HOSTED·AI PLATFORM
Complete AI hosting + GPU cloud stack
Sell multi-tenant GPUaaS for AI model
training · tuning · inference
Designed for service providers, hosted·ai software-defines the GPU, delivering 5x more profit per card vs. GPU passthrough
Turnkey solution for high-margin GPU cloud
Deploys to bare metal in 24h, full stack or standalone
Elastic GPUaaS, bare metal GPU, GPU + VM provisioning
Full suite of orchestration + monetization tools
Secure multi-tenant GPU sharing for max utilization
GPU overcommit for maximum margins
Sell on-premises GPU or wholesale GPU via GPU Mesh
White label UI, REST API, billing + integrations
One-click AI models and apps
With AI model library integration, Ansible playbooks and Bring Your Own Model, your customers can deploy AI instances in seconds.
Software-Defined GPU
Sell GPUaaS with the same dynamic resource allocation, autoscaling and hardware utilization (overcommit) as traditional hosting or cloud.
GPU pooling
Assign GPUs in a cluster to pools. Configure pricing, access and sharing policies to suit different customers & workloads.
100% GPU utilization
hosted·ai distributes workloads to GPU pools dynamically, to maximize utilization of each card.
PAYG GPUaaS
Bill users for the GPU VRAM/TFLOPs consumed, as well as billing for fixed resources or cards.
GPUaaS overcommit
Configure share ratios to enable GPU overselling, and increase margins 5x or more.
Secure multi-tenant GPUaaS
hosted·ai brings full multi-tenancy to GPUs. Serve more customers with less investment in GPU infrastructure, and change the cost/margin equation for GPU cloud.
Isolated GPU sharing
Many users run workloads across a pool of GPUs at the same time. User tasks are isolated and secured from other users.
Transparent to users
Users see one or more virtual GPUs, even when their tasks execute on the same physical GPU as other user tasks.
Elastic GPU resources
Instead of allocating entire GPUs or instances to users, GPU resources (TFLOPS, VRAM) can be provisioned to users from GPU pools.
Dynamic scheduling
Adaptive scheduling optimizes utilization for multiple processes sharing a GPU. Users pay for the GPU resources they consume.
More hosted·ai GPUaaS features
Monetization
Flexible pricing and packaging policies make it easy to bill for your hosted·ai cloud.
GPU
Bill users for the GPU VRAM/TFLOPs consumed, as well as billing for fixed GPU resources, GPU + VM, or bare metal GPU servers
CPU, storage, network
Bill for vCPU cores, storage capacity, bandwidth
Packages
Combine resources into easy-to-consume packages (CPU, GPU, storage, network)
Applications
Combine applications with resources and bill for one-click installs
Regions
Set global or local prices for different datacenter locations
Team + quota management
It’s easy to sell GPUaaS in a way that fits with customer processes and workflows.
Resource pools are allocated to customer teams – teams have their own workspaces, users and permissions/limits
For example, a team might consist of an admin and different developers, each with their own rights and access to resources
Integrations
hosted·ai has a full REST API and integrates with a growing range of service provider billing and customer management systems, including WHMCS
Flexible deployment
hosted·ai can be deployed as a full hyperconverged stack for VMs and GPU/GPUaaS, or as a standalone GPUaaS solution that integrates with your existing cloud or virtualization platform.
Simple, Transparent Pricing
Based on VRAM managed by the platform and consumed by your customers
No upfront licensing fees
24×7 support included