GPU Marketplace
NVIDIA L40SOn-Demand
Estimates only — rates are collected automatically from public provider pages and may be out of date. Prices vary by region, commitment term, and availability, and typically exclude storage, egress, and tax. Confirm current pricing with the provider before purchasing. Last collected .
Company Profile
Legal EntityIBM Cloud®
Parent CompanyIBM
Infrastructure
GPU FleetNVIDIA B300, Intel Gaudi 3, AMD Instinct MI300X, NVIDIA A100, NVIDIA L40s, NVIDIA Blackwell Ultra
Bare MetalBare Metal on VPC
enterprise clients
Compute & Deployment
On-DemandGet on-demand access to a variety of GPUs and AI accelerators, including NVIDIA B300, Intel Gaudi 3 and AMD MI300X.
Spot / InterruptibleSpot instances are highly discounted versions of the standard instances. They are designed to use available compute resources for interruptible or stateless workloads. Spot instances can be preempted (or evicted) at any time.
Bare MetalStorage dense x2de and x3de profiles for Bare Metal on VPC (select availability)
VM-BasedVirtual Servers for VPC
Container-Basedcontainer management tools from Red Hat OpenShift and IBM Code Engine
Terraformsupports creation with Terraform
GPU Hardware
Latest GenGet on-demand access to a variety of GPUs and AI accelerators, including NVIDIA B300, Intel Gaudi 3 and AMD MI300X.
Multi-GPU NodesWhen you use the B300 virtual server profile it is powered by the NVIDIA HGX B300{: external} server and provides high-performance GPU capabilities for AI workloads such as fine tuning and large-scale training.
PCIe vs SXMGPU A100 PCIe virtual server profiles now available (select availability) : The A100 PCIe GPU virtual server profiles are now available in the Washington DC (us-east), Frankfurt (eu-de), London (eu-gb), and Tokyo (jp-tok) regions.
HGX PlatformThe Hopper-1 network profile supports both H100 and H200 instance profiles.
Pricing Model
SubscriptionGet on-demand access to a variety of GPUs and AI accelerators
Hidden Feeshelp you overcome AI's hidden costs and technical hurdles
Performance & Scaling
Multi-Node Trainingprovides high-performance GPU capabilities for AI workloads such as fine tuning and large-scale training.
Perf IsolationIt provides isolated networks for Hopper HGX instances running workloads that require high-bandwidth, low-latency interconnectivity, such as AI training and large-scale simulations.
Developer Experience
OnboardingApply promo code GPU4YOU at checkout; some offerings require creating a support case; accounts with special access required for select availability
CLI ToolingIBM Cloud VPC CLI; operations supported via console, CLI, API, or Terraform
DocumentationRelease notes; API change log; CLI release notes; product and profile documentation (GPU profiles, Accelerated instance profiles, Block Storage docs); console/CLI/API/Terraform usage instructions
Data Center Locations
Coverage
CitiesDallas, Chennai - Airtel, Frankfurt, London, Madrid, Osaka, Sydney, Tokyo, Toronto, Washington, DC, Sao Paulo, Mumbai, Chennai
DallasChennai - AirtelFrankfurtLondonMadridOsakaSydneyTokyoToronto, and WashingtonDC
Compliance Regions
India RegionMumbai - Airtel (`in-mum`),Chennai (`in-che` / `in-che-1`)
Datacenter Locations
Key Strengths
Enterprise focus on security and compliance
customer-controlled encryption keys
broad GPU/accelerator selection and on-demand access
open architecture supporting Red Hat OpenShift and IBM watsonx
full stack server/storage/network offerings
built-in workload protection and automated compliance tools
Known Limitations
Many GPU/profiles are 'select availability' or require special access
some offerings require creating a support case to purchase
spot instances are interruptible/preemptible
instance profiles are only available in certain regions and zones
Additional Information
Core Proposition
The power of choice for your AI deployments
Notable Customers
Harvard University
US Open (United States Tennis Association)
Happiest Minds
Last updated August 2026. Information subject to change.