Serverless GPU Icon

Serverless GPU

Simplify and optimize the use of GPUs accessing your resources remotely from any Kubernetes cluster.
With Serverless GPU, you accelerate the go-to-market of your AI applications by extending available computing power without worrying about servers.
With our GPU servers, fully integrable with any existing IT cluster, you can process your development activities more rapidly and cost-effectively.

SERVERLESS H100

Graphic Processor

NVIDIA H100

Number of GPUs

4

GPU RAM

320 GB

RuntimeClass

seeweb-nvidia-4xh100

Hourly Cost

2.31
?
This order allows you to have at your disposal a cloud platform for creating and destroying all the Serverless GPU you need, even for just one hour
SERVERLESS A100

Graphic Processor

NVIDIA A100

Number of GPUs

GPU RAM

80 GB

RuntimeClass

seeweb-nvidia-1xa100

Hourly Cost

2.25
?
This order allows you to have at your disposal a cloud platform for creating and destroying all the Serverless GPU you need, even for just one hour
SERVERLESS L40S

Graphic Processor

NVIDIA L40S

Number of GPUs

GPU RAM

48 GB

RuntimeClass

seeweb-nvidia-1xl40s

Hourly Cost

0.94
?
This order allows you to have at your disposal a cloud platform for creating and destroying all the Serverless GPU you need, even for just one hour
SERVERLESS RTX A6000

Graphic Processor

NVIDIA Quadro RTX A6000

Number of GPUs

GPU RAM

48 GB

RuntimeClass

seeweb-nvidia-1xa6000

Hourly Cost

0.81
?
This order allows you to have at your disposal a cloud platform for creating and destroying all the Serverless GPU you need, even for just one hour
SERVERLESS L4

Graphic Processor

NVIDIA L4

Number of GPUs

GPU RAM

24 GB

RuntimeClass

seeweb-nvidia-1xl4

Hourly Cost

0.42
?
This order allows you to have at your disposal a cloud platform for creating and destroying all the Serverless GPU you need, even for just one hour

Optimize your inference tasks and easily meet the power needs of your AI workloads with serverless technology.

Without worrying about the underlying infrastructure, you can access cloud resources dedicated to you and your development team, seeing them as if they were local to your cluster, with maximum performance thanks to the powerful GPUs available.

Complete the provisioning of multiple graphics cards and integrate them with any cloud or on-premise environment already running.

  • Serverless GPU: What are the use cases?

    Serverless GPU is the Seeweb cloud-native service that enables you to leverage Kubernetes in AI contexts.

    Your AI applications and inference tasks will be accelerated by the easy and pay-as-you-go provisioning of new cloud GPUs, available within seconds.

    Serverless GPU is ideal for software development teams already using AI infrastructures who need to increase computing power without migrating pipelines or starting over.

    Developed according to the standards of the Microsoft Virtual Kubelet project, Serverless GPU guarantees maximum compatibility with any Kubernetes cluster.

  • How you can use Serverless GPU

    The requirement for using our GPU cloud-native service is the use of Kubernetes.

    To get started, you can download and install the open-source k8sGPU agent to create a virtual node inside your cluster, acting as a bridge to your remote GPUs.

    Once the virtual node is ready, schedule your ML/AI pods as you normally would with any worker node. The k8sGPU agent will manage the remote graphics card provisioning dynamically, executing the pods as if they were local to your cluster.

  • Which Kubernetes distributions are supported?

    Serverless GPU supports most public managed Kubernetes services, such as AKS, EKS, and GKE, as well as on-premise environments like vanilla Kubernetes, OpenShift, Tanzu, and Rancher distributions.

  • How does the billing service work?

    In line with the agile philosophy of serverless technologies, characterized by flexible billing models, your remote, serverless GPUs will be provided on demand. Scale and reduce your resources dynamically, paying only for actual hourly usage.

    Once activated, you will receive periodic reports on actual GPU usage, and real consumption will always be under your control.