Adoption & Ecosystem

Who uses it

The organisations below are listed in the project's ADOPTERS.MD; the cloud providers and a few others are also named in the CNCF graduation announcement.

Organisation	Use case	Source
Alibaba Cloud	Cloud platform	CNCF announcement
Scaleway	Cloud services	CNCF announcement
Gojek	CaraML MLOps platform	CNCF announcement
Red Hat	OpenShift Serverless	ADOPTERS.MD
IBM	Cloud Code Engine	ADOPTERS.MD
Google Cloud	Cloud Run for Anthos	ADOPTERS.MD
VMware	Tanzu Application Platform, Event Broker	ADOPTERS.MD
Bloomberg L.P.	Data science platform	ADOPTERS.MD
Box	Internal serverless PaaS	ADOPTERS.MD
KA-NABELL	E-commerce platform	ADOPTERS.MD
KubeSphere	OpenFunction serving layer	ADOPTERS.MD

ADOPTERS.MD also lists Cisco, Cloudera, Optum, WP Engine, Sinch, Tata Communications, Telekom Deutschland with SVA, and AI-inference users such as Cerebrium, Run:ai, Runhouse, and deepc.ai.

Adoption signals

• knative/serving: 6,063 stars, 1,227 forks, created 2018-01-24 (GitHub API, observed 2026-06-23).
• Contributor scale is roughly 300 or more, based on the GitHub contributors API paginating to page 336 at one contributor per page including anonymous entries (observed 2026-06-23).
• For comparison, knative/eventing had 1,550 stars on the same date; Serving is the repository where stars concentrate.
• CNCF Graduated since 2025-09-11 (CNCF project page).

Ecosystem

• Networking layers: Istio, Contour, Kourier, and Gateway API implementations provide ingress.
• cert-manager integrates through the optional certificate reconciler (cmd/controller/main.go:80).
• OpenTelemetry is the stated direction for metrics and tracing (CNCF announcement).
• knative/eventing with CloudEvents covers the event-driven axis alongside Serving.
• OpenFunction (KubeSphere) builds a FaaS layer on top of Knative Serving (ADOPTERS.MD).

Alternatives

Knative scales on live request concurrency or RPS measured by the queue-proxy, and uses the activator to buffer requests so a Revision can go to and from zero. The alternatives differ mainly in what signal they scale on.

Alternative	Differs by
KEDA	Scales on external event sources and queue depth; pairs with the HPA to reach zero (ThinhDA)
Kubernetes HPA	Scales on CPU and memory; alone it goes 1 to N, not to zero (ThinhDA)
OpenFaaS	RPS-based, adding one instance at a time; lighter-weight FaaS (CNCF blog)
OpenWhisk, Fission, Fn	Adjacent FaaS platforms with their own runtimes (CNCF blog)