Connected
infrastructure
for AI companies

Datum is a venture-backed startup founded in late 2024 by Zac Smith (ex-Equinix, Packet) and Jacob Smith, alongside a founding team from Packet, Voxel, SoftLayer, StackPath, Highwinds, and Zscaler. It provides a network cloud built for agents, targeted at AI companies and alt cloud providers, backed by Amplify Partners, CRV, Cervin Ventures, Encoded Ventures, Ex/Ante, Step Function, and Vine Ventures.

Datum's mission is to help the next 1k clouds by unlocking internet superpowers for every agent and builder. The internet wasn't designed for AI and agents; we're on a mission to fix that. In the future we're building, intelligence is private by default, platforms can scale anywhere, and companies don't have to choose between velocity and control.

Datum was founded on the basis of three trends: First, that software (and its data) was going everywhere. Next, that a new class of providers was being born. And finally, that the 'splinternet' had arrived.

In the future, we believe all companies will need to curate and control how they interact with providers, partners, customers and agents. Datum exists to help them do it.

Datum is designed for Alt Cloud providers (inference clouds, sandboxes, and frontier labs), ISVs and SaaS providers embedding connectivity capabilities, AI-native teams building agent workflows, and tech-forward enterprises that are building for the future.

See alt-cloud.org for a list of what we mean by "Alt Cloud" providers.

Infrastructure for the next 1k clouds. Datum is optimized for agents first and developers next.

See datum.net/use-cases for details.

Datum replaces legacy CDNs, DIY networks, and fragmented infrastructure with a network cloud built for agents — faster, cheaper, and more secure. It combines powerful primitives, global reach, and a neutral strategy, backed by open source. Available as multi-tenant public cloud, as single-tenant Dedicated Cloud (GPU, CPU, and network infrastructure for scaling providers and enterprises who want full control, custom locations, and ideal economics), or as open source (AGPLv3). The control plane is built for both engineers and agents.

"Datum is a modern network-focused cloud with a Kubernetes-style API and first-class CLI. Use Datum in place of a similar platform like Cloudflare to develop, deploy, and protect applications."

A modern edge cloud, backed by open source: a virtual backbone that goes everywhere. Datum helps agents and builders run workloads at the edge and securely connect everything from doorbells to databases, with a consistent runtime designed for Kubernetes platform engineers and AI builders.

Galactic VPC is a free virtual backbone you can start in about 60 seconds. Like a public cloud VPC, but everywhere: it spans clouds and providers, is built with SRv6, and is backed by open source. Attach resources from your desktop to your database — Datum microVMs, ALBs, S3 buckets, Tailscale tailnets, a Mac Mini on your desk, workloads in AWS or GCP, agentic sandboxes at Daytona or Modal, colocation, or GPU infrastructure from a neocloud. Spin up one, dozens, or thousands: agents need VPCs, too.

Datum operates high-performance physical compute and network infrastructure at the key interconnection points around the world. https://www.datum.net/locations/

Datum's control plane is built on the Kubernetes apiserver: resources are declared and reconciled. Agents apply a desired state and read it back with get — the same verbs across compute, networking, DNS, secrets, and IAM. apply has conflict management, so actions are safe to retry.

datumctl uses kubectl as a library, so the familiar verbs behave as expected: api-resources, get, describe, apply, delete, diff, explain. The API is self-describing, so an agent can discover the whole surface without external docs:

• datumctl api-resources — list every resource type.
• datumctl explain <resource>[.<field.path>] — schema and field docs, drilling into any path (e.g. datumctl explain httpproxies.spec).
• datumctl diff — preview what an apply would change before committing.

See Agent Recipes for copy-pasteable task flows built on these verbs.

Usage-based pricing with no seat fees. Datum Cloud is free until billing begins on February 1st, 2027. A credit card is required at signup to keep the platform stable by heading off fraud and abuse. You are notified before billing starts and get a grace period: no unexpected bills, no hidden fees, no per-seat or per-user licensing. If you prefer a demo, just reach out.

Galactic VPC (gVPC) — a virtual backbone that spans providers. Create as many as you want and pay only for data transfer.

Hosted resources

Always free

• datumctl (CLI)
• Coarse-grained permissions and team invites
• Secrets
• Metrics Export
• Audit & Activity Logs
• SSO (passkey, Google, GitHub)
• Service Accounts
• Datum MCP
• AI Assistant (Patch)
• Domains
• Desktop Apps

See datum.net/pricing for the latest rates.

Recent datumctl work supports agentic use:

• Local API proxy (v0.18.0) — datumctl api proxy serves the authenticated API on 127.0.0.1; point any HTTP client at it with credentials attached and auto-refreshed, no token handling.
• Published agent skill (#225) — documents auth patterns, context requirements, and when to use datumctl vs. the broader skills repo, so agents drive the CLI correctly.
• datumctl ai + MCP server mode (v0.15.0) — plain-English resource management that confirms before mutating; also exposable as an MCP server.
• Plugin ecosystem (v0.15.0 → v0.16.0) — datumctl plugin install <name> adds commands (e.g. datumctl compute deploy). Bring-your-own catalogs via datumctl plugin index add; plugin search/browse/install span all catalogs with official-vs-third-party badges and @version pinning. Docs: https://www.datum.net/docs/datumctl.
• Headless login (#211) — datumctl auth login --no-browser gives a device code + link for SSH sessions, VMs, and containers.
• Isolated multi-environment logins (v0.17.0) — sessions stored per environment; datumctl auth switch shows the active account, and refresh/logout stay scoped to the right one.

Datum is backed by open source and tracks all feature work via GitHub enhancements organized into milestones. Currently planned:

• Galactic VPC — bridge cloud environments over a unified private backbone.
• Connectors — manage tunnels and connections into a private gVPC network.
• GSLB — intelligent cross-region traffic routing for availability and latency.
• Edge Apps — deploy full-stack open source and commercial apps at the network edge.
• Object Storage — highly available, scalable S3-compatible storage with zero egress fees (coming soon).
• Interconnects — dedicated, high-speed private connections to AWS and GCP (coming soon).