-only GPU cloud offering Instinct MI300X, MI325X, MI355X and MI455X as bare metal or with managed Kubernetes and Slurm, for AI training and inference. Backed by AMD Ventures.", "hq": "US", "tags": [ "amd", "bare-metal", "gpu-cloud", "roce", "slurm" ], "fields": { "price_asof": "2026-11-09", "interconnect": "RoCE", "fabric": "RoCE", "bare_metal": true, "slurm": true, "kubernetes": true, "regions": "North America (sites not listed on official site)", "gpus": "MI300X, MI325X, MI355X, MI455X", "gpu_h100": false, "gpu_blackwell": false, "gpu_amd": true, "schedulers": "Slurm, Kubernetes", "pricing_model": "per GPU-hour", "provider_type": "gpu-cloud" }, "pricing": { "model": "per GPU-hour", "from": 2.25, "unit": "GPU-hour", "currency": "USD", "note": "AMD MI325X 'starting at' $2.25/GPU-hour; MI355X from $2.95/GPU-hour; on-demand vs reserved not specified; no NVIDIA GPUs" }, "sources": [ { "url": "https://tensorwave.com/", "title": "TensorWave homepage", "accessed": "2026-11-09" }, { "url": "https://tensorwave.com/about", "title": "About TensorWave", "accessed": "2026-11-09" }, { "url": "https://tensorwave.com/products/accelerators/amd-mi325x", "title": "TensorWave AMD MI325X", "accessed": "2026-11-09" }, { "url": "https://tensorwave.com/products/accelerators/amd-mi355x", "title": "TensorWave AMD MI355X", "accessed": "2026-11-09" }, { "url": "https://www.magnetar.com/insights/tensorwave-raises-350-million-series-b-at-155b-valuation", "title": "TensorWave raises $350 million Series B at $1.55B valuation (Magnetar)", "accessed": "2026-11-09" }, { "url": "https://clustermax.semianalysis.com/v3", "title": "SemiAnalysis ClusterMAX 3.0 (Sep 2026)", "accessed": "2026-11-09" } ], "verifiedAt": "2026-11-09" }, { "id": "verda", "name": "Verda", "url": "https://verda.com", "summary": "Helsinki-based self-serve AI cloud selling on-demand and spot GPU instances, instant InfiniBand clusters, serverless containers and managed inference. Renamed from DataCrunch in 2025; EU-owned with EU data centers and a 100% renewable energy claim.", "hq": "FI", "founded": 2020, "tags": [ "eu", "gpu-cloud", "infiniband", "slurm" ], "fields": { "h100_hr": 5.087, "price_asof": "2026-11-11", "interconnect": "InfiniBand", "fabric": "InfiniBand", "slurm": true, "kubernetes": true, "regions": "EU (company site states EU data centers; sites not named on site). Offices in Helsinki, London, San Francisco.", "gpus": "GB300, B300, B200, H200, H100, A100, L40S, RTX PRO 6000", "gpu_h100": true, "gpu_blackwell": true, "gpu_amd": false, "schedulers": "Slurm, Kubernetes", "h100_basis": "On-demand 1x H100 SXM5 80 GB (spot $2.544)", "pricing_model": "on-demand + spot + reserved", "eu_hosted": true, "provider_type": "gpu-cloud" }, "pricing": { "model": "on-demand + spot + reserved", "from": 0.244, "unit": "GPU-hour", "currency": "USD", "note": "'From' is Tesla V100 16GB on-demand. 1x H100 SXM5 80GB $5.087/hr (spot $2.544); 1x H200 SXM5 $6.504; 1x B200 SXM6 $9.229; B300 $12.05; GB300 $13.86. Reserved discounts 2%-25% (1 month to 2 years). USD and EUR toggles on page." }, "sources": [ { "url": "https://verda.com/pricing", "title": "Verda pricing", "accessed": "2026-11-11", "fields": [ "h100_hr", "price_asof", "pricing_model" ] }, { "url": "https://verda.com/", "title": "Verda homepage", "accessed": "2026-11-09" }, { "url": "https://verda.com/company", "title": "Verda company", "accessed": "2026-11-09" }, { "url": "https://verda.com/newsroom", "title": "Verda newsroom", "accessed": "2026-11-09" }, { "url": "https://newsletter.semianalysis.com/p/clustermax-30-the-industry-standard", "title": "SemiAnalysis ClusterMAX 3.0", "accessed": "2026-11-09" } ], "verifiedAt": "2026-11-11" }, { "id": "voltage-park", "name": "Voltage Park", "url": "https://www.voltagepark.com", "summary": "GPU cloud renting NVIDIA HGX H100, H200 and Blackwell capacity on demand or under reserved contracts, with InfiniBand clusters, bare metal and managed Kubernetes. Merged with Lightning AI in January 2026; the Voltage Park brand and products continue under the combined