Skip to content

datum-cloud/datum

Datum is building the internet for AI.

Datum Cloud - Docs - Community Discord - Blog - LinkedIn

Why Datum?

We believe that AI is changing everything — not just how we work and create, but how quickly new applications, agents, and even clouds are being built. Digital leaders today must orchestrate a complex, fragmented web of clouds, specialty providers, customers, and data.

The Internet is built on data center meet-me rooms, where telco providers and hyperscaler clouds talk to each other over real physical cables, called cross-connects. A new connection takes days or weeks of humans moving things around to set up.

We believe the next era of the Internet is already here, and it's growing fast. In the alt-cloud universe, you don't think about virtual machines and VPCs, you think about services. You connect your Vercel app with your Supabase instance all wired up with your Kestra workflow, monitored by your Resolve SRE agent. There isn't a switch or routing table in sight. It's just virtual plumbing to make your business go.

With Datum, cloud and AI-native builders can use the tools they love (like Cursor or a Kubernetes native CLI) to access the internet superpowers that today’s tech giants leverage at scale: authoritative DNS, distributed proxies, global backbones, deterministic routing, cloud on-ramps, and private interconnection.

That's why we're building Datum: to help build 1k clouds in the age of AI.

So what is Datum?

Fully programmable and AI-native

  • Developer- and agent-friendly protocols, interfaces, and workflows
  • Programmatic interconnection between providers and services
  • Security through network-level encryption that's built-in and impossible to break or disable
  • Built using the "operating system for AI" Kubernetes API patterns for operator tooling and familiarity (kubectl, Helm, etc.)

Neutral by design

  • No allegiance to a single cloud, vendor, or region
  • Operates as a trusted, independent layer for alt clouds, incumbents, and digital-first enterprises

Bring your own infra

  • Use Datum’s cloud control plane along with its global network and distributed compute
  • Or run components in your own cloud or infra (e.g., GCP, AWS, NetActuate, Vultr, etc.)

Maximum flexibility

  • Feed full telemetry to your preferred tools (Grafana Cloud, Honeycomb, Datadog, etc.)
  • Support for policy enforcement via SRv6

Some of our favorite features

Declarative management

Datum works just like Kubernetes, because it is Kubernetes. Define your desired infrastructure state and our components reconcile the living system to match. No more syncing or drift.

The Datum control plane is natively compatible with tooling from the Kubernetes ecosystem. Datum APIs are defined as Custom Resources, and resources are managed by operators which can be deployed into any Kubernetes cluster.

Use the tools you're familiar with - kubectl for interacting with API resources via the CLI, kustomize or terraform for configuration management, or any other tool compatible with the Kubernetes API.

Workloads

The Workload resource provides a provider-agnostic way to manage groups of compute instances (VMs or containers). Define instance templates, placement rules (where instances should run across locations/providers), scaling behavior, network attachments, and volume mounts. The responsible infrastructure provider operator handles the provisioning.

Gateways

Leveraging the standard Kubernetes Gateway API (GatewayClass, Gateway,HTTPRoute, etc.), Datum allows you to define how external or internal traffic should connect to your services. Manage TLS certificates, configure routing logic, and control network ingress/egress across the Datum infrastructure.

Components

Datum API server

We deploy a Datum variant of the Kubernetes api-server in the style of the generic control plane (KEP-4080) so that we can leverage the vast ecosystem of libraries and tooling. There is no need to design a bespoke, infrastructure-focused distributed system for you to learn; Kubernetes has the primitives to support it. While the standard api-server operates normally for the cluster itself (think Pods and Deployments), Datum's api-server handles Datum-specific resources like Network and Workload.

Manages networking primitives like Datum VPC Networks (Network,NetworkContext), Subnets (SubnetClaim, Subnet), IP Address Management(IPAM), and network policy concepts (NetworkBinding, NetworkPolicy).

Manages the lifecycle of Workload resources, handling placement logic and the creation of compute instances (WorkloadDeployment, Instance) via infrastructure providers. See the Workloads RFC for design details.

Plugins

Datum Plugins interpret resource definitions such as Workloads and Networks to drive the management of provider specific resources such as Virtual Machines and VPC Networks to meet the declared expectations. Our first example is for Google Cloud Platform (GCP). Supported features include:

  • Deploying Virtual Machine based workload instances with OS images provided via an image library.
  • Deploying sandboxed container based workload instances with any OCI compliant container image.
  • VPC connectivity and IPAM.
  • Attaching instances to one or more networks.

Get Started

The easiest way to understand Datum is to try it! Head over to Datum Cloud, sign up, and follow the Quickstart Guide to begin your journey to a reimagined world of interconnection.

We hope that you will then come and build with us:

License

Datum is primarily licensed under the AGPL v3.0. Specific components mayhave different licenses; please check individual repositories for details.

About

An open network cloud that you can run anywhere, backed by open source.

Topics

Resources

License

Code of conduct

Contributing

Security policy

Stars

Watchers

Forks

Packages

 
 
 

Contributors 14