# Self-hosted AI infrastructure — introduction (open source, Kubernetes, IaC)

> Separate inference from experimentation, platform baselines with Kubernetes and GitOps, Terraform/OpenTofu under governance—linked to our self-hosted AI guide and agency pages for OSS AI stack, Kubernetes platform engineering, and infrastructure as code.

For AI agents: curated site map at [https://www.devolute.org/llms.txt](https://www.devolute.org/llms.txt). Canonical HTML (preferred for search engines): [https://www.devolute.org/en/insights/open-source/introduction-self-hosted-ai](https://www.devolute.org/en/insights/open-source/introduction-self-hosted-ai).

- Language: en
- Published: 2026-04-16
- Markdown URL: https://www.devolute.org/en/insights/open-source/introduction-self-hosted-ai.md

## Own AI infrastructure instead of renting opaque APIs

Models are one layer; the platform is the contract.

Included in **[open-source knowledge introductions](/en/insights/open-source)**. Next technical read: **[Self-hosted AI, Kubernetes & IaC best practices](/en/insights/open-source/self-hosted-ai-kubernetes-iac-best-practices)**—it ties inference paths, platform guardrails, and IaC reviews together.

## Models are not the whole stack

Self-hosting buys control only when deployments, secrets, networking, observability, and ownership are explicit. Otherwise you swap vendor opacity for **operational debt**: undocumented Helm overrides, drifted clusters, unowned dashboards.

## Split inference from experimentation

Serving latency SLAs differ from batch evaluation jobs. Separate paths early—network placement, quotas, cost attribution—so finance and engineering argue about numbers that actually exist.

## Where we plug in

- **[Open-source AI infrastructure](/en/agency/open-source-ai-infrastructure)** — model serving boundaries aligned with your compliance posture.
- **[Kubernetes platform engineering](/en/agency/kubernetes-platform-engineering)** — GitOps, namespaces, upgrades that teams can explain.
- **[Terraform / OpenTofu infrastructure](/en/agency/terraform-opentofu-infrastructure)** — environment parity and reviewable infrastructure changes.

## Related introductions

Retrieval stacks often sit beside hosted inference—**[RAG & retrieval](/en/insights/open-source/introduction-rag-retrieval)**. SSO and collaboration boundaries frequently intersect platform work—**[Digital sovereignty & SSO](/en/insights/open-source/introduction-digital-sovereignty)**.