Lightning AI.", "tags": [ "bare-metal", "gpu-cloud", "infiniband" ], "fields": { "h100_hr": 1.99, "price_asof": "2026-11-09", "interconnect": "InfiniBand, Ethernet", "fabric": "InfiniBand", "bare_metal": true, "kubernetes": true, "regions": "US; six Tier 3+ data centers across four states (company about page)", "gpus": "H100, H200, B200, GB300", "gpu_h100": true, "gpu_blackwell": true, "gpu_amd": false, "schedulers": "Kubernetes", "h100_basis": "\"Starting at\" on-demand price, no contract", "pricing_model": "on-demand + reserved", "provider_type": "gpu-cloud" }, "pricing": { "model": "on-demand + reserved", "from": 1.99, "unit": "GPU-hour", "currency": "USD", "note": "H100 (HGX H100, i.e. SXM) 'starting at $1.99' per GPU-hour without contract; Ethernet and InfiniBand on-demand configurations and B200/GB300 are contact-for-pricing" }, "sources": [ { "url": "https://www.voltagepark.com/", "title": "Voltage Park homepage", "accessed": "2026-11-09" }, { "url": "https://www.voltagepark.com/pricing?tier=on-demand", "title": "Voltage Park pricing", "accessed": "2026-11-09", "fields": [ "h100_hr", "price_asof", "pricing_model" ] }, { "url": "https://www.voltagepark.com/infrastructure", "title": "Voltage Park infrastructure", "accessed": "2026-11-09" }, { "url": "https://www.voltagepark.com/blog/lightning-ai-voltage-park-merger-to-create-the-first-cloud-built-for-ai", "title": "Lightning AI and Voltage Park merger", "accessed": "2026-11-09" }, { "url": "https://clustermax.semianalysis.com/v3", "title": "SemiAnalysis ClusterMAX 3.0 (Sep 2026)", "accessed": "2026-11-09" }, { "url": "https://www.voltagepark.com/about", "title": "Voltage Park about", "accessed": "2026-11-09", "fields": [ "regions" ] } ], "verifiedAt": "2026-11-09" }, { "id": "vultr", "name": "Vultr", "url": "https://www.vultr.com", "summary": "General-purpose cloud provider with 34 regions that sells NVIDIA and AMD GPUs as cloud VMs and bare metal, plus managed Kubernetes. Opened a 20 MW AMD MI355X cluster in Chennai, India in Oct 2026.", "hq": "US", "founded": 2014, "tags": [ "amd", "bare-metal", "gpu-cloud", "slurm" ], "fields": { "h100_hr": 2.99, "price_asof": "2026-11-09", "bare_metal": true, "slurm": true, "kubernetes": true, "regions": "34 cloud data center regions worldwide, latest Chennai, India (Oct 2026, company blog)", "gpus": "B200, H100, GH200, A100, L40S, MI355X, MI325X, MI300X", "gpu_h100": true, "gpu_blackwell": true, "gpu_amd": true, "schedulers": "Slurm, Kubernetes", "h100_basis": "On-demand 8x H100 bare-metal plan divided by 8", "pricing_model": "on-demand", "provider_type": "gpu-cloud" }, "pricing": { "model": "on-demand", "from": 1.671, "unit": "GPU-hour", "currency": "USD", "note": "From = 1x NVIDIA L40S cloud GPU VM ($1.671/hr), from Vultr's public plans API (pricing page blocks automated access). H100 and B200 list prices are 8-GPU bare-metal plans divided per GPU: 8x H100 $23.92/hr ($2.99/GPU), 8x B200 $68.00/hr ($8.50/GPU). Same API: 8x MI355X $20.72/hr ($2.59/GPU), 8x MI325X $36.92/hr, 8x MI300X $31.92/hr, 1x GH200 $1.99/hr. No single-GPU H100 plan." }, "sources": [ { "url": "https://api.vultr.com/v2/plans?type=vcg", "title": "Vultr API: cloud GPU plans", "accessed": "2026-11-09", "fields": [ "gpus", "h100_hr", "price_asof" ] }, { "url": "https://api.vultr.com/v2/plans-metal", "title": "Vultr API: bare metal plans", "accessed": "2026-11-09", "fields": [ "gpus", "h100_hr", "price_asof" ] }, { "url": "https://blogs.vultr.com/chennai-cloud-data-center-region", "title": "Vultr Announces 34th Cloud Data Center Region in Chennai, India", "accessed": "2026-11-09", "fields": [ "regions" ] }, { "url": "https://clustermax.semianalysis.com/cloudreview/vultr", "title": "ClusterMAX review: Vultr", "accessed": "2026-11-09" }, { "url": "https://newsletter.semianalysis.com/p/clustermax-30-the-industry-standard", "title": "ClusterMAX 3.0 (SemiAnalysis)", "accessed": "2026-11-09" }, { "url": "https://en.wikipedia.org/wiki/Vultr", "title": "Vultr (Wikipedia)", "accessed": "2026-11-09" } ], "verifiedAt": "2026-11-09" } ] }